Mi RNomics

Methods in
Molecular Biology 2257
Jens Allmer
Malik Yousef Editors
miRNomics
MicroRNA Biology and
Computational Analysis
Second Edition
METHODS IN MOLECULAR BIOLOGY
Series Editor
John M. Walker
School of Life and Medical Sciences
University of Hertfordshire
Hatfield, Hertfordshire, UK
For further volumes:

http://www.springer.com/series/7651
For over 35 years, biological scientists have come to rely on the research protocols and
methodologies in the critically acclaimed Methods in Molecular Biology series. The series was
the first to introduce the step-by-step protocols approach that has become the standard in all
biomedical protocol publishing. Each protocol is provided in readily-reproducible step-by-
step fashion, opening with an introductory overview, a list of the materials and reagents
needed to complete the experiment, and followed by a detailed procedure that is supported
with a helpful notes section offering tips and tricks of the trade as well as troubleshooting
advice. These hallmark features were introduced by series editor Dr. John Walker and
constitute the key ingredient in each and every volume of the Methods in Molecular Biology
series. Tested and trusted, comprehensive and reliable, all protocols from the series are
indexed in PubMed.
miRNomics
MicroRNA Biology and Computational Analysis
Second Edition
Edited by
Jens Allmer
Medical Informatics and Bioinformatics, Institute for Measurement Engineering and Sensor Technology,
Hochschule Ruhr West, University of Applied Sciences, Mülheim adR, Germany
Malik Yousef
Department of Information System, Galilee Digital Health Research Center (GDH), Zefat Academic College,
Zefat, Israel
Editors
Jens Allmer Malik Yousef
Medical Informatics and Department of Information System, Galilee Digital Health
Bioinformatics, Institute for Research Center (GDH)
Measurement Engineering and Sensor Zefat Academic College
Technology, Hochschule Ruhr West, Zefat, Israel
University of Applied Sciences
Mülheim adR, Germany
ISSN 1064-3745 ISSN 1940-6029 (electronic)

Methods in Molecular Biology
ISBN 978-1-0716-1169-2 ISBN 978-1-0716-1170-8 (eBook)
https://doi.org/10.1007/978-1-0716-1170-8
© Springer Science+Business Media, LLC, part of Springer Nature 2022
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations
and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to
be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty,
expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been
made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Cover image credit: PEDRO GABRIEL NACHTIGALL l Laboratorio Especial de Toxinologia Aplicada, CeTICS,
202 Instituto Butantan, São Paulo, SP, Brazil)
This Humana imprint is published by the registered company Springer Science+Business Media, LLC, part of Springer
Nature.
The registered company address is: 1 New York Plaza, New York, NY 10004, U.S.A.
Dedication
JA dedicates the book to his wife Açalya, his son Lukas Aren, and his daughter Maya Izabel
who was born during the editing of this book.
MY dedicates the book to his wife Sahar, his daughters Mair and Umaymha, and his sons
Anas and Mohammad.
v
Preface
MicroRNAs (miRNAs) are regulators of gene expression and thereby directly involved in
cell homeostasis. Therefore, some part of the book was dedicated to miRNA involvement in
disease and stress response. Dysregulation of miRNAs can lead to disease, which makes these
small molecules interesting biomarkers. On the other hand, miRNAs can be therapeutic
targets via addition/upregulation of cellular miRNAs, administering of synthetic miRNA
mimics, or by downregulating their effect using miRNA decoys.
Since the publication of the first volume in 2014, much has happened concerning the
experimental detection of miRNAs. We dedicated some space to miRNA targeting and the
experimental as well as computational detection of functional interactions. However, spa-
tiotemporal expression patterns of miRNAs make any attempt of determining all functional
interactions in any higher eukaryote a futile endeavor. Therefore, computational detection
of miRNAs, their direct prediction from genomes, and the prediction of their targets remain
important tasks in miRNomics.
Some topics in miRNomics have not evolved much since the miRNomics volume one
while some novel areas of research emerged such as focus on functional interactions.
Nonetheless, all chapters of miRNomics volume two are not just updates of the previous
edition but are novel works. The original idea of creating an overview of the field of
miRNomics with the first volume remained unchanged for this second edition. Even more
prominent than in the first book, each topic addressed in this second volume would deserve
its own textbook. We aim to provide an overview to facilitate entering the complex realm of
miRNomics. The final chapter (Chapter 19) then offers experimental and computational
challenges that would greatly benefit the current body of knowledge in miRNomics once
solved.
The first part of the book introduces the molecular biology involved in miRNomics and
experimental procedures for their further investigation. This is followed by computational
methods for miRNomics including evolutionary considerations. Finally, the outcomes of
miRNA regulation in stress and disease as well as their use as biomarkers and therapeutics are
discussed. For researchers in the field of miRNomics, suitable chapters will be found
throughout the book. For scientists entering the field of miRNomics, this book can be
read cover to cover.
We are very happy that some of the authors of chapters in the first edition of miRNomics
agreed to contribute new works. Others declined due to retirement and change of fields
while former co-authors became lead authors which nicely models the dynamic in the field of
miRNomics. One of the editors also went through many changes during the editing of this
book. Thereby, the process prolonged significantly. We would like to thank the chapter
authors for their patience and their dedication to finalizing this book. We would like to
extend our thanks to our anonymous reviewers and to the authors who helped in reviewing.
Finally, we want to thank the professional editors at Methods in Molecular Biology, first for
inviting us to edit a new volume on miRNomics and second for their patience during the
long editing process.
Mülheim adR, Germany Jens Allmer

Zefat, Israel Malik Yousef
vii
Contents
Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
1 Regulation of MicroRNAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Kemal Ergin and Rahmi Çetinkaya
2 Experimental MicroRNA Detection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Bilge Yaylak and Bünyamin Akgül
3 Functional Annotation of MicroRNAs Using Existing Resources . . . . . . . . . . . . . 57
Harsh Dweep, Louise C. Showe, and Andrew V. Kossenkov
4 Experimental MicroRNA Targeting Validation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Bala Gür Dedeoğlu and Senem Noyan
5 Endogenous miRNA Sponges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
Ayşe Hale Alkan and Bünyamin Akgül
6 MicroRNA Targeting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Hossein Ghanbarian, Mehmet Taha Yıldız, and Yusuf Tutar
7 MicroRNA Databases and Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Tharcı́sio Soares de Amorim, Daniel Longhi Fernandes Pedro,
and Alexandre Rossi Paschoal
8 Computational Detection of Pre-microRNAs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
Müşerref Duygu Saçar Demirci
9 Computational Methods for Predicting Mature microRNAs. . . . . . . . . . . . . . . . . . 175
Malik Yousef, Alisha Parveen, and Abhishek Kumar
10 Computational Detection of MicroRNA Targets. . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
Pedro Gabriel Nachtigall and Luiz Augusto Bovolenta
11 Evolution and Phylogeny of MicroRNAs — Protocols, Pitfalls,
and Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Cristian A. Velandia-Huerto, Ali M. Yazbeck, Jana Schor,
and Peter F. Stadler
12 Ensemble Classifiers for Multiclass MicroRNA Classification . . . . . . . . . . . . . . . . . 235
Luise Odenthal, Jens Allmer, and Malik Yousef
13 MicroRNAs in Genetic Etiology of Human Diseases . . . . . . . . . . . . . . . . . . . . . . . . 255
Melis Olcum, Kemal Ugur Tufekci, and Sermin Genc
14 Role of Exosomal MicroRNAs in Cell-to-Cell Communication . . . . . . . . . . . . . . . 269
Bora Tastan, Emre Tarakcioglu, Yelda Birinci, Yongsoo Park,
and Sermin Genc
15 MicroRNAs and Heat Shock Proteins in Breast Cancer Biology . . . . . . . . . . . . . . 293
Mehmet Taha Yildiz, Lütfi Tutar, Nazlı Irmak Giritlioğlu,
Banu Bayram, and Yusuf Tutar
ix
x Contents
16 Role of MicroRNAs in Extreme Animal Survival Strategies. . . . . . . . . . . . . . . . . . . 311

Hanane Hadj-Moussa, Liam J. Hawkins, and Kenneth B. Storey
17 Computational and Bioinformatics Methods for MicroRNA
Gene Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Ege Riza Karagur, Sakir Akgun, and Hakan Akca
18 The Role of MiRNA in Cancer: Pathogenesis, Diagnosis, and Treatment . . . . . . 375
Erez Uzuner, Gizem Tugçe Ulu, Sevim Beyza Gürler,
and Yusuf Baran
19 44 Current Challenges in miRNomics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
Bünyamin Akgül, Peter F. Stadler, Liam J. Hawkins,
Hanane Hadj-Moussa, Kenneth B. Storey, Kemal Ergin, Rahmi Çetinkaya,
Alexandre R. Paschoal, Pedro G. Nachtigall, Yusuf Tutar, Malik Yousef,
and Jens Allmer
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
Contributors
HAKAN AKCA • Department of Medical Biology, School of Medicine, Pamukkale University,

Denizli, Turkey
BÜNYAMIN AKGÜL • Molecular Biology and Genetics, Izmir Institute of Technology,
Urla, Izmir, Turkey
SAKIR AKGUN • Department of Medical Biology, School of Medicine, Kafkas University,
Kars, Turkey
AYŞE HALE ALKAN • Molecular Biology and Genetics, Izmir Institute of Technology,
Izmir, Turkey
JENS ALLMER • Medical Informatics and Bioinformatics, Institute for Measurement
Engineering and Sensor Technology, Hochschule Ruhr West, University of Applied Sciences,
Mülheim adR, Germany
YUSUF BARAN • Molecular Biology and Genetics, Izmir Institute of Technology, Izmir, Turkey
BANU BAYRAM • Department of Nutrition and Dietetics, Hamidiye Faculty of Health
Sciences, University of Health Sciences, Istanbul, Turkey
YELDA BIRINCI • Department of Molecular Biology and Genetics, Science Faculty, Koç
University, Istanbul, Turkey
LUIZ AUGUSTO BOVOLENTA • Department of Morphology, Institute of Biosciences of Botucatu
(IBB), São Paulo State University (UNESP), Botucatu, Brazil
RAHMI ÇETINKAYA • Department of Basic Oncology and Cancer Biology, Institute of Health
Sciences, Aydın Adnan Menderes University, Aydın, Turkey
THARCÍSIO SOARES DE AMORIM • Department of Computer Science and Bioinformatics and
Pattern Recognition Group, Universidade Tecnologica Federal do Paraná (UTFPR),
Cornélio Procopio, Brazil
BALA GÜR DEDEOĞLU • Biotechnology Institute, Ankara University, Ankara, Turkey
HARSH DWEEP • The Wistar Institute, Philadelphia, PA, USA
KEMAL ERGIN • Department of Histology and Embryology, Medical Faculty, Aydın Adnan
Menderes University, Aydın, Turkey; Department of Basic Oncology and Cancer Biology,
Institute of Health Sciences, Aydın Adnan Menderes University, Aydın, Turkey
SERMIN GENC • Izmir Biomedicine and Genome Center, Izmir, Turkey; Department of
Neuroscience, Institute of Health Science, University of Dokuz Eylul, Izmir, Turkey
HOSSEIN GHANBARIAN • Biotechnology Department & Cellular and Molecular Biology
Research Center, Shahid Beheshti University of Medical Sciences, Tehran, Iran
NAZLI IRMAK GIRITLIOĞLU • Department of Molecular Medicine, Hamidiye Institute of
Health Sciences, University of Health Sciences, Istanbul, Turkey
SEVIM BEYZA GÜRLER • Molecular Biology and Genetics, Izmir Institute of Technology,
Izmir, Turkey
HANANE HADJ-MOUSSA • Department of Biology, Carleton University, Ottawa,
ON, Canada
LIAM J. HAWKINS • Department of Biology, Carleton University, Ottawa, ON, Canada
EGE RIZA KARAGUR • Department of Medical Genetic, School of Medicine, Pamukkale
University, Denizli, Turkey; Department of Medical Biology, School of Medicine,
Pamukkale University, Denizli, Turkey
ANDREW V. KOSSENKOV • The Wistar Institute, Philadelphia, PA, USA
xi
xii Contributors
ABHISHEK KUMAR • Institute of Bioinformatics, Bangalore, India; Manipal Academy of

Higher Education (MAHE), Manipal, Karnataka, India
PEDRO GABRIEL NACHTIGALL • Laboratorio Especial de Toxinologia Aplicada, CeTICS,
Instituto Butantan, São Paulo, SP, Brazil
SENEM NOYAN • Biotechnology Institute, Ankara University, Ankara, Turkey
LUISE ODENTHAL • Bioinformatics/Medical Informatics, University of Bielefeld,
Bielefeld, Germany
MELIS OLCUM • Izmir Biomedicine and Genome Center, Izmir, Turkey
YONGSOO PARK • Department of Molecular Biology and Genetics, Science Faculty,
Koç University, Istanbul, Turkey
ALISHA PARVEEN • Rudolf‑Zenker Institute of Experimental Surgery, Rostock University
Medical Center, Rostock, Germany
ALEXANDRE ROSSI PASCHOAL • Department of Computer Science and Bioinformatics and
Pattern Recognition Group, Universidade Tecnologica Federal do Paraná (UTFPR),
DANIEL LONGHI FERNANDES PEDRO • Department of Computer Science and Bioinformatics
and Pattern Recognition Group, Universidade Tecnologica Federal do Paraná (UTFPR),
MÜŞERREF DUYGU SAÇAR DEMIRCI • Department of Bioinformatics, Faculty of Life and
Natural Sciences, Abdullah Gül University, Kayseri, Turkey
JANA SCHOR • Young Investigators Group Bioinformatics & Transcriptomics, Department of
Molecular Systems Biology, Helmholtz Centre for Environmental Research GmbH – UFZ,
Leipzig, Germany
LOUISE C. SHOWE • The Wistar Institute, Philadelphia, PA, USA
PETER F. STADLER • Bioinformatics Group, Department of Computer Science, and
Interdisciplinary Center for Bioinformatics, University of Leipzig, Leipzig, Germany; Max
Planck Institute for Mathematics in the Sciences, Leipzig, Germany; Institute for
Theoretical Chemistry, University of Vienna, Wien, Austria; Santa Fe Institute, Santa Fe,
NM, USA; Facultad de Ciencias, Universidad Nacional de Colombia,
Sede Bogotá, Colombia
KENNETH B. STOREY • Department of Biology, Carleton University, Ottawa, ON, Canada
EMRE TARAKCIOGLU • Izmir International Biomedicine and Genome Institute, Dokuz Eylul
University, Izmir, Turkey
BORA TASTAN • Izmir International Biomedicine and Genome Institute, Dokuz Eylul
University, Izmir, Turkey
KEMAL UGUR TUFEKCI • Izmir Biomedicine and Genome Center, Izmir, Turkey; Izmir
Democcracy University, Vocational School of Health Services, Izmir, Turkey
LÜTFI TUTAR • Department of Molecular Biology and Genetics, Faculty of Art and Sciences,
Kırşehir Ahi Evran University, Kırşehir, Turkey
YUSUF TUTAR • Division of Biochemistry, Department of Basic Pharmaceutical Sciences,
Hamidiye Faculty of Pharmacy & Division of Molecular Medicine, Hamidiye Institute of
Health Sciences, University of Health Sciences-Turkey, Istanbul, Turkey
GIZEM TUGÇE ULU • Molecular Biology and Genetics, Izmir Institute of Technology,
Izmir, Turkey
EREZ UZUNER • Molecular Biology and Genetics, Izmir Institute of Technology, Izmir, Turkey
CRISTIAN A. VELANDIA-HUERTO • Bioinformatics Group, Department of Computer Science,
and Interdisciplinary Center for Bioinformatics, University of Leipzig, Leipzig, Germany
BILGE YAYLAK • Molecular Biology and Genetics, Izmir Institute of Technology, Izmir, Turkey
Contributors xiii
ALI M. YAZBECK • Bioinformatics Group, Department of Computer Science, and

Interdisciplinary Center for Bioinformatics, University of Leipzig, Leipzig, Germany;
Doctoral School for Science and Technology, Lebanese University, Rafic Hariri University
Campus, Hadath, Lebanon
MEHMET TAHA YILDIZ • Division of Molecular Medicine, Hamidiye Institute of Health
Sciences, University of Health Sciences, Istanbul, Turkey
MALIK YOUSEF • Department of Information System, Galilee Digital Health Research Center
(GDH), Zefat Academic College, Zefat, Israel
Chapter 1
Regulation of MicroRNAs
Kemal Ergin and Rahmi Çetinkaya
Abstract
MicroRNAs are RNAs of about 18–24 nucleotides in lengths, which are found in the small noncoding RNA
class and have a crucial role in the posttranscriptional regulation of gene expression, cellular metabolic
pathways, and developmental events. These small but essential molecules are first processed by Drosha and
DGCR8 in the nucleus and then released into the cytoplasm, where they cleaved by Dicer to form the
miRNA duplex. These duplexes are bound by the Argonaute (AGO) protein to form the RNA-induced
silencing complex (RISC) in a process called RISC loading. Transcription of miRNAs, processing with
Drosha and DGCR8 in the nucleus, cleavage by Dicer, binding to AGO proteins and forming RISC are the
most critical steps in miRNA biogenesis. Additional molecules involved in biogenesis at these stages can
enhance or inhibit these processes, which can radically change the fate of the cell. Biogenesis is regulated by
many checkpoints at every step, primarily at the transcriptional level, in the nucleus, cytoplasm, with RNA
regulation, RISC loading, miRNA strand selection, RNA methylation/uridylation, and turnover rate.
Moreover, in recent years, different regulation mechanisms have been discovered in noncanonical Drosha
or Dicer-independent pathways. This chapter seeks answers to how miRNA biogenesis and function are
regulated through both canonical and non-canonical pathways.
Key words MicroRNA, Biogenesis, RNA-induced silencer complex, Argonaute protein
1 Introduction
MicroRNAs (miRNAs), first identified in Caenorhabditis elegans

and subsequently detected in many eukaryotes, including humans,
are molecules of 18 to 24 nucleotides (~22 nt) long in the small
noncoding RNA class [1]. The human genome contains 1917
annotated hairpin, and 2654 mature miRNAs according to the
latest release of the miRBase database (v22) [2]. miRNAs are
involved in the silencing of RNA and regulation of gene expression
by matching with their mRNA target site(s). It controls protein
expression at the post-transcriptional level and is useful in the
management of many metabolic pathways that are critical for the
cell [3].
miRNAs show base matching with the target region, causing
fragmentation of the target and acting as an effector on it. AGO
Jens Allmer and Malik Yousef (eds.), miRNomics: MicroRNA Biology and Computational Analysis, Methods in Molecular Biology,
vol. 2257, https://doi.org/10.1007/978-1-0716-1170-8_1, © Springer Science+Business Media, LLC, part of Springer Nature 2022
1
2 Kemal Ergin and Rahmi Çetinkaya
proteins, on the other hand, employ several molecules that provide

translational suppression, mRNA deadenylation, and degradation
of mRNA, and act as an “effector” [4]. Generally, the 30 untrans-
lated region (UTR) of mRNAs contain binding sites for miRNAs
[5]. The 50 end region of the miRNA has a vital role in recognizing
the target. This region is called “miRNA seed,” and the sequences
are highly conserved [6]. The miRNA-loaded proteins form a
complex called RISC (RNA-induced silencer complex) to provide
translational suppression. Although the mechanisms of miRNA
biogenesis and function have not been fully elucidated, miRNAs
are involved in many critical cellular processes, including develop-
ment, cellular differentiation, cell division, and cell death [7]. Dis-
ruption in miRNA regulation can affect a wide range of diseases,
including malignant diseases, rheumatoid arthritis, and develop-
mental disorders. Since miRNAs exhibit both tumor suppressor
and oncogene properties in cancer, they can serve as markers and
therapeutic targets in cancer [8–12].
Since miRNA biogenesis does not take place in a single step, the
cell can manage it at each stage. For this reason, many enzymes and
factors play a role in this regulation.
This chapter focuses on current information and open ques-
tions about miRNA regulation and biogenesis. How miRNA tran-
scription is regulated, the step-by-step processes that miRNAs
undergo in the nucleus, molecular core components involved in
this processing, and miRNAs’ journey to the cytoplasm will be
summarized. Next, it takes a look at RISC formation in the cyto-
plasm, and its effects on miRNA biogenesis/function, the interac-
tion between RNA-binding proteins, and miRNA processing.
Finally, it will consider the noncanonical pathway of miRNA
biogenesis.
2 Regulation of miRNA Transcription
Primary machines for miRNA regulation at the transcription pro-

cess were first identified by Lee et al. [9]. They showed that RNA
Pol II and RNA Pol III enzymes are responsible for miRNA tran-
scription in humans and viruses [10].
In mammals, miRNAs are first transcribed as primary miRNA
(pri-miRNA), which are longer transcripts. The resulting transcript
may comprise multiple miRNA stem-loops, and the cap is inserted
via polyadenylation at the 50 end [11]. This pri-miRNA is then
clipped through Drosha for pre-miRNA generation. After this,
pre-miRNA is transferred to the cytoplasm to create a 22 nt duplex.
Then the mature miRNA is associated with several members of the
AGO proteins to form RISC [12]. The promoter regions of the
miRNA genes and the promoter regions of the protein-encoding
Regulation of miRNAs 3
genes have similar histone modifications, enhancer regions, and the

presence of CpG islands. Therefore, transcriptional regulation of
miRNA genes with these factors is carried out in a similar way as the
management of protein-encoding genes [13]. Thus, transcription
factors, enhancers, and chromatin modifications regulate miRNA
transcription. Transcription factors regulate miRNA genes, either
positively or negatively. Expression of miR-34 class miRNAs, which
play an important role in cell cycle and apoptosis, is increased by
p53 [14]. The promoter regions of mir-34 contain palindromic
sequences corresponding to the binding site of the p53 gene
[14]. MyoD induces transcription by binding to the promoter of
the miR-1 gene during myogenesis [15]. MYC induces activation
or suppression of some miRNAs involved in the cell cycle and
apoptosis [12]. NRSF (Neural-Restrictive Silencer Factor), which
is a silencing transcription factor, prevents transcription of miR-9,
miR-124a, and miR-132 miRNAs [16–18]. NRSF specifically
binds MeCP2 to the promoter of miR-124 and inhibits transcrip-
tion of miR-124 in non-neuronal cells. However, NRSF is down-
regulated during neurogenesis, and transcription of miR-124 is
allowed [18]. Epigenetic factors have important roles in the regu-
lation of miRNA genes. For example, methyl-transferases,
DNMT1/3, regulate transcription of miR-9, miR34-b/c,
miR-148a, and let-7 [19]. Besides, some transcription factors reg-
ulate the transcription of miRNAs by generating negative or posi-
tive feedback loops [20]. For example, the transcription factor
Pitx3 forms an autoregulatory loop with miR-133b. Pitx3 induces
transcription of miR-133b. However, increasing miR-133b causes
Pitx3 to be inhibited over time [19, 21]. Similar conditions occur
between Runx1 (Runt Related Transcription Factor-1) and
miR-27a during megakaryopoiesis [22] and between c-MYB and
miR-15a in hematopoiesis [23]. This feedback loop is called a
unilateral negative feedback loop. miR-7 with transcription factor
YAN (ETS binding protein) and miR-200 with ZEB1/2 are exam-
ples of reciprocal negative feedback loops [24, 25].
3 Nuclear Processing of miRNAs
The long primary transcripts produced by the RNA polymerase II

enzyme are subjected to several further processes in the nucleus
before being transferred to the cytoplasm. In the first procedure
catalyzed by an enzyme called Drosha, pri-miRNAs are cleaved
from single-stranded RNA fragments (ssRNA) at the 30 and 50
ends. As a result of this, precursor-miRNA (pre-miRNA) with a
hairpin structure of about 70 nucleotides is obtained [26]. How-
ever, in recent years, it was found that this process was not done by
Drosha alone. Drosha needs a cofactor called DGCR8 (DiGeorge
Critical Region 8) to efficiently process pri-miRNA, and together
they form the microprocessor complex [27]. A disease called

DiGeorge syndrome occurs in the absence of the DGCR8 gene in
humans [12, 28]. Studies have shown that Drosha and DGCR8 are
of critical importance for miRNA production. Therefore, it is sug-
gested that Drosha and DGCR8 must be tightly regulated in the
cell. The second process of miRNAs in the nucleus is performed by
a protein called Exportin 5 (XPO5). This protein transports
miRNA into the cytoplasm, where it is processed further [29].
3.1 Microprocessor The microprocessor complex consisting of the Drosha and DGCR8
Complex cleaves the ssRNA regions of the pri-miRNA to form pre-miRNA.
Pri-miRNA contains 33 bp of stem and ssRNA segments, which act
with DGCR8 [30]. Drosha then cleaves the pri-miRNA from the
ssRNA and dsRNA linkup site. At the same time, the “UG” motif
in the pri-miRNA is recognized by the Drosha, while the DGCR8
recognizes the “UGU” motif for effective cleaving [31].
3.1.1 Drosha Drosha is a nuclear protein (160 kDa) of the RNase III-type
endonuclease family, which act on double-stranded (ds) RNA as
the first step of miRNA processing in the nucleus. Drosha (nuclear
RNase III) is the enzyme responsible for the cleavage of
pri-miRNAs to produce pre-miRNAs forming a hairpin structure
of approximately 70 nucleotides in length [32].
Drosha consists of the following domains; a central domain
required for cleavage activity, the carboxy-terminal domain to
which double-stranded RNAs bind, two consecutive RNase III
domains (RIIIDa and RIIIDb) [32–36]. RNase III domain a
(RIIIDa) and RNase III domain b (RIIIDb) dimerize with each
other. Both RNase III domains bind to DGCR8 and form the
microprocessor complex [33]. Drosha also contains a unique helix
structure called the Bump helix. This structure is essential in the
determination of the pri-miRNA cleavage-off region [34]. Drosha
serves as a molecular ruler with this function, and this is a necessary
feature for efficient pri-miRNA processing.
3.1.2 DGCR8 DGCR8 is a 86 kDa protein that contains an N-terminal domain,

RNA-binding heme domain, two dsRNA binding domains
(dsRBD), and a C-terminal domain [31]. Drosha alone can bind
pri-miRNA but requires DGCR8 for dsRBD activity. DGCR8 has
two dsRBD domains that recognize the ssRNA-dsRNA linkup site.
DGCR8 also interacts with Drosha via the carboxyl end. The heme-
binding domain allows DGCR8 to function as a dimer [35]. It is
thought to be responsible for the recognition of RNA. This field is
required for an active microprocessor complex. The heme-binding
domain allows DGCR8 to enter a suitable conformation
[35]. However, its role in the microprocessor complex is not fully
understood. DGCR8 performs an efficient cleavage by binding to
the terminal loop region of the pri-miRNA and recognizing the
UGU motif [31, 36].
3.1.3 Pri-miRNA The first stage of miRNA maturation is the cleavage of the hairpin
structure from the pri-miRNAs through Drosha. After Drosha-
mediated processing, pre-miRNA is released [32]. Drosha’s sin-
gle-domain dsRBD is unable to form a strong bond with the
substrate and cannot efficiently process pre miRNAs. Other factors
are needed for Drosha to work more effectively [33].
Drosha-mediated cleavage of miRNAs is a critical process for
determining miRNA abundance. The pri-miRNA must be recog-
nized by the microprocessor to be processed [31]. If the micropro-
cessor complex cannot accept the pri-miRNA, then miRNA will not
be functional. Also, this step is critical in the production of miRNAs
because the precise location of the Drosha cleavage site determines
the terminal nucleotide at the 50 or 30 end of the mature miRNA.
The faulty operation of Drosha in cleavage site selection may cause
a change in the miRNA seed sequence and turn to change its
mRNA targets. Therefore, accurate identification of cleavage sites
is very critical for correct maturation. The basal and apical intersec-
tions are essential for the determination of the cleavage site
[37, 38]. However, the connection regions of Drosha and
DGCR8 and how they interact are not yet known. Further studies
are needed to understand how Drosha recognizes pri-miRNAs.
Recently, it was shown that additional elements play a role in
pri-miRNA processing (UG motif, CNNC motif, and UGUG
motif) [12, 37]. These elements are mostly located in the terminal
region of pri-miRNAs, and their binding increase the processing of
pri-miRNAs [37]. For instance, a SNP in the CNNC motif of miR-
15 and miR-16 genes significantly inhibits Drosha-mediated pro-
cessing, and this causing chronic lymphocytic leukemia (CLL)
progression [12, 37, 39]. Studies have shown that they are involved
in other adjuncts in pri-miRNA processing. Once these helper
factors and their roles are identified, pri-miRNA processing will
be better understood.
Drosha-mediated processing of an intronic miRNA encoded in
host mRNA does not affect the function of this mRNA [12]. How-
ever, the miRNA hairpin structure can occur in a region encoding
the gene (exonic site). In this case, Drosha initiates the cleavage
process and causes instability in the host mRNA [9, 12].
3.1.4 Regulation Drosha usually works efficiently with RNA-binding proteins such as
of Drosha and DGCR8 DGCR8 and TRBP. Therefore, the activities and expression of
other proteins accompanying Drosha regulate miRNA processing.
Autoregulation Between DGCR8 (C-terminal domain) interacts with Drosha (middle
Drosha and DGCR8 domain), and stabilizing each other. Drosha cuts the hairpin struc-
tures of DGCR8 mRNAs and reduces DGCR8 production. Thus,
autoregulation between Drosha and DGCR8 regulates miRNA
biogenesis (Fig. 1) [40]. Drosha-mediated processing of miRNA
was found to be significantly reduced by a threefold increase in the
amount of DGCR8 [37]. Therefore, the Drosha and DGCR8 ratio
should be regulated carefully in the cell [37, 41].
Fig. 1 Autoregulation between Drosha and DGCR8. Drosha cuts the hairpin structures of the DGCR8-mRNA.
Thus, the DGCR8-mRNA is rendered unstable. On the other hand, DGCR8 protein regulates Drosha positively
with protein–protein interactions (Figure redrawn and modified after reading Han J et al., 2009)*
Posttranslational Posttranslational modifications (especially phosphorylation, acety-

Modifications of Drosha lation, deacetylation) can regulate protein stability, nuclear locali-
and DGCR8 zation, and microprocessor activity. High rates of phosphorylation
of DGCR8 in the cell increase the action of the microprocessor
complex and the production of miRNAs [42]. The most crucial
cellular pathway that affects miRNA biogenesis and function is the
MAPK signaling pathway. Stress activates the MAPK p38 pathway,
and then p38 phosphorylates Drosha. In this case, Drosha does not
interact with DGCR8 and loses its stability. As a result, apoptosis is
induced because the biogenesis of miRNAs does not occur [36]. It
is reported that activation of the p38 MAPK-MK2 signaling path-
way supports miRNA biogenesis by simplifying the nuclear locali-
zation of p68 [36, 43]. Since p68 is a component of the
microprocessor complex, miRNA biogenesis is enhanced
[43]. Another example is that the ABL protein, a tyrosine kinase,
phosphorylates DGCR8. On DNA damage, the ABL protein phos-
phorylates DGCR8 and enhances the processing of specific
pri-miRNAs [44].
GSK3β pathway, which has a crucial role in glycogen metabo-
lism, is another critical pathway affecting miRNA biogenesis. The
phosphorylation by GSK3β in Ser300 and Ser302 residues, which
are essential for the nuclear localization of Drosha, increases
miRNA biogenesis by stimulating the nuclear localization of
Drosha [45].
Also, the sumoylation of DGCR8 directly affects miRNA bio-
genesis. This process, mediated by SUMO1, occurs at the Lys-707
site of DGCR8 [46]. DGCR8, which undergoes sumoylation,
becomes resistant to degradation by the ubiquitin-proteasome.
Thus, the stability of DGCR8 increases, and its relation to
pri-miRNAs changes. ERK-mediated phosphorylation induces
this type of sumoylation [36, 46].
Fig. 2 RBPs mediated regulation. (a) In a typical miRNA processing step, the pri-miRNA is processed into the
pre-miRNA by the Drosha/DGCR8 complex. (b) p68 and SMAD combine with the Drosha/DGCR8 complex to
enhance the processing of pri-miRNA. (c) p68 and p53 combine with the Drosha/DGCR8 complex to increase
the processing of pri-miRNA. (d) HNRNPA1 binds to the terminal loop of the pri-miRNA, while the KSRP
facilitates the Drosha mediated process by binding to the terminal loop of pri-miRNA. (e) Activation by ER2 and
E2 causes activation of the p68/p72 complex and consequently inhibition of miRNA production. (f) The nuclear
factor (NF) 90/45 complex binds to the stem/loop structure of pri-miRNA, preventing miRNA from being
processed by Drosha/DGCR8. (Figure redrawn and modified after reading Slezak-Prochazka I, 2010 and Ha M
et al., 2014)*
Another modification that changes the affinity of DGCR8

against pri-miRNAs is deacetylation. In this process mediated by
histone deacetylase 1 (HDAC1), Lys residues in the RNA-binding
regions of DGCR8 undergo deacetylation. Thus, the affinity of
DGCR8 against pri-miRNAs is increased [47].
RNA-Binding Proteins RNA-binding proteins (RBPs) can regulate microprocessor activity

Mediated Regulation (Fig. 2). They perform this regulation by interacting with
pri-miRNAs and Drosha. Some miRNAs need p68 and p72 to be
processed by Drosha. SMAD1–3, SMAD5, and p53 interact with
p68 to enhance microprocessor activity [12]. Bone morphogenetic
protein (BMP) and TGF-ß activate R-SMAD proteins, which inter-
act with p68 and pri-miRNAs. This interaction allows the proces-
sing of mir-21 and mir-199a by Drosha [12]. SNIP1 protein,
related to SMAD, can regulate miRNA biogenesis by interacting
with Drosha. The downregulation of SNIP1 reduces the expression
of some miRNAs [48]. Another protein that cooperates with
Drosha and Dicer is TAR DNA-binding protein 43 (TDP43). It
can assist the processing of miRNAs by increasing stability
[12]. Also, ARS2, a component of the nuclear cap-binding protein
complex, regulates pri-miRNA stability by interacting with Drosha

[49]. Alternative splicing factors, hnRNP A1 and KSRP proteins
interact with some pri-miRNAs to increase Drosha-mediated pro-
cessing [12, 50]. Besides, estrogen-α (ER-α) and estrogen recep-
tors (E) have an essential role in miRNA biogenesis. ER-α can
regulate miRNA expression at the transcriptional level by direct
binding to the miRNA promoter [51]. Nuclear factors-90
(NF-90) and -45 (NF-45) interact together with the pri-miRNAs
and prevents the binding of the microprocessor complex to
pri-miRNAs. Accordingly, miRNA production is reduced [52].
4 Nuclear Export
The pre-miRNA must be transported to the cytoplasm to complete

its maturation. This transport is mediated by XPO5 (a member of
the nuclear transport receptor family), which is also responsible for
the transport of tRNA [53, 54]. When the RAN protein passes into
the GTP form in the nucleus, pre-miRNA binds to XPO5. Then
RAN reverts into its GDP form, and the pre-miRNA is released into
the cytoplasm [53].
XPO5 also mediated by the nuclear export of eukaryotic elon-
gation factor 1A (eEF1A) [54]. It regulates the expression of Dicer
necessary for miRNA maturation [55]. The knockdown of XPO5
in HeLa cells caused the Dicer mRNA to accumulate in the nucleus.
Therefore XPO5 is thought to interact with Dicer mRNA and
reduces Dicer’s expression. As a result, XPO5 significantly affects
the miRNA biogenesis and function by modulating the expression
of Dicer while mediating the transport of miRNAs [55]. The regu-
lation of XPO5 was less investigated than other miRNA biogenesis
factors. Further studies with miRNA exports are needed to under-
stand this process in the future better.
4.1 Posttranslational It is reported that posttranslational modifications of XPO5 can

Modifications of XPO5 affect miRNA biogenesis. While XPO5 is not found to be phos-
phorylated in normal hepatocytes, it has been found to be phos-
phorylated at some sites in hepatocellular carcinoma cells [56].
Phosphorylated XPO5 can cooperate with Pin1 (peptidyl-prolyl
cis-trans isomerase NIMA-interacting 1), and this causes changes
in protein conformation [56]. Because of that, pre-miRNAs cannot
be transported from the nucleus to the cytoplasm. Thus XPO5
accumulation occurs and in the nucleus [57]. Posttranslational
modifications of XPO5 can decrease the expression of miRNAs in
hepatocellular carcinoma cells [57]. Acetylation of XPO5 protein
has also been reported [58]. However, the function of XPO5
acetylation remains unknown.
5 Cytoplasmic Processing
After processing in the nucleus, miRNAs transported to the cyto-

plasm via XPO5 are cleaved from the terminal region by Dicer to
form the approximately 22-nt double-stranded RNA duplex [59].
Dicer is commonly found in eukaryotes, including fungi, plants,
and animals [12]. Kanellopoulou et al. demonstrated that Dicer is
critical in miRNA biogenesis [60]. Although the cells survived as a
result of this significant defect in RNAi, they showed significant
defects in cell differentiation. Dicer plays a role in many basic
cellular processes [60]. For this reason, Dicer needs to be tightly
regulated in the cell.
5.1 Structure Dicer forms a molecular dimer with RNase III domains to form a
of Dicer catalytic center similar to Drosha. Dicer is present in almost all
eukaryotic organisms and is conserved mainly in evolutionary
terms [61]. The mammalian Dicer protein consists of the following
domains; DExD (DEAD-like domain)/H-box helicase domain,
PAZ (Piwi/Argonaute/Zwille) and DUF283 (domain of unknown
function) domain, RNase IIIa and RNase IIIb domains, dsRBD
domain (double-strand RNA binding domain) [62]. The terminal
loop of the pre-miRNAs interacts with Dicer’s N-terminal helicase
domain. Thus, Dicer allows the identification and processing of
specific miRNAs.
On the other hand, Dicer’s PAZ domain recognizes the over-
hang sequence on the stem of the pre-miRNA. When precursor
miRNAs are correctly loaded in Dicer, each ribonuclease domain
acting on one strand of the duplex to cut the loop structure
[62]. Park et al. reported that human Dicer binds to the 3’end
and 5’end of the pre-miRNA and determine the cleavage site by the
distance of 22 nucleotides from the 5’end [63]. The exact function
of Dicer’s ATPase/Helicase domain and DUF283 domain is
unknown [64]. However, it is thought that the helicase domain
plays a role in the recognition of the terminal cycle of the precursor
miRNA [65].
5.2 Regulation Dicer proteins generally interact with double-stranded RNA-bind-

of Dicer ing proteins. Two specific RBPs that bind to Dicer are “TAR
RNA-binding Protein (TRBP)” and “Loquacious (Loqs).” TAR
RNA-binding protein is associated with human Dicer, whereas
Loqs is associated with Drosophila Dicer 1. The Loqs has two
distinct isoforms (Loqs-PA and Loqs-PB). Both of these Loqs iso-
forms are important but not essential for miRNA production
[12, 66–68]. Also, a protein called PACT in mammals may interact
with Dicer. However, its role in miRNA biogenesis is unknown
[12, 69, 70]. TRBP can affect target specificity and guide strand
selection by increasing the length of miRNAs [12, 71]. TRBP has
Fig. 3 Regulation of Dicer (a) MAPK / ERK induces phosphorylation of TRBP. TRBP increases the processing of
pre-miRNAs by binding to Dicer. (b) KSRP enhances Dicer-mediated processing by binding to the terminal loop
of the pre-miRNAs. (c) Lin28 is connected to the terminal loop of pre-let-7 and suppresses the Dicer mediated
process (Figure redrawn and modified after Treiber T et al, 2018, Heo I, 2012 and Ha M et al., 2014)*
three different ds-RBP domains. Two of them interact with the

pre-miRNA while the other interacts with Dicer. These domains
facilitate the linking of Dicer to dsRNA (Fig. 3a) [36]. As in
Drosha, RBPs play an essential role in regulation in Dicer. Other
factors that play a role in this regulation are also being investigated.
One of them is KSRP, which interacts with pre-miRNAs to facilitate
Dicer-mediated processing [12] (Fig. 3b). Lin28 inhibits the bind-
ing of Dicer to these miRNAs by causing the uridylation of some
pre-miRNAs [12] (Fig. 3c).
6 RNA-Induced Silencing Complex (RISC)
The RNA duplex, processed by Dicer, is loaded onto the AGO

protein and forms the RNA-induced silencer complex (RISC)
[12, 72]. RISC is a family of heterogeneous molecular complexes
created for silencing genes. The pre-miRNA, processed by Dicer in

the cytoplasm, forms RNA duplex and triggers RISC formation. In
principle, RISC is induced by the emergence of double-stranded
RNA in the cytoplasm [73]. RISC allows the silencing of the target
with different mechanisms. Some of these mechanisms are transla-
tional suppression, mRNA degradation, and DNA elimination
[73]. Although the mechanisms of action of RISC vary slightly
between organisms, there are two standard features: first, all
RISCs contain the AGO protein. AGO selects a strand of miRNA
duplex to produce a mature miRNA. The strand that will form
mature miRNA is called the “guide strand,” and the strand
removed from the complex is called the “passenger” strand. Sec-
ond, each loaded RISC includes a small RNA that provides base
complementarity with its target mRNA [73]. RISC processing is
not a simple binding between these two factors; instead, it takes
place in multiple steps; loading of small RNA duplex into AGO
proteins (pre-RISC), fraying the duplex near the guide 30 end by
N-domain (wedging), and then remove the passenger strand to
form the mature RISC [74].
6.1 AGO Proteins miRNAs interact with Argonaute proteins to form RISC and par-
ticipate in nearly all cellular processes in eukaryotes. AGO proteins
are associated with small RNAs, and they play a role in gene
silencing after transcription [75]. AGO proteins are expressed in
almost every cell and interact with miRNAs or siRNAs.
6.1.1 Structure of AGO AGO proteins were first identified in plants and were thought to be
Proteins the only presence of PIWI and PAZ domains [76]. In the recent 3D
structural studies, AGO protein was found to be two-lobed (N-ter-
minal lobe and C-terminal lobe), which has a central cleft to bind
guide and target RNAs [77]. They have four distinct domains in
these two lobes; amino-terminal (N) domain and PAZ domain at
the N-terminal lobe, MID and PIWI domain at the C-terminal lobe
[77] (Fig. 4). These two lobes are connected by two linkers called
“L1” connecting the N-terminal domain to the PAZ domain and
“L2” connecting the PAZ domain to the MID domain. miRNAs
use the 50 and 30 ends when binding to AGO proteins; The 50 end
interacts with the MID and PIWI domain, the 30 end interacts with
the PAZ domain [74, 77]. Thus, the seed sequence of a miRNA
interacts with the PIWI domain [77].
While all AGO and PIWI proteins have slicer properties in
D. melanogaster, only AGO2 has slicer activity in humans [11, 78,
79]. This feature indicates that the miRNA function is more strictly
regulated in humans. AGO proteins interact with additional factors
and elements to modulate translational suppression, gene silencing,
and mRNA degradation. Mouse stem cells with AGO protein defi-
ciency had defects in miRNA silencing, and these cells were found to
undergo apoptosis [12, 80]. The same results were obtained
Fig. 4 AGO proteins consist of MID, PIWI, N-Terminal, and PAZ domains. The position of the guide strand at the
AGO protein allows the seed sequence to bind to the target site*
in similar studies, and it was concluded that AGO proteins have a

critical role in miRNA biogenesis [12, 80].
6.1.2 Post-translational Post-translational modifications of AGO on different residues sig-

Modifications nificantly affect miRNA activity and biogenesis [78] (Fig. 5). These
and Regulation of AGO post-translational modifications of AGO significantly affect miRNA
activity and biogenesis. The best defined AGO post-translational
modification is phosphorylation. AGO phosphorylation takes place
at three points; Ser387 and Tyr393 residues of the human Ago2,
Tyr529 in the MID domain, and S824-S834 set in PIWI domain
[78]. Ser387 phosphorylation occurs in the L2 region of human
Ago2 [78, 81] (Fig. 5a). L2 Ser387 phosphorylation induces stress
response by stimulating the p38 MAPK pathway [81, 82]. Also,
phosphorylation of Ser387 is mediated by MAPKAPK2 or AKT3,
which leads to P-bodies localization and translational repression
[12]. Horman et al. found that AKT3 protein phosphorylates
AGO Ser387 in HeLa cells [83]. Similarly, Ago2 and GW182
binding protein (LIMD1), needed Ser387 phosphorylation to
interact with Ago2 [78, 84]. These studies have shown that phos-
phorylation of AGO-Ser387 promotes the incorporation of RISC
and improves miRNA function. The phosphorylation of Tyr393
has also been well studied [78, 85]. This phosphorylation is
mediated by EGFR and is stimulated by hypoxic stress (Fig. 5b).
Tyr529 phosphorylation, which is another post-translational
Fig. 5 Post-translational modifications of AGO protein. (a) Ser387 phosphorylation of the L2 region of the AGO
protein promotes miRNA production by promoting RISC formation. Also, the transport of AGO2 to endosomes is
prevented. Ser387 phosphorylation is mediated by MAPKAPK2 and AKT3, (b) Phosphorylation of Tyr393 in the
L2 region decreases the miRNA activity by decreasing the AGO interactions with miRNA, (c) Tyr529
phosphorylation occurs in the miRNA 50 phosphate-binding region of the MID domain. Phosphorylation of
this region prevents the loading of miRNAs into AGO proteins, (d) P700 4-hydroxylation of the PIWI domain
enhances the stability of AGO2, (e) PARylation occurring in any region of the AGO protein reduces miRNA
activity, (f) Other post-translational modifications of AGO proteins have also been described. However, their
functions are not yet known. These; Phosphorylation of S253, T303, and T307 in the PAZ domain; S798
phosphorylation in the PIWI domain (Figure redrawn and modified after reading Gebert LFR, 2019)*
modification of AGO, inhibits miRNA loading [12, 78]

(Fig. 5c). In contrast, Tyr529 Ago2 mutants do not alter the
interaction of miRNA with Dicer [82]. Besides, phosphorylation
of the Ago2 S824-S834 residues controls the transcription of target
mRNAs. Inhibiting this phosphorylation cycle strongly down-
regulated miRNA activity [86, 87].
Hydroxylation of AGO is one of the known post-translational
modifications (Fig. 5d). Hydroxylation of proline 700 regulated
the stability of Ago2 in mice and human cells. Proline 700 is close
to the Trp-binding region in AGO2. Because of that, miRISC can
potentially change the assembly [74, 88].
AGO1–4 proteins were modified by Poly ADP ribose in human
cells [12, 89] (Fig. 5e). The poly ADP-ribose polymer is added to
the residues (Asp, Glu, and Lys) by this process. Poly ADP-ribosy-
lation, which usually modulate the cellular stress response, reduces
translation suppression and endonucleolytic cleavage, which can
impair the target accessibility of AGO1-4 [78, 89, 90]. Other post-
translational modifications (phosphorylation of S253, T303, and
T307 in the PAZ domain; S798 phosphorylation in the PIWI
domain) of AGO proteins have also been described, but their
functions are not yet known [82] (Fig. 5f).
However, AGO proteins can also be modified by different
modification methods. They are stable when loaded with a
miRNA and are unstable when they are empty [91, 92]. Protea-
some-mediated destruction and autophagy may be responsible for
the instability of AGO proteins [12, 93].
6.2 Regulation The RNA duplexes formed after the Dicer process are loaded onto
of miRNA and RISC specific AGO proteins to form RISC. Previously, it was known that
Loading miRNAs were loaded only on AGO1 and siRNAs were loaded on
AGO2. However, recent studies have found that miRNAs are
loaded on both AGO1 and AGO2 [93]. In Drosophila, loading
of miRNAs into AGO proteins is accomplished through the sorting
system [12]. This sorting process takes place according to two
features. The first feature is the loading of the AGO protein accord-
ing to the mismatches in RNA duplexes. The mismatches in the
sequence of miRNA duplex inhibit entry into the AGO2 protein.
Thus the miRNA is directed to the AGO1 protein [94]. The second
feature is the identity of the nucleotide at the 50 end of the guide
chain. While miRNAs containing uracil at the 50 end of the guide
chain are loaded into AGO1, the siRNAs containing cytosine at the
50 end are loaded into AGO2 [12]. In humans, no special rules were
found for loading of AGO1-4 among miRNA classes. Therefore, in
humans, all AGO proteins can bind to both siRNAs and miRNAs
[12]. A complex called RISC-Loading Complex (RLC) must be
formed, before loading siRNA duplexes into AGO2 in Drosophila.
This complex, consisting of Dicer2 and R2D2 proteins, has a
critical role in directing siRNA duplex to AGO2 [12, 95–97]. It
was shown that R2D2 and Dicer-2 form together a stable complex
and are associated the siRNA-generating activity [98]. R2D2 pro-
mote the siRNA passage from Dicer to RISC in Drosophila [98].
Studies of RISC loading in humans have generally been done with
siRNAs and therefore the RISC loading steps of miRNA are not
fully known. Although the role of RLC is not known in humans, a
miRNA Loading Complex (miRLC) was found in mammalian cells,
which act as the primary miRNA loading machinery [99]. Further
studies are needed to understand whether RLC is necessary for
loading siRNAs and miRNAs in humans into AGO proteins.
6.3 miRNA Duplex The miRNA duplex loaded into RISC consists of a passenger strand
Unwinding and Strand and a guide strand. The guide strand (with an unstable terminus at
Selection the 5’ end) is the functional strand showing the base matching to the
target and forms the mature RISC [11, 100]. The thermodynamic
stability of both ends of the duplex is an essential factor because the
unstable strand is already partially opened at physiological
temperature [11]. The N-terminal domain of the AGO protein

stabilizes this open end, and the MID domain fixes the 50 end of
the guide strand, thus obtaining a mature miRNA containing the
seed sequence [36].
Another feature of the human miRNA guide strand is the
tendency to have U-nucleotide at the 50 -end, and more purine
possession in this strand, while the property of the passenger
strand tends to have C-nucleotide at the 50 -end and more pyrimi-
dine [12, 95]. After selecting the guide strand, the passenger strand
must be removed from the complex. The passenger strand is
cleaved by AGO proteins responsible for slicing (AGO2 in humans)
and is removed from the complex with the enzyme C3PO, an
endoribonuclease [96, 101]. C3PO enhances RISC activity by
facilitating the elimination of the passenger strand from the com-
plex [96]. However, since there is only the slicer activity of the
AGO2 protein (AGO-1, -3 and -4 do not have slicer activity) in
humans, this mechanism in which the passenger arm is cleaved is
rarely used [12, 93]. Therefore, other mechanisms are generally
preferred.
7 RNA Sequence Alterations
Modifications, mutations, and changes in the sequence of miRNAs

also regulate the biogenesis and function of miRNAs to a large
extent. There are many ways in which these structural changes
and modifications to the miRNA sequence occur. Such regulation
of miRNAs can affect the targets, expressions, the proliferation of
miRNAs, induce miRNA degradation, and regulate miRNA
turnover.
7.1 SNPs (single-nucleotide polymorphisms) occurring in the miRNA-

Single-Nucleotide encoding genes or the miRNA binding region of the target mRNA
Polymorphism (SNP) significantly alter the miRNA biogenesis and function. An SNP
and miRNA Tailing occurring in the seed sequence, which is particularly important in
the function of miRNA, can alter the miRNA target. Besides, SNPs
in the sequence that constitutes the passenger strand or stem-loop
may reduce the miRNA biogenesis by inhibiting Drosha and Dicer-
mediated processing [6]. miRNA polymorphism may cause many
diseases. It can initiate tumor formation by causing the activation of
oncogenes [102]. For instance, when the pri-miR-146a gene has a
G/C polymorphism, miR-146a expression is reduced, and contri-
buting to genetic susceptibility to papillary thyroid carcinoma
[103]. Sun et al. reported that SNPs could disrupt or improve
miRNA biogenesis and also alter target sites [104].
In addition, the expression of miRNAs can be regulated by

adding RNA tailing to the 3’ ends of miRNAs. There are two
known types of tailing: uridylation and adenylation [5, 12, 105–
114].
In principle, uridylation is most commonly seen in pri-miRNA
and pre-miRNAs. The best-known example of this mechanism is
the uridylation of the Let-7 family by Lin28A protein [5, 12]. Let-7
exhibits very different expression patterns in the cell [105]. For
example, mature let-7 was found to be expressed only in differen-
tiated embryonic stem cells, whereas precursor let-7 was found to
be expressed in undifferentiated cells [106]. These findings indicate
that let-7 is transcriptionally regulated. More recent studies show
that Lin28 inhibits the biogenesis of let-7 [107]. The Lin28A
protein contains the cold shock domain and two chain finger
domains that allow the pre-let-7 to be connected to the terminal
loop [108]. Lin28 can inhibit the biogenesis of let-7 with different
mechanisms. Lin28 regulates both Drosha and Dicer mediated
processing of pri and pre let-7 [109]. It can interfere with
Drosha-mediated operation by connecting to the terminal loop of
pre-let-7. However, considering that Lin28 is localized in the cyto-
plasm, uridylation of pre-let-7 is considered to be a more suitable
option. Pre-let-7’s U-tail added to the 30 end through Lin28 pre-
vents Dicer-mediated processing [109]. Furthermore, Lin28
recruits TUT4 (terminal uridylyltransferase 4) and TUT7 (terminal
uridylyltransferase 7) to elongate the U tail with oligouridylation
and induce degradation of let-7. The tail extended by TUT4 and -7
is degraded by a 30 -50 exonuclease called DIS3L2, thereby inducing
the degradation of let-7 [12, 110] (Fig. 6).
Adenylation is another process where nucleotides are added to
the 30 end of miRNAs. However, adenylation induces miRNA
Fig. 6 LIN28A/B in the cytoplasm prevents Dicer from processing pre-let-7 and induce oligouridylation. In
mammals, TUT4 and 7 catalyze oligouridylation. The oligouridylated pre-let-7 is then degraded by the enzyme
DIS3L2 (Figure redrawn and modified after reading Ha M et al., 2014)*
stabilization and miRNA degradation in some cases. Gld2 (Germ-

line development 2) which is a non-canonical and regulatory cyto-
plasmic poly(A) polymerase, stabilizes miR-122 in liver and mouse
embryonic fibroblasts by monoadenylation (3’terminal) of this
miRNA [78, 111, 112]. Katoh et al. determined that GLD2 can
control the levels of several target mRNAs through regulating the
stability of miR-122 [111]. Contrarily, when the CUGBP1 protein
interacts with miR-122, it recruits poly(A) specific ribonuclease
(PARN) and induces miR-122 degradation. Similar situations are
seen in miR-93 and miR-652-3p [113]. On the other hand, the
monoadenylation of miR-21 by the enzyme PAPD5 induces PARN
mediated miR-21 degradation [114]. 30 adenylation leads to a less
stable miRNA inhibition than 30 uridylation. The RNA tailing is a
necessary process that regulates miRNA activity both positively and
negatively. Precise control of the expression of essential proteins
and enzymes, such as CUGBP1 and PARN, determines how the
RNA tail will influence miRNA.
7.2 RNA Editing RNA editing of the miRNA sequence may alter its target, generat-
ing a new miRNA variation (isomir), and even affect miRNA bio-
genesis. RNA editing refers to changes that occur in the RNA
sequence after transcription [115]. The most common RNA edit-
ing process in miRNAs is deamination, which is catalyzed by ADAR
(adenosine deaminase acting on RNA) enzymes. This enzyme con-
verts the Adenosine nucleotide to Inosine (A to I) [116]. Another
enzyme called CDA (cytidine deaminase acting) converts Cytosine
to Uracil (C to U) [117]. Nevertheless, there is not much informa-
tion available about the C to U editing of miRNAs. ADAR1 and
ADAR2-mediated A to I processing, which usually occurs in pre-
cursor miRNA sequences are most commonly seen in vertebrates
[116]. A to I editing in miRNAs can reduce Drosha and Dicer-
mediated processing and loading to AGO proteins. Editing occur-
ring in the seed sequence may cause the miRNAs to shift to another
target by altering the target specificity [78]. RNA editing in the
stem-loop of pri-miR-142 reduces Drosha mediated processing
[12, 118]. Similarly, the editing of the pri-miR-151 catalyzed by
ADAR1 also reduced the Dicer-mediated processing [12, 119].
RNA editing also has an essential role in cancer. Shoshan et al.
found that overexpression of CREB in melanoma decreased
ADAR1 expression and thus reduced the editing of miR-455-5p.
It modulates unedited miR-455-5p, CPEB1, and tumor suppressors.
This modulation causes tumor growth and metastasis [78, 120].
Studies have shown that the 50 monophosphate group of some
pre-miRNAs has been O-methylated with an enzyme called
BCDIN3D, a human RNA methyltransferase that is not yet well
characterized. BCDIN3D O-methylation leads to loss of the nega-
tive charge of these pre-miRNAs (pre-miR-23b and pre-miR-145)
50 terminal monophosphate group both in vitro and in vivo
Fig. 7 The BCDIN3D enzyme removes the charge of the 50 phosphate of the pre-miRNA. Thus, the
50 phosphate end of the pre-miRNA does not interact with Dicer and reduces the processing of mature
miRNAs*
[12, 121]. This charge is important for interaction with Dicer and
subsequently for miRNA maturation [121]. Dicer-mediated pro-
cessing of these methylated pre-miRNAs, which have lost their
negative charge, has been reduced in vitro (Fig. 7). Inhibition of
the enzyme BCDIN3D increases the Dicer-mediated processing of
pre-miR-145 and mature miR-145. Moreover, this enzyme is pre-
dicted to contribute to carcinogenesis because BCDIN3D’s reduc-
tion destroys tumorigenic phenotypes in breast cancer cells
[121]. It is thought that a large number of other miRNAs are
also methylated by similar enzymes.
7.3 Regulation The abundance of miRNAs in the cell is tightly controlled at the
of the Stability, transcriptional and post-translational levels [12, 122, 123]. These
Degradation, regulatory mechanisms are known to modulate miRNA activity and
and Turnover of miRNA function. However, the regulation of the abundance of miRNA by
the degradation of mature miRNAs is among the less studied sub-
jects. The degradation and turnover of miRNAs are mostly unknown.
Previously, miRNAs were thought to be extremely stable in the cell.
The fact that mature miRNAs still exist hours (or even days) after
inhibition of miRNA production led to this idea [122]. But it has
been found that different factors change the miRNA turnover by
miRNA degradation and stabilization. A miRNA turnover is a tool
that allows for the regulation of the miRNA level in the cell according
to developmental, physiological, and environmental conditions
[123]. Each miRNA has a different and specific turnover ratio
[78]. However, miRNA turnovers are controlled by many factors.
Most of the mRNAs are added 50 Cap and 30 poly (A) tail to prevent
degradation by exonucleases because unprotected ends cause the
mRNA to be vulnerable to exonucleolytic activity.
A similar situation can be seen in pri-miRNAs. However,
mature miRNAs do not have any 30 poly (A) tails or 50 Cap. In
plants, a protein called HEN1 (Huan Enhancer1) methylates the 3’
ends of miRNAs and siRNAs, protecting them against 3’ end
uridylation and exonucleolytic degradation. This methylation
takes place before double-stranded RNAs are loaded into
AGO [123].
miR-29a/b, which plays a role in the regulation of the cell

cycle, is an important example of understanding the miRNA turn-
over [122]. Although miR-29b level increases in mitotic cells, it has
a stable level throughout the cell cycle. This stability is due to the
presence of uracils between nucleotides 9 and 11 of miR-29b. The
presence of uracils accelerate the turnover, which allows miR-29b
to degrade rapidly, resulting in cell cycle regulation. The lack of
uracil repeats of miR-29a causes it to have a longer half-life than
miR-29b [122, 124]. This example shows that degradation can
control the turnover ratio of miRNAs. Viral infections can also
regulate the stability of miRNA. It was observed that the
miR-27a/b level decreased rapidly after cytomegalovirus infection
in mice [125]. The factors produced by the cytomegalovirus induce
the degradation of miR-27a/b and reduce miR-27a/b levels in the
host cell [125].
Although each miRNA has its specific stability and turnover
rate, this varies according to the miRNA sequence. Adenylation of
miR-122, specific to human hepatocytes, by Gld2, protects it
against exonucleolytic activities [126]. A similar situation is also
observed in immunity. Cytokine mRNAs need to be strictly regu-
lated to maintain the immune response without causing inflamma-
tion. Zcchc11, a ribonucleotidyltransferase interacting with toll-
like signaling pathway, is required for preserving the poly(A) tail
length and stability of cytokine IL-6 (interleukin-6) transcripts
[127]. It is reported that Zcchc11 regulates IL-6 production by
uridylating cytokine-targeting microRNAs like mir-26a [127].
On the other hand, studies with the identification of enzymes
involved in miRNA turnover have increased. SDN1, a exoribonu-
clease which is encoded by ‘Small RNA-Degrading Nuclease’
(SDN) genes in Arabidopsis thaliana, mediate degradation of
mature miRNAs by processing on single-stranded miRNAs and
be precise to methyl alteration on the 3’ terminal ribose of miRNAs
[12, 128]. The knockdown of SDN genes has led to increased levels
of miRNAs. In C. elegans, the degradation of miRNAs is mediated
by XRN enzymes [12, 129]. However, to achieve this degradation,
the XRN enzyme must be able to reach the 50 end of the miRNA.
Therefore, miRNAs loaded into RISC should be released from this
complex first. The mature miRNA, which is released, is degraded by
the XRN enzyme [129]. miRNA turnover can be obtained by
stabilization and degradation. Furthermore, miRNA turnover is
critical in the regulation of miRNA function.
7.4 miRNA Sequence The ability of miRNAs to bind to their targets depends on the
Modifications specificity of the seed sequence. On the other hand, miRNA bio-
genesis is also regulated by modifications of secondary structures of
RNAs. Changes in the sequence of a miRNA, and in particular in
the seed sequence, can significantly affect the miRNA biogenesis
and turnover.
Processing of miRNAs by Drosha or Dicer, RNA editing, or

addition of nucleotides to miRNAs can generate structures called
’IsomiRs’ that vary from the original miRNA sequence, causing
changes in the length or sequence of mature miRNAs [78, 130].
The primary process that determines the changes in the length
and sequence of the miRNA, constituting the isomiRs, is processed
by Drosha and Dicer [12]. The miRNA variations generated by the
Drosha process were detected in miRNA-142 and miRNA-342 in
T cells [131]. Variations occurring in the region of the Dicer
cleavage generally modulate the seed sequence of miRNA. Thus
the choice of guide strands alters substantially [71]. Essentially, the
primary mechanism of the 50 -isomiR production is the processing
by Drosha or Dicer. Rarely, adding or removing nucleotides to the
50 end of mature miRNAs can change the position of the seed
sequence [78, 130]. Cleavage-directed 50 end variation significantly
altered the miRNA targets in flies and humans [67, 132,
133]. However, 50 isomiR abundance varies in each cell type. In
addition, certain isomiRs are dominant in specific cell types
[78]. On the other hand, 30 isomiRs are different in terms of
stability and turnover [78].
8 Viral Regulators of miRNA
Viral miRNAs, unlike proteins, are nonimmunogenic and do not

cause an immune response in the host, and require less coding
capacity than proteins. Also, miRNAs can affect multiple targets
to varying degrees, not just a single target. Viral miRNAs can assist
in creating a cellular environment that provides viral replication by
using a gene regulation mechanism that is protected within the host
cell. Over 200 viral miRNAs have been characterized in polyoma-
viruses, ascoviruses, and adenoviruses, as well as herpesviruses
[134]. In this way, viruses can significantly affect miRNA function.
For example, the hepatitis B virus reduces the expression of
miR-122 in liver cancer cells [135]. The best-studied miRNA–
virus interactions are the hepatitis C virus (HCV, a single-stranded
RNA virus that mediates liver infection) and miR-122 (liver-specific
microRNA) [78, 135, 136]. The miR-122 interaction of the 50 end
of the hepatitis C virus genome (RNA) causes increased viral infec-
tion in the host and abundance of the virus genome [136]. The
increased risk of HCV infection and hepatocellular carcinoma
development reveals the function of miR-122. Another virus, the
bovine viral diarrhea virus, regulates host miRNAs by interactions
with miRNAs. The 30 UTR of the genome of this virus binds with
let-7 and miR-17 to enhance viral replication. Besides, miR-17
stabilizes the virus genome and improves its translation after bind-
ing to the viral genome [137]. Herpesviridae family can also regu-
late the host miRNA. For example, the Herpes virus saimiri
modulates miRNA function by expressing uracyl-rich RNAs

(HSURs), which have binding sites for host miRNAs
[12, 78]. Cazalla et al. found that HSUR1 reduced miR-27a levels
in infected T cells by degradation. Thus, the removal of repression
over mRNAs that are suppressed by miR-27 causes activation of T
cells [138]. Furthermore, it was found that HSUR1 carrying a
nonfunctional miR-27a binding domain had fewer titers. This
result shows that miR-27a should be degraded for active herpesvi-
rus infection to occur [138].
9 Localization of miRNAs
Although miRNAs are mainly localized in the cytoplasm, they were

also found in the nucleus, nucleolus and mitochondria [139].
9.1 Nuclear Although the last processing steps for the formation of mature
Localization of miRNAs miRNA are performed in the cytoplasm, some mature miRNAs
were found in the nucleus. For example, it was found that
miR-122 was expressed in hepatocytes which migrate to the
nucleus [140]. Meister et al. reported that approximately 20% of
miR-21 was localized in the nucleus [141]. The hexanucleotide
localization signal at the 30 end of miR-29b provides its nuclear
localization [78, 142]. More recent studies have shown that most
miRNAs are localized in both the nucleus and the cytoplasm
[143, 144]. Biogenesis and maturation of mir-709 have been
shown to inhibit miR-15a/16-1 formation in the nucleus in mice.
This showed that expression and maturation of miRNA can be
regulated by another miRNA, and this could be evidence for
‘miRNA hierarchy’ [145] Also, in a study conducted in
C. elegans, it was found that ALG-1 (AGO homolog), increased
the production of miRNAs [78, 146]. Raised levels of mature-let-7
in the cytoplasm cause ALG1 to bind to these let-7 s. The resulting
complex then passes to the nucleus and binds to pri-let-7 and
removes the repression on the miRNA biogenesis. This creates a
positive feedback loop [146]. Despite the studies, we have very
little knowledge about the functions of miRNAs localized in the
nucleus.
9.2 Extracellular miRNAs are also be localized in extracellular environments like

miRNA blood serum, plasma, cell culture medium, and other different
biofluids [147]. The discovery of these circulating miRNAs in
exosomes and liquid biopsies suggests that they may contribute to
intercellular signaling [78]. Besides, extracellular miRNAs may
have a regulatory role in some cancer types. For example, breast
cancer cell-derived exosomes have components that are susceptible
to miRNA maturation and include pre-miRNAs and essential
proteins. These exosomes can induce the proliferation of cells

present in culture [78, 148].
A question about exosomal miRNAs is how these miRNAs are
sorted into exosomes. Ahadi et al. suggested that lncRNAs play a
role in this event [149]. miRNA motifs called EXOmotif are
thought to be important in the sorting of miRNAs to exosomes
by interacting with exosomal RNA-binding proteins [78]. How-
ever, more studies are needed to understand better the mechanisms
and functions of these circulating exosomal miRNAs.
10 Regulation of miRNA Target Sites
Modifications in the target sites of miRNAs can also regulate the

activity and functions of miRNAs. A modification occurring in the
miRNA target region may alter target-site complementarity
[78]. For example, N6-methyladenosine modification in RNA
does not recognize AGO protein. This indicates that adenosine
methylation occurring in the target mRNA can regulate miRNA
targeting [150]. Therefore the formation of mRNA isoforms may
alter the miRNA target sites. It also allows miRNAs to alter mRNA
sensitivity specific to cell type [151].
Similarly, the RNA-binding proteins modulate the miRNA
target sites. For example, the Pumilio protein (PUM1) binds and
modifies the 30 UTR of the mRNA of p27. Pumilio reveals the
target regions of this mRNA for miR-221 and miR-222, allowing
silent cells to enter the cell cycle again [152]. In another study, it
was observed that the interaction of AGO2 with mRNAs increased
by a riboprotein called AUF1 (AU-rich element-binding factor 1)
in HeLa cells [78, 153].
11 Noncanonical Pathways
The miRNA biogenesis also has three noncanonical pathways:

Microprocessor complex-independent, Dicer-independent, and
terminal uridylyl transferase (TUTase)-dependent.
11.1 The essential components of the noncanonical pathway are the

Microprocessor- mirtrons (Fig. 8). Mirtrons are a class of miRNAs of the
Independent Pathway microprocessor-independent pathway and are located in the introns
of host genes [154]. Mirtrons may function as pre-miRNAs after
splicing and therefore are not subjected to any process by the
microprocessor complex. After splicing, it is directly transported
to the cytoplasm and processed by Dicer [155]. However, before it
is transported to the cytoplasm, it is exposed to several more
processes in the nucleus. Once the host mRNA is splicing, a struc-
ture called “lariat” is formed which is not functional
Fig. 8 The microprocessor or Drosha/DGCR8 complex-independent noncanonical pathway comprises “Mir-

trons.” The Mirtrons, which are formed by the splicing process, form the structures called “lariat,” and they
form the stem–loop structure by the “debranching” process. Tailed mirtrons contain longer sequences at the
50 or 30 end. The nuclease or nuclear exosome cuts these tails. In contrast, some miRNAs, along with the
7-methylguanosine-cap are transported directly into the cytoplasm via XPO-1 and without the Drosha/DGCR8
procedure (Figure redrawn and modified after reading Treiber T et al., 2019 and Ha M et al., 2014)*
[12]. Therefore, the lariat structure is debranched and folds to form

the known pre-miRNA structure. As a result, the stem–loop struc-
ture is formed and is ready to be transported to the cytoplasm
[156] (Fig. 8). Some mirtrons have more sequences at their ends,
and these form more extended structures than other mirtrons.
Therefore, these mirtrons need to be excised by exonucleases
[36]. The enzyme responsible for trimming the 30 end is the
“Exosome Complex,” while the enzyme involved in trimming the
other end is not yet known [157].
Endogenous dsRNA transcripts, another class of miRNA, are
also not subjected to Drosha-mediated processing [158]. These
miRNAs are transcribed by RNA polymerase II and have a
7-methylguanosine (m7G) cap [36]. In the canonical pathway,
the cap of pri-miRNAs is cut by the microprocessor complex,
while the “m7G” cap is not cut in the noncanonical pathway.
Instead, this “m7G” cap is recognized by the cap-binding complex
and Exportin-1, making it easy to transport to the cytoplasm [159].
The miRNAs transported to the cytoplasm are subjected to Dicer-
mediated processing. Then, the m7G cap only selects the 3p arm
and loads it into the RISC. In this way, the 5p arm is not expressed,
and only the 3p arm is expressed [159] (Fig. 8).
Fig. 9 In the presence of a “La chaperone,” pre-tRNA forms mature tRNAs, whereas, in the absence of “La,”
tRNA forms the miRNA structure. (Figure redrawn and modified after reading Hasler D et al. 2016)*
The microprocessor complex independent but Dicer-

dependent miRNAs consist of tRNAs, precursor tRNAs, small
nucleolar RNAs (snoRNAs), and other noncoding small RNAs
[12]. A precursor tRNA requires a chaperone protein called “La”
so that it can be folded correctly [160]. Hasler D et al. reported that
in the absence of ‘La chaperone’ the misfolded structure is pro-
cessed by Dicer and may form a miRNA [160] (Fig. 9). Moreover,
many of these pre-tRNA fragments are installed on RISC, and their
functions are still not elucidated. Also, interestingly they reported
that one specific pre-tRNA (pre-tRNA-IIe) can generate both
tRNA and miRNA even when La exists [160].
11.2 Most miRNAs must be processed by Dicer to mature, but some

Dicer-Independent miRNAs are produced in a Dicer-independent way. The most obvious
Pathway example of this is the processing of miR-451, which is an erythrocyte-
specific miRNA. Although not processed by Dicer, it is expressed at
high levels. miR-451 is transported to the cytoplasm after processing
by Drosha [161]. However, since it has a very compact structure
(~18 bp), it cannot be processed by Dicer in the cytoplasm. Instead,
miR-451 is loaded into the AGO2 protein, and the 30 strand is cleaved
using the catalytic activity of AGO2 [12, 36]. After processing by
AGO2, a 30 nucleotide-length structure is formed. In addition,
PARN (poly(A)-specific ribonuclease) was required to cut the 3’
overhang of miR-451. However, it was found that trimming of pre-
miRNAs cleaved by AGO2 was not necessary for target silencing.
This has shown that RISC is functional with miRNAs which are
longer than the mature form [12, 162] (Fig. 10).
Fig. 10 Dicer-independent pathway. Precursor miRNA-451 is processed by the microprocessor complex and
transferred to the cytoplasm (mir-451 could be transported to the cytoplasm by XPO5, but there is no evidence
for this). It is then uploaded to AGO2 for cleavage and trimmed by PARN to form mature miR-451 (Figure
redrawn and modified after reading Ha M et al., 2014)*
Fig. 11 In the TUTase-dependent pathway, which is another noncanonical pathway, pre-miRNAs with a short
30 overhang undergo monouridylation with the enzymes TUT-2, -4, and -7. Thus, the Dicer process
pre-miRNAs and then mature miRNA is obtained (Figure redrawn and modified after reading Heo I et al.,
2012)*
11.3 Terminal Some pre-miRNAs have a shorter 30 overhang due to the structure
Uridylyl Transferase of pri-miRNAs, which are the precursors of these. Pre-miRNAs
(TUTase)-Dependent with a short overhang cannot be efficiently processed by Dicer
Pathway and cannot form a mature miRNA. Because of that, the 30 overhang
of these pre-miRNAs is subjected to monouridylation by TUTases
(e.g., TUT2, 4, and 7) [12, 126]. Thus, the 30 overhang is
extended and processed efficiently by Dicer [126] (Fig. 11).
12 Conclusions and Future Perspectives
There are many missing parts of the miRNA biogenesis, which is a

large picture. MiRNAs encoded from exonic and intronic regions
are mostly unknown. However, when the promoter regions are
detected, and how they are localized, the regulation of miRNA
transcription will be better understood. The connection regions
of Drosha and DGCR8 and how they interact with each other are
not well explained, which causes many questions to remain unan-
swered. More proteins that interact with Drosha are thought to be
involved in miRNA regulation [163]. Even the mysteries about the
structure of the Drosha protein are not solved as well. This limits
our understanding of the stages in miRNA maturation. Respond-
ing to how Dicer and TRBP interact with each other in humans and
how this complex interacts with other molecules involved in other
maturation stages of miRNA will combine many parts of the puz-
zle. In this context, the crystal structure and interaction regions of
RNA-binding proteins should be examined at the molecular level.
Also, post-translational modifications, especially acetylation of
XPO5, need to be better elucidated.
The links between miRNA tail and miRNA degradation are also
mostly unknown, as in oligouridylated pre-let-7 degradation by the
DIS3L2 enzyme. Moreover, it is necessary to identify the nucleases
to understand miRNA degradation better.
Methylation of miRNAs can inhibit maturation. Currently,
only BCDIN3D is known to inhibit the binding of Dicer by
methylating pre-miRNA. However, the identification of other
enzymes involved in miRNA methylation will provide a further
understanding of the factors involved in regulation.
As noncanonical pathways of miRNA biogenesis are discov-
ered, different factors that modulate miRNA biogenesis have been
uncovered. It is predicted that there will be more alternative ways in
the future. Which pathways will be preferred in miRNA biogenesis
in different cellular stresses and finding out which adjunct factors
will play a role in these alternative ways will significantly change our
understanding of miRNA regulation.
However, with the removal of these question marks, many of
these parts will be completed, and it will undoubtedly revolutionize
our understanding of miRNA regulation.
*All the figures were redrawn and modified by Rahmi Çetin-
kaya after reading the references (especially [12], [36], [40], [78],
[109], [110], [126], [160] and [163]).
References
1. Perron MP, Provost P (2008) Protein inter- 5. Bartel DP (2009) MicroRNAs: target recog-
actions and complexes in human microRNA nition and regulatory functions. Cell 136
biogenesis and function. Front Biosci (2):215–233
13:2537–2547 6. Cai Y et al (2009) A brief review on the
2. Kozomara A, Birgaoanu M, Griffiths-Jones S mechanisms of miRNA regulation. Genomics
(2019) miRBase: from microRNA sequences Proteomics Bioinformatics 7(4):147–154
to function. Nucleic Acids Res 47(D1): 7. Macfarlane LA, Murphy PR (2010) Micro-
D155–D162 RNA: biogenesis, function and role in cancer.
3. Ambros V (2004) The functions of animal Curr Genomics 11(7):537–561
microRNAs. Nature 431(7006):350–355 8. Lu J et al (2005) MicroRNA expression pro-
4. Huntzinger E, Izaurralde E (2011) Gene files classify human cancers. Nature 435
silencing by microRNAs: contributions of (7043):834–838
translational repression and mRNA decay. 9. Lee Y et al (2004) MicroRNA genes are tran-
Nat Rev Genet 12(2):99–110 scribed by RNA polymerase II. EMBO J 23
(20):4051–4060
10. Pfeffer S et al (2005) Identification of micro- 25. Park SM et al (2008) The miR-200 family
RNAs of the herpesvirus family. Nat Methods determines the epithelial phenotype of cancer
2(4):269–276 cells by targeting the E-cadherin repressors
11. Kim VN, Han J, Siomi MC (2009) Biogenesis ZEB1 and ZEB2. Genes Dev 22(7):894–907
of small RNAs in animals. Nat Rev Mol Cell 26. Lee Y et al (2002) MicroRNA maturation:
Biol 10(2):126–139 stepwise processing and subcellular localiza-
12. Ha M, Kim VN (2014) Regulation of micro- tion. EMBO J 21(17):4663–4670
RNA biogenesis. Nat Rev Mol Cell Biol 15 27. Denli AM et al (2004) Processing of primary
(8):509–524 microRNAs by the microprocessor complex.
13. Ozsolak F et al (2008) Chromatin structure Nature 432(7014):231–235
analyses identify miRNA promoters. Genes 28. Shiohama A et al (2003) Molecular cloning
Dev 22(22):3172–3183 and expression analysis of a novel gene
14. He L et al (2007) A microRNA component of DGCR8 located in the DiGeorge syndrome
the p53 tumour suppressor network. Nature chromosomal region. Biochem Biophys Res
447(7148):1130–1134 Commun 304(1):184–190
15. Chen JF et al (2006) The role of microRNA-1 29. Yi R et al (2003) Exportin-5 mediates the
and microRNA-133 in skeletal muscle prolif- nuclear export of pre-microRNAs and short
eration and differentiation. Nat Genet 38 hairpin RNAs. Genes Dev 17(24):3011–3016
(2):228–233 30. Han J et al (2006) Molecular basis for the
16. Hwang JY et al (2014) The gene silencing recognition of primary microRNAs by the
transcription factor REST represses miR-132 Drosha-DGCR8 complex. Cell 125
expression in hippocampal neurons destined (5):887–901
to die. J Mol Biol 426(20):3454–3466 31. Nguyen TA et al (2015) Functional anatomy
17. Packer AN et al (2008) The bifunctional of the human microprocessor. Cell 161
microRNA miR-9/miR-9* regulates REST (6):1374–1387
and CoREST and is downregulated in Hun- 32. Lee Y et al (2003) The nuclear RNase III
tington’s disease. J Neurosci 28 Drosha initiates microRNA processing.
(53):14341–14346 Nature 425(6956):415–419
18. Visvanathan J et al (2007) The microRNA 33. Han J et al (2004) The Drosha-DGCR8 com-
miR-124 antagonizes the anti-neural REST/ plex in primary microRNA processing. Genes
SCP1 pathway during embryonic CNS devel- Dev 18(24):3016–3027
opment. Genes Dev 21(7):744–749 34. Kwon SC et al (2016) Structure of human
19. Krol J, Loedige I, Filipowicz W (2010) The DROSHA. Cell 164(1–2):81–90
widespread regulation of microRNA biogene- 35. Senturia R et al (2010) Structure of the
sis, function and decay. Nat Rev Genet 11 dimerization domain of DiGeorge critical
(9):597–610 region 8. Protein Sci 19(7):1354–1365
20. Sacar Demirci MD, Yousef M, Allmer J 36. Treiber T, Treiber N, Meister G (2019) Reg-
(2019) Computational prediction of func- ulation of microRNA biogenesis and its cross-
tional MicroRNA-mRNA interactions. Meth- talk with other cellular pathways. Nat Rev Mol
ods Mol Biol 1912:175–196 Cell Biol 20(1):5–20
21. Kim J et al (2007) A MicroRNA feedback 37. Auyeung VC et al (2013) Beyond secondary
circuit in midbrain dopamine neurons. Sci- structure: primary-sequence determinants
ence 317(5842):1220–1224 license pri-miRNA hairpins for processing.
22. Ben-Ami O et al (2009) A regulatory inter- Cell 152(4):844–858
play between miR-27a and Runx1 during 38. Zeng Y, Yi R, Cullen BR (2005) Recognition
megakaryopoiesis. Proc Natl Acad Sci U S A and cleavage of primary microRNA precursors
106(1):238–243 by the nuclear processing enzyme Drosha.
23. Zhao H et al (2009) The c-myb proto-onco- EMBO J 24(1):138–148
gene and microRNA-15a comprise an active 39. Calin GA, Dumitru CD, Shimizu M et al
autoregulatory feedback loop in human (2002) Frequent deletions and down-
hematopoietic cells. Blood 113(3):505–516 regulation of micro-RNA genes miR15 and
24. Li X, Carthew RW (2005) A microRNA med- miR16 at 13q14 in chronic lymphocytic leu-
iates EGF receptor signaling and promotes kemia. Proc Natl Acad Sci U S A 99
photoreceptor differentiation in the Drosoph- (24):15524–15529. https://doi.org/10.
ila eye. Cell 123(7):1267–1277 1073/pnas.242606799
40. Han J et al (2009) Posttranscriptional cross- translation to the cytoplasm. EMBO J 21

regulation between Drosha and DGCR8. Cell (22):6205–6215
136(1):75–84 55. Bennasser Y et al (2011) Competition for
41. Gregory RI et al (2004) The microprocessor XPO5 binding between dicer mRNA,
complex mediates the genesis of microRNAs. pre-miRNA and viral RNA regulates human
Nature 432(7014):235–240 dicer levels. Nat Struct Mol Biol 18
42. Herbert KM et al (2013) Phosphorylation of (3):323–327
DGCR8 increases its intracellular stability and 56. Sun HL et al (2016) ERK activation globally
induces a progrowth miRNA profile. Cell Rep downregulates miRNAs through phosphory-
5(4):1070–1081 lating Exportin-5. Cancer Cell 30
43. Hong S et al (2013) Signaling by p38 MAPK (5):723–736
stimulates nuclear localization of the micro- 57. Li J et al (2018) Pin1 impairs microRNA
processor component p68 for processing of biogenesis by mediating conformation change
selected primary microRNAs. Sci Signal 6 of XPO5 in hepatocellular carcinoma. Cell
(266):ra16 Death Differ 25(9):1612–1624
44. Tu CC et al (2015) The kinase ABL phos- 58. Wu K et al (2018) The role of Exportin-5 in
phorylates the microprocessor subunit MicroRNA biogenesis and cancer. Genomics
DGCR8 to stimulate primary microRNA pro- Proteomics Bioinformatics 16(2):120–126
cessing in response to DNA damage. Sci Sig- 59. Bernstein E et al (2001) Role for a bidentate
nal 8(383):ra64 ribonuclease in the initiation step of RNA
45. Tang X et al (2011) Glycogen synthase kinase interference. Nature 409(6818):363–366
3 beta (GSK3beta) phosphorylates the RNA- 60. Kanellopoulou C et al (2005) Dicer-deficient
ase III enzyme Drosha at S300 and S302. mouse embryonic stem cells are defective in
PLoS One 6(6):e20391 differentiation and centromeric silencing.
46. Zhu C et al (2015) SUMOylation at K707 of Genes Dev 19(4):489–501
DGCR8 controls direct function of primary 61. Zhang R et al (2018) Comprehensive evolu-
microRNA. Nucleic Acids Res 43 tionary analysis of the major RNA-induced
(16):7945–7960 silencing complex members. Sci Rep 8
47. Wada T, Kikuchi J, Furukawa Y (2012) His- (1):14189
tone deacetylase 1 enhances microRNA pro- 62. MacRae IJ, Doudna JA (2007) Ribonuclease
cessing via deacetylation of DGCR8. EMBO revisited: structural insights into ribonuclease
Rep 13(2):142–149 III family enzymes. Curr Opin Struct Biol 17
48. Yu B et al (2008) The FHA domain proteins (1):138–145
DAWDLE in Arabidopsis and SNIP1 in 63. Park JE et al (2011) Dicer recognizes the 50
humans act in small RNA biogenesis. Proc end of RNA for efficient and accurate proces-
Natl Acad Sci U S A 105(29):10073–10078 sing. Nature 475(7355):201–205
49. Gruber JJ et al (2009) Ars2 links the nuclear 64. Qin H et al (2010) Structure of the Arabidop-
cap-binding complex to RNA interference sis thaliana DCL4 DUF283 domain reveals a
and cell proliferation. Cell 138(2):328–339 noncanonical double-stranded RNA-binding
50. Trabucchi M et al (2009) The RNA-binding fold for protein-protein interaction. RNA 16
protein KSRP promotes the biogenesis of a (3):474–481
subset of microRNAs. Nature 459 65. Tsutsumi A et al (2011) Recognition of the
(7249):1010–1014 pre-miRNA structure by Drosophila Dicer-1.
51. Jiang CF et al (2016) Estrogen regulates Nat Struct Mol Biol 18(10):1153–1158
miRNA expression: implication of estrogen 66. Chendrimada TP et al (2005) TRBP recruits
receptor and miR-124/AKT2 in tumor the dicer complex to Ago2 for microRNA
growth and angiogenesis. Oncotarget 7 processing and gene silencing. Nature 436
(24):36940–36955 (7051):740–744
52. Sakamoto S et al (2009) The NF90-NF45 67. Fukunaga R et al (2012) Dicer partner pro-
complex functions as a negative regulator in teins tune the length of mature miRNAs in
the microRNA processing pathway. Mol Cell flies and mammals. Cell 151(4):912
Biol 29(13):3754–3769 68. Liu X et al (2007) Dicer-1, but not loqua-
53. Lund E et al (2004) Nuclear export of micro- cious, is critical for assembly of miRNA-
RNA precursors. Science 303(5654):95–98 induced silencing complexes. RNA 13
54. Bohnsack MT et al (2002) Exp5 exports (12):2324–2329
eEF1A via tRNA from nuclei and synergizes
with other transport pathways to confine
69. Lee HY et al (2013) Differential roles of LIM-domain-containing proteins. Cell Rep

human dicer-binding proteins TRBP and 20(1):173–187
PACT in small RNA processing. Nucleic 85. Yang M et al (2014) Dephosphorylation of
Acids Res 41(13):6568–6576 tyrosine 393 in argonaute 2 by protein tyro-
70. Lee Y et al (2006) The role of PACT in the sine phosphatase 1B regulates gene silencing
RNA silencing pathway. EMBO J 25 in oncogenic RAS-induced senescence. Mol
(3):522–532 Cell 55(5):782–790
71. Lee HY, Doudna JA (2012) TRBP alters 86. Quevillon Huberdeau M et al (2017) Phos-
human precursor microRNA processing phorylation of Argonaute proteins affects
in vitro. RNA 18(11):2012–2019 mRNA binding and is essential for
72. Hammond SM et al (2001) Argonaute2, a microRNA-guided gene silencing in vivo.
link between genetic and biochemical analyses EMBO J 36(14):2088–2106
of RNAi. Science 293(5532):1146–1150 87. Golden RJ et al (2017) An Argonaute phos-
73. Pratt AJ, MacRae IJ (2009) The phorylation cycle promotes microRNA-
RNA-induced silencing complex: a versatile mediated silencing. Nature 542
gene-silencing machine. J Biol Chem 284 (7640):197–202
(27):17897–17901 88. Qi HH et al (2008) Prolyl 4-hydroxylation
74. Sheu-Gruttadauria J, MacRae IJ (2017) regulates Argonaute 2 stability. Nature 455
Structural foundations of RNA silencing by (7211):421–424
Argonaute. J Mol Biol 429(17):2619–2639 89. Leung A et al (2012) Poly(ADP-ribose) reg-
75. Hutvagner G, Simard MJ (2008) Argonaute ulates post-transcriptional gene regulation in
proteins: key players in RNA silencing. Nat the cytoplasm. RNA Biol 9(5):542–548
Rev Mol Cell Biol 9(1):22–32 90. Leung AK et al (2011) Poly(ADP-ribose) reg-
76. Swarts DC et al (2014) The evolutionary ulates stress responses and microRNA activity
journey of Argonaute proteins. Nat Struct in the cytoplasm. Mol Cell 42(4):489–499
Mol Biol 21(9):743–753 91. Smibert P et al (2013) Homeostatic control
77. Niaz S (2018) The ago protein: an overview. of Argonaute stability by microRNA availabil-
Biol Chem 399(6):525–547 ity. Nat Struct Mol Biol 20(7):789–795
78. Gebert LFR, MacRae IJ (2019) Regulation of 92. Martinez NJ, Gregory RI (2013) Argonaute2
microRNA function in animals. Nat Rev Mol expression is post-transcriptionally coupled to
Cell Biol 20(1):21–37 microRNA abundance. RNA 19(5):605–612
79. Miyoshi K, Tsukumo H, Nagami T, Siomi H, 93. Forstemann K et al (2007) Drosophila micro-
Siomi MC (2005) Slicer function of Drosoph- RNAs are sorted into functionally distinct
ila Argonautes and its involvement in RISC argonaute complexes after production by
formation. Genes Dev 19(23):2837–2848 dicer-1. Cell 130(2):287–297
80. Su H et al (2009) Essential and overlapping 94. Czech B et al (2009) Hierarchical rules for
functions for mammalian Argonautes in Argonaute loading in Drosophila. Mol Cell
microRNA silencing. Genes Dev 23 36(3):445–456
(3):304–317 95. Hu HY et al (2009) Sequence features asso-
81. Zeng Y et al (2008) Phosphorylation of Argo- ciated with microRNA strand selection in
naute 2 at serine-387 facilitates its localization humans and flies. BMC Genomics 10
to processing bodies. Biochem J 413 (413):1–11
(3):429–436 96. Liu Y et al (2009) C3PO, an endoribonu-
82. Rudel S et al (2011) Phosphorylation of clease that promotes RNAi by facilitating
human Argonaute proteins affects small RISC activation. Science 325
RNA binding. Nucleic Acids Res 39 (5941):750–753
(6):2330–2343 97. Pham JW et al (2004) A Dicer-2-dependent
83. Horman SR et al (2013) Akt-mediated phos- 80s complex cleaves targeted mRNAs during
phorylation of argonaute 2 downregulates RNAi in drosophila. Cell 117(1):83–94
cleavage and upregulates translational repres- 98. Liu Q et al (2003) R2D2, a bridge between
sion of MicroRNA targets. Mol Cell 50 the initiation and effector steps of the Dro-
(3):356–367 sophila RNAi pathway. Science 301
84. Bridge KS et al (2017) Argonaute utilization (5641):1921–1925
for miRNA silencing is determined by 99. Liu X et al (2012) Precursor microRNA-
phosphorylation-dependent recruitment of programmed silencing complex assembly
pathways in mammals. Mol Cell 46 114. Boele J et al (2014) PAPD5-mediated 30 ade-

(4):507–517 nylation and subsequent degradation of
100. Schwarz DS et al (2003) Asymmetry in the miR-21 is disrupted in proliferative disease.
assembly of the RNAi enzyme complex. Cell Proc Natl Acad Sci U S A 111
115(2):199–208 (31):11467–11472
101. Matranga C et al (2005) Passenger-strand 115. Brennicke A, Marchfelder A, Binder S (1999)
cleavage facilitates assembly of siRNA into RNA editing. FEMS Microbiol Rev 23
Ago2-containing RNAi enzyme complexes. (3):297–316
Cell 123(4):607–620 116. Nishikura K (2016) A-to-I editing of coding
102. Mishra PJ et al (2008) MiRSNPs or and non-coding RNAs by ADARs. Nat Rev
MiR-polymorphisms, new players in micro- Mol Cell Biol 17(2):83–96
RNA mediated regulation of the cell: introdu- 117. Salter JD, Bennett RP, Smith HC (2016) The
cing microRNA pharmacogenomics. Cell APOBEC protein family: united by structure,
Cycle 7(7):853–858 divergent in function. Trends Biochem Sci 41
103. Jazdzewski K et al (2008) Common SNP in (7):578–594
pre-miR-146a decreases mature miR expres- 118. Yang W et al (2006) Modulation of micro-
sion and predisposes to papillary thyroid car- RNA processing and expression through
cinoma. Proc Natl Acad Sci U S A 105 RNA editing by ADAR deaminases. Nat
(20):7269–7274 Struct Mol Biol 13(1):13–21
104. Sun G et al (2009) SNPs in human miRNA 119. Kawahara Y et al (2007) RNA editing of the
genes affect biogenesis and function. RNA 15 microRNA-151 precursor blocks cleavage by
(9):1640–1651 the dicer-TRBP complex. EMBO Rep 8
105. Pasquinelli AE et al (2000) Conservation of (8):763–769
the sequence and temporal expression of let-7 120. Shoshan E et al (2015) Reduced adenosine-
heterochronic regulatory RNA. Nature 408 to-inosine miR-455-5p editing promotes
(6808):86–89 melanoma growth and metastasis. Nat Cell
106. Suh, Mi-Ra, et al., Human embryonic stem Biol 17(3):311–321
cells express a unique set of microRNAs. Dev 121. Xhemalce B, Robson SC, Kouzarides T
Biol, 2004. 15;270(2): p. 488–98 (2012) Human RNA methyltransferase
107. Newman MA, Thomson JM, Hammond SM BCDIN3D regulates microRNA processing.
(2008) Lin-28 interaction with the Let-7 pre- Cell 151(2):278–288
cursor loop mediates regulated microRNA 122. Ruegger S, Grosshans H (2012) MicroRNA
processing. RNA 14(8):1539–1549 turnover: when, how, and why. Trends Bio-
108. Mayr F et al (2012) The Lin28 cold-shock chem Sci 37(10):436–446
domain remodels pre-let-7 microRNA. 123. Sanei M, Chen X (2015) Mechanisms of
Nucleic Acids Res 40(15):7492–7506 microRNA turnover. Curr Opin Plant Biol
109. Heo I et al (2008) Lin28 mediates the termi- 27:199–206
nal uridylation of let-7 precursor MicroRNA. 124. Zhang Z et al (2011) Uracils at nucleotide
Mol Cell 32(2):276–284 position 9-11 are required for the rapid turn-
110. Faehnle CR, Walleshauser J, Joshua-Tor L over of miR-29 family. Nucleic Acids Res 39
(2014) Mechanism of Dis3l2 substrate recog- (10):4387–4395
nition in the Lin28-let-7 pathway. Nature 514 125. Marcinowski L et al (2012) Degradation of
(7521):252–256 cellular mir-27 by a novel, highly abundant
111. Katoh T et al (2009) Selective stabilization of viral transcript is important for efficient virus
mammalian microRNAs by 30 adenylation replication in vivo. PLoS Pathog 8(2):
mediated by the cytoplasmic poly e1002510
(A) polymerase GLD-2. Genes Dev 23 126. Heo I et al (2012) Mono-uridylation of
(4):433–438 pre-microRNA as a key step in the biogenesis
112. Burns DM, D’Ambrogio A, Nottrott S, Rich- of group II let-7 microRNAs. Cell 151
ter JD (2011) CPEB and two poly (3):521–532
(A) polymerases control miR-122 stability 127. Jones MR et al (2009) Zcchc11-dependent
and p53 mRNA translation. Nature uridylation of microRNA directs cytokine
473:105–108 expression. Nat Cell Biol 11(9):1157–1163
113. Katoh T, Hojo H, Suzuki T (2015) Destabi- 128. Ramachandran V, Chen X (2008) Degrada-
lization of microRNAs in human cells by 30 tion of microRNAs by a family of exoribonu-
deadenylation mediated by PARN and cleases in Arabidopsis. Science 321
CUGBP1. Nucleic Acids Res 43 (5895):1490–1492
(15):7521–7534
129. Chatterjee S, Grosshans H (2009) Active 144. Jeffries CD, Fried HM, Perkins DO (2011)
turnover modulates mature microRNA activ- Nuclear and cytoplasmic localization of neural
ity in Caenorhabditis elegans. Nature 461 stem cell microRNAs. RNA 17(4):675–686
(7263):546–549 145. Tang R et al (2012) Mouse miRNA-709
130. Neilsen CT, Goodall GJ, Bracken CP (2012) directly regulates miRNA-15a/16-1 biogen-
IsomiRs--the overlooked repertoire in the esis at the posttranscriptional level in the
dynamic microRNAome. Trends Genet 28 nucleus: evidence for a microRNA hierarchy
(11):544–549 system. Cell Res 22(3):504–515
131. Wu H et al (2009) Alternative processing of 146. Zisoulis DG et al (2012) Autoregulation of
primary microRNA transcripts by Drosha microRNA biogenesis by let-7 and Argo-
generates 50 end variation of mature micro- naute. Nature 486(7404):541–544
RNA. PLoS One 4(10):e7566 147. Sohel MH (2016) Extracellular/circulating
132. Llorens F et al (2013) A highly expressed MicroRNAs: release mechanisms, functions
miR-101 isomiR is a functional silencing and challenges. Achieve Life Sci 10
small RNA. BMC Genomics 14:104 (2):175–186
133. Karali M et al (2016) High-resolution analysis 148. Melo SA et al (2014) Cancer exosomes per-
of the human retina miRNome reveals isomiR form cell-independent microRNA biogenesis
variations and novel microRNAs. Nucleic and promote tumorigenesis. Cancer Cell 26
Acids Res 44(4):1525–1540 (5):707–721
134. Skalsky RL, Cullen BR (2010) Viruses, 149. Ahadi A et al (2016) Long non-coding RNAs
microRNAs, and host interactions. Annu harboring miRNA seed regions are enriched
Rev Microbiol 64:123–141 in prostate cancer exosomes. Sci Rep 6:24922
135. Bandiera S et al (2015) miR-122--a key factor 150. Schirle NT et al (2015) Water-mediated rec-
and therapeutic target in liver disease. J Hepa- ognition of t1-adenosine anchors Argonaute2
tol 62(2):448–457 to microRNA targets. elife 4:e07646
136. Jopling CL, Schutz S, Sarnow P (2008) 151. Nam JW et al (2014) Global analyses of the
Position-dependent function for a tandem effect of different cellular contexts on micro-
microRNA miR-122-binding site located in RNA targeting. Mol Cell 53(6):1031–1043
the hepatitis C virus RNA genome. Cell 152. Kedde M et al (2010) A Pumilio-induced
Host Microbe 4(1):77–85 RNA structure switch in p27-3’ UTR con-
137. Scheel TK et al (2016) A broad RNA virus trols miR-221 and miR-222 accessibility.
survey reveals both miRNA dependence and Nat Cell Biol 12(10):1014–1020
functional sequestration. Cell Host Microbe 153. Min KW et al (2017) AUF1 facilitates
19(3):409–423 microRNA-mediated gene silencing. Nucleic
138. Cazalla D, Yario T, Steitz JA (2010) Down- Acids Res 45(10):6064–6073
regulation of a host microRNA by a Herpes- 154. Ruby JG, Jan CH, Bartel DP (2007) Intronic
virus saimiri noncoding RNA. Science 328 microRNA precursors that bypass Drosha
(5985):1563–1566 processing. Nature 448(7149):83–86
139. Leung AKL (2015) The whereabouts of 155. Westholm JO, Lai EC (2011) Mirtrons:
microRNA actions: cytoplasm and beyond. microRNA biogenesis via splicing. Biochimie
Trends Cell Biol 25(10):601–610 93(11):1897–1904
140. Foldes-Papp Z et al (2009) Trafficking of 156. Okamura K et al (2007) The mirtron pathway
mature miRNA-122 into the nucleus of live generates microRNA-class regulatory RNAs
liver cells. Curr Pharm Biotechnol 10 in Drosophila. Cell 130(1):89–100
(6):569–578 157. Flynt AS et al (2010) MicroRNA biogenesis
141. Meister G et al (2004) Human Argonaute2 via splicing and exosome-mediated trimming
mediates RNA cleavage targeted by miRNAs in Drosophila. Mol Cell 38(6):900–907
and siRNAs. Mol Cell 15(2):185–197 158. Chong MM et al (2010) Canonical and alter-
142. Hwang HW, Wentzel EA, Mendell JT (2007) nate functions of the microRNA biogenesis
A hexanucleotide element directs microRNA machinery. Genes Dev 24(17):1951–1960
nuclear import. Science 315(5808):97–100 159. Xie M et al (2013) Mammalian 50 -capped
143. Liao JY et al (2010) Deep sequencing of microRNA precursors that generate a single
human nuclear and cytoplasmic small RNAs microRNA. Cell 155(7):1568–1580
reveals an unexpectedly complex subcellular 160. Hasler D et al (2016) The lupus autoantigen
distribution of miRNAs and tRNA 30 trailers. La prevents mis-channeling of tRNA
PLoS One 5(5):e10563
fragments into the human MicroRNA path- 162. Yoda M et al (2013) Poly(A)-specific ribonu-
way. Mol Cell 63(1):110–124 clease mediates 30 -end trimming of
161. Yang JS et al (2010) Conserved vertebrate Argonaute2-cleaved precursor microRNAs.
mir-451 provides a platform for dicer- Cell Rep 5(3):715–726
independent, Ago2-mediated microRNA 163. Slezak-Prochazka I, Kroesen BJ, Van Den
biogenesis. Proc Natl Acad Sci U S A 107 Berg A (2010) MicroRNAs, macrocontrol:
(34):15163–15168 regulation of miRNA processing. RNA 16
(6):1087–1095
Chapter 2
Experimental MicroRNA Detection Methods

Bilge Yaylak and Bünyamin Akgül
Abstract
MicroRNAs (miRNAs) are considerably small yet highly important riboregulators involved in nearly all
cellular processes. Due to their critical roles in posttranscriptional regulation of gene expression, they have
the potential to be used as biomarkers in addition to their use as drug targets. Although computational
approaches speed up the initial genomewide identification of putative miRNAs, experimental approaches
are essential for further validation and functional analyses of differentially expressed miRNAs. Therefore,
sensitive, specific, and cost-effective microRNA detection methods are imperative for both individual and
multiplex analysis of miRNA expression in different tissues and during different developmental stages.
There are a number of well-established miRNA detection methods that can be exploited depending on the
comprehensiveness of the study (individual miRNA versus multiplex analysis), the availability of the sample
and the location and intracellular concentration of miRNAs. This review aims to highlight not only
traditional but also novel strategies that are widely used in experimental identification and quantification
of microRNAs.
Key words miRNA, microRNA, Detection, RT-PCR, Quantification, Experimental
1 Introduction
miRNAs are short, single-stranded noncoding RNAs of 18 to

25 nucleotides that play vital roles as micromanagers of gene
expression through promoting mRNA degradation or translational
repression [1]. The canonical miRNA biogenesis pathway in human
is a two-step process that involves two ribonuclease III endonu-
cleases, namely, Drosha and Dicer, which perform the nuclear and
cytoplasmic cleavage events, respectively. Firstly, the primary
miRNA (pri-miRNA) is transcribed from a miRNA gene to
pri-miRNA by RNA polymerase II and is processed first into a
precursor miRNA (pre-miRNA) in the nucleus [2]. Then, ribonu-
clease III endonuclease Drosha, the double-stranded RNA binding
protein DGCR8 (DiGeorge Syndrome Critical Region 8), and its
associated proteins as a microprocessor complex further processes
the transcript into a ~70-nucleotide stem-loop precursor called
pre-miRNA. Pre-miRNAs are immediately exported to the cytosol
33
34 Bilge Yaylak and Bünyamin Akgül
by the Exportin-5/RanGTP complex. In cytoplasm, final proces-

sing events are carried out by Dicer, resulting in the production of
miRNA duplexes [3]. The strand that is more unstable at its 50 -end
is incorporated into miRNA-induced silencing complex (miRISC)
with Argonaute family of protein (AGO1-4 in humans). miRISC
assembly is initiated through the glycine–tryptophan protein of
182 kDa (GW182), which then recruits other effector proteins
such as PAN2/PAN3 and CCR4-NOT [4–6]. In the noncanonical
pathway, miRNA biogenesis can be independent of either Drosha/
DGRC8 or Dicer with different sets of proteins carrying out the
biogenesis [7–9]. Based on the complementarity between the
miRNA and its target, the target mRNA is destabilized or transla-
tionally silenced [10].
Up to date, nearly 2000 miRNAs have been identified (miR-
Base) in human [11]. Remarkably, as a single mRNA can be regu-
lated by multiple miRNAs, a single miRNA may base-pair with and
regulate more than one target mRNA [12, 13]. Moreover, the
targeting process is stochiometric so there is a competition among
miRNAs with multiple targets. It has been reported that almost all
KEGG pathways in human can be regulated by miRNAs. In fact,
there is an elaborate regulation of miRNA activity both at the
transcriptional and posttranscriptional level to maintain cellular
homeostasis [1, 10, 14]. The literature abounds in a number of
important diseases associated with the aberrant expression of miR-
NAs [15]. Thus, not only to understand the molecular mechanisms
that underlie the functions of miRNAs but also to delineate the
contribution of miRNA activity in a disease pathology, it is essential
to measure the expression levels of miRNAs both in vivo and
in vitro [16]. Accurate and sensitive quantification of miRNAs
may facilitate the accurate diagnosis of diseases, establishing these
short RNA transcripts as valuable molecular markers [17]. Once
the molecular mechanisms of miRNA-mediated regulation of vari-
ous biological pathways is uncovered, miRNA-based therapies may
be developed to manipulate these biological processes.
Identification and quantification of miRNAs are challenging
due to their short length, O-methyl 30 modifications, the existence
of miRNA isomers, and high homology within miRNA families
[18]. Thus, new miRNA detection methods are emerging to over-
come these challenges. This chapter encompasses both conven-
tional and novel miRNA detection and quantification methods,
and their strengths and weaknesses.
2 Conventional Methods for miRNA Detection and Quantification
2.1 Northern Blotting Northern blotting constitutes the gold standard conventional
approach for validating miRNA expression data because it does
not require any sophisticated equipment or technical knowledge.
Experimental microRNA Detection 35
Table 1
Northern blot based methods for microRNA detection
Method Level of Detail Advantages References

0
3 -digoxigenin-labeled RNA 50 ng Equal sensitivity, short exposure [24]
oligo probes (DIG) times, longer shelf life, safe.
(compared with 32P-labeled
probes)
LNA probes 2.5 μg High sensitivity, time saver [21]
EDC (1-ethyl-3- 10,000 times short Increased stability and sensitivity. [25]
(3-dimethylaminopropyl) exposure time (compared with UV cross-link
carbodiimide)
DIG-labeled LNA 0.05 fmol Very high sensitivity, safe [26]
oligonucleotide probes and
membranes cross-linked
with 1-ethyl-3-
(3-dimethylaminopropyl)
carbodiimide (EDC)
High-resolution northern blot Simultaneous Fill the pre-miRNA detection gaps [27]
analysis of of other protocols
pre-miRNA and
mature miRNAs
It has been historically used to differentiate RNAs based on the

difference in RNA length [19]. The analysis of RNA by Northern
blotting entails three steps. The first step involves the seperation of
RNAs based on their size through agarose gel electrophoresis. As
single-stranded polynucleotides, most cellular RNAs have the ten-
dency to form secondary structures through intramolecular base-
pairing that prevent the fractionation of RNAs according to their
size. Thus, a denaturing agent, such as formaldehyde or glyoxal/
DMSO, is used to disrupt the secondary structures so that the
electrophoresis becomes directly proportional to the size of
RNAs. In the second step, fractionated RNAs are transferred onto
a nylon membrane ensuring the same pattern as in the gel. RNAs
must be cross-linked to the membrane to prevent their loss during
subsequent manipulations. In the thirt step, a labeled antisense
probe is hybridized to the filter containing the fractionated
RNAs. The hybridization conditions must be emprically tested to
optimize the probe:target hybridization. Nonspecifically bound
labeled probes are then washed away through the use of washing
buffers of varying stringency [20].
Northern blotting is typically used to determine the abundace
and the size of miRNA (mature versus precursor) due to its practi-
cality and reliability [21]. Although the fractionation of RNAs on
an agarose or polyacrylamide gel can be cumbersome, it allows for
the simultaneous detection of miRNAs of varying sizes and
different types such as pri-miRNAs, pre-miRNAs or mature miR-

NAs. The simplicity of the technique facilitates the analysis of
miRNA expression almost in any laboratory setting. However,
there are several drawbacks associated with this method. First of
all, this method does not support high-throughput analyses of
miRNA expression. Northern blotting also fails to distinguish
between two miRNAs with the same or similar molecular weight.
For example, two different miRNAs composed of slightly different
nucleotides on the same lane will have the same or similar molecular
weight and northern blotting might classify them as one due to
cross-hybridization [22]. Further, different miRNA-specific labeled
antisense RNA or ssDNA should be prepared for the analysis of
different miRNAs, requiring the casting of as many gels as the
number of miRNAs to be analyzed. Another disadvantage of this
technique is its low sensitivity, requiring the use of a relatively large
amount of starting material or a more sensitive detection method
that might involve hazardous radioactive isotopes. The need for a
high amount of starting material may be cumbersome under the
circumstances where the source RNA can be difficult to obtain,
such as precious patient tissues or mutant samples. Additionally,
this method is relatively time-consuming as it requires fractionation
of RNAs on a gel followed by the transfer of RNAs onto a filter and
prolonged hybridization and washing steps [23].
The conventional northern blotting technique has been
improved by optimizing cross-linking methods, electrophoresis
steps and probe design strategies (Table 1). Ramkissoon and col-
leagues reported the use of 30 -digoxigenin-labeled RNA oligo
probes as a nonradioactive alternative to 32P-labeled probes since
both oligo probes were equally sensitive in detecting miRNAs as
low as 50 ng in quantity [24]. In 2004, Valoczi et al. used locked
nucleic acid (LNA™) probes for miRNA detection through north-
ern blotting. Compared to the use of traditional oligonucleotide
probes, this approach has overcome two major disadvantages of
traditional northern blotting, poor sensitivity, and time consump-
tion. With the use of LNA™-based probes, hybridization, washing
and signal detection steps could be completed within 4 h [21]. As
an alternative hybridization method, EDC [1-ethyl-3-(3-dimethy-
laminopropyl) carbodiimide] has been suggested by Pall et al. to
increase hybridization stability [25]. Further improvement in a
sensitive and nonradioactive northern blot method was reported
by Kim et al. [26]. They used digoxygenin-labeled LNA™ oligo-
nucleotide probes and membranes cross-linked with 1-ethyl-3-
(3-dimethylaminopropyl) carbodiimide (EDC). This approach
reduces the exposure time (nearly 1000-fold) and the amount of
input RNA required, facilitating the visualization of miRNA as low
as 0.05 fmol. Koscianska et al. reported a high-resolution northern
blot as a means for the analysis of mature miRNAs and their
pre-miRNAs [27]. By optimizing the electrophoresis step, the
resolution of the blot was enhanced to a near-nucleotide level.

Considering the small size of miRNAs, the highest-possible resolu-
tion is required to discriminate between miRNAs that differ solely
with one or a few nucleotides in size. This is highly important in the
validation of data obtained from high-throughput approaches such
as deep-sequencing. It should be noted that only the most abun-
dant miRNA variants can be detected by high-resolution northern
blotting [28]. Perhaps the most important criterion in miRNA
detection by Northern blotting is the development and use of a
sensitive probe-design strategy. For example, LNA™-modified
RNA probes offer tenfold higher sensitivity compared with the
traditional oligo probes. Additionally, the sensitivity of northern
blotting is affected by the the type of cross-linking. Therefore, EDC
[1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride]
has been used for RNA-nylon membrane cross-linking instead of
conventional methods such as UV cross-linking, which is less effi-
cient for small RNA detection [23].
2.2 PCR-Based Reverse transcriptase quantitative polymerase chain reaction

miRNA Amplification (RT-qPCR) is a conventional nucleic acid amplification method
widely used in gene expression studies and accepted as the ‘gold
standard’ for miRNA profiling [29]. This approach is based on the
conversion of miRNAs to cDNA followed by real-time measure-
ment of product amplification. Basically, the cycle threshold value
of each template and the logarithm of the starting quantity are
directly proportional to each other in RT-qPCR. When compared
to qualitative northern blotting, RT-qPCR provides absolute quan-
tification [22]. The fundamental variables in miRNA detection
methodology namely, size, cost, and precision are balanced in this
method. Due to its low cost, convenience and especially precision,
it is selected as the method of choice for validating the expression of
candidate miRNAs identified by other high-throughput approaches
such as microarray or RNA-seq [30]. Although this approach is
typically limited to the analysis of individual miRNAs, a careful
primer design (e.g., labeling with various fluorescent dyes) may
facilitate the analysis of multiple miRNAs in a single reaction. It is
also important to keep in mind that this technique allows only for
the analysis of known miRNAs, without any possibility of novel
miRNA discovery [22].
A number of different approaches have been reported in the
PCR-based analysis of mature miRNA expression. As standard and
quantitative PCR methods require a template whose size is at least
twice as big as the size of a standard forward or reverse primer, the
size of mature miRNAs is too short to be amplified through the
standard PCR procedures. Thus, all PCR-based miRNA analyses
require miRNA size extension to facilitate reverse transcription
followed by PCR amplification. Special assay reagents and primers
(forward primer and reverse primer) are also optimized to increase
Fig. 1 Stem-loop primer based miRNA detection with the TaqMan probe by RT qPCR
miRNA specificity and assay accuracy. Stem-loop RT is a feasible

approach to properly amplify mature miRNAs in which the target
miRNA binds to a stem-loop probe for cDNA synthesis (Fig. 1)
[31, 32]. However, the stem-loop method requires specific reverse
transcription primers and a unique probe for every miRNA assay,
which adds to the cost of the analysis of a large number of miRNAs.
Gaur et al. reported the use of stem-loop primers following reverse
transcription, which increases the sensitivity of the assay
[33]. Another commonly used method to extent miRNA size
takes advantage of tailing by a stretch of A residues through poly-
A polymerase (PAP) followed by oligo(dT)-based cDNA prepara-
tion (Fig. 2) [34]. The stem-loop method has superiority due to its
Fig. 2 Poly(A)-tailing-based method for miRNA quantification
high sensitivity but the poly-A tail method surpasses the stem-loop
method in its specificity. However, both poly-A tail and stem-loop
methods can be performed to detect highly abundant miRNAs at
comparable expression levels [32]. The poly-A tail method can be
desirable if the amount of the starting material is low but it is
suitable neither for detecting 20 -oxymethyl-modified plant miR-
NAs nor for distinguishing mature miRNAs from pre-miRNAs
[23]. Mature miRNAs can be quantified rapidly by RT-PCR by
using deoxyuridine-incorporated oligonucleotides and heminested
primers in conjunction with a fast thermal cycler. This method is
capable of detecting miRNAs from minute amounts of RNA as low

as 1 pg, or from only 10 cells [35].
PCR-based techniques are typically employed to analyze
mature miRNAs. However, they can also be used to distinguish
mature miRNAs from their precursor mRNAs and highly homolo-
gous family members by taking advantange of RT oligonucleotides
with a secondary structure and heminested reverse PCR primers
[35]. Specifically designed bidirectional extension sequences may
be hybridized to cDNA after reverse transcription. This hybrid is
then used as a template in the PCR reaction for the specific detec-
tion of miRNAs. Compared to the poly-A-tailing-based method,
this type of extension has various advantages such as more clear gel
electrophoresis images, lower CT values and distinct peak visualiza-
tion in sequencing chromatograms [36]. Yet another sensitive and
cost-effective method, called two-tailed RT-PCR, was reported in
2017, which is unique due to the use of target-specific primers
[37]. These primers consist of two hemiprobes complementary to
two distinct parts of the targeted miRNA, connected by a hairpin
structure [37].
There are two options for the detection of PCR products:
SYBR chemistry or TaqMan® [38]. SYBR green is a nonspecific
fluorescent dye that intercalates any dsDNA and its fluorescence is
proportional to the amount of dsDNA, making it possible to
quantitate the amount of the final dsDNA PCR product. The
SYBR green fluorescent dye binds to any dsDNA without the
requirement for additional probes, simplifying the assay setup and
reducing the detection costs. However, one major drawback of the
SYBR green is its cancerogenicity that stems from its intercalating
property [39]. Thus, caution must be practiced while handling
solutions that contain this substance. More importantly, false-
positive signals might arise from dsDNA sequences produced
from off-target templates. Although most off-target products may
be easily spotted based on their different sizes, the off-target pro-
ducts of sizes similar to those of the target product might compli-
cate the interpretation of the results. Typically, a melting curve
analysis is advised at the end of the PCR run to identify potential
nontarget sequences [40].
TaqMan® chemistry was developed to elevate the specificity of
qPCR by eliminating potential off-target signals [41]. This detec-
tion method involves the use of a sequence-specific fluorescent
reporter that generates fluorescent signals only after hybridization
of the probe with a complementary sequence. Prior to amplifica-
tion, the reporter and quencher are in close proximity to each other
in TaqMan® technology. At this stage, energy gets transferred from
the reporter to the quencher (fluorescence resonance energy trans-
fer), suppressing the permanent increase in the fluorescence signal
from the reporter. After the denaturation step, the probe binds
miRNA template specifically and DNA polymerase with a 50
nuclease activity starts to chew on the reporter while creating a new

complementary sequence. This process in turn separates the 50
reporter dye from the 30 quenching dye, resulting in an increase
in the fluorescence signal proportional to the amount of the tem-
plate nucleic acid. The use of a target-specific fluorogenic primer
ensures the quantitation of only target sequences, disregarding the
nontarget sequences (Fig. 1). If the reporter probes are labeled
with different fluorescent dyes, multiple targets may be amplified
in a single assay without the requirement for post-PCR processing.
However, the need for an additional primer with a fluorescence dye
increases the cost [41].
Droplet digital PCR (ddPCR) improved the precision and
robustness of PCR-based techniques by partitioning individual
analyte miRNA molecules and PCR reagents into individual
water-oil emulsion droplets [42]. PCR amplification is performed
within each nanoliter-sized droplet that requires a dramatically
reduced amount of sample. The massive partitioning of the sample
generates data points from thousands of droplets, increasing the
statistical power of the data and eliminating the potential bias of
PCR. Additionally, absolute quantification can be attained without
the requirement for calibration standarts or a reference gene as
required by conventional qPCR methods. This method can be
used in combination with other approaches to escalate the sensitiv-
ity of detection. For example, the use of ligation-based PCR in
conjunction with ddPCR has led to the precise quantification of
miRNAs of low abundance or within single cells [43]. Further
improvements in the digital PCR technology resulted in the dis-
covery of a chip-based quantStudio 3D digital PCR technology
that uses a sealed chip technology for absolute quantification of
miRNAs in a streamlined and robust manner [44].
2.3 In Situ Most in vitro approaches are quite precise and practical with respect
Hybridization to the analysis of miRNA expression in test tubes or on chips.
However, in vitro analyses do not provide information about the
spatial distribution or intracellular location of miRNAs. For spatio-
temporal expression or intracellular transport studies, it is essential
to examine the miRNA expression in vivo. In situ hybridization
(ISH) is a hybridization-based detection method that can be used
to visualize or quantify miRNAs inside the cell [45]. Detection of
miRNAs at the cellular level can be applied to histological samples
as well in order to analyze the existing tissues in an inventory.
Basically, in situ hybridization-based miRNA detection provides a
single cell level information about the physiological function of
miRNA at the spatial location. As with many other miRNA detec-
tion methods, the short length of miRNAs makes it difficult to
generate a probe with high specificity. Thus, traditional RNA or
DNA probes typically display low affinity and specificity to short
targets. From the currently existing ISH-based miRNA detection
Fig. 3 Schematic representation of the fluorescent and chemogenic detection of miRNAs by in situ
hybridization
methods, the most widely accepted method involves the use of

LNA™ modified DNA probes for ISH-based miRNA detection
[46]. The use of LNA™ modified DNA probes increases the speci-
ficity by enhancing their hybridization efficiency with RNA mole-
cules, which can also be applied to formalin-fixed and paraffin-
embedded tissues [47]. LNA™ probes that undergo special chem-
ical modifications facilitate hybridization at higher melting temper-
ature that results in greater thermodynamic stability [47, 48]. Soe
et al. reported a modified version of ISH-based miRNA detection
method that can be employed to analyze low copy number miR-
NAs. Specifically, probes that contain 20 -O-methyl RNAs (2OMe)
and LNA™ at every third base (2:1 ratio) enhance miRNA detec-
tion sensitivity [49]. Similar to their use in Northern blotting,
DIG-labeled LNA™ probes are useful in ISH detection of miRNAs
with improved sensitivity. DIG-labeled probe is recognized by an
alkaline phosphatase (AP)-conjugated antibody that can be used to
quantify the template miRNA amount [50]. Detection of the signal
can be also attained by fluorescence in situ hybridization (FISH)
method. The FISH method is modified for miRNA detection by
using LNA™ probes labeled with DIG at their 50 ends. After
hybridization and washing steps, an anti-DIG antibody, a HRP
(horseradish peroxidase)-conjugated secondary antibody and the
HRP substrate cyanine-5-conjugated tyramide are added onto the
sample to obtain a signal that can be recorded by fluorescent
microscopy [51] (Fig. 3).
Fig. 4 Overview of miRNA profiling by biotin-labeling based microarray
2.4 miRNA A DNA microarray is a platform that contains thousands of micro-

Microarrays scopic spots on a solid surface or beads, which is used to genotype a
bulk of the genome or to quantitatively analyze the expression
pattern of a large number of genes simultaneously (Fig. 4)
[52]. Microarray-based expression analysis is one of the traditional
hybridization-based yet high-throughput strategies for the quanti-
tative analysis of miRNA candidates under any cellular conditions
[53]. Thus, miRNAs whose expression patterns correlate with
biological pathways can be used as biological markers for diseases
[54]. Although the microarray is useful for the simultaneous
profiling a large set of miRNAs, it only permits the analysis of a
known set of miRNAs, preventing the discovery of novel miRNAs.
Additionally, further validation is required to eliminate false-
positive signals that may stem from cross-hybridization [22].
Current microarray-based strategies have been improved in the
probe design and miRNA labeling to overcome the challenges
associated with the low abundance and short length of miRNAs
[23]. Acquisition of Tm(melting temperature)-normalized probe
sets is quite challenging for the genomewide expression profiling of
miRNAs. LNA™-modified capture probes were designed as an
alternative to eradicate this problem [55]. Additionally, exogenous
and endogenous positive and negative control probes are included
in the chip design to increase sensitivity and accuracy during hybri-
dization and detection [23]. Because labeling is a key step of
microarray, various labeling approaches have been developed that

include direct labeling with a fluorescent dye as well as indirect
labeling of a reverse transcript or PCR product of miRNA rather
than miRNA itself. Although direct labeling is adventageous in that
it prevents artificial errors introduced by PCR reaction or reverse
transcription, indirect labeling might be useful while working with
scarce and unstable miRNAs [56].
Inherent to the nature of miRNAs, there are several challenges
that need to be taken into account while using traditional micro-
array platforms. Firstly, oligo(dT)-based reverse transcription can-
not be used to label miRNAs because mature miRNAs lack a poly
(A) tail. The presence of a hydroxyl (-OH) group at the 30 -end may
be exploited to tail mature miRNAs with terminal deoxynucleotide
transferase (TdT) and poly(A) polymerase (PAP). In this method, a
poly(U) tail with either biotin-conjugated or amine-modified
nucleotides is first appended to the 30 end of the mature miRNAs
by PAP enzyme. Then, labeled miRNAs are hybridized to miRNA
specific probes on chips. The major advantage of this labeling
technique is highly sensitive miRNA detection through the
extented amine- or biotin-modified tail [57]. Regardless of the
types of nucleotides used to tail the mature miRNAs, labeling by
tailing leads to variations in the labeling efficiency due to variations
in the tail length. More homogenously labeled 30 -ends can be
attained by directly labeling the 30 end of the miRNAs with Cy3,
using T4 RNA ligase [58]. In this approach, T4 RNA ligase can be
stimulated by dimethylsulfoxide to minimize the interference of
sequence and structure differences among miRNAs, further
improving the efficiency of labeling [23]. Importantly, most plants
are methylated at the 20 position of the 30 end of the miRNAs,
which reduces ligation efficiency [58]. Several commercial plat-
forms such as Agilent (stem-loop sequence-based) and Exiqon
(LNA™-based) probes have been designed to eliminate melting
problems of miRNA probes that originate from a short length of
miRNAs [58]. Chemical alkylation-based labeling and platinum
coordination chemistry with nucleic acids are alternative chemical
approaches of miRNA labeling [59]. However, these chemical-
based methods are nonresponsive to 30 end modifications [60].
Microarray is a very useful method for high-throughput
miRNA quantification. However, the short size of miRNAs com-
plicates the specificity of probes with a possibility of cross-
hybridization or background signals. Therefore, next-generation
sequencing has been developed to circumvent these disadvantages
of the microarray, which has also paved the way for genomewide
discovery of novel miRNAs [61].
2.5 Next Generation Small RNA-seq takes advantage of the high-throughput capability
Sequencing of next-generation sequencing to discover novel miRNAs in addi-
tion to the analysis of curated miRNAs [62]. It is a comprehensive
approach that has the highest resolution possible [22]. In this

approach, small RNAs of the desired size are first purified from a
polyacrylamide gel to enrich for the small RNAs [63]. 50 and 30
adapters are then ligated to the purified RNAs before cDNA syn-
thesis followed by sequencing. This method is highly sensitive as it
can detect miRNAs expressed at low levels as long as sufficient
sequencing depth can be attained. Perhaps one of the most impor-
tant advantages of this approach is that, unlike all other detection
methods, it permits novel miRNA discovery. Additionally,
RNA-seq is superior in distinguishing miRNAs from the same
family or those that have high homology. Due to its high resolution
at the nucleotide level, RNA-seq is also quite useful in genomewide
screening of miRNA editing events [64]. Recent studies have
shown that there is an extensive heterogeneity both at the 50 - and
30 -ends of miRNAs, generating miRNA isoforms of different func-
tions [65]. RNA-seq easily identifies all variations in miRNA
sequences, including editing and posttranscriptional additions
[66]. The major disadvantage of this method is that it requires
sophisticated equipments and expertise in bioinformatics data anal-
ysis to process the massive NGS data [62]. Although acquisition of
large data sets at the nucleotide resolution level can be attractive,
the process is very long and not useful for routine clinical tests.
PCR-based library preparation may introduce bias especially if the
templates are rich in their GC contents. Thus, GC-rich miRNAs
may be underrepresented in the final reads [61].
3 MicroRNA Detection Based on PCR-Free Signal Amplification
miRNA detection with northern blotting, RT-PCR, microarray,

and NGS have limitations despite their extensive use in miRNA
profiling. Microarray and NGS are very sensitive and comprehen-
sive yet costly. RT-PCR and northern blotting are low throughput
and do not facilitate novel miRNA discovery. Furthermore, north-
ern blotting is a laborious and cumbersome methodology that
requires very specific labeling. Isothermal amplification-, rolling
cycle amplification- and nuclease-mediated amplification-based
miRNA identifications are PCR-free signal amplification methods
that circumvent many of the limitations associated with conven-
tional methods [67].
3.1 Exponential Exponential isothermal amplification reaction (EXPAR) is a novel

Isothermal method used to quantify miRNA expression levels in attomolar
Amplification sensitivity through a two-stage exponential amplification reaction
for miRNA Detection and a time-resolved fluorescence sensor in real samples [68]. In this
method, a nicking enzyme is used to cleave the polymerase-
extended product at a specific site, releasing relatively smaller pro-
ducts at an exponential rate that can serve as primers in the second
reaction in a similar fashion. The advantages of this approach

include the use of isothermal conditions, high amplification effi-
ciency, and relatively faster amplification rates. Basically, this
method allows for amplification of a trace amount of target miR-
NAs [67]. Under isothermal conditions, a wide range of miRNA
amount can be selectively quantitated at the 10 orders of magnitute
without any requirement for a modified DNA probe [69]. It is so
selective in discriminating among the same miRNA family members
that two miRNAs with a single nucleotide difference can be accu-
rately discriminated. With these advantages, it offers an important
alternative to traditional approaches for the detection of miRNAs.
EXPAR can rapidly amplify short oligonucleotides within minutes
and exhibits good amplification efficiency and rapid kinetics of
amplification compared to PCR through its isothermal
[67, 70]. Unfortunately, the use of SYBR Green I for the detection
of reaction products does not permit the simultaneous analysis of
multiple amplicons. A novel time-resolved fluorescence method
that uses anthanide ion compounds as a signal indicator was devel-
oped to overcome this problem [69].
Li’s group designed a typical EXPAR in which the target
miRNA was used as a trigger to initiate amplification [71]. Firstly,
the template sequence is designed in such a way that it possesses
two complementary sequences (upstream and downstream) to the
target miRNA and a nicking site. The downstream complementary
sites are released by cleavage after amplification and serve as a
trigger for downstream amplifications. Therefore, the target
miRNA is amplified exponentially and detected by fluorescence
dye SYBR Green I [71] (Fig. 5).
Fig. 5 Schematic illustration of isothermal amplification for miRNA detection

3.2 Rolling Cycle Rolling circle amplification (RCA) is another well-known isother-
Amplification mal reaction widely used for DNA and RNA amplification with
for miRNA Detection high specificity and sensitivity, eliminating the need for optimiza-
tion of rather short primer sequences required in PCR-based
approaches. RCA uses a padlock probe that is circularized following
the binding of a target sequence complementary to the 50 - and
30 -ends of the padlock probe [72]. The target miRNA binds to the
ends of the padlock probe and circularizes it, which is then ligated
and extended by a T4 RNA ligase 2 and phi 29 DNA polymerase,
respectively. Tandem repeats of the target miRNAs are then cleaved
by a nickase, producing a source of target miRNA for the
subsequent RCA reactions. The inclusion of a fluorescent-labeled
third primer improves the RCA-based detection of miRNAs
[73]. There are several advantages associated with this approach.
First of all, the shortness of miRNA sequence is no longer an issue
as short miRNAs are highly efficient in inducing padlock probe
circularization. Second, the target miRNA used to circularize the
padlock probe seves as a primer to synthesize the reaction product,
which remains localized with the target miRNA. The target–prod-
uct colocalization facilitates the use of this technique for the in situ
detection of miRNAs [74]. If a nickase is skipped in the reaction,
the long tandem repeat products can be exploited in various detec-
tion methods such as fluorescence, colorimetry or gel electropho-
resis [73, 75]. Signal-based quantifications enhance sensitivity
without amplifying the number of target miRNAs so it provides a
more direct route to quantify miRNAs [76]. However,
amplification-based methods provide a much higher sensitivity
especially in miRNA detection studies because short miRNAs are
more suitable for ligation reactions.
The sensitivity of RCA was improved by designing dumble-
shaped DNA seal probes with a toehold on its loop [77]. The
binding of the target miRNA onto the toehold domain of the
dumble-shaped probe induces the circularization of the probe,
initiating the RCA process. This approach improves the sensitivity
in that even a single nucleotide mismatch between the target
miRNA and the toehold domain results in cessation of circulariza-
tion. The long and cascaded products may be exploited for in situ
imaging of miRNAs in cells or various tissues [74]. The sensitivity
of RCA can also be increased by using nicking endonucleases that
cut one strand of double-stranded DNA in a sequence-specific
manner [78]. There are several options available in this technique.
Liu et al. (2013) used a linear rolling circle approach following the
target-miRNA-mediated circularization of the padlock probe.
Then, multiple padlock probes were hybridized to the cascaded
linear product for cleavage by a nickase that generates new primers
to be used in the subsequent RCA reactions [78].
3.3 Nuclease- Certain endonucleases and exonucleases specifically cut DNA

Mediated within the DNA:RNA duplexes. This property of nucleases can be
Amplification exploited in amplification studies when coupled with nuclease-
for miRNA Detection mediated degradation of DNA probes. In a simple, sensitive and
specific manner, fluorescent, colorimetric or electrochemical meth-
ods can be used to detect and quantify the nuclease-degraded DNA
probes [79, 80]. The use of different fluorophores makes it possible
to analyze multiple miRNAs in solutions with the aid of a multi-
color sensor. Duplex-specific nuclease is another example of nucle-
ase that selectively digests DNA in the double-stranded DNA:
miRNA duplexes but not single-stranded DNA [81]. This
approach is widely used to detect miRNAs in combination with
TaqMan™ probes, molecular beacons or nanoparticles. Addition-
ally, enzyme-free methods have been developed in which the target
miRNA directly binds a DNA hairpin probe carrying a fluorophore
at its sticky end [82].
4 miRNA Detection Relying on Nanomaterials and/or Nanotechnology
4.1 Conventional Nanomaterials are promising tools in detection of nucleic acids due
Nanoparticles to their sensitivity and efficiency in particular when they are com-
bined with electrical approaches, microarrays or surface plasmon
resonance (SPR) [83]. Nanoparticles are highly flexible and modu-
lar structures and have several desirable features such as biocompat-
ibility, small size with a high surface area, and good connectivity
that help improve the intensity of the signal dramatically. The
examples of nanostructures include gold nanoparticles, silver nano-
particles, magnetic nanoparticles, carbon-based nanomaterials,
quantum dots, and metal-organic frameworks [84]. Identification
and evaluation of miRNAs by nanomaterials involves a molecular
recognition occasion, typically a transduction process coupled to
the hybridization of a nucleic acid probe with a miRNA strand. The
conventional base pairing between the probe and the target miRNA
results in a measurable signal. Transduction process is the funda-
mental difference between conventional and nanotechnology-
based methods, in which the special physicochemical properties of
nanoparticles are vital in improving the signal readout [85].
Magnetosensors, mangenic beads coated with p19 protein, are
able to detect endogenous miRNAs and as low as fmol of synthetic
targets [86]. The p19 protein selectively binds to short RNA
duplexes within the range of miRNAs, serving as a target miRNA
recognition receptor. This approach has been further improved by
using streptavidin to capture biotin-labeled miRNAs. The Strep-
HRP polymer system is then conjugated to the p19beads-miRNA
system to generate the signal read-out. Without a requirement for
the target amplification steps, this technique presents itself as a
simple and convenient method that can be combined with other
methods such as multiplexed target detection and arrays [87].
Gold nanoparticles (AuNPs) have specific optical properties

and they are employed as carriers of miRNA in cell transfection
due to their low toxicity [88]. Polyalanine nanowire network was
developed as a sensitive electrical biosensor to detect miRNAs
directly in a spectrum from 10 fM to 20 pM [89]. Isoniazid-capped
OsO2 nanoparticles were used to detect miRNA ultrasensitively
through its electrocatalytic properties [90]. Likewise, ruthenium
oxide nanoparticles were developed as highly sensitive biosensors
for miRNA detection [91]. Electrochemical detection of miRNA
was performed by using conducting magnetic microbeads and
ferrocene-capped AuNPs generated only in the presence of a
miRNA target [92]. Moreover, nanoparticle-assisted signal ampli-
fication strategy was performed to detect miRNAs through cou-
pling fluorescent metal nanoclusters [93]. Further, Takalkar et al.
developed a sensitive approach for visual detection of 10 pM of
miRNA through AuNP-coated lateral flow strip biosensor and silica
nanorod label [94]. In 2017, a novel class of intracellular nanop-
robe called gold nanoparticle loaded split-DNAzyme probe was
generated to detect miRNAs [95].
4.2 Silicon Nanowire High sensitvity and fast response of silicon nanowire biosensors
Biosensors (SiNW) have advantages in miRNA detection as a label-free
method [96]. These biosensors are highly sensitive, selective and
responsive [96]. One fM (femtomol) of miRNA could be detected
by silicon nanowire biosensors coupled with a peptide nucleic acid
probe [97, 98]. Poly-SiNW was developed to quantify miRNA
directly with high specificity and sensitivity where a ssDNA was
used as a probe to detect as low as 1 fM of target miRNA [96].
5 Other Methods for miRNA Detection
Surface plasmon resonance (SPR) allows for site-specific, rapid, and

sensitive detection of miRNAs [99]. DNA, RNA, or antibody-
based assay and SPR approaches are combined to analyze miRNAs
in less than 30 min. In addition to detection within such a short
time, this method facilitates the detection of miRNAs as low as
2 pM [99]. SERS (surface-enhanced raman scattering) can detect
extremely low amount of analytes. As a label-free approach, SERS
may be combined with silver nanodrop substances to detect miR-
NAs based on distinct binding patterns of miRNAs to ssRNA,
RNA:DNA duplex and a thiolated ssDNA [100]. Electrochemical
biosensors have been developed to improve selectivity and sensitiv-
ity while lowering the cost. Gao et al. reported the first electro-
chemical miRNA detection method [101]. Isoniazid-substituted
osmium complex was oxidized electrocatalytically and sensitivity
was improved 2000-fold as measured by amperometry rather than
direct voltammetry [101].
Although conventional microarray is a powerful tool to analyze

many miRNAs simultaneously, it is not as potent in discriminating
highly homologous sequence variants such as pri-mRNAs and
mature miRNAs. To address these problems, microarray-based
assay, called ligase-assisted sandwich hybridization assay was devel-
oped [102]. This method involves two different probes, one that is
immobilized on to a solid surface and complementary to the 50 end
of the target miRNA and the other being a stem-loop primer
complementary to the 30 end of the target miRNA. Hybridization
between the target miRNA and the immobilized probe recruits the
fluorescently labeled stem-loop primer, which is then ligated to the
immobilized primer for the signal detection [103]. This approach is
highly convenient for the detection and quantification of miRNAs
within 3 h in total RNA samples extracted from blood, shortening
the time required to complete the conventional microarray proto-
col that takes about 24 h [102, 104].
6 Concluding Remarks
The field of miRNA-mediated regulation of gene expression is

expanding with the discovery of various miRNA variants, such as
miRNA isoforms or edited miRNAs. The use of miRNAs for thera-
peutics purposes also adds to the growing interest in this field.
Thus, it is highly important to continue to develop new detection
methods that can facilitate the detection and quantification of all
types of miRNAs with high sensitivity and accuracy at a relatively
reasonable cost. Northern blotting and qPCR-based approaches
are quite useful, convenient and fairly cost-effective in the analyses
of a defined number of miRNAs. With a higher sensitivity and
speed, qPCR is typically preferred over northern blotting especially
if the size of the RNA is not a concern. Microarray technology is a
good high-throughput alternative at a reasonable cost for the anal-
ysis of known miRNAs. RNA sequencing surpasses all other meth-
ods in its ability for the discovery of novel RNA transcripts despite
its cost and requirement for sophisticated equipment and expertise
in bioinformatics. Additionally, PCR-free amplification-based
miRNA detection methods have been developed to incerase sensi-
tivity of miRNA detection to overcome low abundancy (Table 2).
Recently, nanotechnology-based detection methods hold great
promise not only for detection but also purification of miRNAs.
Table 2
Comparison of experimental miRNA detection methods
Sensitivity
and High-
Method specificity Advantages Disadvantages throughput Quantitative
Northern 1. Low 1. Cheap 1. Sample No Semiquantitative
blotting sensitivity 2. More spesific degradation
2. with LNA 2. Radiolabelling
Medium probes carcinogenic risk
spesificity 3. Time-
consuming
4. To validate
known miRNA
RT-PCR 1. 1. Easy to 1. To validate Medium Yes
Sensitive perform known miRNA
2. 2. Cheap 2. Amplification
Medium problems based
specificity on short lenght of
miRNA
3. Primer
designing
challenges
Microarray 1. Middle 1. Easy to 1. Cross- Yes Yes
sensitivity perform when hybridization
2. Lower compared to 2. Only sets of
specificty NGS known miRNAs
than 2. Simultaneous 3. Background
RT-PCR quantification of signals
large sets of
miRNA
In situ 1. 1. Spesific to 1. Spesific probe No Semiquantit
hybridization Sensitive type of the cell design is
(with 2. Intracellular challenge
LNA location of
probes) miRNA
Deep- 1. Middle 1. Novel 1. Expensive Yes Yes
sequencing sensitivity miRNA 2. Huge data
2. Very identification needs to be
spesific 2. High processed
discrimination bioinformatically,
among miRNA needs expertise
families 3. Time
consuming
PCR-free 1. High 1. Compatible 1. Avoided PCR No Yes
amplification sensitivity with diversified mediated biased
methods 2. Middle detection results
spesificity methods
(fluorescence,
colorimetry)
Acknowledgments
This work was supported by the Scientific and Technical Research

Council of Turkey, TUBITAK (Project No. 104T144 to BA). BA
would like to thank all the past and current members of the Non-
coding RNA Laboratory for their contribution to the miRNA
research.
References
1. Li S, Liu L, Zhuang X et al (2009) Micro- 13. De Planell-Saguer M, Rodicio MC (2013)

RNAs inhibit the translation of target mRNAs Detection methods for microRNAs in clinic
on the endoplasmic reticulum in Arabidopsis. practice. Clin Biochem 46:869–878
Cell 153:562–574 14. Hamid SM, Akgül B (2014) Master regula-
2. Macfarlane LA, Murphy PR (2010) Micro- tors of post-transcriptional gene expression
RNA: biogenesis, function and role in cancer. are subject to regulation. Methods Mol Biol
Curr Genomics 11:537–561 1107:303–310
3. Michlewski G, Caceres JF (2019) Post- 15. Dong H, Lei J, Ju H et al (2012) Target-cell-
transcriptional control of miRNA biogenesis. specific delivery, imaging, and detection of
RNA 25:1–16 intracellular microRNA with a multifunc-
4. Iwasaki S, Kobayashi M, Yoda M et al (2010) tional SnO2 nanoprobe. Angew Chem Int
Hsc70/Hsp90 chaperone machinery med- Ed Engl 51:4607–4612
iates ATP-dependent RISC loading of small 16. Cissell KA, Deo SK (2009) Trends in micro-
RNA duplexes. Mol Cell 39:292–299 RNA detection. Anal Bioanal Chem
5. Krol J, Loedige I, Filipowicz W (2010) The 394:1109–1116
widespread regulation of microRNA biogene- 17. Degliangeli F, Pompa PP, Fiammengo R
sis, function and decay. Nat Rev Genet (2014) Nanotechnology-based strategies for
11:597–610 the detection and quantification of micro-
6. O Brien J, Hayder H, Zayed Y et al (2018) RNA. Chemistry 20:9476–9492
Overview of microRNA biogenesis, mechan- 18. Leshkowitz D, Horn-Saban S, Parmet Y et al
isms and circulation. Front Endocr 9:402s (2013) Differences in microRNA detection
7. Cheloufi S, Dos Santos CO, Chong MM et al levels are technology and sequence depen-
(2010) A dicer-independent miRNA biogen- dent. RNA 19:527–538
esis pathway that requires ago catalysis. 19. Alwine JC, Kemp DJ, Stark GR (1977)
Nature 465:584–589 Method for detection of specific RNAs in aga-
8. Xie M, Li M, Vilborg A et al (2013) Mamma- rose gels by transfer to
lian 50 -capped microRNA precursors that diazobenzyloxymethyl-paper and hybridiza-
generate a single microRNA. Cell tion with DNA probes. Proc Natl Acad Sci
155:1568–1580 U S A 74:5350–5354
9. Yang JS, Maurin T, Robine N et al (2010) 20. Josefsen K, Nielsen H (2011) Northern blot-
Conserved vertebrate mir-451 provides a ting analysis. Methods Mol Biol 703:87–105
platform for dicer-independent, ago2- 21. Valoczi A, Hornyik C, Varga N et al (2004)
mediated microRNA biogenesis. PNAS Sensitive and specific detection of microRNAs
107:15163–15168 by northern blot analysis using LNATM-
10. Chu CY, Rana TM (2006) Translation repres- modified oligonucleotide probes. Nucleic
sion in human cells by microRNA-induced Acids Res 32:175
gene silencing requires RCK/p54. Proc Natl 22. Ouyang T, Liu Z, Han Z et al (2019) Micro-
Acad Sci U S A 4:e210 RNA detection specificity: recent advances
11. Kozomara A, Griffiths-Jones S (2014) miR- and future perspective. Anal Chem
Base: annotating high confidence microRNAs 91:3179–3186
using deep sequencing data. Nucleic Acids 23. Dong H, Lei J, Ding L et al (2013) Micro-
Res 42:68–73 RNA: function, detection, and bioanalysis.
12. Hammond SM (2015) An overview of micro- Chem Rev 113:6207–6233
RNAs. Adv Drug Deliv Rev 87:314
24. Ramkissoon SH, Mainwaring LA, Ogasawara highly accurate miRNA quantification.
Y et al (2006) Hematopoietic-specific micro- Nucleic Acids Res 45:e144
RNA expression in human cells. Leuk Res 38. Tajadini M, Panjehpour M, Javanmard SH
30:643–647 (2014) Comparison of SYBR green and Taq-
25. Pall GS, Hamilton AJ (2008) Improved Man methods in quantitative real-time poly-
northern blot method for enhanced detection merase chain reaction analysis of four
of small RNA. Nat Protoc 3:1077–1084 adenosine receptor subtypes. Adv Biomed
26. Kim SW, Li Z, Moore PS et al (2010) A Res 3:85
sensitive non-radioactive northern blot 39. Dragan AI, Pavlovic R, McGivney JB (2012)
method to detect small RNAs. Nucleic Acids SYBR green I: fluorescence properties and
Res 38:e98 interaction with DNA. J Fluoresc
27. Koscianska E, Starega-Roslan J, Sznajder LJ 22:1189–1199
et al (2011) Northern blotting analysis of 40. Ruiz-Villaba A, Pelt-Verkuil E, Gunst QD
microRNAs, their precursors and RNA inter- et al (2017) Amplification of nonspecific pro-
ference triggers. Mol Cell Biol 12:14 ducts in quantitative polymerase chain reac-
28. Chiang HR, Schoenfeld LW, Ruby JG et al tions (qPCR). Biomol Detect Quantif
(2010) Mammalian microRNAs: experimen- 14:7–18
tal evaluation of novel and previously anno- 41. Kim C, Iseki H, Herbas MS et al (2007)
tated genes. Genes Dev 24:992–1009 Development of TaqMan-based real-time
29. Na J, Shin GW, Son HG et al (2017) Multi- PCR assays for diagnostic detection of Babesia
plex quantitative analysis of microRNA bovis and Babesia bigemina. Am J Trop Med
expression via exponential isothermal amplifi- Hyg 77:837–841
cation and conformation-sensitive DNA sepa- 42. Hindson CM, Chevillet JR, Briggs HA et al
ration. Sci Rep 7:11396 (2013) Absolute quantification by droplet
30. Aravin A, Tuschl T (2005) Identification and digital PCR versus analog real-time PCR.
characterization of small RNAs involved in Nat Methods 10:1003–1005
RNA silencing. FEBS Lett 579 43. Tian H, Sun Y, Liu C et al (2016) Precise
(26):5830–5840 quantitation of microRNA in a single cell
31. Chen C, Ridzon DA, Broomer AJ et al (2005) with droplet digital PCR based on ligation
Real-time quantification of microRNAs by reaction. Anal Chem 88:11384–11389
stem–loop RT–PCR. Nucleic Acids Res 33: 44. Conte D, Verri C, Borzi C et al (2015) Novel
e179 method to detect microRNAs using chip-
32. Mou G, Wang K, Xu D et al (2013) Evalua- based quantStudio 3D digital PCR. BMC
tion of three RT-qPCR-based miRNA detec- Genomics 16:849
tion methods using seven rice miRNAs. Biosci 45. Mitchell V, Beauvillain JC, Mazzuca M
Biotechnol Biochem 77:1349–1353 (1992) Combination of immunocytochemis-
33. Gaur A, Jewell DA, Liang Y et al (2007) try and in situ hybridization in the same semi-
Characterization of microRNA expression thin sections: detection of met-enkephalin
levels and their biological correlates in and pro-enkephalin mRNA in the hypotha-
human cancer cell lines. Cancer Res lamic magnocellular dorsal nucleus of the
67:2456–2468 guinea-pig. J Histochem Cytochem
34. Shi R, Chiang VL (2005) Facile means for 40:581–585
quantifying microRNA expression by real- 46. Song R, Ro S, Yan W (2010) In situ hybridi-
time PCR. BioTechniques 39:519–525 zation detection of microRNAs. Methods
35. Wan G, Lim QE, Too HP (2010) High- Mol Biol 629:287–294
performance quantification of mature micro- 47. Hanna JA, Wimberly H, Kumar S et al (2012)
RNAs by real-time RT-PCR using Quantitative analysis of microRNAs in tissue
deoxyuridine-incorporated oligonucleotides microarrays by in situ hybridization. BioTech-
and hemi-nested primers. RNA niques 52:235–245
16:1436–1445 48. Singh SK (1998) LNA™ (locked nucleic
36. Kim KJ, Kwak J, Lee JH et al (2017) Real- acids): synthesis and high-affinity nucleic acid
time qRT-PCR assay for the detection of miR- recognition. Chem Commun 4:455–456
NAs using bi-directional extension sequences. 49. Soe MJ, Moller T, Dufva M et al (2011) A
Anal Biochem 536:32–35 sensitive alternative for microRNA in situ
37. Androvic P, Valihrach L, Elling J et al (2017) hybridizations using probes of 20 -O-methyl
Two-tailed RT-qPCR: a novel method for RNA + LNA™. J Histochem Cytochem
59:661–672
50. Zhang D, Xie L, Jin Y (2017) In situ detection breast cancer heterogeneity. Nucleic Acids
of microRNAs: the art of microRNA research Res 43:9158–9175
in human diseases. J Cytol Histol 3:13 66. Gong J, Wu Y, Zhang X et al (2014) Compre-
51. Shi Z, Johnson J, Stack MS (2012) Fluores- hensive analysis of human small RNA
cence in situ hybridization for microRNA sequencing data provides insights into expres-
detection in archived oral cancer tissues. J sion profiles and miRNA editing. RNA Biol
Oncol 2:903581 11:1375–1385
52. Bumgarner R (2013) DNA microarrays: 67. Tian T, Wang J, Zhou X (2015) A review:
types, applications and their future. Curr Pro- micro RNA detection methods. Org Biomol
toc Mol Biol 22:Unit–22.1 Chem 13:2226–2238
53. Liu CG, Calin GA, Meloon B et al (2004) An 68. Jia H, Li Z, Liu C et al (2010) Ultrasensitive
oligonucleotide microchip for genome-wide detection of microRNAs by exponential iso-
microRNA profiling in human and mouse tis- thermal amplification. Angew Chem Int Ed
sues. Proc Natl Acad Sci U S A Engl 49(32):5498
101:9740–9744 69. Wang K, Zhang K, Lv Z et al (2014) Ultra-
54. Stenvang J, Kauppinen S (2008) MicroRNAs sensitive detection of microRNA with isother-
as targets for antisense-based therapeutics. mal amplification and a time-resolved
Expert Opin Biol Therapy 8:59–81 fluorescence sensor. Biosens Bioelectron
55. Castoldi M, Schmidt S, Benes V (2006) A 57:91–95
sensitive array for microRNA expression 70. Zhang Y, Zhang CY (2012) Sensitive detec-
profiling (miChip) based on locked nucleic tion of microRNA with isothermal amplifica-
acids (LNATM). RNA 12:913920 tion and a single-quantum-dot-based
56. Castoldi M, Benes V, Muckenthaler MU nanosensor. Anal Chem 84:224–231
(2009) How to assay microRNA expression. 71. Deng R, Zhang K, Li J (2017) Isothermal
In: Rajesh K, Rossi J, Gaur R (eds) Regulation amplification for microRNA detection: from
of Gene Expression by Small RNA. CRC the test tube to the cell. Am Chem Soc
Press, pp 216–237 50:1059–1068
57. Li W, Ruan K (2009) Micro RNA detection 72. Nilsson M, Malmgren H, Samiotaki M et al
by microarray. Anal Bioanal Chem (1994) Padlock probes: circularizing oligonu-
394:1117–1124 cleotides for localized DNA detection. Sci-
58. Thomson JM, Parker JS, Hammond SM ence 265:2085–2088
(2007) Microarray analysis of miRNA gene 73. Cheng Y, Zhang X, Li Z et al (2009) Highly
expression. Methods Enzymol 427:107–122 sensitive determination of microRNA using
59. Pritchard CC, Cheng HH, Tewari M (2012) target-primed and branched rolling-circle
MicroRNA profiling: approaches and consid- amplification. Angew Chem Int Ed Engl
erations. Nat Rev Genet 13:358–369 48:18
60. Wark AW, Lee HJ, Corn RM (2008) Multi- 74. Deng R, Tang L, Tian Q et al (2014)
plexed detection methods for profiling micro- Toehold-initiated rolling circle amplification
RNA expression in biological samples. Angew for visualizing individual microRNAs in situ
Chem 47:644–652 in single cells. Angew Chem Int Ed Engl 53:9
61. Liu P (2018) MicroRNA expression analysis: 75. Jonstrup SP, Koch J, Kjems J (2006) A micro-
next-generation sequencing. Methods Mol RNA detection system based on padlock
Biol 1783:171–183 probes and rolling circle amplification. RNA
62. Baker M (2010) MicroRNA profiling: separ- 12:1747–1752
ating signal from noise. Nat Methods 76. Chapin SC, Doyle PS (2011) Ultrasensitive
7:687–692 multiplexed microRNA quantification on
63. Hafner M, Landgraf P, Ludwig J et al (2008) encoded gel microparticles using rolling circle
Identification of microRNAs and other small amplification. Anal Chem 83:7179–7185
regulatory RNAs using cDNA library 77. Zhou Y, Huang Q, Gao J et al (2010) A
sequencing. Methods 44:3–12 dumbbell probe-mediated rolling circle
64. Alon S, Eisenberg E (2013) Identifying RNA amplification strategy for highly sensitive
editing sites in miRNAs by deep sequencing. microRNA detection. Nucleic Acids Res 38:
Methods Mol Biol 1038:159–170 e156
65. Telonis AG, Loher PL, Jing Y (2015) Beyond 78. Liu H, Li L, Duan L (2013) High specific and
the one locus one miRNA paradigm: micro- ultrasensitive isothermal detection of micro-
RNA isoforms enable deeper insights into RNA by padlock probe-based exponential
rolling circle amplification. Anal Chem samples of glioma patients via combination of
85:7941–7947 conducting magnetic microbeads and
79. Shen W, Deng H, Ren Y et al (2013) A real- ferrocene-capped gold nanoparticle/strepta-
time colorimetric assay for label-free detection vidin conjugates. Biosens Bioelectron
of microRNAs down to sub-femtomolar 86:502–507
levels. Chem Commun 49:4959–4961 93. Li RD, Wang Q, Yin BC et al (2016) Enzyme-
80. Yang C, Dou B, Shi K et al (2014) Multi- free detection of sequence-specific micro-
plexed and amplified electronic sensor for RNAs based on nanoparticle-assisted signal
the detection of microRNAs from cancer amplification strategy. Biosens Bioelectron
cells. Anal Chem 86:11913–11918 77:995–1000
81. Shagin DA, Rebrikov DV, Kozhemyalo VB 94. Takalkar S, Xu H, Chen J et al (2016) Gold
(2002) A novel method for SNP detection nanoparticle coated silica nanorods for sensi-
using a new duplex-specific nuclease from tive visual detection of microRNA on a lateral
crab hepatopancreas. Genome Res flow strip biosensor. Anal Sci 32:617–622
12:1935–1942 95. Wu Y, Huang J, Yang X et al (2017) Gold
82. Yang L, Liu C, Ren W et al (2012) Graphene nanoparticle loaded split-DNAzyme probe
surface-anchored fluorescence sensor for sen- for amplified miRNA detection in living
sitive detection of microRNA coupled with cells. Anal Chem 89:8377–8383
enzyme-free signal amplification of hybridiza- 96. He J, Zhu J, Changguo G et al (2015) Label-
tion chain reaction. Am Chem Soc Appl free direct detection of miRNAs with poly-
Mater Interfaces 4:6450–6453 silicon nanowire biosensors. Proc Natl Acad
83. Yu HP, Hsiao YL, Pan HY et al (2011) Micro- Sci U S A 10:e0145160
RNA quantification using DNA polymerase 97. Zhang GJ, Chua JH, Chee RE et al (2009)
and pyrophosphate quantification. Anal Bio- Label-free direct detection of miRNAs with
chem 419:228–233 silicon nanowire biosensors. Biosens Bioelec-
84. Masud MK, Umer M, Hossain SA et al (2018) tron 24:2504–2508
Nanoarchitecture frameworks for electro- 98. Dorvel BR, Reddy B, Go J et al (2012) Silicon
chemical miRNA detection. Trends Biochem nanowires with high-k hafnium oxide dielec-
Sci 44:433–452 trics for sensitive detection of small nucleic
85. Fiammengo R (2017) Can nanotechnology acid oligomers. ACS Nano 6:6150–6164
improve cancer diagnosis through miRNA 99. Sipova H, Zhang S, Dudley AM et al (2010)
detection? Biomark Med 11(1):69–86 Surface plasmon resonance biosensor for
86. Campuzano S, Torrente-Rodrigues RM, rapid label-free detection of microribonucleic
Lopez-Hernandez E (2014) Magnetobiosen- acid at subfemtomole level. Anal Chem
sors based on viral protein p19 for microRNA 82:10110–10115
determination in cancer cells and tissues. 100. Driskell JD, Tripp R (2010) Label-free SERS
Angew Chem Int Ed Engl 53:6168–6171 detection of microRNA based on affinity for
87. Tian Y, Tian Y, Luo X et al (2014) Identifica- an unmodified silver nanorod array substrate.
tion and characterization of microRNAs Chem Commun 46:3298–3300
related to salt stress in broccoli, using high- 101. Gao Z, Yu YH (2007) A microRNA biosensor
throughput sequencing and bioinformatics based on direct chemical ligation and electro-
analysis. BMC Plant Biol 14:226 chemically amplified detection. Sensors
88. Crew E, Tessel MA, Rahman S (2012) Micro- Actuators Chem 121:552–559
RNA conjugated gold nanoparticles and cell 102. Ueno T, Funatsu T (2014) Label-free quanti-
transfection. Anal Chem 84:26–29 fication of microRNAs using ligase-assisted
89. Fan Y, Chen X, Trigg AD et al (2007) Detec- sandwich hybridization on a DNA microarray.
tion of microRNAs using target-guided for- Proc Natl Acad Sci U S A 9:e90920
mation of conducting polymer nanowires in 103. Chou CC, Chen CH, Lee TT et al (2004)
nanogaps. J Am Chem Soc 129:5437–5443 Optimization of probe length and the num-
90. Gao Z, Yang Z (2006) Detection of micro- ber of probes per gene for optimal microarray
RNAs using electrocatalytic nanoparticle tags. analysis of gene expression. Nucleic Acids Res
Anal Chem 78:1470–1477 32:e99
91. Peng Y, Gao Z (2011) Amplified detection of 104. Iizuka R, Ueno T, Funatsu T (2016) Detec-
microRNA based on ruthenium oxide tion and quantification of microRNAs by
nanoparticle-initiated deposition of an insu- ligase-assisted sandwich hybridization on a
lating film. Anal Chem 83:820–827 microarray. Methods Mol Biol 1368:53–65
92. Lu Z, Tang H, Wu D (2016) Amplified vol-
tammetric detection of miRNA from serum
Chapter 3
Functional Annotation of MicroRNAs Using Existing

Resources
Harsh Dweep, Louise C. Showe, and Andrew V. Kossenkov
Abstract
MicroRNAs (miRNAs) are endogenous small noncoding RNAs that are involved in most biological
signaling pathways, including the cell cycle, apoptosis, proliferation, immune response, metabolism as
well as in biological processes including organ development and in human diseases like cancers. During
the past two decades, high-throughput transcriptomic profiling using next generation sequencing and
microarrays have been extensively utilized to identify differentially expressed miRNAs across different
conditions and diseases. A natural extension of miRNA identification is to the process of functionally
annotating known or predicted gene targets of those miRNAs and, by inference, revealing their potential
influences on diverse biological pathways and functions. In this chapter, we provide a stepwise guideline on
how to perform functional enrichment analyses on miRNAs of interest using publicly available resources
such as miRWalk2.0.
Key words MicroRNAs, Databases, miRWalk2.0, Functions, Pathways, Diseases, Prediction, Valida-
tion, Workflow, Analysis
1 Introduction
MicroRNAs (miRNAs) are the shortest molecules (18–25 nucleo-

tides in size) in the noncoding RNA (ncRNAs) families. They are
highly conserved among different organisms [1] and are known to
regulate the expression of their target messenger RNAs (mRNAs)
by binding to sequences in the 50 untranslated regions (50 -UTR),
the coding sequences (CDS), and/or sequences in the 30 -UTR of
target mRNAs [2–12].
Most miRNAs are produced via the canonical biogenesis path-
way. RNA polymerase II (RNA pol II) binds the promoter region of
a miRNA gene to produce a primary miRNA (pri-miRNA) tran-
script (can be longer than 1000 nt). Drosha, a nuclear protein
belonging to the RNase III-type endonucleases family, processes
the pri-miRNA transcript into a long hairpin structure (60–120 nt
in length) to form the precursor miRNA (pre-miRNA). This
57
58 Harsh Dweep et al.
pre-miRNA is then transported into the cytoplasm by the protein

exportin 5 (EXP5) and the RAN-GTP complex. In the cytoplasm,
the pre-miRNA is cleaved by Dicer to a double-stranded ~18–25 nt
product, consisting of a miRNA and a miRNA* strand. One of the
miRNA strands (guide strand) is loaded onto RISC to assemble a
miRNA–RISC complex. It is this complex that regulates translation
by binding with the specific target sequences [1]. The remaining
strand (the passenger strand) is degraded. The mechanism of
miRNA-directed gene regulation is accomplished by two mechan-
isms: targeted message cleavage and translational repression. If the
interaction among a miRNA and its target is a perfect or near-
perfect complementarity, then the target mRNA can be cleaved
and degraded. If the base-pairing between miRNA and mRNA is
imperfect, it results in translational repression rather than cleavage.
In animals, base pairing is primarily imperfect leading predomi-
nantly to translational repression [13, 14].
The miRBase release 22.1 (October 2018) documents a total
of 38,589 miRNAs from 721 species. Of these, 2656 miRNAs are
annotated in human [15]. The putative miRNA–target interactions
resulting from different prediction programs suggest that approxi-
mately 30–90% of all human mRNAs can potentially be regulated
by miRNAs. These observations suggest that each miRNA is pre-
dicted to be capable of targeting hundreds of mRNAs, although
coexpression of miRNA and target must be considered as well. In
addition, a single mRNA can contain binding sites for several
miRNAs [16–18]. Various algorithms and databases have been
developed to explore the putative interactions between miRNAs
and mRNAs of interest [17, 19, 20]. Among these resources, miR-
Walk (http://zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk)
[20] and miRWalk2.0 (http://zmf.umm.uni-heidelberg.de/
mirwalk2) [17] list putative miRNA binding sites on all the target
regions (promoter, 50 -UTR, CDS, and 30 -UTR) of a gene by
combining different prediction programs and datasets. Most
importantly, miRWalk2.0 supplies putative and validated miRNA–
target interactions with the help of several novel and unique search
methods.
miRNAs are known to play crucial roles in most biological
signaling pathways as well as in a variety of human diseases [21–
25] and medical conditions [26]. The potential biomedical impor-
tance of miRNAs to understanding and treating these conditions
has led to an expansion of studies of miRNA regulated translation.
High-throughput transcriptomic profiling platforms have now
been used to identify differentially expressed miRNAs (DEMs)
and genes (DEGs) in a wide variety of biological systems. After
identifying differentially expressed DEMs, the next task is to iden-
tify their known or predicted gene targets and the potential associa-
tions of those targets with diverse biological pathways and
Functional Annotation of miRNAs 59
Fig. 1 Overview of the contents of miRWalk2.0. The miRNA-target information hosted by miRWalk2.0 is
broadly categorized into predicted and experimentally verified miRNA target genes. The predicted information
is generated by integrating putative miRNA-target binding sites resulted from 13 different datasets including
the miRWalk2.0 algorithm. Similarly, the validated data is collected by combining information obtained via an
automated text-mining search in PubMed and datasets from public resources such as miRTarBase. This
collated data is publicly available to the scientific community with the help of an intuitive and a well-structured
information retrieval system. The primary sequence for this hairpin was manually designed such that the
selected elements were guaranteed to be present in one hairpin. The sequence was folded using RNAShapes
functions. This chapter provides a stepwise guideline by carrying

out these functional enrichment analyses using a variety of existing
resources including miRWalk2.0 and DAVID [27].
2 Protocols
miRWalk2.0, is an extensively utilized fountainhead of the miRNA

research community that hosts approximately 949 million miRNA–
target interactions, less than 1% of them verified, with the help of an
intuitive and a well-organized information retrieval system. The
information retrieval system of miRWalk2.0 has two major modules
(Fig. 1), the Predicted Target module (PTM; see Subheading 2.1)
and the Validated Target module (VTM; see Subheading 2.2).
These two modules contain different search interfaces through
which users can easily obtain miRNA–target interactions by enter-
ing their identifiers of interest (e.g., miRNA names). Altogether,
these search interfaces offer the largest available collection (putative
and experimentally validated) on the associations of miRNAs with
genes, noncoding RNAs (such as lncRNAs), regulators, protein
classes, Online Mendelian Inheritance in Man (OMIM) disorders,
pathways, and human phenotype-, disease- and gene-ontologies.
MiRWalk2.0 also supplies an additional functionality for users,
allowing them to generate customized datasets on their miRNAs
of interest. These datasets are very useful to conduct large-scale
stand-alone enrichment analysis using external tools such as
DAVID [27] and IPA [28] (see Subheadings 2.3–2.5).
2.1 Basic Protocol 1: The Predicted Target Module (PTM) of miRWalk2.0 provides
miRNA Target putative miRNA–target interactions information with the help of
Prediction and several novel and unique features as well as existing external
Functional Enrichment resources, making it a one-stop resource for all kinds of miRNA-
Analysis Using the related data. A few examples of these features are (a) miRNA bind-
PTM of miRWalk2.0 ing sites within the complete sequence of protein-coding genes,
(b) comparison of putative targets gathered from the 13 different
prediction datasets, and (c) statistical enrichment analysis of puta-
tive targets for one or more miRNA(s) of interest within OMIM
database disorders, gene regulators, protein classes, pathways, and
gene ontologies. In this section, we provide a step-by-step guide-
line on how to carry out a complete analysis from miRNA target
retrieval to functional enrichment analysis (gene ontologies, path-
ways, etc.) using miRWalk2 on data from a previously published rat
miRNA study. It is noteworthy that the multiple miRNA batch
search methods of miRWalk2.0 permit users to search a maximum
of 20 identifiers (miRNAs) in a single query. To overcome this
issue, we also outline an additional guideline (see Subheadings 2.3
and 2.4) to perform the functional enrichment analysis on putative
and verified target-genes of miRNAs (n > 20 miRNAs) using other
enrichment analysis tools (such as DAVID).
2.1.1 Necessary l Hardware and software: A computer with an Internet connec-

Resources tion with an installed web browser (for instance, Google
Chrome or Firefox) is required.
l Material: A list of miRNAs of interest. As an example, we are
using a previously published miRNA study [26] in which two
miRNA biomarkers (rno-miR-34a-5p and rno-miR-455-3p)
were discovered to be involved in the early detection of
thioacetamide-induced liver cancer. For simplicity, we will only
focus on the target genes of rno-miR-34a-5p.
l Method: http://zmf.umm.uni-heidelberg.de/apps/zmf/
mirwalk2/miRretsys-self.html.
All the search methods including the “miRNA information
retrieval system” of miRWalk2.0 are self-explanatory and designed
in four major parts. Users can predict the functional implications of
their miRNA(s) of interest by performing a customized or default
query using the “miRNA information retrieval system” with the
following steps:
1. Open the web browser and either type or copy-paste “http://

zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk2/
miRretsys-self.html” in the search box and hit the “Enter”
button. This URL opens the “microRNA information retrieval
system” search interface of miRWalk2.0. On this page, choose
a species (e.g., Human, Rat), a database (e.g., miRBase), and
identifier types (e.g., miRNA name) from the given drop-down
menus and either copy-paste or upload a list of identifiers
(a maximum of 20 miRNAs are permitted in a query; see
Subheading 2.3 for more than 20 miRNAs).
To successfully execute a query, it is very important to
properly select a species, database and identifiers of interest.
Also, users must provide the identifiers of interest in accor-
dance with the maximum limit (n < 20).
2. Several checkboxes are available to access additional informa-
tion about the input miRNAs. Hence, select one or more
checkboxes to obtain additional data (such as miRNA informa-
tion table, similar miRNAs and stem-loop structure).
3. From the prediction parameters, select the seed starting posi-
tion of the miRNA from its mature 50 end (from 1 to 6), check
one or more region(s) of rat genes on which users want to scan
possible binding sites of the provided miRNAs (a maximum of
10 kb, that is, 10,000 is allowed for the promoter region),
enter a minimum seed length of miRNA (the default value is
7 nt), and/or a p-value to reduce the number of target genes.
Then, pick at least 2 out of 13 prediction datasets for a meta-
analysis of miRNA–target interactions.
4. The next block contains parameters for functional enrichment
analysis. Users can choose KEGG [29], WikiPathways [30], or
Panther [31] for pathway databases followed by inputting a p-
value (default is set to 0.05) and selecting a multiple testing
method (Bonferroni and Holm, that is, BH is set to default) to
filter the output results. In addition, users can perform binding
site enrichment analysis of the queried miRNAs against gene
ontologies (biological processes (GOBP), molecular functions
(GOMF), cellular components (GOCC)), gene (CNV, TF,
etc.), and panther protein classes (such as kinase, receptors, or
all) by selecting the appropriate checkboxes.
5. Lastly, hit the “Search” button to execute the constructed
query. A formatted HTML output page is returned with hyper-
links leading to additional information on the queried miRNAs
(Fig. 2a).
2.1.2 Interpretation of The output interface of the “miRNA information retrieval system”
the miRNA Information provides a multilayered display of data including sequences; acces-
Retrieval System Results sion numbers; other miRNAs having similar seed regions;
Fig. 2 A schematic diagram of miRNA-based method of miRWalk2.0. (a) Select rat, miRBase, miRNA for step1,
provide “rno-miR-34a-5p” in the textbox area and choose defaults for the remaining steps. Click “Search”
button to run the query. (b) Result page for queried miRNA. From top to bottom (right side), first and second
tables show putative sites predicted by miRWalk2.0 and meta-analysis whereas third and last tables display
enriched pathways and GOBPs for rno-miR-34a-5p target genes
alignment of miRNAs belonging to the same family; miRNA-host-

gene; putative target genes; and significantly enriched pathways,
ontologies, and gene and protein classes on the queried miRNAs.
The output page returned by the “miRNA-based” search for rno--
miR-34a-5p depicts results in four sections (tables, Fig. 2b). The
mature miRNA information (the first section, that is, basic infor-
mation on queried miRNAs, Fig. 2b) contains hyperlinks to seven
additional pages: “miRNA Table,” “Similar miRNAs,” “Similar
Seeds,” “miRNA Family,” “miRNA Alignment,” “Host gene of
miRNA,” and “External links.” Using these hyperlinks, users can
acquire basic information about the queried miRNAs such as
mature sequences, family, other miRNAs having identical mature
sequences, seed region, alignment profiles of all the members of
input miRNA family, genomic location, and external databases links
to gather further data. All of these data can be easily downloaded by
clicking on the “Download Table” link given on the top of
every page.
The second section, “pre-miRNA information,” contains links
to three pages through which a user can get precursor data on the
input miRNAs (Fig. 2b).
The putative target genes table (third section: miRNA-target

interaction data, Fig. 2b) offers links to retrieve putative miRNA
binding sites on the complete sequence of all known protein-
coding genes of a selected species. These binding sites are generated
using the miRWalk2.0 algorithm and 12 other prediction datasets.
The first row (miRWalk2.0 predictions) holds putative miRNA
binding sites links on the promoter, 50 -UTR, CDS, and 30 -UTR
regions. Clicking on these links will retrieve the putative binding
site predictions generated by the miRWalk2.0 algorithm on the rat
genome. The output result pages of all four regions are organized
in a similar fashion. Therefore, we have selected 30 -UTR results
tables for interpretation (Fig. 2). The tables presented in Fig. 2 are
subsets of the original tables downloaded from miRWalk2.0. The
first table, shown on the top right side (marked in red) in Fig. 2,
displays the top 4 target genes with seven columns providing links
to external databases. The first three columns (Gene, EntrezID,
and RefseqID) host information on predicted target genes
(mRNAs) whereas the remaining columns contain data about tar-
get sites such as miRNA binding site lengths, start and end posi-
tions of binding sites and their p-values. The first record (row)
describes that rno-miR-34a-5p is predicted to target the 30 -UTR
of Colgalt1 gene with a seed length of 12 nt (binding site coordi-
nates: 1876 to 1887) with a p-value of 0.0001. Also, four additional
links at the top of the table permit users to download information
on putative target genes (gene symbols and EntrezIDs), mRNAs
(RefseqIDs), the complete table of interactions and a comprehen-
sive atlas of binding sites with different starting positions (position
1–6) of a miRNA seed.
Similar to the miRWalk2.0 algorithm’s data, the comparison of
13 different miRNA-target prediction datasets section (meta-
analysis prediction, Fig. 2b) offers users four links to collect puta-
tive target genes of queried miRNAs (Fig. 2). The first three
columns of the second table (right side, marked in orange) host
similar information as described for the miRWalk2.0 algorithm’s
results, while the last column, that is, “SUM” describes the number
of algorithms which have detected interactions between rat genes
and queried miRNAs. The columns displayed between the third
and the last column (“1” denotes “target found” while “0” depicts
“target not found”) depict which of the selected algorithms have
predicted a given miRNA-target interaction. Moreover, links are
available to download these tables.
The last (fourth) section, that is, “functional annotations
(Fig. 2b)” contains the key information, that is, possible implica-
tions of queried miRNAs with biological signaling pathways, gene
ontologies, and different classifications belonging to genes and
proteins. The section may have a varied number of rows (links)
due to the selection of one or more checkboxes at the time of
parameter selection. This section can be mainly divided into four

parts: (a) the pathways table (third table on the right side, marked
in blue) provides all the statistically significant pathways whose
members are found as putative targets of queried miRNAs
(KEGG pathways table, Fig. 2). miRWalk2.0 supplies three possible
choices for pathways; thus, a user can select KEGG, Wiki, or
Panther gene sets for pathway enrichment. The pathway results
table contains five columns. The first row indicates that rno-miR-
34a-5p was enriched for its binding sites on the representative
members of the “pathways in cancer” with a p-value of 1.74e-07
and an FDR value of 3.2e-05. A link is provided under the “Path-
Name” column leading to the miRNA-gene-pathway overview.
Other relevant information in the remaining columns are given
on the output page. (b) Gene ontologies (GO) table (marked in
green) displays significantly enriched GOBP, GOMF and GOCC
for the input miRNAs the GOBP enrichment table is shown
(Fig. 2b). (c) the gene classes table depicts statistically enriched
gene classes, and (iv) the Panther protein classification holds infor-
mation on protein classes enriched for the binding sites of queried
miRNAs. These tables are arranged in a similar fashion as Pathways
and GOBP tables. These results can be downloaded by a simple
click on the “Download” link. The result tables shown in Fig. 2 are
simplified and formatted versions, therefore, readers are recom-
mended to use the original tables for the interpretation of data—
as the original tables are loaded with additional information.
2.2 Basic Protocol 2: miRWalk2.0 gathers experimentally validated miRNA interactions

Gathering Published associated with genes (information aggregated from the existing
miRNA-Target resources, e.g., miRTarBase [32]) and diseases, cell lines, pathways,
Interactions Using organs and miRNA processing proteins. Users can fetch verified
miRWalk2.0 and miRNA–target interactions with the help of search methods imple-
miRTarBase mented under the miRWalk2.0’s VTM. These methods are
organized similar to the interfaces of the PTM. This protocol out-
lines the steps required to obtain validated target genes for one or
more miRNAs using two popular resources: miRWalk2.0 and miR-
TarBase databases.
2.2.1 Necessary l Methods: miRWalk2.0 (http://zmf.umm.uni-heidelberg.de/

Resources, Hardware and apps/zmf/mirwalk2/miRpub.html) and miRTarBase (http://
Software, and Material: mirtarbase.mbc.nctu.edu.tw/php/search.php).
See Subheading 2.1
2.2.2 Validated miRNA One can collect verified miRNA targets from the VTM of miR-
Targets Search Interface of Walk2.0 using these steps:
miRWalk2.0
1. Go to
http://zmf.umm.uni-heidelberg.de/apps/zmf/
mirwalk2/miRpub.html and pick a species, database and an
identifier type appropriately. Then either copy-paste or upload

a list of identifiers of interest (see Subheading 2.1).
2. Select additional information tables to get basic information on
the queried miRNAs. Note, that this step is optional.
3. Click on the “SEARCH” button to run the default or a custo-
mized query.
2.2.3 Interpretation of Information about the validated target-genes of queried miRNAs is

Validated miRNA Target- displayed in a formatted HTML page containing three blocks [33]:
Genes Obtained Using (a) basic and other sequence related information on input miRNAs,
miRWalk2.0 (b) this block hosts information on the precursors of input miR-
NAs, and (c) the last block contains a link which describes all the
validated target-genes on the user’s input miRNAs. This table has
five columns. The first four columns hold information about the
queried miRNAs and their experimentally verified targets along
with their identifiers, whereas, the last column provides a link to
the PubMed database through which users can gather further
information about the reported interaction via studying the refer-
enced article. This information can be downloaded by clicking the
“Download Table” link provided at the top of the HTML page.
2.2.4 miRTarBase’s miRTarBase [32] is a database hosting published miRNA–target

Target-Genes interactions. It documents 422,517 interactions between 4076
miRNAs and 23,054 genes for 23 species. miRTarBase collects
these interactions by manually surveying relevant literature pub-
lished on functional studies involving miRNAs. This resource
supplies the most recent experimentally verified miRNAs data.
Therefore, we briefly describe the steps needed to collect verified
target-gene information on one or more miRNAs of interest from
miRTarBase release 7.0. These steps are as follows:
1. Go to http://mirtarbase.mbc.nctu.edu.tw/php/search.php,
click on the “By miRNA” tab and pick a species of interest,
provide either the miRNA or family name (Fig. 3a). On the
same page, there are other search methods (such as target gene,
pathway, experimental method, and an advanced search tab) for
users, but this protocol is focused on the “By miRNA” search
tab. Enter “rno-miR-34a” in the given search box.
2. Click on the “Submit” button to run the query.
2.2.5 Interpretation of The “miRNA-based” search displays the results in a HTML format-
Validated Targets of hsa- ted table with 25 records per page. Figure 3b depicts target genes for
miR-34a Found in hsa-miR-34a-5p obtained from miRTarBase. The first 5 columns
miRTarBase contain data on the queried miRNA and its published target genes
and the next seven columns show information on the methods
(reporter assay, western blots, qPCR, microarrays, NGS, pSILAC,
Fig. 3 Overview of miRTarBase’s user interface. (a) depicts the miRNA-based search interface. To obtain the
verified target genes, first select “human” from the species drop-down menu, tick “miRNA ID” radio button
and provide “miR-34a-5p” in the search textbox area. Click on the “Submit” button to run the input query. (b)
shows the experimentally verified targets of hsa-miR-34a-5p together with information on techniques adopted
for validation and number of publications reported a given miRNA–target interaction. (c) miRTarBase provides
links to download the entire content of miRTarBase. Also, it offers species-specific interactions files for
download
and other) applied to validate a given miRNA–target interaction.

The last two columns summarize how many methods (out of the
seven techniques) have verified a given target gene and in how many
studies (articles). Also, it is possible to download the entire infor-
mation stored in miRTarBase by the “Download” page at http://
mirtarbase.mbc.nctu.edu.tw/php/download.php (Fig. 3c).
2.3 Alternative Advanced genomic-profiling technologies including microarrays

Protocol 1: Functional and NGS, generally produce a large amount of information (more
Enrichment Analysis than 1000 potential candidates: genes and/or miRNAs). The big-
Using DAVID gest challenge for investigators is to accumulate the predicted and
validated data on hundreds to thousands of potential candidate
genes/miRNAs. Because searching hundreds of miRNAs in a single
query could reduce the overall performance or even sometimes
could result in the breakdown of a server (database) hosting
miRNA contents. To avoid such events, most databases do not
permit users to accomplish the functional enrichment analysis on all

the significant miRNAs in a single query. To help large-scale enrich-
ment analysis using external tools (e.g., DAVID [27]), some data-
bases including miRWalk2.0 and miRTarBase allow users to
download the entire (or customized) datasets. For example, miR-
Walk2.0 has a functionality named “Customized dataset builder”
(http://zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk2/cus
tom.html) by which users can create a customized list of putative
targets for one or more miRNAs of interest. Additionally, miR-
Walk2.0 supplies ready to use miRNA data files in two formats
(GSEA and Rdata) at http://zmf.umm.uni-heidelberg.de/apps/
zmf/mirwalk2/holistic.html.
Similarly, the entire content of miRTarBase can be down-
loaded: http://mirtarbase.mbc.nctu.edu.tw/php/download.php.
This alternative protocol succinctly describes steps to accom-
plish a large-scale functional enrichment analysis for thousands of
target genes (n ¼ 2049) of rno-miR-34a-5p using the Database for
Annotation, Visualization and Integrated Discovery (DAVID 6.8
[27]). DAVID permits up to 3000 genes in a single query, however,
EASE (a stand-alone, Windows Desktop software) and its related
annotation files can be downloaded from https://david.ncifcrf.
gov/ease/ease1.htm to locally perform functional enrichment
analysis on any number of genes. DAVID provides a comprehensive
set of functional annotation tools for users to elucidate possible
implications on a large list of genes. For any given gene list, the
DAVID tool can determine associated biological information
including gene ontology (GO) terms, diseases, pathways, protein
domains, and tissue-specific expression patterns.
2.3.1 Necessary l Hardware and software: see Subheadings 2.1 and 2.2.
Resources l Material: A list of target genes of one or more miRNAs of
interest. Here we use 2049 putative targets of rno-miR-34a-5p
collected from the comparative platform of miRWalk2.0 (see
Subheading 2.1).
l Method: https://david.ncifcrf.gov/.
The necessary steps required to elicit enriched functional pat-
terns on any given gene list (putative target genes of rno-miR-34a-
5p), are the following:
1. Go to https://david.ncifcrf.gov/, click on the “Functional
Annotation” link from the shortcut to DAVID tools box
which redirects users to query search page (https://david.
ncifcrf.gov/summary.jsp, Fig. 4a). Then either copy-paste or
upload a list of genes. There is an option to upload a
multilist file.
2. Select an identifier type (Fig. 4a) from the given drop-
down menu.
Fig. 4 Functional enrichment analysis using DAVID database. (a) describes steps needed to run functional
enrichment analysis. Briefly, either copy-paste or upload a file of target genes, choose “Entrez_Gene_ID” for
identifier type, pick “Gene List” for list type and click “Submit List” button. The submission of input identifiers
switches users to “List” tab and opens functional annotation parameters. Press the “Clear All” button, click on
pathways toggle and select “KEGG_PATHWAY” and then run the enrichment analysis by clicking on the
“Functional Annotation Clustering” button (top right portion). (b) depicts the top 10 significantly enriched KEGG
pathways obtained on the target genes of rno-miR-34a-5p. These results suggest that rno-miR-34a-5p is a
key regulator of cancer and other associated biological pathways
3. Choose the type of input file. There are two choices: gene list
and background list. For this protocol, select gene list radio
button.
4. Click on the “Submit List” button. This submission opens the
“Gene List Manager and Annotation Summary Results” tab
(Fig. 4). Pick your species of interest (e.g., rat).
5. Click on the “Clear All” button to unselect default settings and
expand the “Pathways” toggle menu by clicking on it and
ticked KEGG pathways checkbox.
6. Press the “Functional Annotation Clustering” button to get
the results. This shows one or more annotated clusters having
overlapping genes for the gene list. Also, users can download
nonclustered annotations by clicking the button given at the
bottom of results page.
7. To obtain other enriched biological themes (such as GO and
tissue expression), repeat step 5 to clear the default settings and
choose one or more gene-sets of desired biological themes:
“Disease,” “Functional Categories,” “General Annotations,”
“Literature,” “Main Accessions,” “Protein Domains,” “Pro-

tein Interactions,” and “Tissue Expression.” A few popular
gene sets are “KEGG_PATHWAY” from pathways, “INTER-
PRO” from protein domains, “OMIM_DISEASE” from dis-
ease, and “UP_TISSUE” from expression categories. Follow
step 6 to download the results.
Users can adopt the aforementioned steps to interrogate target
genes of one or more miRNAs of interest against any biological
themes (for example GOBP).
2.3.2 Interpretation of The result pages of DAVID’s annotations are organized in a similar
DAVID’s Enrichment fashion and each page presents the enriched functional findings in a
Results HTML formatted table. Here, we chose the KEGG pathways table
to interpret the results. The KEGG table displays enriched path-
ways for the putative targets of rno-miR-34a-5p with the data
associated with 11 columns (Fig. 4b). The Category, Term, RT,
Genes, and Count columns show information include biological
terms and the genes list of interest. The remaining columns: “LT”
(list total), “PH” (population hits), PT (population total), % (per-
centage), P-Value (Fisher’s exact test) and Benjamini (correction
for multiple testing method) provide gene counts, data on the
overlap between queried and all known genes, the percentage of
queried genes mapped to biological terms, modified Fisher’s Exact
test p-value and adjusted p-value. The smaller the p- and Benjamini
values identify the more significantly enriched pathways. Users can
download this and other biological theme tables by clicking on the
“Download File” link. More information on the output result
tables can be found here: https://david-d.ncifcrf.gov/helps/func
tional_annotation.html#E3.
2.4 Alternative Ingenuity Pathway Analysis (IPA, https://www.

Protocol 2: Functional qiagenbioinformatics.com/products/ingenuity-pathway-analysis/
Enrichment Analysis ), a web-based commercial software [28] (a free trial version can be
by IPA requested from its homepage). IPA enables researchers to under-
stand human biological signaling systems by analyzing hundreds to
thousands of potential candidates derived from omics experiments,
such as RNA-seq, small RNA-seq, microarrays including miRNAs
and SNPs, metabolomics, and proteomics. IPA hosts manually
curated information on biological signaling with the help of pow-
erful search methods. These methods allow researchers to obtain
the associated biological context (regulators, relationships,
mechanisms, functions and pathways) on their genes list of interest.
2.4.1 Necessary l Hardware and software: see Subheadings 2.1 and 2.2.
Resources l Material: see Subheading 2.3. Generally, most of the databases
accept only specific identifiers (miRNA names, mature IDs,
etc.), however, IPA allows users to supply additional information
Fig. 5 Functional enrichment analysis using IPA software. This outlines the workflow required to perform
functional enrichment analysis on the target genes of rno-miR-34a-5p against the biological themes
documented inside IPA software. Succinctly, login to your IPA account, create a project folder and upload
the dataset of interest. Then, submit the core analysis to find out possible functions of the uploaded dataset
including expression fold-change and p-value. The expression

fold-change information helps to provide a better overview of
activated and inhibited biological patterns. Therefore, we have
added a second column named “FC” (fold change) with an
arbitrary value (-2 fold) for all putative target genes of rno--
miR-34a-5p resulting from miRWalk2.0. We have considered a
negative fold change value because the rno-miR-34a-5p was
significantly upregulated in the miRNA study [26] and, previous
studies suggest that the mechanism of action of most miRNAs is
to negatively regulate their target genes [1]. For example, if a
miRNA is upregulated then expressions of its targets should be
downregulated in a given biological condition and vice-versa
[34, 35]. This interpretation becomes more complicated with
transcriptional pausing where the major effect is at the level of
protein expression. In this case downstream effects of the pro-
tein changes may be detected in levels of mRNA regulated by the
affected protein.
l Method: https://www.qiagenbioinformatics.com/products/
ingenuity-pathway-analysis/.
Users can uncover significantly enriched biological systems for
the target genes of their miRNAs of interest by IPA software using
the following steps:
1. Go to IPA website (https://www.qiagenbioinformatics.com/

products/ingenuity-pathway-analysis/) and install IPA utility.
Once IPA is installed, launch it by double clicking on the IPA
icon (Fig. 5) and provide your username and password. Those
who already have IPA access can directly switch to step 2.
2. It is highly recommended to create a project name
(e.g. Book_chapter) to organize all the related information in
a single folder. To accomplish this, users can create a project
folder by right clicking on “My Project” given at the top-left
corner, that is, Project Manager.
3. Upload an input dataset either from the file menu or by right
clicking on “Dataset Files” under the “Book_chapter” project.
IPA also allows batch upload for more than one dataset.
4. IPA core analysis consists of diseases/functions, networks,
upstream regulators, pathways and toxicological related gene
sets. A core analysis can either be performed from the window
opened after successfully uploading an input dataset or by right
clicking on the uploaded dataset under “Dataset Files” of the
project folder. The core analysis launching functionality opens a
window that requires users to choose the type of core analysis.
After selecting “Expression Analysis” from the drop-down
menu followed by clicking on the “Next” button launches
the main analysis window, that is, “Create Expression Analysis”
(Fig. 5). This window requires users to select either the default
or customized settings for the core expression analysis. For
first-time IPA’s users, we run this core analysis with the default
settings, however, users are encouraged to customize the core
analysis by changing one or more parameters (Fig. 5).
5. Click on the “RUN ANALYSIS” button to submit the query
to the IPA server. IPA notifies the users by email and shows a
popup once the results are ready for review.
2.4.2 Interpretation of Figure 6 shows the summary tab of biological themes obtained on
IPA Enrichment Results putative target genes of rno-miR-34a-5p using IPA core (expres-
sion) analysis. The summary tab displays top findings of canonical
pathways, upstream regulators, disease and bio functions, regulator
effects, and networks along with their p-values, and scores (if any).
The canonical pathways tab shows all the signaling and metabolic
pathways including those that are predicted to be activated and
inhibited in the queried dataset. The pathway window divides the
results into a graphical representation (bar chart; upper portion)
and a tabular view (lower portion). The bottom left corner (marked
in red, is a customized version of pathway results) in Fig. 6 depicts
both graphical as well as network views of the signaling pathways
predicted for rno-miR-34a-5p targets. The graphical portion dis-
plays a vertical bar chart with four different colors: red (positive Z-
Fig. 6 Overview of the results obtained from IPA. IPA displays the functional enrichment results with the help of
several tabs such as summary, pathways (red color), upstream regulators (orange color), diseases and
functions (blue color) and networks (green color). The summary page shows the top findings of each biological
themes along with several values such as p-value, overlap and status. Each and every biological theme’s
results can be further annotated in detail by simply clicking on their appropriate tabs
Score), blue (negative Z-Score), gray (no activity patterns avail-

able), and white (Z-Score ¼ 0). The tabular view provides detailed
information ( p-value, Z-Score, State, molecules, etc.) of a given
pathway. These results can be customized and downloaded as an
image or text file.
The Upstream Regulator Analysis in IPA estimates which tran-
scriptional regulators have a significant number of their known
targets in the list and predicts the regulators’ activation state (acti-
vated or inhibited) based on the behavior of the targets. IPA can
then be used to visualize a holistic view of regulators-targets net-
works to gain detailed insight how these regulators interact with
each other and with their targets to fine-tune gene regulatory net-
works in the studied condition. The bottom portion (marked in
orange) of Fig. 6 is a tabular view of the results documented under
the “Upstream Regulators” tab. This table supplies regulator name,
expression fold-change (FC), molecule type, predicted activity state
of the regulator, activation Z-Score, p-value, and target gene infor-
mation. Users can view any predicted regulatory network by click-
ing on the “Mechanistic Network” link given in the last column.
The color of the lines (edges) reflects the expected direction of
effect between the two molecules. Color of the molecule reflects
the z-score calculated from the dataset which indicates activation or
inhibition of that molecule. Figure 6 shows a customized view of

the upstream regulator results (table marked in orange) in which
the first row contains miR-34, a microRNA, is predicted to be
activated within the target genes of miR-34a-5p with p ¼ 2.1e-13
and activation Z-Score ¼ 4.79).
The IPA Downstream Effects Analysis predicts increases or
decreases in downstream biological activities (diseases and func-
tions) occurring in the tissues or cells being studied. Taking the
expression patterns of the queried genes into consideration, IPA
calculates the likely effect (increase or decrease) for diseases and cell
biological processes such as cancer, apoptosis, and cell migration.
Figure 6 describes the “Diseases and Functions” tab results in a
hierarchical heatmap and table view. In the hierarchical heatmap of
diseases and functions, orange squares predict increases and blue
squares predict decreases for function or disease. Users can drill
down to their areas of interest by clicking on the colored squares.
After one click through to a single square, a window appears that
illustrates how each gene in the queried dataset contributes to the
predicted function or disease. The functions and diseases as well as
toxicological functions predicted on the target genes of miR-34a-
5p are shown in Fig. 6 (marked in blue; right side).
Like pathway, regulators and function, the “Networks” tab
supplies top diseases and functions predicted for the queried
genes (Fig. 6, top right corner portion marked in green). Further
information on IPA functionalities and how to utilize them can be
found at http://tv.qiagenbioinformatics.com/tag/ipa%20tutorial
and http://resources.qiagenbioinformatics.com/manuals/
ingenuitypathwayintegration/current/Ingenuity_Pathway_Analy
sis.pdf.
2.5 Alternative This alternative protocol outlines other key features of miRWalk2.0
Protocol 3 that can be used to obtain miRNA–lncRNA (long ncRNA),
miRNA–mitochondrial genome, and pathway–gene–miRNA inter-
action information.
2.5.1 miRNA–ncRNA Many studies have demonstrated that miRNAs can hybridize with
Interaction Data other non-coding RNAs (ncRNAs), especially long ncRNAs [36–
38]. These interactions are of great interest to the scientists as this
information may help to uncover the complexity of miRNA-
induced regulatory networks and their critical roles in maintaining
the cell integrity by controlling other ncRNAs. Considering the
necessity of this information, miRWalk2.0 offers a “miRNA-
ncRNA-based” search page (http://zmf.umm.uni-heidelberg.de/
apps/zmf/mirwalk2/mir-mir-self.html). Both the query as well as
the result interfaces of “miRNA:ncRNA interaction information
retrieval system” are organized similar to a previous search method
(see Subheading 2.1) through which users can collect miRNA–
ncRNA binding sites prediction and other basic information.
2.5.2 Mitochondrial Users can utilize the “Mitochondrial gene-miRNA interaction

Gene–miRNA information retrieval system” search page (http://zmf.umm.uni-
Interaction Data heidelberg.de/apps/zmf/mirwalk2/mito-self.html) to gather pre-
dicted interactions between mitochondrial genes and miRNAs.
2.5.3 Pathway-Gene- The “Pathway information retrieval system” (http://zmf.umm.

miRNA Interactions Data uni-heidelberg.de/apps/zmf/mirwalk2/path-self.html) is very
helpful for those researchers who are interested in exploring novel
regulatory circuits (biomarker miRNAs) for one or more signaling
pathway(s) of interest in a disease or condition. With this search
interface, users can determine predicted and published gene-
miRNA interactions belonging to one or more queried pathways
(a maximum of one pathway is allowed). It also provides a list of
miRNAs that are enriched for their binding sites within the genes
associated with the queried pathway. All the result tables of this
retrieval system are hyperlinked with miRBase, Gene, RefSeq,
KEGG, and WikiPathways for further annotation.
“Gene ontology” (http://zmf.umm.uni-heidelberg.de/apps/
zmf/mirwalk2/go-self.html), “Gene and protein class” (http://
zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk2/gp-class-self.
html), “Chromosome Targets” (http://zmf.umm.uni-heidelberg.
de/apps/zmf/mirwalk2/chr-self.html), “OMIM” (http://zmf.
umm.uni-heidelberg.de/apps/zmf/mirwalk2/omim-self.html),
“Disease ontologies” (http://zmf.umm.uni-heidelberg.de/apps/
zmf/mirwalk2/do.html), and “Human phenotype ontologies”
(http://zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk2/hpo.
html) retrieval systems are organized in a similar fashion to “Path-
way-gene-miRNA” search and result pages. With these search
methods, a user can obtain gene–miRNA interactions on a selected
gene ontology, a class, a chromosome, an OMIM disorder, a Dis-
ease Ontology and a Human Phenotype Ontology of interest by
choosing a species, choosing only one ontology term, inputting
seed length and/or p-value, and gene regions, external programs
and then click “SEARCH” button. The result tables of these above
search methods can be easily downloaded.
3 Conclusion
Microarray and sequencing technologies have become an integral

part of expression profiling pipelines to identify potentially impor-
tant candidate genes and miRNAs associated with the process
under investigation. Often, these profiling techniques generate
large number of significant genes and/or miRNAs ranging from a
few hundreds to thousands. Since the discovery of Lin-4, the first
miRNA, several databases with diverse miRNA data have been
developed and hosted to encourage the identification of miRNA
targets and their potential roles in biological systems. Most impor-

tantly, miRNAs have become prime candidates for consideration as
diagnostic and therapeutic targets of human diseases because of
their stability and easy detection in different biological fluids [39–
41]. However, understanding the biological implications of the
detection of miRNAs in association with specific diseases or
biological conditions requires the identification of the potential
miRNAs’ targets and the significance of those targets to diverse
biological conditions. Unfortunately, there is no single database
that allows the user to carry out target prediction and functional
analyses on thousands of miRNAs in a single query. However, the
resources like miRWalk2.0 and miRTarBase offer key features by
which one can generate customized datasets for putative and vali-
dated miRNA targets. These datasets can be used to perform a
large-scale functional enrichment for one or more miRNAs using
academic and commercial tools. This chapter provides a stepwise
guideline through which users can conduct the enrichment analysis
on any number of genes and/or miRNAs. The workflow is imple-
mented in two steps: (1) Collection of target genes on miRNAs of
interest using miRWalk2.0 or miRTarBase databases and (2) inter-
rogation of the collected target genes against the biological themes
documented in DAVID database or Ingenuity IPA software. This
guideline is not only ideal for exploring miRNA–gene interactions
but can also be used to explore the potential interactions of miR-
NAs with mitochondrial genes, lncRNAs and other miRNAs
including cross-kingdom gene regulation.
4 Outlook
miRNAs play a crucial role in organ development and in human

diseases by regulating one or more biological signaling pathways.
Although, many miRNA tools and databases have been developed
to host miRNA interactions associated with genes, diseases, gene
ontologies, signaling pathways, tissues/organs, and cell lines, there
is no single tool/database that can offer a multistep integrative
analysis starting from NGS data upload, quality control check,
identification of differentially expressed miRNAs and their target-
genes prediction with different programs followed by gene ontol-
ogies and pathways enrichment. To fill this gap, future miRNA
bioinformatic tools should integrate the following features [42]:
l An up-to-date miRNA knowledge (e.g., isomiRs and their tar-
gets) retrieval system.
l Functionality to analyze NGS miRNA data involving all steps,
starting from the quality control till functional characterization.
l Multilevel omics data integration functionality such as single cell

RNA sequencing data on human organs and diseases to further
dissect miRNA roles at cell types level.
Acknowledgments
We wish to acknowledge support from NCI U01 CA200495-04

and The Wistar Discovery Fund (to L. C. Showe). Additional
support included R50 CA211199-03 (to A. V. Kossenkov) and
NCI Cancer Center Support Grant (CCSG) P30 CA010815
(to H. Dweep, A. V. Kossenkov, and L. C. Showe).
References
1. Bartel DP (2004) MicroRNAs: genomics, bio- 11. Tay Y, Zhang J, Thomson AM et al (2008)
genesis, mechanism, and function. Cell MicroRNAs to Nanog, Oct4 and Sox2 coding
116:281–297 regions modulate embryonic stem cell differ-
2. Dewing AS, Rueli RH, Robles MJ et al (2012) entiation. Nature 455:1124–1128
Expression and regulation of mouse selenopro- 12. Zhou H, Rigoutsos I (2014) MiR-103a-3p
tein P transcript variants differing in targets the 50 UTR of GPRC5A in pancreatic
non-coding RNA. RNA Biol 9:1361–1369 cells. RNA 20:1431–1439
3. Elcheva I, Goswami S, Noubissi FK et al 13. Dalmay T (2013) Mechanism of miRNA-
(2009) CRD-BP protects the coding region mediated repression of mRNA translation.
of betaTrCP1 mRNA from miR-183-mediated Essays Biochem 54:29–38
degradation. Mol Cell 35:240–246 14. Park JH, Shin C (2014) MicroRNA-directed
4. Fang Z, Rajewsky N (2011) The impact of cleavage of targets: mechanism and experimen-
miRNA target sites in coding sequences and tal approaches. BMB Rep 47:417–423
in 30 UTRs. PLoS One 6:e18067 15. Kozomara A, Birgaoanu M, Griffiths-Jones S
5. Forman JJ, Coller HA (2010) The code within (2019) miRBase: from microRNA sequences
the code: microRNAs target coding regions. to function. Nucleic Acids Res 47:D155–D162
Cell Cycle 9:1533–1541 16. Brennecke J, Stark A, Russell RB et al (2005)
6. Forman JJ, Legesse-Miller A, Coller HA Principles of microRNA-target recognition.
(2008) A search for conserved sequences in PLoS Biol 3:e85
coding regions reveals that the let-7 microRNA 17. Dweep H, Gretz N (2015) miRWalk2.0: a
targets dicer within its coding sequence. Proc comprehensive atlas of microRNA-target inter-
Natl Acad Sci U S A 105:14879–14884 actions. Nat Methods 12:697
7. Hausser J, Syed AP, Bilen B et al (2013) Analy- 18. Gaynullina D, Dweep H, Gloe T et al (2015)
sis of CDS-located miRNA target sites suggests Alteration of mRNA and microRNA expres-
that they can effectively inhibit translation. sion profiles in rat muscular type vasculature
Genome Res 23:604–615 in early postnatal development. Sci Rep
8. Kim DH, Saetrom P, Snove O Jr et al (2008) 5:11106
MicroRNA-directed transcriptional gene 19. Dweep H, Sticht C, Gretz N (2013) In-silico
silencing in mammalian cells. Proc Natl Acad algorithms for the screening of possible micro-
Sci U S A 105:16230–16235 RNA binding sites and their interactions. Curr
9. Place RF, Li LC, Pookot D et al (2008) Genomics 14:127–136
MicroRNA-373 induces expression of genes 20. Dweep H, Sticht C, Pandey P et al (2011)
with complementary promoter sequences. miRWalk--database: prediction of possible
Proc Natl Acad Sci U S A 105:1608–1613 miRNA binding sites by “walking” the genes
10. Schnall-Levin M, Rissland OS, Johnston WK of three genomes. J Biomed Inform
et al (2011) Unusually effective microRNA 44:839–847
targeting within repeat-rich coding regions of 21. Bushati N, Cohen SM (2007) microRNA func-
mammalian mRNAs. Genome Res tions. Annu Rev Cell Dev Biol 23:175–205
21:1395–1403
22. Dai Q, Zhao J, Qi X et al (2014) MicroRNA 33. Parveen A, Gretz N, Dweep H (2016) Obtain-
profiling of rats with ochratoxin A nephrotoxi- ing miRNA-target interaction information
city. BMC Genomics 15:333 from miRWalk2.0. Curr Protoc Bioinformatics
23. Durand C, Roeth R, Dweep H et al (2011) 55:12.15.11–12.15.27
Alternative splicing and nonsense-mediated 34. Dweep H, Sticht C, Kharkar A et al (2013)
RNA decay contribute to the regulation of Parallel analysis of mRNA and microRNA
SHOX expression. PLoS One 6:e18115 microarray profiles to explore functional regu-
24. Dweep H, Georgiou GD, Gretz N et al (2013) latory patterns in polycystic kidney disease:
CNVs-microRNAs interactions demonstrate using PKD/Mhm rat model. PLoS One 8:
unique characteristics in the human genome. e53780
An interspecies in silico analysis. PLoS One 8: 35. Pandey P, Qin S, Ho J et al (2011) Systems
e81204 biology approach to identify transcriptome
25. Papagregoriou G, Erguler K, Dweep H et al reprogramming and candidate microRNA tar-
(2012) A miR-1207-5p binding site polymor- gets during the progression of polycystic kid-
phism abolishes regulation of HBEGF and is ney disease. BMC Syst Biol 5:56
associated with disease severity in CFHR5 36. Lai EC, Wiel C, Rubin GM (2004) Comple-
nephropathy. PLoS One 7:e31021 mentary miRNA pairs suggest a regulatory role
26. Dweep H, Morikawa Y, Gong B et al (2017) for miRNA:miRNA duplexes. RNA
Mechanistic roles of microRNAs in hepatocar- 10:171–175
cinogenesis: a study of thioacetamide with mul- 37. Yoon JH, Abdelmohsen K, Gorospe M (2014)
tiple doses and time-points of rats. Sci Rep Functional interactions among microRNAs
7:3054 and long noncoding RNAs. Semin Cell Dev
27. Huang da W, Sherman BT, Lempicki RA Biol 34:9–14
(2009) Systematic and integrative analysis of 38. Younger ST, Pertsemlidis A, Corey DR (2009)
large gene lists using DAVID bioinformatics Predicting potential miRNA target sites within
resources. Nat Protoc 4:44–57 gene promoters. Bioorg Med Chem Lett
28. Kramer A, Green J, Pollard J Jr et al (2014) 19:3791–3794
Causal analysis approaches in ingenuity path- 39. Baldassarre A, Felli C, Prantera G et al (2017)
way analysis. Bioinformatics 30:523–530 Circulating microRNAs and bioinformatics
29. Kanehisa M (2019) Toward understanding the tools to discover novel diagnostic biomarkers
origin and evolution of cellular organisms. Pro- of pediatric diseases. Genes (Basel) 8:234
tein Sci 28:1947–1951 40. Ingenito F, Roscigno G, Affinito A et al (2019)
30. Slenter DN, Kutmon M, Hanspers K et al The role of exo-miRNAs in cancer: a focus on
(2018) WikiPathways: a multifaceted pathway therapeutic and diagnostic applications. Int J
database bridging metabolomics to other Mol Sci 20:4687
omics research. Nucleic Acids Res 46: 41. Weber JA, Baxter DH, Zhang S et al (2010)
D661–D667 The microRNA spectrum in 12 body fluids.
31. Mi H, Thomas P (2009) PANTHER pathway: Clin Chem 56:1733–1741
an ontology-based pathway database coupled 42. Akgöl B, Stadler PF, Hawkins LJ, Hadj-
with data analysis tools. Methods Mol Biol Moussa H, Storey KB, Ergin K, Çetinkaya R,
563:123–140 Paschoal AR, Nachtigall PG, Tutar Y, Yousef
32. Chou CH, Shrestha S, Yang CD et al (2018) M, Allmer J (2021) 44 Current challenges in
miRTarBase update 2018: a resource for exper- miRNomics. In: Allmer J, Yousef M (eds) miR-
imentally validated microRNA-target interac- Nomics: microRNA biology and computa-
tions. Nucleic Acids Res 46:D296–D302 tional analysis. Methods in molecular biology,
vol 2257. Springer, New York
Chapter 4
Experimental MicroRNA Targeting Validation

Bala Gür Dedeoğlu and Senem Noyan
Abstract
microRNAs (miRNAs) have recently been recognized as a new dimension of posttranscriptional regulation.
It is well defined that most human protein-coding genes are regulated by one or more miRNAs. Therefore,
it is crucial to identify genes targeted by the miRNAs to better understand their functions. Although
bioinformatics tools have the ability to identify target candidates it is still essential to identify physiological
targets by experimental approaches. Currently, the majority of miRNA-target experimental validation
approaches assess the changes in target expression in mRNA or protein level upon miRNA upregulation
or downregulation. Additionally, finding out direct physical interactions between miRNAs and their targets
is also among the experimental techniques. In this chapter we reviewed the existing experimental techni-
ques for miRNA target identification by considering their advantages and potential drawbacks.
Key words miRNA, miRNA–target interaction, Cross-linking and immunoprecipitation (CLIP),

RNA immunoprecipitation (RIP)
1 Introduction
MicroRNAs (miRNAs) are short, noncoding RNAs that are known

to post-transcriptionally regulate gene expression through
sequence-specific base pairing with target mRNAs [1]. In general
the distribution of miRNA genes are within annotated transcripts
and can be exonic, intronic, or a combination of genome and they
are evolutionarily conserved [2, 3]. With a wide variety of expres-
sion profiles, miRNAs contribute to gene control in developmental
timing, embryogenesis, growth control, and programmed cell
death. On the other hand disruption of miRNA function may
cause development of many human diseases such as complex
genetic diseases, neurological disorders, and cancer [4, 5].
In animals, the canonical miRNA biogenesis can be divided
into two major steps: at first precursor miRNA (pri-miRNA)
genes are transcribed by RNA polymerase II and cleavage occurs
by Drosha in the nucleus. Pre-miRNA hairpins are exported by
Exportin-5 to the cytoplasm and are processed into 22-nucleotide
duplex miRNAs by Dicer. The mature miRNAs are incorporated
79
80 Bala Gür Dedeoğlu and Senem Noyan
into the RNA-induced silencing complex (RISC), which can medi-

ate downregulation of target gene by translational inhibition or
mRNA degradation [6]. This mechanism is rigidly controlled
by feedback regulation [7]. Although most miRNAs are generated
by this pathway, some animal pre-miRNA–like hairpins are made by
splicing of short hairpin introns (referred to as Mirtrons) [8].
Since the discovery of miRNAs, many bioinformatics algo-
rithms and web-based programs have been developed showing
target interactions and to relate miRNAs with diseases
[9, 10]. For instance, TargetScan scores based on the complemen-
tarity between seed region and target sequence, DIANA and PicTar
define targets via thermodynamic interactions, and miRWalk algo-
rithm searches for seven nucleotides long seeds based on Watson–
Crick complementarity on the target mRNA and when it identifies
a perfect base-pairing of the seed region it extends the length of the
miRNA seed until it come across a mismatch [11, 12]. Various
databases support miRNA target predictions based on sequence
complementarity to target sites with significance on perfect com-
plementarity in the seed region. TarBase [13], miRanda [14] and
microRNA.org [15] gather validated and/or bioinformatically pre-
dicted miRNA–target interactions. On the other hand, some target
prediction methods such as RNAhybrid are established based on
calculations of mRNA secondary structure and hybridization
between miRNA and target mRNA [16]. Additionally, some data-
bases such as miR2Disease, PhenomiR, and HMDD consider the
relationship between miRNAs, target genes and human diseases
[17–19].
Although the developments in bioinformatics approaches still
have the potential to predict the miRNA targets with increasing
confidence, as very well reviewed by Steinkraus et al. most common
algorithms (miRanda, PITA, and TargetScan) still display a rela-
tively high false negative (44–82%) and false positive (46–63%) rate
[20, 21]. Therefore, it is inevitable to identify the physiological
targets of the miRNAs experimentally.
2 Experimental Principles of miRNA–Target Recognition
To modulate miRNA expression in the cells and find out related

target mRNAs, miRNA mimics or antagomiRs have been
improved. Regulation of miRNA level in the cells could be accom-
plished by transient transfection or stable introduction of miRNA
expression vector constructs and afterward regulated mRNA tar-
gets are identified via experimental methods directly or indirectly.
2.1 Regulating Regulation of intracellular miRNA levels is one of the key steps for
miRNA Expression identification of miRNA targets experimentally. There are mainly
two approaches to manipulate their expression levels.
Experimental Methods for miRNA-Target Gene Validation 81
Downregulation for the upregulation of targets or overexpression

to find out potential targets among downregulated mRNAs. Anti-
sense oligonucleotides (ASO) and locked nucleic acid (LNA) tech-
nologies are highly preferred to inhibit miRNA expression while
miRNA mimics and expression vectors are used for restoring or
upregulating miRNA levels.
RNAi technology mediated by small interfering RNA (siRNA)
is a powerful technique to knock-down sequence-specific gene
expression [22, 23]. Inhibition of miRNAs via chemically modified
siRNAs, anti-miR oligonucleotides, is a broadly used way to bind
mature miRNAs and block their activity. Targeting miRNAs with
antisense oligonucleotides (ASOs) is an effective way for masking
the effects of miRNAs. ASO is a strategy for controlling the inhibi-
tion of miRNAs in vitro and in vivo [24]. This inhibition can be
provided with complementary miRNA oligonucleotide sequences
to repress miRNA production or function. Therefore, researchers
also refer to these oligonucleotides as antagomiRs.
LNA-modified oligonucleotides, that are structurally different
from ASOs, are used to block specific miRNA activity. Using for
functional analysis studies third generation based on triple-RNA
strand mimics designed to imitate mature miRNAs may contain
LNA-enhanced passenger strand without off-target miRNA activ-
ity. Biotinylated LNA miRNA mimics are a potent method for
identifying targets in RNA pull-down investigations. LNA was
developed in the late 1990s and contains 20 -oxygen and the 4-
0
-carbon atoms, which form a methylene bridge structure together
[25]. To improve targeting affinity of LNAs chemical modifications
can be performed at the sugar–phosphate backbone that can pro-
mote miRNA degradation in cells [26, 27].
Two different approaches are used to enhance miRNA expres-
sion level within specific cellular environments: transfection of syn-
thetic miRNA mimics, or miRNA-expressing DNA constructs.
miRNA mimics are chemically modified double-stranded RNA
molecules to be protected from nuclease degradation and are simi-
lar to the endogenous miRNA–miRNA*duplex. The addition of
methyl or fluorine groups and also aromatic compounds is known
to improve the pharmacodynamics of miRNA mimics
[28, 29]. They may be used as a powerful replacement strategy to
understand miRNA–target relationships if cells express low level of
the endogenous miRNA of interest.
Transient transfection is one of the delivery methods of miRNA
mimics into cultured cells and this approach is a fast and easy way to
understand functions of endogenous miRNAs. One of the impor-
tant concerns of transient transfection is the concentration of the
miRNA mimic used. It can be optimized by qRT-PCR and this
strategy can be beneficial when transfection efficiency is
problematic.
Additionally, miRNA mimics or miRNA inhibitors cannot bind

to cell membranes because of being negatively charged molecules
[30]. This disadvantage led the development of successful nucleic
acid delivery systems such as viral and nonviral vectors, lipid- and
polymer-based carriers, and miRNA expressing constructs in both
cellular and animal models [31]. Lipid- and polymer-based vehicles
are used as miRNA carriers to deliver them into target tumor cells.
Trang et al. indicated that novel lipid-based miRNA formulation is
a promising approach for the treatment of lung cancer. This data
showed that therapeutic delivery of miR-34a mimics using a neutral
lipid emulsion (NLE) led to the expression of miR-34a in tumor
tissues and inhibited tumor growth in NSCLC mouse model
through repression of mRNA targets [32]. To maintain the long
term expression of transfected mimics or miRNA inhibitors, vector-
mediated constructs that can control host genome stably were
developed for gene expression or silencing [33]. Retroviral vectors
are one example for effective delivery tools ensuring long-term
expression of shRNAs in cells. They are preferred in gene therapy
applications since they do not induce strong immune
responses [34].
Due to the resistance of some cells to transfection and because
synthetic miRNA transient transfection could have off-target
effects in the host genome [35, 36], lentiviral vectors have become
an alternative approach. They are used to produce even more stable,
long-term inhibition as they integrate into the genome. Viral deliv-
ery systems such as adenoviruses or adeno-associated viruses
(AAVs) and lentivirus vectors encoding miRNA mimics or
miRNA antagonists enhance the systemic delivery efficacy for can-
cer therapy and they are potent miRNA delivery tools in vivo. For
example, systemic lentiviral delivery of miR-146 suppresses meta-
static potential in breast cancer cells via downregulated NF-kappaB
signaling [37]. Researchers could also utilize inorganic miRNA
vectors within nanoparticles for miRNA-based cancer therapy. A
study showed that conjugation of gold nanoparticles with
miR-130b successfully delivered into tumors [38].
2.2 Experimental In miRNA research, it is critically important to be able to determine

Techniques for miRNA–target interactions. Some experimental techniques for the
Identifying miRNA identification of miRNA targets contain the analyses of global
Targets transcript expression by high-throughput techniques such as
microarray or next generation sequencing after transfection of
miRNA mimic or antagonist into cells. The interaction between
miRNAs and their targets following miRNA delivery could be
confirmed with protein expression assays such as Western blotting.
However, this relationship between miRNA and target gene pro-
ducts with regulatory functions is defined as indirect. Because of
this, the experimental approaches could be classified into two main
approaches according to the regulation mechanism of the miRNA;
direct or indirect. Especially capture-based technologies in which

RNA binding proteins, which binds to target and the miRNA of
interest simultaneously, are used for coimmunoprecipitation and
luciferase reporter assay that analyze the direct interaction of
miRNA and its target could be clustered under the direct
approaches [39]. On the other hand, qRT-PCR, Western blot,
and array technologies can be used to detect the regulative effect
of miRNAs indirectly. We focused in this review on direct
approaches to identify physical interactions between miRNA and
its target genes.
2.3 Capture-Based miRNA mediated gene silencing requires the interaction of miRNA
Technologies for and the RNA-induced silencing complex (miRISC) to show func-
Detection of Direct tion on its target mRNAs. Co-immunoprecipitation or pull-down
miRNA–Target Binding with the components of miRISC is one of the strategies to deter-
Events mine these mRNAs bound by miRISC. The core functional com-
ponent of miRISC is represented by members of the highly
2.3.1 RNA conserved Argonaute (Ago) family of proteins, which directly con-
Immunoprecipitation (RIP) tact both the miRNAs and their cognate target RNAs
Based Approaches [40]. Ago-bound coimmunoprecipitated RNAs can be analyzed
by coupling to qRT-PCR, microarray (RIP-Chip), or next-
generation sequencing (RIP-Seq) [41–45]. RIP-chip was first
described by Keene et al. in 2004 [46] and followed by RIP-seq
in 2008 [47] just after next-generation sequencing was established.
Since both RIP-chip and RIP-seq depend on the RBP-RNA inter-
action, to preserve this interaction purification protocols with low
stringency conditions are required. This requirement is one of the
reasons for the risk of high false positive discovery rates in both
techniques. The solution is to stabilize the interactions in the cells
to increase the stringency of the purification and to get precise
results consequently. Additionally, the direct interactions could be
determined by these protocols but they do not give the precise
binding site of the interaction [48].
2.4 Crosslinking and In order to stabilize the RNA–protein binding, which also enables
Immunoprecipitation the capture of more transient interactions, recent techniques utilize
(CLIP)-Based ultraviolet irradiation for the cross-linking of the RNA to the
Approaches protein before the immunoprecipitation followed by digestion of
the unprotected mRNA to determine the precise location of the
target site [49–52]. These crosslinking based techniques include
high-throughput sequencing of RNAs isolated by crosslinking
immunoprecipitation (HITS-CLIP), photoactivatable
ribonucleoside-enhanced CLIP (PAR-CLIP), and individual-
nucleotide resolution CLIP (iCLIP).
HITS-CLIP is an important technique since it was used in the
first study, which confirmed that miRNAs preferentially bind their
targets through the 3’-UTR and the importance of CDS-mediated
interactions [50, 53]. Although HITS-CLIP could map the
miRNA binding site it is not able to determine the precise position

of crosslinking between the RNA and protein [49]. It allows
mapping of protein binding sites with a resolution of nearly
100 nucleotides. On the contrary, low efficiency of 254 nm RNA–
protein crosslinking has been pointed out to be one of the limita-
tions of the technique.
Photoactivatable ribonucleoside-enhanced CLIP (PAR-CLIP)
has been proposed as an alternative to short wave UV induced
crosslinking of natural nucleic acids by using 365 nm-UV light to
induce efficient cross-linking. A higher efficiency of nucleic acid–
protein crosslinking is achieved in PAR-CLIP by incubating
cultured cells with a photoactivatable nucleosides such as
4-thiouridine. The incorporation of 4-thiouridine causes thymidine
to cytidine transitions in cross-linked sites and they are detected as
mutations in sequencing, which allows marking sites of direct
interaction [52, 54].
To increase the resolution of the technique to one nucleotide
individual nucleotide resolution CLIP (iCLIP) was developed as an
alternative. When the nucleotides are cross-linked to peptides
reverse transcriptase stops at that point and remain modified after
proteinase K digestion. The resulting short cDNAs lack a 50 site
needed for PCR amplification. In conventional CLIP library prep-
aration these cDNAs are lost. To prevent this loss, and include these
small cDNAs in the library, an alternative strategy, which relies on
the ligation of a special adapters promoting circularization and
subsequent linearization by restriction endonuclease digestion is
employed. This alternative method enables all RT cDNA products
including those originating from truncated transcripts at cross-
linked sites to be included in the CLIP library [54, 55].
2.5 Bait Approaches Another alternative approach to investigate the miRNA–target

interaction is using mRNAs or modified miRNAs as baits to pull
down miRNAs or to capture mRNA targets respectively.
Transfection of the cells with synthetic miRNA duplexes
labeled with biotin at the 30 end of the sense miRNA strand was
one of the first bait approaches used by Orom et al. [56]. Similarly,
Hassan et al. developed an experimental approach using formalde-
hyde to stabilize ternary complexes between endogenous miRNAs,
mRNAs, and RBPs. This affinity based assay is for the isolation of a
specific mRNA with its bound miRNAs using a biotinylated DNA
antisense capture oligonucleotide specific to the mRNA of interest,
which they called miR-CATCH [57, 58].
One alternative approach for miRNA pull-down assays is called
LAMP (Labeled microRNA pull-down assay), which utilizes digox-
igenin (DIG)-labeled pre-miRNA oligonucleotides that are mixed
with cell extracts enabling immunoprecipitation with anti-DIG
antibodies before analysis of the coimmunoprecipitated mRNAs
[53, 59].
miR-TRAP (miRNA target affinity purification) is an approach,

which is alternative to biotin labeling in which the miRNA is
conjugated to psoralen (Pso) to produce a highly photoreactive
probe. When the cells are exposed to UVA radiation, the Pso
moiety on the miRNA mimic reacts with uridine on target
mRNAs. This reaction enables the bound complex to be stringently
purified by biotin–streptavidin affinity purification, which
diminishes the recovery of nonspecific targets [60].
Tandem Affinity Precipitation Target identification (TAP-Tar)
is a biochemical approach that depends on sequential pull down
first via the Argonaute moiety and then via the miRNA [61]. The
mRNA–miRNA complex is first pulled down with anti-FLAG anti-
bodies and then purified on streptavidin beads that prevent high
levels of background that is one of the limitations of one-step pull
down procedures.
2.6 Approaches to Once an miRNA’s potential target sites have been determined by
Verify Physical computational predictions or via experiments, the next step is
Interaction of miRNAs assessing the functionality of the interaction. One of the most
with Their Targets common methods for target validation is the use of reporter assays.
Then mRNA and protein expression levels are analyzed to validate
2.6.1 Reporter Assay the regulatory effect of miRNA.
The dual-luciferase reporter assay is generally used for miR-
NA’s functional validation of predicted targets since 1999
[62]. This system utilizes firefly and Renilla luciferase, which are
well known to enhance experimental accuracy.
A common strategy is to determine whether an miRNA has the
capacity to modulate gene expression through reporter vector
within the 30 -UTR of a gene of interest (GOI). The experimental
approach consists of cloning the wild type or mutant forms 30 -UTR
of the GOI downstream of the luciferase reporter to generate the
reporter construct. Next, each construct transiently cotransfected
with the miRNA into host cells and the reporter expression is
measured after optimized transfection conditions. According to
changes in luciferase activity it can be interpreted whether an
miRNA could bind to the UTR sequence and repress its expression
[63]. In this system, which is defined as Off-System, miRNA
expression in cells causes a decrease in the bioluminescence signal
by binding to the UTR hence blocking the translation of the
luciferase reporter gene [64, 65]. However, this system has some
limitations such as loss of signal by cell death or nonspecific regula-
tions of the luciferase promoter.
The bioluminescence signal reduction caused by the nonspe-
cific effects has allowed the development of positive molecular
imaging systems (ON-systems) [66]. These systems are generally
based on oligonucleotide beacons, which are stem–loop hairpin
probes containing an antisense hybridization sequence conjugated
to a fluorescent dye at one end and a quencher at the other end. In
the presence of a complementary miRNA sequence, it opens the

hairpin probe and physically separates the fluorophore from the
quencher which then allows for fluorescence activity [67]. Unfor-
tunately, molecular beacons do not have the ability to distinguish
between nonfunctional miRNA and Ago-loaded functional
miRNA yet.
Northern blot analysis as a tool to study RNA expression is
used; however, its findings are often insensitive for mature miRNAs
[68]. Since miRNAs are not like mRNAs, new techniques are
needed to quantify their expression. These include high-
throughput real-time PCR and cDNA microarrays [69]. The real-
time PCR method to quantify both precursor and mature miRNAs
have already been developed [70, 71]. Additionally, proteomic
approaches propose the opportunity to investigate for large-scale
studies. Mass spectrometry (MS) might be a substitute technique
to measure multiplexed miRNA quantification [72]. Liquid
chromatography-tandem mass spectrometry (LC-MS/MS) is one
of the accepted mass spectrometric analysis approaches and it is
based on absolute quantification of a protein of interest at the
peptide level [73]. Li X et al reported a novel method for miRNA
quantification by a combination of liquid chromatography-
electrospray ionization tandem mass spectrometry (LC-ESI-MS/
MS) and multistage signal amplification (MSA) [74].
2.6.2 Protein Profiling- The terminologies “protein arrays” and “antibody arrays” are occa-
Based Approaches sionally used interchangeably but they are different methods. Pro-
tein arrays are used to examine protein interactions with other
molecules such as DNA or drugs via immobilized recombinant
proteins while antibody arrays are used for quantifying protein
expression profiling using immobilized antibodies. RPPAs (Reverse
Phase Protein Array) serve as a valuable way to analyze relative
abundances for candidate proteins. On account of its high-
throughput potential, it would represent molecular picture of the
cell [75]. RPPA platform contains protein lysates spotting onto
slides and each slide is incubated with specific antibody to capture
relative expression of protein. With high specificity and sensitivity,
straightforward sample handling and preparation, also rapid pro-
tein extraction, the RPPA platform provides more benefits com-
pared to other proteomics methods [75]. Therefore the
combination of miRNA regulation with RPPA could be used to
determine miRNA targets specific to miRNA of interest
[76, 77]. Antibody array, which is defined as a high-throughput
ELISA (enzyme-linked immunosorbent assay) based platform, has
also been adapted to miRNA research owing to detection of multi-
ple proteins at the same time [78]. A recent study reported that
PPFIA1/PARP1/NF-κB-P65/KIT signal transduction was altered
by miR-181a overexpression in K562 cells. This study employed
Cancer Signaling Phospho-Antibody Array containing 269 specific
antibodies to identify direct miRNA targets and altered signaling

pathways after miR-181a mimic expression [79].
Opposed to small-scale studies, translational profiling studies
purpose to evaluate the effect of aberrant miRNA expression
changes at protein level. Proteome profiling approaches is given
several instances such as 2D-DIGE, SILAC, iTRAQ, and ICAT.
Two-dimensional difference gel electrophoresis (2D DIGE) is a
traditional strategy to quantify expression changes in relative pro-
tein abundance. It depends on protein samples tagged with differ-
ent fluorescent dyes simultaneously on the same gel and analyzed
by mass spectrometry [80]. This method has been utilized to
determine TPM1 (tumor suppressor tropomyosin 1) as the puta-
tive target of miR-21 in MCF-7 cells [81]. Another specific peptide
labeling method for absolute quantification by tandem mass spec-
trometry is Isobaric tags for relative and absolute quantitation
(iTRAQ) [82]. Yang et al. obtained global proteomic profiling
using an iTRAQ based strategy to identify targets of miR-21 in
MCF-7 breast cancer cells [83]. Several studies were reported using
these techniques to determine the targets of specific miRNAs of
interest. In one study, proteome profiling of miR-1 overexpressed
HeLa cells was performed using SILAC (stable isotope labeling
with amino acids in cell culture), which is a high-throughput
approach to measure relative protein abundance to detect miR-1
targets [84]. Bargaje et al. also reported SILAC-based proteome
profiling to detect the few functional miRNA targets in miR-29a
over-expressed cells [85].
3 Conclusion
The integration of computational techniques and recent experi-

mental approaches will help the researchers to further understand
the cellular roles of miRNAs, by exploring their targets, finding out
the networks and circuits that they are regulating, which could
enable them to be used in various diseases as therapeutics.
References
1. Williams AE (2008) Functional aspects of ani- 4. Im H-I, Kenny PJ (2012) MicroRNAs in neu-
mal microRNAs. Cell Mol Life Sci 65:545–562 ronal function and dysfunction. Trends Neu-
2. Rodriguez A, Griffiths-Jones S, Ashurst JL et al rosci 35:325–334
(2004) Identification of mammalian micro- 5. Lujambio A, Lowe SW (2012) The microcos-
RNA host genes and transcription units. mos of cancer. Nature 482:347–355
Genome Res 14:1902–1910 6. Filipowicz W, Bhattacharyya SN, Sonenberg N
3. Olena AF, Patton JG (2010) Genomic organi- (2008) Mechanisms of post-transcriptional
zation of microRNAs. J Cell Physiol regulation by microRNAs: are the answers in
222:540–545 sight? Nat Rev Genet 9:102–114
7. Fabbri M (2014) Non-coding RNAs and can-
cer. Springer, New York, pp 1–284
8. Westholm JO, Lai EC (2011) Mirtrons: micro- 26. Davis S, Propp S, Freier SM et al (2009) Potent
RNA biogenesis via splicing. Biochimie inhibition of microRNA in vivo without degra-
93:1897–1904 dation. Nucleic Acids Res 37:70–77
9. Hamzeiy H, Allmer J, Yousef M (2014) 27. Davis S, Lollo B, Freier S et al (2006)
Computational methods for microRNA target Improved targeting of miRNA with antisense
prediction. Methods Mol Biol 1107:207–221 oligonucleotides. Nucleic Acids Res
10. Allmer J (2014) Computational and bioinfor- 34:2294–2304
matics methods for microRNA gene predic- 28. Prakash TP, Allerson CR, Dande P et al (2005)
tion. Methods Mol Biol 1107:157–175 Positional effect of chemical modifications on
11. Mazière P, Enright AJ (2007) Prediction of short interference RNA activity in mammalian
microRNA targets. Drug Discov Today 12 cells. J Med Chem 48:4247–4253
(11-12):452–458 29. Choung S, Kim YJ, Kim S et al (2006) Chemi-
12. Dweep H, Sticht C, Pandey P et al (2011) cal modification of siRNAs to improve serum
miRWalk--database: prediction of possible stability without loss of efficacy. Biochem Bio-
miRNA binding sites by “walking” the genes phys Res Commun 342:919–927
of three genomes. J Biomed Inf 44:839–847 30. Garzon R, Marcucci G, Croce CM (2010) Tar-
13. Sethupathy P, Corda B, Hatzigeorgiou AG geting microRNAs in cancer: rationale, strate-
(2006) TarBase: a comprehensive database of gies and challenges. Nat Rev Drug Discov
experimentally supported animal microRNA 9:775–789
targets. RNA 12(2):192–197 31. Lu PY, Xie F, Woodle MC (2005) In vivo
14. John B, Enright AJ, Aravin A et al (2004) application of RNA interference: from func-
Human microRNA targets, vol 2 tional genomics to therapeutics. Adv Genet
15. Betel D, Wilson M, Gabow A et al (2008) The 54:117–142
microRNA.org resource: targets and expres- 32. Trang P, Wiggins JF, Daige CL et al (2011)
sion. Nucleic Acids Res 36:149–153 Systemic delivery of tumor suppressor micro-
16. Krüger J, Rehmsmeier M (2006) RNAhybrid: RNA mimics using a neutral lipid emulsion
microRNA target prediction easy, fast and flex- inhibits lung tumors in mice. Mol Ther
ible. Nucleic Acids Res 34:451–454 19:1116–1122
17. Jiang Q, Wang Y, Hao Y et al (2009) miR2Di- 33. Brummelkamp TR, Bernards R, Agami R
sease: a manually curated database for micro- (2002) A system for stable expression of short
RNA deregulation in human disease. Nucleic interfering RNAs in mammalian cells. Science
Acids Res 37:98–104 296:550–553
18. Ruepp A, Kowarsch A, Theis F (2012) Pheno- 34. Lois C, Refaeli Y, Qin XF et al (2001) Retro-
miR: microRNAs in human diseases and viruses as tools to study the immune system.
biological processes. Methods Mol Biol Curr Opin Immunol 13:496–504
822:249–260 35. Khan AA, Betel D, Miller ML et al (2009)
19. Li Y, Qiu C, Tu J et al (2014) HMDD v2.0: a Transfection of small RNAs globally perturbs
database for experimentally supported human gene regulation by endogenous microRNAs.
microRNA and disease associations. Nucleic Nat Biotechnol 27:549–555
Acids Res 42:1070–1074 36. Huang H, Zhang C, Wang B et al (2018)
20. Lee T, Wang N, Houel S et al (2015) Dosage Transduction with lentiviral vectors altered
and temporal thresholds in microRNA proteo- the expression profile of host microRNAs. J
mics. Mol Cell Proteomics 14:289–302 Virol 92:e00503–e00518
21. Rajewsky N (2006) microRNA target predic- 37. Bhaumik D, Scott GK, Schokrpur S et al
tions in animals. Nat Genet 38 Suppl:S8–S13 (2008) Expression of microRNA-146 sup-
presses NF-kappaB activity with reduction of
22. Kim D, Rossi J (2008) RNAi mechanisms and metastatic potential in breast cancer cells.
applications. BioTechniques 44:613–616 Oncogene 27:5643–5647
23. Almeida R, Allshire RC (2005) RNA silencing 38. Crew E, Tessel MA, Rahman S et al (2012)
and genome regulation. Trends Cell Biol MicroRNA conjugated gold nanoparticles and
15:251–258 cell transfection. Anal Chem 84:26–29
24. Esau CC, Monia BP (2007) Therapeutic 39. Sacar Demirci MD, Yousef M, Allmer J (2019)
potential for microRNAs. Adv Drug Deliv Computational prediction of functional
Rev 59:101–114 microRNA-mRNA interactions. Methods Mol
€
25. GrUnweiler A, Hartmann RK (2007) Locked Biol 1912:175–196
nucleic acid oligonucleotides. BioDrugs
21:235–243
40. Meister G (2013) Argonaute proteins: func- 55. Huppertz I, Attig J, D’Ambrogio A et al
tional insights and emerging roles. Nat Rev (2014) iCLIP: protein-RNA interactions at
Genet 14:447–459 nucleotide resolution. Methods 65:274–287
41. Hendrickson DG, Hogan DJ, McCullough 56. Orom UA, Lund AH (2010) Experimental
HL et al (2009) Concordant regulation of identification of microRNA targets. Gene
translation and mRNA abundance for 451:1–5
hundreds of targets of a human microRNA. 57. Hassan T, Smith SGJ, Gaughan K et al (2013)
PLoS Biol 7:e1000238 Isolation and identification of cell-specific
42. Fasanaro P, Greco S, Lorenzi M et al (2009) An microRNAs targeting a messenger RNA using
integrated approach for experimental target a biotinylated anti-sense oligonucleotide cap-
identification of hypoxia-induced miR-210. J ture affinity technique. Nucleic Acids Res 41:
Biol Chem 284:35134–35143 e71
43. Jovanovic M, Reiter L, Clark A et al (2012) 58. Vencken S, Hassan T, McElvaney NG et al
RIP-chip-SRM--a new combinatorial large- (2015) miR-CATCH: microRNA capture
scale approach identifies a set of translationally affinity technology. Methods Mol Biol
regulated bantam/miR-58 targets in 1218:365–373
C. elegans. Genome Res 22:1360–1371 59. Hsu R-J, Tsai H-J (2011) Performing the
44. Easow G, Teleman AA, Cohen SM (2007) Iso- labeled microRNA pull-down (LAMP) assay
lation of microRNA targets by miRNP immu- system: an experimental approach for high-
nopurification. RNA 13:1198–1204 throughput identification of microRNA-target
45. Meier J, Hovestadt V, Zapatka M et al (2013) mRNAs. Methods Mol Biol 764:241–247
Genome-wide identification of translationally 60. Baigude H, Li Z, Zhou Y et al (2012)
inhibited and degraded miR-155 targets using miR-TRAP: a benchtop chemical biology strat-
RNA-interacting protein-IP. RNA Biol egy to identify microRNA targets. Angew
10:1018–1029 Chem Int Ed Engl 51:5880–5883
46. Penalva LOF, Tenenbaum SA, Keene JD 61. Nonne N, Ameyar-Zazoua M, Souidi M et al
(2004) Gene expression analysis of messenger (2010) Tandem affinity purification of miRNA
RNP complexes. Methods Mol Biol target mRNAs (TAP-Tar). Nucleic Acids Res
257:125–134 38:e20
47. Cloonan N, Forrest ARR, Kolle G et al (2008) 62. Tuschl T, Zamore PD, Lehmann R et al (1999)
Stem cell transcriptome profiling via massive- Targeted mRNA degradation by double-
scale mRNA sequencing. Nat Methods stranded RNA in vitro. Genes Dev
5:613–619 13:3191–3197
48. Darnell RB (2010) HITS-CLIP: panoramic 63. Kuhn DE, Martin MM, Feldman DS et al
views of protein-RNA regulation in living (2008) Experimental validation of miRNA tar-
cells. Wiley Interdiscip Rev RNA 1:266–286 gets. Methods 44:47–54
49. Hafner M, Landthaler M, Burger L et al (2010) 64. Ko MH, Kim S, Hwang DW et al (2008) Bioi-
Transcriptome-wide identification of maging of the unbalanced expression of micro-
RNA-binding protein and microRNA target RNA9 and microRNA9* during the neuronal
sites by PAR-CLIP. Cell 141:129–141 differentiation of P19 cells. FEBS J
50. Chi SW, Zang JB, Mele A et al (2009) Argo- 275:2605–2616
naute HITS-CLIP decodes microRNA-mRNA 65. Ko HY, Hwang DW, Lee DS et al (2009) A
interaction maps. Nature 460:479–486 reporter gene imaging system for monitoring
51. Konig J, Zarnack K, Rot G et al (2010) iCLIP microRNA biogenesis. Nat Protoc
reveals the function of hnRNP particles in 4:1663–1669
splicing at individual nucleotide resolution. 66. Liu W-L, Owen DP, Fisher KD et al (2009)
Nat Struct Mol Biol 17:909–915 Establishment of a positive-readout reporter
52. Thomson DW, Bracken CP, Goodall GJ system for siRNAs. J RNAi Gene Silencing
(2011) Experimental strategies for microRNA 5:331–338
target identification. Nucleic Acids Res 67. Santangelo P, Nitin N, Bao G (2006) Nanos-
39:6845–6853 tructured probes for RNA detection in living
53. Martinez-Sanchez A, Murphy CL (2013) cells. Ann Biomed Eng 34:39–50
MicroRNA target identification-experimental 68. Schmittgen TD, Lee EJ, Jiang J et al (2008)
approaches. Biology (Basel) 2:189–205 Real-time PCR quantification of precursor and
54. Popova VV, Kurshakova MM, Kopytova DV mature microRNA. Methods 44:31–38
(2015) Methods to study the RNA-protein 69. Liu C-G, Calin GA, Meloon B et al (2004) An
interactions. Mol Biol (Mosk) 49:472–481 oligonucleotide microchip for genome-wide
microRNA profiling in human and mouse tis- 77. Leivonen S-K, Sahlberg KK, Makela R et al
sues. Proc Natl Acad Sci U S A 101:9740–9744 (2014) High-throughput screens identify
70. Schmittgen TD, Jiang J, Liu Q et al (2004) A microRNAs essential for HER2 positive breast
high-throughput method to monitor the cancer cell growth. Mol Oncol 8:93–104
expression of microRNA precursors. Nucleic 78. Yuan Y, Lin Z-T, Wang H et al (2017) Protein
Acids Res 32:e43 arrays I: antibody arrays. Methods Mol Biol
71. Chen C, Ridzon DA, Broomer AJ et al (2005) 1654:261–269
Real-time quantification of microRNAs by 79. Gu C, Liu Y, Yin Z et al (2019) Discovery of
stem-loop RT-PCR. Nucleic Acids Res 33: the oncogenic Parp1, a target of bcr-abl and a
e179 potential therapeutic, in mir-181a/PPFIA1
72. Christians U, Klepacki J, Shokati T et al (2012) signaling pathway. Mol Ther Nucleic Acids
Mass spectrometry-based multiplexing for the 16:1–14
analysis of biomarkers in drug development 80. Beckett P (2012) The basics of 2D DIGE.
and clinical diagnostics- how much is too Methods Mol Biol 854:9–19
much? Microchem J 105:32–38 81. Zhu S, Si M-L, Wu H et al (2007) MicroRNA-
73. Xu F, Zhou W, Cao J et al (2017) A combina- 21 targets the tumor suppressor gene tropomy-
tion of DNA-peptide probes and liquid osin 1 (TPM1). J Biol Chem
chromatography-tandem mass spectrometry 282:14328–14336
(LC-MS/MS): a quasi-targeted proteomics 82. Wiese S, Reidegeld KA, Meyer HE et al (2007)
approach for multiplexed microRNA quantifi- Protein labeling by iTRAQ: a new tool for
cation. Theranostics 7:2849–2862 quantitative mass spectrometry in proteome
74. Li X, Zhao J, Xu R et al (2020) Mass spectro- research. Proteomics 7:340–350
metric quantification of microRNAs in 83. Yang Y, Chaerkady R, Beer MA et al (2009)
biological samples based on multistage signal Identification of miR-21 targets in breast can-
amplification. Analyst 145:1783–1788 cer cells using a quantitative proteomic
75. Creighton CJ, Huang S (2015) Reverse phase approach. Proteomics 9:1374–1384
protein arrays in signaling pathways: a data 84. Vinther J, Hedegaard MM, Gardner PP et al
integration perspective. Drug Des Devel Ther (2006) Identification of miRNA targets with
9:3519–3527 stable isotope labeling by amino acids in cell
76. Seviour EG, Sehgal V, Lu Y et al (2016) Func- culture. Nucleic Acids Res 34:e107
tional proteomics identifies miRNAs to target a 85. Bargaje R, Gupta S, Sarkeshik A et al (2012)
p27/Myc/phospho-Rb signature in breast and Identification of novel targets for miR-29a
ovarian cancer. Oncogene 35:691–701 using miRNA proteomics. PLoS One 7:
e43243
Chapter 5
Endogenous miRNA Sponges

Ayşe Hale Alkan and Bünyamin Akgül
Abstract
MicroRNAs (miRNAs) are a class of noncoding RNAs of 17–22 nucleotides in length with a critical
function in posttranscriptional gene regulation. These master regulators are themselves subject to regula-
tion both transcriptionally and posttranscriptionally. Recently, miRNA function has been shown to be
modulated by exogenous RNA molecules that function as miRNA sponges. Interestingly, endogenous
transcripts such as transcribed pseudogenes, long noncoding RNAs (lncRNAs), circular RNAs (circRNAs)
and mRNAs may serve as natural miRNA sponges. These transcripts, which bind to miRNAs and competi-
tively sequester them away from their targets, are naturally existing endogenous miRNA sponges, called
competing endogenous RNAs (ceRNAs). Here we present a historical background of miRNAs, exogenous
and endogenous miRNA sponges as well as some examples of endogenous miRNA sponges involved in
regulatory mechanisms associated with various diseases, developmental stages, and other cellular processes.
Key words miRNA, miRNA sponge, Competing endogenous RNA, lncRNA, circRNAs, Pseudogene
1 Introduction
MicroRNAs (miRNAs) are evolutionarily conserved small noncod-

ing RNAs (ncRNAs) of 17–22 nucleotide in length that regulate
gene expression posttranscriptionally [1–3]. The regulatory role of
miRNAs comes about through binding primarily to the 30 untrans-
lated regions (UTR) of messenger RNAs (mRNA), 50 UTR, or
coding region, resulting in mRNA instability or translational
repression [4, 5]. In addition to their significance in disease pathol-
ogy, miRNAs attract attention due to their potential for use as
biomarkers since they exist in extracellular fluids which makes
them readily accessible [6, 7].
miRNAs were discovered for the first time in 1993 through the
study of lin-4 in Caenorhabditis elegans [3, 8]. It was shown that
lin-4 regulates temporal development of C. elegans larvae by down-
regulation of the LIN-14 protein, which plays a role in the devel-
opmental transition into the second larval stage [9]. Later, it was
understood that the lin-4-mediated downregulation of the LIN-14
protein involves an interaction between the 30 UTR of the lin-14
91
92 Ayşe Hale Alkan and Bünyamin Akgül
mRNA and lin-4. Additionally, the lin-4 mRNA was shown to be

transcribed but not translated [3]. These findings paved the way
for the discovery of a plethora of miRNAs in other animal models.
Indeed, the discovery of let-7 and its homologues indicated that
the regulatory role of miRNAs in gene expression is not limited to
C. elegans. In 2002, a miRNA database, miRbase, was established,
providing a repository of information pertinent to miRNA
sequences, annotation, nomenclature, and target prediction
[10, 11]. The most recent version of this database, miRbase
22, contains 38,589 miRNA entries pertaining to 271 organisms
[12]. Although some of these miRNAs have been extensively char-
acterized, more studies are needed to uncover the functions of
many other miRNAs as well as miRNA–protein and miRNA–
RNA interactions.
2 Biogenesis of miRNAs
miRNA biogenesis involves post- or cotranscriptional processing of

RNA polymerase II and III transcripts either in canonical or non-
canonical pathways [5, 13]. In the canonical pathway, a large pri-
mary transcript (pri-miRNA) is generated through the
transcription of a miRNA gene by RNA polymerase II [14]. The
pri-miRNA is then converted into a 70- to 120-nucleotide-long
precursor RNA (pre-miRNA) through a multiprotein complex, the
microprocessor [15]. The microprocessor is composed of the
DiGeorge Syndrome Critical Region 8 (DGCR8), an RNA binding
protein, and Drosha, a ribonuclease III enzyme [16]. After
pre-miRNAs are exported to the cytoplasm through an exportin
5 (XPO5)/RanGTP complex, Dicer-1 which is an RNase III
enzyme cooperates with dsRNA-binding proteins, protein kinase
RNA activator, and transactivation response RNA binding protein
to generate 18- to 23-nucleotide-long duplexes [16–21]. Subse-
quently, thermodynamic asymmetry of the duplex and stability of
base pairing at the 50 end determine the separation mode of miRNA
strands [22]. Of the two strands, the original 50 end of the
pre-miRNA hairpin forms the 5p strand while the 30 end of the
precursor generates the 3p strand [23]. These two terms have
been added to the miRBase nomenclature for clarity, for example,
hsa-miR-539-5p and hsa-miR-539-3p. The strand with the less
stable 50 -end is typically loaded onto the miRNA-induced silencing
complex (miRISC). The functional miRISC complex is formed
through the binding of the miRNA to the AGO proteins as well
as the glycine-tryptophan protein of 182 kDa (GW182) and some
other accessory proteins within the RISC complex [24]. The strand
contributing to the RISC formation is called guide strand. The
other strand pairing at the 50 end rather stably constitutes the
passenger strand that is released during the incorporation of the
Endogenous miRNA Sponges 93
guide strand into RISC. After its release, the passenger strand is
typically degraded. In some cases, the passenger strand may bind to
AGO and possess miRNA regulatory functionality [25–27].
Non-canonical pathways lack a distinct type of proteins taking
part in the canonical pathway such as Drosha, Dicer, Exportin
5, and Ago2 [28, 29]. Advances in RNA sequencing technology
revealed a number of small RNAs with resemblance to miRNAs
with respect to their structure and function. Interestingly, the
biogenesis of these non-canonical miRNAs does not involve some
critical steps of the canonical one and thus these RNAs are desig-
nated as noncanonical miRNAs [30–32]. These non-canonical
pathways are composed of Drosha/DGCR8-independent or
Dicer-independent pathways. In the Drosha/DGCR8-
independent pathway, pre-miRNAs serve as Dicer substrates as
Drosha is not involved in the biogenesis. Exportin 1 is used in the
transportation of these pre-miRNAs to the cytoplasm
[33, 34]. However, Drosha is involved in the processing of miR-
NAs from endogenous short hairpin RNA (shRNA) transcripts in
the Dicer-independent pathway. Here, AGO2 is needed for the
maturation of pre-miRNAs in the cytoplasm since their length is
not appropriate to serve as Dicer substrates [35].
3 miRNA Mode of Action
miRNA modes of action include target mRNA degradation

through deadenylation or decapping and translational repression
[36, 37]. The extent of complementarity between miRNA and the
30 UTR of target mRNA determines which of the two mechanisms
will apply [38, 39]. Although the majority of interactions between
miRNAs and their target mRNAs take place in the 30 UTR, miR-
NAs can bind to other sites on mRNAs such as 50 UTRs, coding
sequences and promoter regions. Whereas gene expression is
silenced through interactions between miRNAs and 50 UTRs or
coding regions, miRNA binding to promoter regions will stimulate
transcription [40–43]. There are also other dynamics affecting the
mode of action of miRNA such as methylation, gene polymorph-
isms, gene amplification, deletion of Dicer, and translocations
[44, 45].
4 miRNA Sponges
Although the miRNA-mediated regulation of gene expression is

highly important for cellular homeostasis, just as important is the
careful regulation of miRNA expression. Since aberrant miRNA
expression may lead to a variety of diseases [46], the cell devotes
multiple layers of regulation to carefully modulate the intracellular
miRNA abundance [47]. Regulation of miRNA genes depends

upon their physical location in the genome. As the majority of
miRNA genes is located in a transcription unit similar to protein-
coding genes [48], these types of miRNAs are heavily subject to
transcriptional regulation [49]. Due to the existence of miRNA
genes within the intronic sequences of protein-coding or noncod-
ing sequences, miRNA biogenesis may be regulated during the
processing of pri-miRNAs [50]. It was uncovered that this regula-
tion comprises cleavage of the DGCR8 50 UTR hairpin by Drosha
in the microprocessor complex resulting in a miRNA [51]. miRNA
activity may also be regulated during the export of pre-miRNA to
the cytoplasm [52], the cleavage by Dicer [53, 54], editing [55],
miRNA stability [5], or miRISC localization [1]. Interestingly,
miRNA function may be modulated posttranscriptionally through
a competition between the target mRNA and nontarget mRNAs,
namely sponges [56].
miRNA sponges present a separate layer of regulation that
dictates the posttranscriptional fate of miRNAs [57]. miRNA
sponges are basically RNA transcripts that harbor multiple binding
sites for miRNAs and thus possess the ability to modulate the
miRNA–target interactions [56, 58–60]. RNA-mediated seques-
tration of miRNAs disconnects the miRNA from its target mRNA,
derepressing the target mRNA for upregulated expression and/or
stability [57, 58, 61–63]. Historically, exogenous miRNA sponges
were used artificially to examine and validate miRNA–target inter-
actions [57]. The first in vivo use of a miRNA sponge was reported
in Drosophila melanogaster with a loss-of-function activity similar to
exogenous sponges [62]. In another study, a miR-223 sponge was
documented in mouse [63].
Endogenous RNAs with sponging activity, endogenous
miRNA sponges, have been reported in plants and prokaryotes as
well [64–69]. These RNAs are a type of competitive endogenous
RNA (ceRNA) [56]. The ceRNA hypothesis was proposed by
Pandolfi et al. [70, 71] in which an RNA transcript with miRNA
recognition elements (MREs) generates a regulatory network by
harboring multiple MREs. The RNA transcripts act as decoys and
bind miRNAs which suit their target sites, thereby upregulating the
activity of other RNAs possessing similar MREs through competi-
tion [70–75]. These endogenous miRNA sponges are composed of
endogenously transcribed pseudogenes, long noncoding RNAs,
and circular RNAs in addition to mRNAs (Fig. 1) [56, 65, 67, 68].
Although the miRNA–ceRNA interactions typically result in
the sequestration of miRNAs (Fig. 2), target-directed miRNA
sequestration might also lead to miRNA degradation through a
process called target-directed miRNA degradation (TDMD)
[76]. While a miRNA is degraded following the posttranscriptional
modification of the miRNA sequence upon the miRNA–target
interaction, sponges sequester and inhibit miRNA activity without
Fig. 1 Different types of miRNA sponges. lncRNA, long noncoding RNA; circRNA, circular RNA
Fig. 2 Mode of action of miRNA sponges. A miRNA binds to its target mRNA and posttranscriptionally
suppresses its function in the absence of miRNA sponges such as lncRNAs or circRNAs. However, miRNA
sponges sequester the miRNA away from its target mRNA, resulting in derepression of the target mRNA
influencing miRNA stability [77–80]. In the TDMD mechanism,

miRNAs remain in a stable state within RISC in the case of a weak
complementarity between the miRNA and its target. However, 30
trimming, tailing, and degradation proceed fairly quick in the case
of an extensive complementarity between the miRNA and its target
RNA [77]. It is thought that the endogenous target RNAs that

stimulate TDMD are rather scarce. More work is needed to under-
stand the requirements for TDMD and the target mRNAs that
induce miRNA degradation [81].
4.1 mRNAs Act Although the binding of a miRNA to the 30 UTR of a target mRNA
as a miRNA Sponge typically regulates the posttranscriptional fate of the target mRNA,
recent evidence suggests that such interactions may influence the
miRNA fate as well. For example, certain mRNAs serve as miRNA
sponges by harboring multiple miRNA binding sites in their 30
UTR (Fig. 2) [82, 83]. Perhaps, one of the best examples of a
protein coding transcript that sponges miRNAs is the phosphatase
and tensin homolog gene (PTEN), a gene with tumor suppressor
activity. Interestingly, there are many other protein-coding ceRNAs
that harbor the same miRNA target sites as PTEN such as SER-
INC1 [67], VAPA [69], CNOT6L [67], and ZEB2 [67, 84]. The
PTEN mRNA competes with these transcripts for binding to miR-
NAs [74]. Such an interaction results in the upregulation of the
PTEN mRNA as the ZEB2 mRNA sponges miRNAs that normally
bind to the 30 UTR of and trigger the degradation of the PTEN
mRNA [69]. In another example, the 30 UTR of the versican
sponges miR-199a-3p and miR-144 and subsequently elevates
the translation of RB1 and PTEN in breast carcinoma cells by
sequestering activity of miR-199a-3p and miR-144 [85]. The
versican-mediated ceRNA function regulates the expression of ver-
sican, CD34, and fibronectin in HCC cells [86]. The 30 UTR of
CD44 is yet another example of a ceRNA that competes for hsa--
miR-216a-5p, hsa- miR-330-3p, and hsa-miR-608 and thereby
upregulates CD44 and CDC42 in the breast cancer cell line
MT-1. This intricate regulatory mechanism results in the inhibition
of cell proliferation and tumor- formation, promotion of angiogen-
esis, and the induction of apoptosis [87, 88]. In a study by Wu et al.
(2017), X-linked inhibitor of apoptosis protein (XIAP) was shown
to act as a miRNA sponge and to increase the expression of XIAP
and FSCN1 through its 30 -UTR by sponging miR-29a-5p [72]. In
a more recent study, the 30 UTR of SATB1 was reported as a
miRNA sponge for has-miR-495-3p [73].
4.2 LncRNAs Act LncRNAs are a well-characterized class of noncoding RNAs that
as miRNA Sponges are longer than 200 nucleotides. H19 was identified as one of the
first examples of lncRNAs and is involved in genomic imprinting
[75]. The fact that H19 was not translated despite the presence of
small open reading frames urged the researchers to call this tran-
script a lncRNA. The expression of this noncoding transcript had a
significant role in the embryonic development, demonstrating a
functional link between a lncRNA and a cellular phenotype.
Subsequent studies revealed that lncRNAs are involved in the reg-
ulation of complex transcriptional and posttranscriptional gene
regulatory networks [89, 90]. Through these regulatory pathways,

lncRNAs can modulate various cellular processes such as develop-
ment, differentiation, and cell fate as well as disease pathology.
Additionally, they act as ceRNAs where they control the expression
of genes by serving as miRNA sponges [91].
There are numerous examples of lncRNAs with ceRNA activity
(Fig. 2). HULC, hepatocellular carcinoma upregulated long non-
coding RNA, is the first lncRNA to be reported as a ceRNA with
miRNA sponging activity [78]. HULC sponges hsa-miR-372-5p,
causing derepression of its target gene, PRKACB, which is a kinase
that targets the cAMP response element binding protein (CREB)
[70, 92]. Metallothionein 1 pseudogene 3 (MT1P3) is a lncRNA
that is upregulated in the megakaryocytes of type 2 diabetes (T2D)
patients. Zhou et al. (2018) showed that platelet activation is
regulated by MT1P3, which sponges hsa-miR-126. This sequestra-
tion in turn results in the upregulation of p2y12, a Gi-coupled
receptor mainly expressed in platelets under T2D condition
[93]. Metastasis associated lung adenocarcinoma transcript-1
(MALAT-1) is another example of a lncRNA with ceRNA activity.
It is highly expressed in different types of cancers. Wang et al.
(2018) found that MALAT1 sponges hsa-miR-383, hsa-miR-15,
hsa-miR-205, and hsa-miR-375. Since these miRNAs target
CCND1 and CCND2, MALAT1-mediated sponging of miRNAs
results in the derepression of CCND1 and CCND2 expression and
regulation of cell proliferation during recurrent pregnancy loss
[91]. Recently, KCNQ1OT1 was shown to act as a sponge for
hsa-miR-214. Interestingly, the competition between
KCNQ1OT1 and BMP2 for binding to hsa-miR-214 is critical
for regulation of osteogenic differentiation of bone marrow mes-
enchymal stem cells [94].
4.3 Pseudogenes Act Pseudogenes are DNA sequences deprived of the protein-coding
as miRNA Sponges potential due to the presence of a frame shift mutation and prema-
ture stop codon although they might possess the other typical
features of a protein-coding transcript [95, 96]. The first pseudo-
gene was reported in 1977, which was the oocyte-type 5S RNA of
Xenopus laevis [97]. This transcript had a truncation at its 50 -end in
addition to 14-nt mismatches, disabling its potential to code for a
protein compared to its protein-coding counterpart
[97, 98]. Although the functions of pseudogenes have been ques-
tioned for a long time, the existence of a high sequence homology
between pseudogenes and their homologous coding transcripts was
recognized to evolve into a gene regulatory mechanism in which,
the pseudogene serves as a molecular decoy for a miRNA or an
RNA-binding protein [74].
PTENP1 is a highly homologous processed pseudogene of
PTEN, a tumor suppressor that is well documented to negatively
regulate the AKT/PKB pathway. PTENP1 shares a 98% sequence
identity with the PTEN mRNA and can serve as a prognostic factor
due to its involvement in tumor progression [68, 70]. Despite the
lack of protein-coding capacity, the high homology to the PTEN
mRNA maintains the ability to serve as a template for binding
various miRNAs just like the coding PTEN transcript. The compe-
tition between PTEN and PTENP1 transcripts for the binding of
miRNAs such as miR-17, miR-21, miR-214, miR-19, and miR-26
determine the intracellular fate of the PTEN mRNA [70, 74]. Simi-
larly, the BRAF pseudogene, BRAFP1, is another example of a
pseudogene that elevates the expression of its coding transcript,
BRAF, by sponging murine miRNAs mmu-miRs-134, -543, and
-653 and human miRNAs hsa-miRs-30a, -182, -876 [99]. BGAP1
is the pseudogene of GBA, a lysosomal glucocerebrosidase that
possesses binding sites for miR-22-3p [100]. Straniero et al.
(2017) reported that overexpression of the 30 UTR of the GBAP1
transcript results in an increase in the amount of the GBA mRNA
by sequestering miR-22-3p away from the GBA mRNA [100].
4.4 CircRNAs Act Circular RNAs (circRNAs) are one of the most recent types of
as miRNA Sponge non-coding RNAs that lack a 50 cap and 30 poly(A) tail due to
their covalently closed circular structures [101–104]. They can
regulate gene expression both transcriptionally and posttranscrip-
tionally. First examples of circRNAs were reported in 1976 in
several plant viroids and were assumed to be splicing errors
[105]. Later, other circRNAs were documented that originate
from the testis-determining gene Sry [106]. These RNAs were
disregarded for a long time as most attention was devoted to the
characterization of linear RNAs, such as mRNAs or lncRNAs, in
transcriptomics studies. With the advances in RNA sequencing
technology and bioinformatics tools, recent transcriptomics studies
identified thousands of circular RNAs with a potential for transla-
tional capacity and miRNA sponge activity [103, 107]. In recent
years, it has been shown that circRNA-mediated sponging of miR-
NAs has the potential for regulating several processes such as cancer
and apoptosis, and probably many other cellular phenotypes
[102, 107].
ciRS-7, which is also known as CDR1-AS, is the best known
example of a circRNA that harbors multiple binding sites for the
same miRNA. [108]. The presence of 63 conserved binding sites
for miR-7 qualifies this circRNA as one of the best examples of a
biologically important miRNA sponge that is involved in the regu-
lation of midbrain development in zebrafish. Other examples
include sry, cir-ITCH and circ_0001946 [94–96]. Carrying
16 putative target sites for miR-138, sry circRNA has a regulatory
role in transcription. A circRNA spanning several exons of ubiquitin
(Ub) protein ligase (E3) (ITCH) regulates the expression level of
the ITCH mRNA by sponging hsa-miR-7-5p, hsa-miR-17-5p, and
hsa-miR-214-3p. The ITCH protein then stimulates
ubiquitination and degradation of phosphorylated Dvl2 followed

by the inhibition of the Wnt/β-catenin pathway [109]. A poorly
characterized circ_0001946 is localized in the cytoplasm of lung
adenocarcinoma cells and functions as a molecular sponge for
miR-135a-5p, resulting in the upregulation of Sirtuin
expression [110].
miRNAs are an important group of regulatory ncRNAs that micro-

manage gene expression especially at the posttranscriptional level. A
great progress has been made toward the miRNA-mediated regu-
lation of gene expression and the field has now shifted toward how
miRNAs themselves are regulated. Recent findings suggest that the
regulation of miRNAs is not limited to transcriptional regulation
and stability as endogenous sponges present themselves as endoge-
nous competitors for target RNAs that sequester miRNAs away
from their targets. This mechanism has gained increasing attention
recently because the cell can regulate the target RNA fate without
having to differentially express miRNAs. The fate of the target RNA
can be simply regulated by sponging miRNAs. The sponge serves as
a simple on-and-off switch for the miRNA. The regulation this way
could possibly be less costly to the cell not to mention the speed of
the regulatory response.
Many different types of RNA transcripts can serve as endoge-
nous miRNA sponges that include, but not limited to, transcribed
pseudogenes, long noncoding RNAs, recently discovered circular
RNAs, and mRNAs (Fig. 2). Although the presence of a multiple
miRNA binding site is preferable to call an RNA transcript a
sponge, it is not a prerequisite. Studies conducted with endogenous
miRNA sponges so far showed that these natural miRNA sponges
have significant roles in distinct pathways ranging from develop-
ment to various diseases including cancer and diabetes. The current
efforts are focused on the identification of phenotype-specific
sponges as these sponges have the potential to serve as prognostic
markers or drug targets. However, elucidation of this complex
network of interactions among miRNAs, ceRNAs and target
RNAs is required to better understand the molecular mechanisms
of development and disease pathology [73]. For example, although
the basis of complementarity appears to be similar between the
sponge-mediated miRNA sequestration and miRNA-mediated
suppression of target RNAs, the effect on miRNA-bound RNA is
quite different. It needs to be resolved what precisely determines
(target RNA sequences or proteins in the complex) the functional
outcome of miRNA–RNA interactions. Another interesting ques-
tion is whether or not the chemical structure of miRNAs is the same
when they are bound to their repressed target or sponging
molecules. Along the same line, the identity of the protein com-
plexes that are associated with miRISCs during sponging needs to
be delineated. Because testing these hypothesis experimentally is
laborious and takes a long time, it would be timely and helpful to
develop algorithms that would distinguish the type of miRNA–
RNA interactions from the perspective of sponging versus gene
silencing.
Acknowledgments
This work was supported by the Scientific and Technical Research

Council of Turkey, TUBITAK (Project No. 104T144 to BA). BA
would like to thank all the past and current members of the Non-
coding RNA Laboratory for their contribution to the miRNA
research. We would like to thank Dr. İpek Erdoğan for critically
editing the manuscript.
References
1. Akgül B, Erdoğan İ (2018) Intracytoplasmic 10. Reinhart BJ, Slack FJ, Basson M et al (2000)
re-localization of miRISC complexes. Front The 21-nucleotide let-7 RNA regulates devel-
Genet 9:115–132 opmental timing in Caenorhabditis elegans.
2. Bartel DP (2018) Metazoan microRNAs. Cell Nature 403:901–906
173:20–51 11. Slack FJ, Basson M, Liu Z et al (2000) The
3. Lee RC, Feinbaum RL, Ambros V (1993) LIN-41 RBCC gene acts in the C. elegans
The C. elegans heterochronic gene lin-4 heterochronic pathway between the let-7 reg-
encodes small RNAs with antisense comple- ulatory RNA and the LIN-29 transcription
mentarity to lin-14. Cell 75:843–854 factor. Mol Cell 5:659–669
4. Broughton JP, Lovci MT, Huang JL et al 12. Kozomara A, Birgaoanu M, Griffiths-Jones S
(2016) Pairing beyond the seed supports (2018) miRBase: from microRNA sequences
microRNA targeting specificity. Mol Cell to function. Nucleic Acids Res 47:
64:320–333 D155–D162
5. Ha M, Kim VN (2014) Regulation of micro- 13. O’Brien J, Hayder H, Zayed Y et al (2018)
RNA biogenesis. Nat Rev Mol Cell Biol Overview of microRNA biogenesis, mechan-
15:509–524 isms of actions, and circulation. Front Endo-
6. Wang J, Chen J, Sen S (2016) MicroRNA as crinol (Lausanne) 9:402
biomarkers and diagnostics. J Cell Physiol 14. Lee Y, Kim M, Han J et al (2004) MicroRNA
231:25–30 genes are transcribed by RNA polymerase
7. Hayes J, Peruzzi PP, Lawler S (2014) Micro- II. EMBO J 23:4051–4060
RNAs in cancer: biomarkers, functions and 15. Lee Y, Ahn C, Han J et al (2003) The nuclear
therapy. Trends Mol Med 20:460–469 RNase III Drosha initiates microRNA proces-
8. Wightman B, Ha I, Ruvkun G (1993) Post- sing. Nature 425:415–419
transcriptional regulation of the heterochro- 16. Denli AM, Tops BBJ, Plasterk RHA et al
nic gene lin-14 by lin-4 mediates temporal (2004) Processing of primary microRNAs by
pattern formation in C. elegans. Cell the microprocessor complex. Nature
75:855–862 432:231–235
9. Bhaskaran M (2015) MicroRNAs: history, 17. Chendrimada TP, Gregory RI, Kumaraswamy
biogenesis, and their evolving role in animal E et al (2005) TRBP recruits the dicer com-
development and disease. Vet Pathol plex to Ago2 for microRNA processing and
51:759–774 gene silencing. Nature 436:740–744
18. Lee Y, Hur I, Park S-Y et al (2006) The role of generate a single microRNA. Cell
PACT in the RNA silencing pathway. EMBO 155:1568–1580
J 25:522–532 34. Yang J-S, Lai EC (2011) Alternative miRNA
19. Lee Y (2002) MicroRNA maturation: step- biogenesis pathways and the interpretation of
wise processing and subcellular localization. core miRNA pathway mutants. Mol Cell
EMBO J 21:4663–4670 43:892–903
20. Yi R (2003) Exportin-5 mediates the nuclear 35. Yang J-S, Maurin T, Robine N et al (2010)
export of pre-microRNAs and short hairpin Conserved vertebrate mir-451 provides a plat-
RNAs. Genes Dev 17:3011–3016 form for dicer-independent, Ago2-mediated
21. Okada C, Yamashita E, Lee SJ et al (2009) A microRNA biogenesis. Proc Natl Acad Sci
high-resolution structure of the pre-microrna 107:15163–15168
nuclear export machinery. Science 36. Yekta S (2004) MicroRNA-directed cleavage
326:1275–1279 of HOXB8 mRNA. Science 304:594–596
22. Iwasaki S, Katsuma S, Tomari Y et al (2010) 37. Filipowicz W, Bhattacharyya SN, Sonenberg
Hsc70/Hsp90 chaperone machinery med- N (2008) Mechanisms of post-transcriptional
iates ATP-dependent RISC loading of small regulation by microRNAs: are the answers in
RNA duplexes. Mol Cell 39:292–299 sight? Nat Rev Genet 9:102–114
23. Yoda M, Kawamata T, Paroo Z et al (2010) 38. Huntzinger E, Izaurralde E (2011) Gene
ATP-dependent human RISC assembly path- silencing by microRNAs: contributions of
ways. Nat Struct Mol Biol 17:17–24 translational repression and mRNA decay.
24. Krol J, Loedige I, Filipowicz W (2010) The Nat Rev Genet 12:99–110
widespread regulation of microRNA biogene- 39. Ipsaro JJ, Joshua-Tor L (2015) From guide to
sis, function and decay. Nat Rev Genet target: molecular insights into eukaryotic
11:597–610 RNA-interference machinery. Nat Struct
25. Kawamata T, Tomari Y (2010) Making RISC. Mol Biol 22:20–28
Trends Biochem Sci 35:368–376 40. Xu W, San Lucas A, Wang Z et al (2014)
26. Czech B, Zhou R, Erlich Y et al (2009) Hier- Identifying microRNA targets in different
archical rules for Argonaute loading in dro- gene regions. BMC Bioinformatics 15:S4
sophila. Mol Cell 36:445–456 41. Forman JJ, Legesse-Miller A, Coller HA
27. Okamura K, Liu N, Lai EC (2009) Distinct (2008) A search for conserved sequences in
mechanisms for microRNA strand selection coding regions reveals that the let-7 micro-
by drosophila Argonautes. Mol Cell RNA targets dicer within its coding sequence.
36:431–444 Proc Natl Acad Sci 105:14879–14884
28. Ruby JG, Jan CH, Bartel DP (2007) Intronic 42. Zhang J, Zhou W, Liu Y et al (2018) Onco-
microRNA precursors that bypass Drosha genic role of microRNA-532-5p in human
processing. Nature 448:83–86 colorectal cancer via targeting of the 50 UTR
29. Babiarz JE, Ruby JG, Wang Y et al (2008) of RUNX3. Oncol Lett 15:7215–7220
Mouse ES cells express endogenous shRNAs, 43. Dharap A, Pokrzywa C, Murali S et al (2013)
siRNAs, and other microprocessor- MicroRNA miR-324-3p induces promoter-
independent, dicer-dependent small RNAs. mediated expression of RelA gene. PLoS
Genes Dev 22:2773–2785 One 8:e79467
30. Okamura K, Hagen JW, Duan H et al (2007) 44. Tomankova T, Petrek M, Kriegova E (2010)
The Mirtron pathway generates microRNA- Involvement of microRNAs in physiological
class regulatory RNAs in drosophila. Cell and pathological processes in the lung. Respir
130:89–100 Res 11:159
31. Berezikov E, van Tetering G, Verheul M et al 45. Nana-Sinkam SP, Hunter MG, Nuovo GJ et al
(2006) Many novel mammalian microRNA (2009) Integrating the MicroRNome into the
candidates identified by extensive cloning study of lung disease. Am J Respir Crit Care
and RAKE analysis. Genome Res Med 179:4–10
16:1289–1298 46. Erdoğan İ, Coşacak Mİ, Nalbant A et al
32. Cheloufi S, Santos COD, Chong MMW et al (2018) Deep sequencing reveals two Jurkat
(2010) A dicer-independent miRNA biogen- subpopulations with distinct miRNA profiles
esis pathway that requires Ago catalysis. during camptothecin-induced apoptosis.
Nature 465:584–589 Turk J Biol 42(2):113–122
33. Xie M, Li M, Vilborg A et al (2013) Mamma- 47. Hamid SM, Akgül B (2014) Master regula-
lian 50 -capped microRNA precursors that tors of posttranscriptional gene expression are
subject to regulation. Methods Mol Biol in vivo by lentiviral vectors. Nat Methods
1107:303–310 6:63–66
48. Rodriguez A (2004) Identification of mam- 64. Lalaouna D, Carrier M-C, Semsey S et al
malian microRNA host genes and transcrip- (2015) A 30 external transcribed spacer in a
tion units. Genome Res 14:1902–1910 tRNA transcript acts as a sponge for small
49. Schanen BC, Li X (2011) Transcriptional reg- RNAs to prevent transcriptional noise. Mol
ulation of mammalian miRNA genes. Geno- Cell 58:393–405
mics 97:1–6 65. Franco-Zorrilla JM, Valli A, Todesco M et al
50. Kim Y-K, Kim VN (2007) Processing of intro- (2007) Target mimicry provides a new mech-
nic microRNAs. EMBO J 26:775–783 anism for regulation of microRNA activity.
51. Saçar Demirci MD, Yousef M, Allmer J Nat Genet 39:1033–1037
(2019) Computational prediction of func- 66. Migault M, Donnou-Fournet E, Galibert M
tional microRNA–mRNA interactions. Meth- et al (2017) Definition and identification of
ods Mol Biol 1912:175–196 small RNA sponges: focus on miRNA seques-
52. Lee EJ, Baek M, Gusev Y et al (2007) System- tration. Methods 117:35–47
atic evaluation of microRNA processing pat- 67. Tay Y, Kats L, Salmena L et al (2011) Coding-
terns in tissues, cell lines, and tumors. RNA independent regulation of the tumor suppres-
14:35–42 sor PTEN by competing endogenous
53. Michael MZ, O’ Connor SM, Pellekaan H, mRNAs. Cell 147:344–357
van NG et al (2003) Reduced accumulation 68. Poliseno L, Salmena L, Zhang J et al (2010) A
of specific microRNAs in colorectal neoplasia. coding-independent function of gene and
Mol Cancer Res 1:882–891 pseudogene mRNAs regulates tumour biol-
54. Obernosterer G (2006) Post-transcriptional ogy. Nature 465:1033–1038
regulation of microRNA expression. RNA 69. Karreth FA, Tay Y, Perna D et al (2011) In
12:1161–1167 vivo identification of tumor- suppressive
55. Kawahara Y, Megraw M, Kreider E et al PTEN ceRNAs in an oncogenic BRAF-
(2008) Frequency and fate of microRNA edit- induced mouse model of melanoma. Cell
ing in human brain. Nucleic Acids Res 147:382–395
36:5270–5280 70. Yang C, Wu D, Gao L et al (2016) Competing
56. Bak RO, Mikkelsen JG (2014) miRNA endogenous RNA networks in human cancer:
sponges: soaking up miRNAs for regulation hypothesis, validation, and perspectives.
of gene expression. Wiley Interdiscip Rev Oncotarget 7:13479–13490
RNA 5:317–333 71. Salmena L, Poliseno L, Tay Y et al (2011) A
57. Ebert MS, Neilson JR, Sharp PA (2007) ceRNA hypothesis: the Rosetta stone of a
MicroRNA sponges: competitive inhibitors hidden RNA language? Cell 146:353–358
of small RNAs in mammalian cells. Nat Meth- 72. Wu Q, Yan H, Tao S-Q et al (2017) XIAP
ods 4:721–726 30 -untranslated region as a ceRNA promotes
58. Ebert MS, Sharp PA (2010) MicroRNA FSCN1 function in inducing the progression
sponges: progress and possibilities. RNA of breast cancer by binding endogenous
16:2043–2050 miR-29a-5p. Oncotarget 8:16784–16800
59. Kluiver J, Gibcus JH, Hettinga C et al (2012) 73. Sun L, Liu L, Yang J et al (2018) SATB1
Rapid generation of microRNA sponges for 30 -UTR and lncRNA-UCA1 competitively
microRNA inhibition. PLoS One 7:14–21 bind to miR-495-3p and together regulate
the proliferation and invasion of gastric can-
60. Reichel M, Li J, Millar AA (2011) Silencing cer. J Cell Biochem 120(4):6671–6682
the silencer: strategies to inhibit microRNA
activity. Biotechnol Lett 33:1285–1292 74. Thomson DW, Dinger ME (2016) Endoge-
nous microRNA sponges: evidence and con-
61. Otaegi G, Pollock A, Sun T (2012) An opti- troversy. Nat Rev Genet 17:272–283
mized sponge for microRNA miR-9 affects
spinal motor neuron development in vivo. 75. Xing Q, Huang Y, Wu Y et al (2018)
Front Neurosci 5:1–9 Integrated analysis of differentially expressed
profiles and construction of a competing
62. Loya CM, Lu CS, Van Vactor D et al (2009) endogenous long non-coding RNA network
Transgenic microRNA inhibition with spatio- in renal cell carcinoma. PeerJ 6:e5124
temporal specificity in intact organisms. Nat
Methods 6:897–903 76. Ghini F, Rubolino C, Climent M et al (2018)
Endogenous transcripts control miRNA levels
63. Gentner B, Schira G, Giustacchini A et al and activity in mammalian cells by target-
(2009) Stable knockdown of microRNA
directed miRNA degradation. Nat Commun 91. Wang Y, Liu HZ, Liu Y et al (2018) Down-
9:3119 regulated MALAT1 relates to recurrent preg-
77. Ameres SL, Horwich MD, Hung JHJ-H et al nancy loss via sponging miRNAs. Kaohsiung J
(2010) Target RNA-directed trimming and Med Sci 34:503–510
tailing of small silencing RNAs. Science 92. Wang J, Liu X, Wu H et al (2010) CREB
328:1534–1539 up-regulates long non-coding RNA, HULC
78. Fuchs Wightman F, Giono LE, Fededa JP expression through interaction with
et al (2018) Target RNAs strike back on microRNA-372 in liver cancer. Nucleic Acids
microRNAs. Front Genet 9:435 Res 38:5366–5383
79. Kleaveland B, Shi CY, Stefano J et al (2018) A 93. Zhou M, Gao M, Luo Y et al (2018) Long
network of noncoding regulatory RNAs acts non-coding RNA metallothionein 1 pseudo-
in the mammalian brain. Cell 174:350–362. gene 3 promotes p2y12 expression by spong-
e17 ing miR-126 to activate platelet in diabetic
80. Mata l, de M, Gaidatzis D, Vitanescu M et al animal model. Platelets 00:1–8
(2015) Potent degradation of neuronal miR- 94. Wang C-G, Liao Z, Xiao H et al (2019)
NAs induced by highly complementary tar- LncRNA KCNQ1OT1 promoted BMP2
gets. EMBO Rep 16:500–511 expression to regulate osteogenic differentia-
81. Friedman RC, Farh KK-H, Burge CB et al tion by sponging miRNA-214. Exp Mol
(2008) Most mammalian mRNAs are con- Pathol 107:77–84
served targets of microRNAs. Genome Res 95. Shidhi PR, Suravajhala P, Nayeema A et al
19:92–105 (2015) Making novel proteins from pseudo-
82. Chi SW, Zang JB, Mele A et al (2009) Argo- genes. Bioinformatics 31:33–39
naute HITS-CLIP decodes micro- 96. Lafontaine I, Dujon B (2010) Origin and fate
RNA–mRNA interaction maps. Nature of pseudogenes in Hemiascomycetes: a com-
460:479–486 parative analysis. BMC Genomics 11:260
83. Bartel DP (2009) MicroRNAs: target recog- 97. Jacq C, Miller JR, Brownlee GG (1977) A
nition and regulatory functions. Cell pseudogene structure in 5S DNA of Xenopus
136:215–233 laevis. Cell 12:109–120
84. Poliseno L, Pandolfi PP (2015) PTEN 98. Sen K, Ghosh TC (2013) Pseudogenes and
ceRNA networks in human cancer. Methods their composers: delving in the “debris” of
77–78:41–50 human genome. Brief Funct Genomics
85. Lee DY, Jeyapalan Z, Fang L et al (2010) 12:536–547
Expression of Versican 30 -untranslated region 99. Karreth FA, Reschke M, Ruocco A et al
modulates endogenous microRNA functions. (2015) The BRAF pseudogene functions as a
PLoS One 5:e13599 competitive endogenous RNA and induces
86. Fang L, Du WW, Yang X et al (2013) Versican lymphoma in vivo. Cell 161:319–332
30 -untranslated region (30 -UTR) functions as 100. Straniero L, Rimoldi V, Samarani M et al
a ceRNA in inducing the development of (2017) The GBAP1 pseudogene acts as a
hepatocellular carcinoma by regulating ceRNA for the glucocerebrosidase gene
miRNA activity. FASEB J 27:907–919 GBA by sponging miR-22-3p. Sci Rep
87. Jeyapalan Z, Deng Z, Shatseva T et al (2011) 7:1–13
Expression of CD44 30 -untranslated region 101. Kulcheski FR, Christoff AP, Margis R (2016)
regulates endogenous microRNA functions Circular RNAs are miRNA sponges and can
in tumorigenesis and angiogenesis. Nucleic be used as a new class of biomarker. J Biotech-
Acids Res 39:3026–3041 nol 238:42–51
88. Rutnam ZJ, Yang BB (2012) The non-coding 102. Han C, Seebacher NA, Hornicek FJ et al
30 UTR of CD44 induces metastasis by reg- (2017) Regulation of microRNAs function
ulating extracellular matrix functions. J Cell by circular RNAs in human cancer. Oncotar-
Sci 125:2075–2085 get 8:64622–64637
89. Jarroux J, Morillon A, Pinskaya M (2017) 103. Song YZ, Li JF (2018) Circular RNA hsa_-
Long non coding RNA biology. Springer Sin- circ_0001564 regulates osteosarcoma prolif-
gapore, Singapore eration and apoptosis by acting miRNA
90. Brannan CI, Dees EC, Ingram RS et al (1990) sponge. Biochem Biophys Res Commun
The product of the H19 gene may function as 495:2369–2375
an RNA. Mol Cell Biol 10:28–36 104. Chen L-L, Yang L (2015) Regulation of cir-
cRNA biogenesis. RNA Biol 12:381–388
105. Nigro JM, Cho KR, Fearon ER et al (1991) RNAs with regulatory potency. Nature
Scrambled exons. Cell 64:607–613 495:333–338
106. Capel B, Swain A, Nicolis S et al (1993) Cir- 109. Abdelmohsen K, Panda AC, De S et al (2015)
cular transcripts of the testis-determining Circular RNA ITCH has inhibitory effect on
gene Sry in adult mouse testis. Cell ESCC by suppressing the Wnt/β-catenin
73:1019–1030 pathway. Oncotarget 5:1–11
107. Bilge Y, Erdoğan İ, Akgül B (2019) Tran- 110. Yao Y, Hua Q, Zhou Y (2018) CircRNA
scriptomics analysis of circular RNAs differen- has_circ_0006427 suppresses the progression
tially expressed in apoptotic HeLa cells. Front of lung adenocarcinoma by regulating
Genet 10:176 miR-6783–3p/DKK1 axis and inactivating
108. Memczak S, Jens M, Elefsinioti A et al (2013) Wnt/β-catenin signaling pathway. Biochem
Circular RNAs are a large class of animal Biophys Res Commun 111:1367–1375
Chapter 6
MicroRNA Targeting
Hossein Ghanbarian, Mehmet Taha Yıldız, and Yusuf Tutar
Abstract
MicroRNAs (miRNAs) are small noncoding elements that play essential roles in the posttranscriptional
regulation of biochemical processes. miRNAs recognize and target multiple mRNAs; therefore, investigat-
ing miRNA dysregulation is an indispensable strategy to understand pathological conditions and to design
innovative drugs. Targeting miRNAs in diseases improve outcomes of several therapeutic strategies thus,
this present study highlights miRNA targeting methods through experimental assays and bioinformatics
tools. The first part of this review focuses on experimental miRNA targeting approaches for elucidating key
biochemical pathways. A growing body of evidence about the miRNA world reveals the fact that it is not
possible to uncover these molecules’ structural and functional characteristics related to the biological
processes with a deterministic approach. Instead, a systemic point of view is needed to truly understand
the facts behind the natural complexity of interactions and regulations that miRNA regulations present.
This task heavily depends both on computational and experimental capabilities. Fortunately, several miRNA
bioinformatics tools catering to nonexperts are available as complementary wet-lab approaches. For this
purpose, this work provides recent research and information about computational tools for miRNA
targeting research.
Key words miRNA, miRNA Targeting, Bioinformatics tools
1 Introduction
MicroRNAs (miRNAs) are single-stranded noncoding endogenous

mediators that modulate gene expression through RNA interfer-
ence (RNAi) and, therefore, determining and comparing miRNA
content under normal and pathological states is of growing interest
in the scientific community to investigate the etiology of several
diseases.
miRNAs control translation by base-pairing with their target
mRNAs. Thus, RNAi mediated miRNA silencing targets specific
genes based on sequence complementarity. This defines how miR-
NAs silence and function play pivotal roles in systemic diseases and
molecular mechanisms.
Databases and algorithms have been employed to elucidate the
miRNA–disease association in conjunction with clinical outcomes.
105
106 Hossein Ghanbarian et al.
Biomedical significance of predicting novel miRNA–disease asso-

ciations provides novel biomarkers and drug design strategies.
2 Experimental miRNA Target Identification
Many cellular pathways and processes, in healthy or abnormal con-

ditions, are regulated by miRNAs, which directly target more than
one-third of all genes. Almost all pathways have miRNA involve-
ment [1–11]. Several studies reveal that specific miRNAs might
individually regulate one to hundreds of targets [12, 13]. The
miRNA-loaded RISC (RNA-induced silencing complex) interacts
with the 30 UTR of target transcripts. Five to eight nucleotides from
the 50 end of the mature miRNA, known as the seed sequence,
interact with the complementary sequence which is located in the
30 UTR of target mRNAs [14] and thereby marks the target
mRNAs for degradation or inhibition of translation [15–
17]. Based on seed pairing, prediction of miRNA–mRNA interac-
tions is approachable [18–20]. Beyond the canonical seed matches,
an alternative miRNA binding region has recently identified as “G-
bulge sites” (or G-U wobble), which include a G-bulge position,
about 5–6 nucleotides of the seed region [21]. It has been pre-
dicted that around 15% of gene regulations by microRNAs are
mediated through functional bulge interactions [22, 23]. While in
animals, most microRNAs are thought to be involved in the degra-
dation of their target mRNAs [24], they might have two other
modes of action to regulate target-gene expression: translational
inhibition, and transcriptional gene silencing.
Identifying the targets of microRNAs is crucial for further
analysis of miRNA regulatory networks in cellular processes [25–
27]. Although various bioinformatics methods have been used for
miRNA target prediction [28], many predicted targets might be
false positives or false-negatives; consequently, genuine miRNA
targets can be missed and not possible to identify without experi-
mental evidence [12, 13, 22, 28–31]. Therefore, unbiased experi-
mental strategies are necessary to overcome the complexity of
identifying endogenous miRNA targets and their functions. While
each process has its pros and cons [32], gene expression profiling,
pull-down, and immunoprecipitation methods are currently avail-
able for miRNA target identification. On the other hand, functional
interactions are under spatiotemporal control and it will not be
possible to generate all pairs of putative miRNA–target interactions
to investigate them experimentally.
2.1 Gene Expression The actual microRNA target identification among the vast number
Profiling of predicted targets indicates usefulness of bioinformatics tools.
MicroRNA targets are currently identified through several experi-
mental approaches based on quantifying expression changes of
miRNA Targeting 107
target genes at the RNA and the protein levels following inhibition
or overexpression of specific miRNA. As posttranscriptional regu-
lators, the RNA level is only indicative while the protein level
should be the gold standard.
2.1.1 Transcriptome Since microRNAs are involved in degrading their targets mRNAs
Profiling (or inhibit the translation of their target mRNAs), high-
throughput techniques including microarray and sequencing are
the conventional experimental approaches for analyzing the expres-
sion of mRNAs in cells [33]. In this regard, miR-1 and miR-124a
mimics were initially transfected into Hela cells and down-
regulated mRNAs analyzed by microarray, indicating that several
mRNA transcripts down-regulated after the miRNAs over expres-
sion. The down-regulated transcripts were then analyzed for direct
targets according to miR/mRNA seed region interactions. Thereby
88% and 76% of down-regulated transcripts were revealed as direct
targets of miR-1 and miR-124, respectively [34]. It should be
noted that a fraction of miRNA will lead to target degradation at
transcriptional level control, but protein level needs to be assessed.
While similar studies supported mRNA downregulation following
the induction of miRNA expression [35, 36], the inhibition of
miRNA action by antisense short oligoribonucleotides “antago-
miRs” caused the upregulation of a high number of transcripts
[24, 37]. Probably, applying miRNA inhibitors and mimic simulta-
neously for specific microRNAs and comparing both increased and
decreased transcripts may assist with miRNA target validation,
however HITS-CLIP and similar approaches are the gold standard
for the detection and validation of miRNA–target
interactions [38].
An alternative to the studying gene expression using microar-
rays is performing next-generation sequencing (NGS) technology
such as RNA-seq. The latter allows the detection of unexpected
RNA species, while the former does not [38]. However, transcrip-
tome profiling using high-throughput methods has two limita-
tions. While predominate genes and pathways affected by a
miRNA, high-throughput methods does not distinguish between
direct targets as differential gene expression is observed amongst a
pool of indirect changes in transcript abundance. Also, after tran-
scriptome profiling, intensive bioinformatics analysis requires direct
target detection based on miRNA–mRNA seed region matches
[39]. According to the NGS analysis, it has been demonstrated
that around 80% of miRNA targets are regulated at the RNA level
[40]. However, transcriptome profiling cannot reveal those miRNA
targets, which are only regulated at the translational level, calling
for proteomics to support target identification.
2.1.2 Proteomic Since a microRNA’s regulatory roles modulate the level of mRNA
Approaches translation [12], proteomic approaches are required to assay the
ultimate effect of miRNAs and identifying their targets. SILAC
(stable isotope labeling by amino acids in cell culture) is an estab-
lished biochemical method in which newly synthesized proteins are
labeled by growing cells in medium containing heavy isotopes
[41]. Protein expression changes are then measured by mass spec-
trometry after modulation via miRNAs [13, 41]. The regulatory
role of several microRNAs including miR-1, miR-143, and
miR-181 on protein expression has been studied by this method
[41, 42]. Targeting may be monitored with enzymatic and chemi-
cal labeling, with reference peptide and label free techniques besides
metabolic SILAC labeling [43]. Over expression of microRNAs in
different cell types shows the repression of hundreds of proteins
directly or indirectly [12, 13]. However, biochemical approaches,
such as SILAC, are unable to determine the direct miRNA targets;
thus, bioinformatics methods are additionally needed for further
analysis.
2.1.3 Translation Translationally active mRNAs are actively associated with elongat-
Profiling “Polysome ing ribosomes. The mRNA fragments that are attached to the
Profiling” ribosomes, can be recovered through ribosome-profiling and can
be identified by deep sequencing [44]. In the polysome profiling
technique, miRNA transfected cells are treated with cycloheximide
to arrest the translating ribosomes [40]. Nonattached mRNAs are
then digested with RNase I and monosomes are purified through a
sucrose gradient centrifugation. Next, ribosomal RNA fragments
(~30 nucleotides) and total mRNA, in parallel, are isolated after
miRNA overexpression and identified by high-throughput
sequencing to determine the mRNA translational repression
based on the miRNA mechanism of action and lowered mRNA
levels account for the decreased protein production.
Similarly, to identifying the relative contribution of microRNA-
mediated translational repression and mRNA degradation, poly-
some profiling has been performed in HEK-293 T cells with over-
expression of miR-124, where it revealed around 600 putative
targets [45]. Accordingly, the mode of action of several other
microRNAs, including miR-1, miR-155, and miR-223 have been
identified based on the polysome profiling approach, showing the
expression and repression of genes with at least one predicted
microRNA binding site in their 30 UTR [40]. According to these
kinds of studies, miRNAs regulate gene expression mainly through
mRNA destabilization, where at least 84% of the targeted mRNAs
are degraded [40]. While ribosomal profiling approaches are quan-
titative and high throughput, similar to proteomic approaches, they
are unable to elucidate fully miRNAs’ target sites, since it cannot
provide detailed information about miRNA–mRNA interactions.
miRNA Targeting 109
2.2 Immuno- Different biochemical approaches have been developed to identify

precipitation the direct targets of specific miRNAs. Since all miRNAs bind to the
target mRNAs in the RISC complexes, coimmunoprecipitation of
the RISC proteins along with target mRNAs is the crucial point in
these approaches. To identify direct miRNA targets, cells are initi-
ally lysed following transfection with a miRNA mimic (miR over-
expression) or a miRNA inhibitor may increase accuracy Target
mRNAs, along with RISC, are coimmunoprecipitated using spe-
cific antibodies against the RISC components such as Argonaute
(AGO) [30, 46–50] or TNRC6 [48] proteins. For further quanti-
fying the captured RNAs, biochemical methods are followed by
high-throughput techniques such as RIP-Seq (ribonucleoprotein
immunoprecipitation followed by high-throughput sequencing) or
RIP-ChIP (Ribonucleoprotein Immunoprecipitation followed by
microarray chip analysis) [51, 52]. For instance, to identify targets
of miR-1 and miR-124, the RIP-ChIP assay has been applied in
HeLa or HEK-293 T cells after miRNA overexpression
[49, 50]. To do this; along with miR-1 or miR-124 mimics, an
exogenous epitope-tagged Argonaute (AGO2 or AGO1) expressed
in the cells, and the miRISC complex are immunoprecipitated
using an antibody against the epitope tag. After analyzing by micro-
array, 75% of the miR-124 and 70% of the miR-1 targets found in
mRNAs coimmunoprecipitated with Argonaute. Thus, the percen-
tages refer to the enrichment significance over what would be
expected by chance. Unlike other high-throughput approaches,
RIP-Seq/RIP-ChIP methods capture miRNA–mRNA target inter-
actions and allow the identification of the loading status of RISC
and the amount of mRNA and which mRNAs are targeted. How-
ever, the interaction between the AGO proteins and the miRNA–
mRNA target might be not stable during the coimmunoprecipita-
tion process. Dissociation of the miRISC complex during the
coimmunoprecipitation process is not only a potential drawback
in target identification but also may artificially facilitate interactions
between RNA and proteins in cell lysis, which may not necessarily
exist in vivo [53].
In order to stabilize the cellular RNA–protein interactions,
ultraviolet (UV) irradiations are used to cross-link the miRNA–
mRNA target duplex to the AGO proteins before
immunoprecipitation. This miRNA identification method, named
high-throughput sequencing of RNAs isolated by cross-linking
immunoprecipitation (HITS-CLIP), can provide the spectrum of
target mRNAs for each miRNA currently coexpressed [27, 47,
54]. In this case, first, UV exposure cross-links RNA–protein com-
plexes, and miRISC then precipitated through the immunoprecipi-
tation method. In the end, while the unbound RNA digests,
miRISC-protected RNA fragments are purified and analyzed by
HT-Seq. HITS-CLIP successfully identifies mapping of the
miRNA–mRNA interaction sites within both the UTR, coding
sequence, actual target, client miRNA specificity, and the multiplic-

ity in the genome [35, 55]. Performing HITS-CLIP shows that
miRNAs preferentially bind their targets through the 30 UTR,
while 25% of these interactions are mapped to open reading frames
(ORFs) and only 1% to the 50 UTR [55].
Although HITS-CLIP sheds resolves the location of the
mRNA targeted region to ~50 nucleotides [56], the precise posi-
tion of the cross-link between the RNAs and the RNA binding
proteins are not clear [48]. To identify the accurate location of the
cross-link, the PAR-CLIP method (photoactivatable-ribonucleo-
side-enhanced cross-linking and immunoprecipitation) has been
developed. This method uses photoreactive ribonucleoside analogs
such as 4-thiouridine, improving RNA recovery by 100- to 1000-
fold compared to the HITS-CLIP [48]. Upon UV irradiation,
4-thiouridine is incorporated into RNAs, resulting in thymidine
(T) to cytidine (C) transitions more frequently in cross-linked
sites, thereby marking direct interaction sites and inducing efficient
cross-linking [48]. Apart from precisely indicating the location of
targeting, this approach is also capable of identifying those targets
that only affected at the translational level. Using the PAR-CLIP
method, more miRNA targeting sites within the coding regions
were detected [47]. Similar to other high-throughput approaches,
a complex bioinformatics analysis still needs to verify direct
miRNA–mRNA interactions from both HITS-CLIP and
PAR-CLIP data, because of their indirect target identification
approach [57, 58].
Another analytical technique in the case of miRNA–mRNA
interaction is cross-linking ligation and sequencing of hybrids
(CLASH: cross-linking, ligation, and sequencing of hybrids).
CLASH is a great tool to predict the DNA–DNA, protein–DNA,
RNA–protein, and RNA–RNA interactions. By using this method,
18,000 miRNA–mRNA interactions have been analyzed [59, 60].
2.3 Pull-Down To improve miRNA target identification, several pull-down meth-

Methods ods have been developed. miRNAs labeled with chemical tags are
useful to identify pulled down species. LAMP (Labeled microRNA
pull-down assay) is one of the pull-down methods in which digox-
igenin (DIG)-labeled pre-miRNAs are used as probes to capture
target mRNAs [61]. Briefly, the DIG-labeled mature miRNA probe
is generated by Dicer cleavage following incubation of the
DIG-labeled pre-miRNA with cell extracts. The mature probe is
then incorporated into the RISC complex to bind their target
mRNA. Finally, the compound of DIG-labeled miRNA and
bound mRNA is precipitated by anti-DIG for further analysis.
While a novel target, hand2, has been identified for miRNA-1
using this method, DIG effects on miRNA function and the trans-
fection possibility of the labeled pre-miRNA into living cells remain
unclear.
miRNA Targeting 111
In an alternative method, the biotin-labeled synthetic miRNA

duplexes are used to directly capture miR-associated targets
[38]. After transfection, the 30 end-labeled miRNAs are incorporate
into miRISC in living cells, and the miRNA–mRNA complexes are
then captured from cell lysates by streptavidin beads for further
analysis to reveal miRNA–mRNA interactions. This technique has
been used to identify the targets for several miRNAs, including
miR-10a, miR-124, and miR-34a [61, 62]. For instance, using
biotin tagged miR-10a, Lund et al. have shown that miR-10a
induces ribosome mRNA translation through interaction with
their 50 UTR [61]. Although the biotin-labeled miRNA method
directly binds to the associated targets by incorporation into miR-
ISC complex and enabling pull down targets of a single miRNA,
capture efficiency of this method through seed matches needs to be
improved. To further enhance the capture efficiency, an alternative
biotin-labeled method, termed miR-TRAP (miRNA target affinity
purification), has been developed based on the creation of a cova-
lent link between a miRNA and its target mRNA. Photoreactive
molecules are conjugated to a biotin-labeled miRNA and covalently
bound to target mRNAs in the miRISC complex upon irradiation
[62]. The miRNA–mRNA complex is then purified through bio-
tin–streptavidin affinity purification. This procedure has been
employed for target identification of several miRNAs including
miR-29a and miR-135b [63].
Beyond their high efficiency and specificity, miRNA biotinyla-
tion methods need to be improved in respect to the incorporation
efficiency of labeled miRNA into RISC, since 30 -biotinylation
diminishes this step [64, 65]. To exclude the side effect of biotin,
Hall et al. have developed a new method, miRNA cross-linking and
immunoprecipitation (miR-CLIP), through changing the biotiny-
lation sites on miRNAs and combining this method with CLIP
[64]. The miR-CLIP process has been used for target identification
of miR-106a with a miR-probe labeled with biotin and a photo-
reactive molecule at the mid site. The miRNA–mRNA duplex in
RISC is immunoprecipitated with an AGO2 antibody, followed by
enrichment of the miRNA-associated targets on streptavidin beads.
While this method improves the identification accuracy, it still has
some limitations because of low capture efficiency, and also it is not
universal. To establish a comprehensive strategy; a novel method
has been developed based on the bioorthogonal ligation method
[66], a biocompatible ligation strategy for the labeling of biomo-
lecules [67]. Instead of biotin, miRNAs are labeled with tetrazole, a
small bioorthogonal group, at the 30 -end, so-called photoclickable
miRNA.
Regarding the conjugational possibility of the bioorthogonal
groups with different tags, the biotin molecule can attach to tetra-
zole through a photoclick reaction [68], resulting in a 30 -end
labeled miRNA without affecting its biological function. Based on
this method, photoclickable miR-122 probes have been success-

fully used for target identification. While this method has been
successfully introduced as a universal strategy, its capture efficiency
may be improved through combination with cross-linking
processes.
3 Experimental Validation of miRNA Targets
There is a wealth of bioinformatics tools predicting the targets of

the miRNAs (for additional information refer to [69]). While it is
not possible to experimentally test all predicted possible miRNA–
target interactions, it needs to be kept in mind that experimental
methods are needed to validate selected miRNA target predictions.
In this section, we first briefly explain the most important algo-
rithms that can predict miRNA targets and later on describe the
experimental methods to validate the results of the computational
predictions.
3.1 Computer Imperfect complementarity, complex regulatory networks at pro-

Algorithms to Predict tein and mRNA levels make target prediction challenging [70–
miRNA Targets 73]. However, bioinformatics and computational biology try to
help to predict the target mRNA, but there is no clear correlation
to define the rules of miRNA target recognition and target predic-
tion algorithms. Further, miRNA only partially overlaps with its
mRNAs, so it is quite sophisticated to predict the targets. Generally,
the sequence of 30 untranslated region (30 -UTR) mRNA use to find
the pray mRNA by these bioinformatics tools [29]. Throughout
evolution, the 50 end of the miRNAs, from nucleotide two to eight,
are conserved to interact with the 30 -UTR region of mRNAs. This
interaction is quite essential for miRNA functionality [74]. Compu-
tational biology helps to predict the potential targets of a miRNA,
and then follow-up wet-lab experiments confirm/validate the result
in vivo or in vitro [22]. Actually, initial studies to determine miRNA
target site before prediction tools available came from wet-lab
experiments by employing site-directed mutagenesis. Initial char-
acterization of let-7 and lin-4 provided a pattern which inspired
target prediction algorithms [75]. The bioinformatics tools predict
the matching based on the three criteria, seed sequence [35, 76],
target sites [35, 76–78], and thermodynamic rules [79]. Also, in
recent years, data mining and machine learning have been used to
predict target sites [80, 81]. Further, there are tools that uses just
sequences in order to predict miRNA targets [82, 83]. There are
several search engines to find the target mRNAs for one single
miRNA. Search engines such as DIANA microT-CDS (http://
www.microrna.gr/microT-CDS) [84], TargetScan (http://www.
targetscan.org) [85], PicTar (http://pictar.org) [86], miRanda
(http://microrna.sanger.ac.uk) [87], and ElMMo (http://www.
miRNA Targeting 113
mirz.unibas.ch) [88] are resources to access miRNA targets. The

algorithms can predict the best miRNAs for a sequence of a gene.
Although databases allow for both directions, target predictions are
always from miRNA to target and the targets are precalculated on a
per miRNA basis. It is better to use multiple algorithms to predict
miRNAs of a sequence, as they provide different results based on
various criteria, and in many cases, there are few overlaps in pre-
dicting miRNAs by two different bioinformatics algorithms
[89]. Many of the miRNAs predicted by these algorithms fail to
attach the target mRNA in vivo or in vitro [90]. The three-
dimensional structure of the mRNA and the its exposure angle to
the miRNA play a crucial role for mRNA–miRNA interactions
[91]. Among all of the above bioinformatics algorithms, only
DIANA microT-CDS [92], and TargetScan [85] have updated
their tools recently. To increase sensitivity and specificity of the
tools, it is imperative to prove the target genes in vitro and
in vivo. Further, the quality of pre-miRNA target prediction was
reported [93].
To prove the target genes, experimental molecular-based meth-
ods are most common. Techniques such as western blot, RT-PCR,
Northern blot, RNA sequencing, microarray, pull-down, immuno-
precipitation, and enzyme-linked immunosorbent assays are fre-
quently used to confirm the target genes [94].
3.2 Thermodynamic The structure of the target mRNA is quite essential for the ability of
Rules for miRNA– a miRNA to bind the target. Complicated structures of the mRNA
mRNA Interaction is thermodynamically a significant hurdle for miRNA to find the
Validation right position on the target mRNA and energy is released through
breaking up the mRNA structure and binding of the miRNA. For
example, the approachability of the heart- and neural crest
derivatives-expressed protein 2 (Hand2) transcription factor for
miRNA-1 is crucial for the interaction [91]. One site to calculate
the free energy (ΔG) of miRNA–mRNA interactions is mFold
[95]. It is interesting that the ΔG at binding regions in comparison
to the flanking areas. The variability of free energy is more obvious
moving from 50 to 30 of the mRNA that reveals the importance of
not only the target site but the flanking regions in miRNA–mRNA
interactions and interaction energy depends the length of miRNA.
The more the ΔG difference is at both flanking parts and the target
site, the more preferable is the bond to the miRNA [96, 97].
3.3 Reporting One procedure used to confirm miRNA–mRNA interactions is the

Methods reporter system. In this system, the entire 30 -UTR region of the
target gene is cloned into a plasmid with luciferase or green fluores-
cent gene. Luciferase or fluorescent activity measured in transfected
cells after 24–48 h of transfection then demonstrates the miRNA–
mRNA interaction [98]. Fluorescence decreases as a result of
miRNA binding to the target mRNA and thereby the interaction
can be inferred. This procedure was applied to miRNA-16.

MiRNA-16 effects in A549 cells, showing its role in regulating
Cyclin D1 (CCND1), Cyclin D3 (CCND3), Cyclin E1
(CCNE1), and CDK6 [99]. Cloning the 30 -UTR as a whole is a
tedious job, so, to validate a target gene, only the regions near the
miRNA regulatory element (MRE) can be considered. MRE is a
short sequence of nucleotides, around 60 base pairs (bp). Also,
even a small part of the MRE with ~22 bp can be applied to
prove the target mRNA. The cost of this method is less than
using all the surrounding regions of the MRE; synthesis of short
sequences is cheaper than longer ones [94]. In addition, an opti-
mized light switch luciferase assay system with 700 artificial miRNA
target reporters as positive controls can show the mRNA target of a
miRNA in the 30 -UTR [100]. The Luciferase reporter system with
4891 nonbiased cDNA sequences in the 30 -UTR makes the
30 -UTR library even more efficient to reveal the target mRNAs
not only at the 30 end but the whole length of the mRNA. The
system also helps to find the protein targets of a newly detected
miRNA [101].
3.4 miRNA–mRNA Another method used to experimentally prove the miRNA–mRNA

Interactions interaction is coexpression of the miRNA and mRNA in one cell.
The total RNA of the cell must be extracted after coexpression to
analyze the repression of the interested mRNA by a single miRNA
in the cells [102]. Another analytical technique to evaluate an
miRNA–mRNA interaction is CLASH, which is detailed above.
3.5 Western Blot Western blotting can help evaluate an miRNA–mRNA interaction
in a cell by monitoring protein abundance of a target with available
antibodies. Overexpression of a miRNA may enhance regulation of
a protein in the cell due to its interaction with the mRNA in
question. Using pSILAC (pulsed stable isotope labeling with
amino acids) revealed the role of this miRNA on the protein level
[103]. Knocking down miRNA expression followed by proteome
analysis can be another procedure to check the influence of a
particular miRNA on the protein level. Quantitative-mass-spec-
trometry based on using SILAC (stable isotope labeling with
amino acids in cell culture) was used to demonstrate the effects of
upregulating some miRNAs such as miRNA-124, miRNA-1, and
miRNA-181, and deleting miRNA-223 on the proteome of Hela
cells and neutrophils, respectively. Thus, protein levels change in
the cells without altering mRNA levels [12].
miRNA Targeting 115
4 Web-Based Bioinformatics Tools for miRNA Target Predictions
The importance of wet-lab efforts providing experimentally sup-

ported target data is evident for testing and improving bioinfor-
matics tools [104] especially taking the increasing volume of
miRNA data (currently ~4000 miRNAs from 23 species and exper-
imentally validated 422,517 miRNA–target interactions on miR-
TarBase) into account. Wide range of computational methods and
algorithms for computational miRNA research has been developed,
including web-based bioinformatics tools. A simple search with a
keyword of, for example “miRNA tools,” in the PubMed platform
results in almost 2800 papers published over the last 10 years.
Similarly, a search for web-based miRNA tools on the OMICtools
platform (https://omictools.com) [105] yields ~300 bioinformat-
ics tools applicable for variety of miRNA studies [105].
A wide variety of bioinformatics tools are available with pur-
poses ranging from the precise prediction of possible miRNAs,
miRNA target genes, possible miRNA precursors, miRNA func-
tion, miRNA–disease associations, novel miRNA prediction, and
miRNA–mRNA gene networks [106–112].
Although the vast majority of the programs are Unix/Linux
using the command line interface, a number of Windows compati-
ble programs either with graphical or web user interface are avail-
able. Free access with no limitations especially makes these tools
preferable among scientists who are less familiar with
computations.
Although there are many excellent reviews [113–117] and
book sections [118–121] available pointing out the same issue
with more detailed information about the programs such as algo-
rithms they use, pros and cons covering similar programs, we
practically aimed to provide a list of freely available, easy to use
web-based miRNA tools requiring no advanced computer skills
(Table 1). The tools have been selected based on a few criteria
such as availability, ease of use, availability of tutorials and/or help
sections, and the number of references to the paper presenting the
tool. Downloadable tools, modules or packages for varying
operating systems and open source software like Bioconductor
[158] are not included in the list. It should be noted that the
availability and functionality of the tools might change over time.
5 Conclusion
The human genome project revealed a large portion of the

sequence of the genome. A sequenced genome may not, however,
provide lead to a deeper understanding of the gene regulation
mechanism. It merely provides a parts list, which still needs to be
Table 1
List of some web-based miRNA research tools
Name Features Organisms Link References

canEvolve Query functionalities are H. sapiens www.canevolve.org [122]
designed to fulfill most
frequent analysis needs of
cancer researchers with a
view to generate novel
hypotheses. It stores gene,
microRNA (miRNA) and
protein expression profiles,
copy number alterations for
multiple cancer types, and
protein-protein interaction
information
CoMeTa The CoMeTa tool aims at the H. sapiens http://cometa.tigem. [123]
inference of miRNA targets it/index.php
and miRNA-regulated gene
networks by integrating
expression data from
hundreds of cellular and
tissue conditions
ComiR Predicts whether a given H. sapiens http://www.benoslab. [124]
mRNA is targeted by a set of M. musculus pitt.edu/comir/
miRNAs. ComiR uses D. melanogaster
miRNA expression to C. elegans
improve and combine
multiple miRNA targets for
each of the four prediction
algorithms: miRanda, PITA,
TargetScan, and mirSVR.
The composite scores of the
four algorithms are then
combined using a support
vector machine trained on
Drosophila Ago1 IP data
ComTAR ComTAR allows users to Plants http://rnabiology.ibr- [125]
analyze the variations of conicet.gov.ar/
known miRNA targets comtar/
during evolution and to
predict previously unknown
interactions by focusing on
the conservation of the
potential targeting
cWords Uses a method designed for H. sapiens http://servers.binf.ku. [126]
regulatory motif discovery in M. musculus dk/cwords/
differential case–control D. melanogaster
mRNA expression datasets C. elegans
(continued)
miRNA Targeting 117
Table 1
(continued)

DIANA tools Provides algorithms, databases H. sapiens https://diana.e-ce.uth. [84]
and software for interpreting gr/home
and archiving data in a
systematic framework
ranging from the analysis of
expression regulation from
deep sequencing data, the
annotation of miRNA
regulatory elements and
targets to the interpretation
of the role of ncRNAs in
various diseases and
pathways
EpimiR Stores the experimentally H. sapiens http://www.jianglab. [127]
validated mutual regulations M. musculus cn/EpimiR/
between epigenetic R. norvegicus
modifications (including G. gallus
DNA methylation, histone C. familiaris
acetylation, H3K4me3 and A. thaliana
H3K27me3, etc.) and
miRNAs
metaMIR metaMIR is a microRNA H. sapiens http://rna.informatik. [128]
(miRNA) framework to uni-freiburg.de/
predict interactions in metaMIR/Input.jsp
human between miRNAs
and clusters of genes. The
user provides a set of genes
to be targeted, and
optionally genes not to be
targeted. Taking data from a
reference database of
previously established
predictive algorithms,
metaMIR will return
miRNA candidates predicted
to coregulate genes among
those entered by analyzing
all possible subset
combinations
microRNA. This is a comprehensive H. sapiens http://www.microrna. [129]
org resource of microRNA M. musculus org/microrna/
target predictions and R. norvegicus home.do
expression profiles. Allows D. melanogaster
user to run prediction for C. elegans
microRNA targets and view
target downregulation
scores among experimentally
observed expression patterns
(continued)
Table 1
(continued)

miRCancer Provides comprehensive H. sapiens http://mircancer.ecu. [130]
collection of microRNA edu/
(miRNA) expression profiles
in various human cancers
which are automatically
extracted from published
literatures in PubMed. It
utilizes text mining
techniques for information
collection
MiRcode Provides “whole H. sapiens http://www.mircode. [131]
transcriptome” human org/index.php
microRNA target
predictions based on the
comprehensive GENCODE
gene annotation, including
>10,000 long noncoding
RNA genes
miRDB miRDB is an online database H. sapiens http://mirdb.org/ [132]
for miRNA target prediction R. norvegicus
and functional annotations. G. gallus
All the targets in miRDB M. musculus
were predicted by a C. familiaris
bioinformatics tool,
MirTarget, which was
developed by analyzing
thousands of miRNA–target
interactions from high-
throughput sequencing
experiments
mirDIP Finds microRNAs that target a H. sapiens http://ophid.utoronto. [133]
gene, or genes targeted by a ca/mirDIP/
microRNA, in Homo sapiens
miRecords miRecords is a resource for H. sapiens http://c1.accurascience. [134]
animal miRNA–target M. musculus com/miRecords/
interactions consisting of D. melanogaster
experimentally validated C. elegans
miRNA targets and the R. norvegicus
predicted targets component G. gallus
of miRecords which is an C. familiaris
integration of predicted D. rerio
miRNA targets produced by O. aries
11 established miRNA target
prediction programs
(continued)
miRNA Targeting 119
Table 1
(continued)

miRmap Comprehensively searches and H. sapiens https://mirmap.ezlab. [135]
predicts microRNA target P. troglodytes org/
repression strength and B. taurus
covers all four approaches G. gallus
(thermodynamic, D. rerio
evolutionary, probabilistic, M. musculus
and sequence-based R. norvegicus
features) using eleven Opossum
predictor features, three of
which are novel
miRNAminer miRNAminer is a web-based H. sapiens http://groups.csail.mit. [136]
tool used for homologous M. musculus edu/pag/
miRNA gene search in R. norvegicus mirnaminer/
several species. Given a P. troglodytes
search query, candidate C. familiaris
homologs are identified B. taurus
using BLAST search and C. elegans
then tested for their known G. gallus
miRNA properties, such as M. mulatta
secondary structure, energy, M. domestica
alignment, and O. anatinus
conservation, in order to
assess their fidelity
miRNAsong This tool generates miRNA 219 species https://www2.med. [137]
sponge constructs for muni.cz/histology/
specific miRNAs and miRNA miRNAsong/index.
families/clusters and tests php
them for potential binding
to miRNAs in selected
organisms. Currently,
miRNAsong allows for
testing of sponge constructs
in 219 species covering
35,828 miRNA sequences
miRNet Offers statistical, visual and H. sapiens https://www.mirnet. [138]
network-based approaches ca/
to help researchers
understand miRNAs
functions and regulatory
mechanisms. The
miRNA–target interaction
data were downloaded from
11 well-annotated databases
miRror Provides analysis tools for the H. sapiens http://www.proto.cs. [139]
cooperative regulation by M. musculus huji.ac.il/mirror/
microRNAs on gene sets and R. norvegicus
pathways D. melanogaster
C. elegans
D. rerio
(continued)
Table 1
(continued)

miRSystem miRSystem is a database which H. sapiens http://mirsystem.cgm. [140]
integrates seven well known M. musculus ntu.edu.tw/index.
miRNA target gene php
prediction programs:
DIANA, miRanda,
miRBridge, PicTar, PITA,
rna22, and TargetScan
miRTar It is a tool that enables H. sapiens http://mirtar.mbc.nctu. [141]
biologists easily to identify edu.tw/human/
the biological functions and tutorial.php
regulatory relationships
between a group of known/
putative miRNAs and
protein coding genes. It also
provides perspective of
information on the miRNA
targets on alternatively
spliced transcripts
miRTar2GO Program uses a method that – http://www.mirtar2go. [142]
predicts miRNA targets by org/index.html
allocating CLIPed regions of
the mRNA 30 UTRs to
miRNA seed regions
miRWalk 2.0 Supplies the biggest available H. sapiens http://mirwalk.umm. [143]
collection of predicted and M. musculus uni-heidelberg.de/
the experimentally verified R. norvegicus
miRNA–target interactions
and documents miRNA
binding sites within the
complete sequence of a gene,
but also combines this
information with a
comparison of binding sites
resulting from 12 existing
miRNA–target prediction
program
oncomiRDB Database includes H. sapiens http://lifeome.net/ [144]
experimentally verified database/
oncogenic and tumor- oncomirdb/
suppressive miRNAs for
both the computational
analysis and experimental
study on miRNA regulatory
networks and functions in
cancer
(continued)
miRNA Targeting 121
Table 1
(continued)

Pharmaco- Allows for the exploration of H. sapiens http://www.pharmaco- [145]
miR the full potential of miRNA mir.org/
pharmacogenomics by
searching the human
genome for
[miRNAs–genes–drugs]
associations with variable
search options, allowing for
different miRNA target
predicting algorithms and
searching only relevant
subgroups of genes or
miRNAs
PhenomiR The PhenomiR database H. sapiens http://mips.helmholtz- [146]
provides information about muenchen.de/
differentially regulated phenomir/
miRNA expression in
diseases and other biological
processes
PicTar PicTar is an algorithm for the Vertebrates https://pictar.mdc- [147]
identification of microRNA M. musculus berlin.de/
targets. This searchable D. melanogaster
website provides details (30 C. elegans
UTR alignments with
predicted sites, links to
various public databases,
etc.)
PITA This tool allows one to run the H. sapiens https://genie. [148]
PITA algorithm on the M. musculus weizmann.ac.il/
choice of UTRs and D. melanogaster pubs/mir07/mir07_
microRNAs. PITA starts by C. elegans prediction.html
scanning the UTR for
potential microRNA targets
(using the supplied seed
matching tools) and then
scores each site
psRNATarget The psRNATarget was Plants http://plantgrn.noble. [149]
specifically developed to org/psRNATarget/
identify target transcripts of
plant regulatory small RNAs
(sRNA), including most of
microRNAs (miRNAs) and a
subset of small interfering
RNAs (phasiRNAs),
through i) analyzing
(continued)
Table 1
(continued)

complementary matching
between sRNA and target
using a predefined scoring
schema; and ii) evaluating
target site accessibility by
calculating unpaired energy
(UPE)
RNA22 v2 Uses a pattern-based approach H. sapiens https://cm.jefferson. [150]
for the discovery of M. musculus edu/rna22/
microRNA binding sites and D. melanogaster Interactive/
their corresponding C. elegans
microRNA–mRNA
complexes
RNAhybrid The tool is primarily meant as a H. sapiens https://bibiserv.cebitec. [151]
means for microRNA target D. melanogaster uni-bielefeld.de/
prediction by finding the C. elegans rnahybrid?
minimum free energy id¼rnahybrid_view_
hybridization of a long submission
(target) and a short (query)
RNA
Semirna Uses putative target to search Plants http://www. [152]
for miRNAs in a genome, or bioinfocabd.upo.es/
a known miRNA to search semirna/semirna.
for orthologs html
SomamiR SomamiR is a database of H. sapiens http://compbio.uthsc. [153]
cancer somatic mutations in edu/SomamiR/
microRNAs (miRNA) and
their target sites that
potentially alter the
interactions between
miRNAs and competing
endogenous RNAs (ceRNA)
including mRNAs, circular
RNAs (circRNA), and long
noncoding RNAs (lncRNA)
TAPIR Designed for the prediction of Plants http://bioinformatics. [154]
plant microRNA targets. psb.ugent.be/
The server offers the webtools/tapir/
possibility to search for plant
miRNA targets using a fast
and a precise algorithm. The
precise option is much
slower but guarantees to find
less perfectly paired
miRNA–target duplexes.
Furthermore, the precise
(continued)
miRNA Targeting 123
Table 1
(continued)

option allows the prediction
of target mimics, which are
characterized by a
miRNA–target duplex
having a large loop, making
them undetectable by
traditional tools
TargetScan Predicts biological targets of H. sapiens http://www.targetscan. [155]
miRNAs by searching for the M. musculus org/vert_72/
presence of conserved 8mer, R. norvegicus
7mer, and 6mer sites that P. troglodytes
match the seed region of C. familiaris
each miRNA B. taurus
C. elegans
G. gallus
M. mulatta
D. melanogaster
D. rerio
X. laevis
Opossum
ViralmiR Identifies the microRNA Viruses http://csb.cse.yzu.edu. [156]
precursor in virus tw/viralmir/
VIRmiRNA is the first Viruses http://crdd.osdd.net/ [157]
specialized resource of servers/virmirna/
experimentally proven virus- index.html
encoded miRNAs and their
associated targets. This
database would enhance the
understanding of viral/host
gene regulation and may also
prove beneficial in the
development of antiviral
therapeutics
extracted from the large sequence. miRNAs are part of the puzzle
and along with innovative computer algorithms and unique experi-
mental designs scientist are now enabled to investigate another part
of the regulatory complexity next to gene regulation via transcrip-
tion factors [159, 160]. This work summarizes current research
methodologies in respect to miRNA targeting. The interactome of
these molecules under healthy and disease states provides clues for
disease treatment and drug design.
References
1. Hamzeiy H, Suluyayla R, Brinkrolf C, Widespread changes in protein synthesis
Janowski SJ, Hofestaedt R, Allmer J (2017) induced by microRNAs. Nature 455
Visualization and analysis of microRNAs (7209):58–63
within KEGG pathways using VANESA. J 14. Jinek M, Doudna JA (2009) A three-
Integr Bioinform 14(1):20160004 dimensional view of the molecular machinery
2. Hamzeiy H, Suluyayla R, Brinkrolf C, of RNA interference. Nature 457
Janowski SJ, Hofest€adt R, Allmer J (2018) (7228):405–412
Visualization and Analysis of miRNAs Impli- 15. Doench JG, Petersen CP, Sharp PA (2003)
cated in Amyotrophic Lateral Sclerosis Within siRNAs can function as miRNAs. Genes Dev
Gene Regulatory Pathways. Stud Health 17(4):438–442
Technol Inform 253:183–187 16. Doench JG, Sharp PA (2004) Specificity of
3. Calin GA, Croce CM (2006) MicroRNA- microRNA target selection in translational
cancer connection: the beginning of a new repression. Genes Dev 18(5):504–511
tale. Cancer Res 66(15):7390–7394 17. Humphreys DT, Westman BJ, Martin DI,
4. Griffiths-Jones S, Grocock RJ, Van Dongen S, Preiss T (2005) MicroRNAs control transla-
Bateman A, Enright AJ (2006) miRBase: tion initiation by inhibiting eukaryotic initia-
microRNA sequences, targets and gene tion factor 4E/cap and poly (A) tail function.
nomenclature. Nucleic Acids Res 34 Proc Natl Acad Sci U S A 102
(suppl_1):D140–D144 (47):16961–16966
5. He L, He X, Lim LP, De Stanchina E, Xuan Z, 18. Bartel DP (2009) MicroRNAs: target recog-
Liang Y et al (2007) A microRNA component nition and regulatory functions. Cell 136
of the p53 tumour suppressor network. (2):215–233
Nature 447(7148):1130–1134 19. Lewis BP, Burge CB, Bartel DP (2005) Con-
6. He L, Thomson JM, Hemann MT, served seed pairing, often flanked by adeno-
Hernando-Monge E, Mu D, Goodson S sines, indicates that thousands of human
et al (2005) A microRNA polycistron as a genes are microRNA targets. Cell 120
potential human oncogene. Nature 435 (1):15–20
(7043):828–833 20. Long D, Lee R, Williams P, Chan CY,
7. Johnson SM, Grosshans H, Shingara J, Ambros V, Ding Y (2007) Potent effect of
Byrom M, Jarvis R, Cheng A et al (2005) target structure on microRNA function. Nat
RAS is regulated by the let-7 microRNA fam- Struct Mol Biol 14(4):287–294
ily. Cell 120(5):635–647 21. Chi SW, Hannon GJ, Darnell RB (2012) An
8. Lu J, Getz G, Miska EA, Alvarez-Saavedra E, alternative mode of microRNA target recog-
Lamb J, Peck D et al (2005) MicroRNA nition. Nat Struct Mol Biol 19(3):321
expression profiles classify human cancers. 22. Alexiou P, Maragkakis M, Papadopoulos GL,
Nature 435(7043):834–838 Reczko M, Hatzigeorgiou AG (2009) Lost in
9. Mayr C, Bartel DP (2009) Widespread translation: an assessment and perspective for
shortening of 30 UTRs by alternative cleavage computational microRNA target identifica-
and polyadenylation activates oncogenes in tion. Bioinformatics 25(23):3049–3055
cancer cells. Cell 138(4):673–684 23. Sethupathy P, Megraw M, Hatzigeorgiou AG
10. O’Donnell KA, Wentzel EA, Zeller KI, Dang (2006) A guide through present computa-
CV, Mendell JT (2005) c-Myc-regulated tional approaches for the identification of
microRNAs modulate E2F1 expression. mammalian microRNA targets. Nat Methods
Nature 435(7043):839–843 3(11):881–886
11. Wienholds E, Kloosterman WP, Miska E, 24. Elmen J, Lindow M, Silahtaroglu A, Bak M,
Alvarez-Saavedra E, Berezikov E, de Bruijn E Christensen M, Lind-Thomsen A et al (2008)
et al (2005) MicroRNA expression in zebra- Antagonism of microRNA-122 in mice by
fish embryonic development. Science 309 systemically administered LNA-antimiR leads
(5732):310–311 to up-regulation of a large set of predicted
12. Baek D, Villén J, Shin C, Camargo FD, Gygi target mRNAs in the liver. Nucleic Acids Res
SP, Bartel DP (2008) The impact of micro- 36(4):1153–1162
RNAs on protein output. Nature 455 25. Bracken CP, Scott HS, Goodall GJ (2016) A
(7209):64–71 network-biology perspective of microRNA
13. Selbach M, Schwanh€ausser B, Thierfelder N, function and dysfunction in cancer. Nat Rev
Fang Z, Khanin R, Rajewsky N (2008) Genet 17(12):719
miRNA Targeting 125
26. Hausser J, Zavolan M (2014) Identification expression studies. Bioinformatics 26

and consequences of miRNA–target interac- (24):3131–3132
tions—beyond repression of gene expression. 40. Guo H, Ingolia NT, Weissman JS, Bartel DP
Nat Rev Genet 15(9):599–612 (2010) Mammalian microRNAs predomi-
27. Demirci MDS, Yousef M, Allmer J (2019) nantly act to decrease target mRNA levels.
Computational prediction of functional Nature 466(7308):835–840
microRNA–mRNA interactions. In: Compu- 41. Vinther J, Hedegaard MM, Gardner PP,
tational biology of non-coding RNA. Andersen JS, Arctander P (2006) Identifica-
Humana Press, New York, NY, pp 175–196 tion of miRNA targets with stable isotope
28. M Witkos T, Koscianska E, J Krzyzosiak W labeling by amino acids in cell culture. Nucleic
(2011) Practical aspects of microRNA target Acids Res 34(16):e107
prediction. Curr Mol Med 11(2):93–109 42. Yang Y, Chaerkady R, Kandasamy K, Huang
29. Bentwich I (2005) Prediction and validation TC, Selvan LDN, Dwivedi SB et al (2010)
of microRNAs and their targets. FEBS Lett Identifying targets of miR-143 using a
579(26):5904–5910 SILAC-based proteomic approach. Mol Bio-
30. Easow G, Teleman AA, Cohen SM (2007) Syst 6(10):1873–1882
Isolation of microRNA targets by miRNP 43. Allmer J (2012) Existing bioinformatics tools
immunopurification. RNA 13(8):1198–1204 for the quantitation of post-translational
31. Sethupathy P, Corda B, Hatzigeorgiou AG modifications. Amino Acids 42(1):129–138
(2006) TarBase: a comprehensive database of 44. Ingolia NT, Ghaemmaghami S, Newman JR,
experimentally supported animal microRNA Weissman JS (2009) Genome-wide analysis
targets. RNA 12(2):192–197 in vivo of translation with nucleotide resolu-
32. Li J, Zhang Y (2019) Current experimental tion using ribosome profiling. Science 324
strategies for intracellular target identification (5924):218–223
of microRNA. ExRNA 1(1):6 45. Hendrickson DG, Hogan DJ, McCullough
33. Thomas M, Lieberman J, Lal A (2010) Des- HL, Myers JW, Herschlag D, Ferrell JE,
perately seeking microRNA targets. Nat Brown PO (2009) Concordant regulation of
Struct Mol Biol 17(10):1169 translation and mRNA abundance for
34. Lim LP, Lau NC, Garrett-Engele P, hundreds of targets of a human microRNA.
Grimson A, Schelter JM, Castle J et al PLoS Biol 7(11):e1000238
(2005) Microarray analysis shows that some 46. Beitzinger M, Peters L, Zhu JY, Kremmer E,
microRNAs downregulate large numbers of Meister G (2007) Identification of human
target mRNAs. Nature 433(7027):769–773 microRNA targets from isolated argonaute
35. Grimson A, Farh KKH, Johnston WK, protein complexes. RNA Biol 4(2):76–84
Garrett-Engele P, Lim LP, Bartel DP (2007) 47. Chi SW, Zang JB, Mele A, Darnell RB (2009)
MicroRNA targeting specificity in mammals: Argonaute HITS-CLIP decodes micro-
determinants beyond seed pairing. Mol Cell RNA–mRNA interaction maps. Nature 460
27(1):91–105 (7254):479–486
36. Linsley PS, Schelter J, Burchard J, 48. Hafner M, Landthaler M, Burger L,
Kibukawa M, Martin MM, Bartz SR et al Khorshid M, Hausser J, Berninger P et al
(2007) Transcripts targeted by the (2010) Transcriptome-wide identification of
microRNA-16 family cooperatively regulate RNA-binding protein and microRNA target
cell cycle progression. Mol Cell Biol 27 sites by PAR-CLIP. Cell 141(1):129–141
(6):2240–2252 49. Hendrickson DG, Hogan DJ, Herschlag D,
37. Xu G, Fewell C, Taylor C, Deng N, Ferrell JE, Brown PO (2008) Systematic iden-
Hedges D, Wang X et al (2010) Transcrip- tification of mRNAs recruited to argonaute
tome and targetome analysis in MIR155 2 by specific microRNAs and corresponding
expressing cells using RNA-seq. RNA 16 changes in transcript abundance. PLoS One
(8):1610–1622 3:5
38. Nicolas FE, Pais H, Schwach F, Lindow M, 50. Karginov FV, Conaco C, Xuan Z, Schmidt
Kauppinen S, Moulton V, Dalmay T (2008) BH, Parker JS, Mandel G, Hannon GJ
Experimental identification of microRNA- (2007) A biochemical approach to identifying
140 targets by silencing and overexpressing microRNA targets. Proc Natl Acad Sci U S A
miR-140. RNA 14(12):2513–2520 104(49):19291–19296
39. Le Brigand K, Robbe-Sermesant K, Mari B, 51. Baroni TE, Chittur SV, George AD, Tenen-
Barbry P (2010) MiRonTop: mining micro- baum SA (2008) Advances in RIP-chip analy-
RNAs targets across large scale gene sis: RNA-binding protein
immunoprecipitation-microarray profiling. targetome uncovers a lincRNA H19–miR-

In: Post-transcriptional gene regulation. 106a interaction. Nat Chem Biol 11(2):107
Humana Press, pp 93–108 65. Okamura K, Phillips MD, Tyler DM,
52. Keene JD, Komisarow JM, Friedersdorf MB Duan H, Chou YT, Lai EC (2008) The regu-
(2006) RIP-Chip: the isolation and identifica- latory activity of microRNA* species has sub-
tion of mRNAs, microRNAs and protein stantial influence on microRNA and 30 UTR
components of ribonucleoprotein complexes evolution. Nat Struct Mol Biol 15(4):354
from cell extracts. Nat Protoc 1(1):302 66. Li J, Huang L, Xiao X, Chen Y, Wang X, Zhou
53. Mili S, Steitz JA (2004) Evidence for reasso- Z et al (2016) Photoclickable MicroRNA for
ciation of RNA-binding proteins after cell the intracellular target identification of Micro-
lysis: implications for the interpretation of RNAs. J Am Chem Soc 138
immunoprecipitation analyses. RNA 10 (49):15943–15949
(11):1692–1694 67. Sletten EM, Bertozzi CR (2009) Bioortho-
54. Licatalosi DD, Mele A, Fak JJ, Ule J, gonal chemistry: fishing for selectivity in a sea
Kayikci M, Chi SW et al (2008) HITS-CLIP of functionality. Angew Chem Int Ed 48
yields genome-wide insights into brain alter- (38):6974–6998
native RNA processing. Nature 456 68. Lim RK, Lin Q (2011) Photoinducible
(7221):464–469 bioorthogonal chemistry: a spatiotemporally
55. Zisoulis DG, Lovci MT, Wilbert ML, Hutt controllable tool to visualize and perturb pro-
KR, Liang TY, Pasquinelli AE, Yeo GW teins in live cells. Acc Chem Res 44
(2010) Comprehensive discovery of endoge- (9):828–839
nous Argonaute binding sites in Caenorhab- 69. Nachtigall PG, Bovolenta LA (2021)
ditis elegans. Nat Struct Mol Biol 17(2):173 Computational detection of microRNA tar-
56. Jiang Q, Feng MG, Mo YY (2009) Systematic gets. In: Allmer J, Yousef M (eds) miR-
validation of predicted microRNAs for cyclin Nomics: microRNA biology and
D1. BMC Cancer 9(1):194 computational analysis. Methods in molecular
57. Uhl M, Houwaart T, Corrado G, Wright PR, biology, vol 2257. Springer, New York
Backofen R (2017) Computational analysis of 70. Elton TS, Yalowich JC (2015) Experimental
CLIP-seq data. Methods 118:60–72 procedures to identify and validate specific
58. Wang T, Xiao G, Chu Y, Zhang MQ, Corey mRNA targets of miRNAs. EXCLI J 14:758
DR, Xie Y (2015) Design and bioinformatics 71. Huang Y, Zou Q, Song H, Song F, Wang L,
analysis of genome-wide CLIP experiments. Zhang G, Shen X (2010) A study of miRNAs
Nucleic Acids Res 43(11):5263–5274 targets prediction and experimental valida-
59. Helwak A, Kudla G, Dudnakova T, Tollervey tion. Protein Cell 1(11):979–986
D (2013) Mapping the human miRNA inter- 72. Kuhn DE, Martin MM, Feldman DS, Terry
actome by CLASH reveals frequent nonca- AV Jr, Nuovo GJ, Elton TS (2008) Experi-
nonical binding. Cell 153(3):654–665 mental validation of miRNA targets. Methods
60. Plotnikova OM, Skoblov MY (2018) Effi- 44(1):47–54
ciency of the miRNA–mRNA interaction pre- 73. Stark A, Brennecke J, Russell RB, Cohen SM
diction programs. Mol Biol 52(3):467–477 (2003) Identification of Drosophila micro-
61. Ørom UA, Nielsen FC, Lund AH (2008) RNA targets. PLoS Biol 1(3):E60
MicroRNA-10a binds the 50 UTR of ribo- 74. Brennecke J, Stark A, Russell RB, Cohen SM
somal protein mRNAs and enhances their (2005) Principles of microRNA–target recog-
translation. Mol Cell 30(4):460–471 nition. PLoS Biol 3(3):e85
62. Lal A, Thomas MP, Altschuler G, Navarro F, 75. Hamzeiy H, Allmer J, Yousef M (2014)
O’Day E, Li XL et al (2011) Capture of Computational methods for microRNA tar-
microRNA–bound mRNAs identifies the get prediction. In: miRNomics: microRNA
tumor suppressor miR-34a as a regulator of biology and computational analysis. Humana
growth factor signaling. PLoS Genet 7(11): Press, Totowa, NJ, pp 207–221
e1002363 76. Lewis BP, Shih IH, Jones-Rhoades MW, Bar-
63. Baigude H, Li Z, Zhou Y, Rana TM (2012) tel DP, Burge CB (2003) Prediction of mam-
miR-TRAP: a benchtop chemical biology malian microRNA targets. Cell 115
strategy to identify microRNA targets. (7):787–798
Angew Chem Int Ed 51(24):5880–5883 77. Mahen EM, Watson PY, Cottrell JW, Fedor
64. Imig J, Brunschweiger A, Brümmer A, MJ (2010) mRNA secondary structures fold
Guennewig B, Mittal N, Kishore S et al sequentially but exchange rapidly in vivo.
(2015) miR-CLIP capture of a miRNA PLoS Biol 8(2):e1000307
miRNA Targeting 127
78. Broughton JP, Lovci MT, Huang JL, Yeo GW, 91. Zhao Y, Samal E, Srivastava D (2005) Serum
Pasquinelli AE (2016) Pairing beyond the response factor regulates a muscle-specific
seed supports microRNA targeting specificity. microRNA that targets Hand2 during cardio-
Mol Cell 64(2):320–333 genesis. Nature 436(7048):214–220
79. Yue D, Liu H, Huang Y (2009) Survey of 92. Reczko M, Maragkakis M, Alexiou P,
computational algorithms for microRNA tar- Grosse I, Hatzigeorgiou AG (2012) Func-
get prediction. Curr Genomics 10 tional microRNA targets in protein coding
(7):478–492 sequences. Bioinformatics 28(6):771–776
80. Vlachos IS, Hatzigeorgiou AG (2013) Online 93. Demirci MDS, Baumbach J, Allmer J (2017)
resources for miRNA analysis. Clin Biochem On the performance of pre-microRNA detec-
46(10–11):879–900 tion algorithms. Nat Commun 8(1):1–9
81. Yousef M, Jung S, Kossenkov AV, Showe LC, 94. Choi C, Han J, Tran NTT, Yoon S, Kim G,
Showe MK (2007) Naı̈ve Bayes for micro- Song S et al (2017) Effective experimental
RNA target predictions—machine learning validation of miRNA targets using an
for microRNA targets. Bioinformatics 23 improved linker reporter assay. Genet Res
(22):2987–2992 99:e2
82. Yousef M, Levy D, Allmer J (2018) Species 95. Zuker M (2003) Mfold web server for nucleic
categorization via microRNAs based on acid folding and hybridization prediction.
30 UTR target sites using sequence features. Nucleic Acids Res 31(13):3406–3415
In: Proceedings of the 11th international 96. Ghoshal A, Shankar R, Bagchi S, Grama A,
joint conference on biomedical engineering Chaterji S (2015) MicroRNA target predic-
systems and technologies, vol 4, pp 112–118 tion using thermodynamic and sequence
83. Yousef M, Khalifa W, Acar İE, Allmer J (2017) curves. BMC Genomics 16(1):999
Distinguishing between microRNA targets 97. Ghoshal A, Zhang J, Roth MA, Xia KM,
from diverse species using sequence motifs Grama AY, Chaterji S (2018) A distributed
and K-mers. In: Bioinformatics, pp 133–139 classifier for microRNA target prediction
84. Paraskevopoulou MD, Georgakilas G, with validation through TCGA expression
Kostoulas N, Vlachos IS, Vergoulis T, Reczko data. IEEE/ACM Trans Comput Biol Bioin-
M et al (2013) DIANA-microT web server form 15(4):1037–1051
v5. 0: service integration into miRNA func- 98. Oh SW, Do Won Hwang DSL (2013) In vivo
tional analysis workflows. Nucleic Acids Res monitoring of microRNA biogenesis using
41(W1):W169–W173 reporter gene imaging. Theranostics 3
85. Garcia DM, Baek D, Shin C, Bell GW, (12):1004
Grimson A, Bartel DP (2011) Weak seed- 99. Liu Q, Fu H, Sun F, Zhang H, Tie Y, Zhu J
pairing stability and high target-site abun- et al (2008) miR-16 family induces cell cycle
dance decrease the proficiency of lsy-6 and arrest by regulating multiple cell cycle genes.
other microRNAs. Nat Struct Mol Biol 18 Nucleic Acids Res 36(16):5391
(10):1139 100. Aldred SF, Collins P, Trinklein N (2011)
86. Lall S, Grün D, Krek A, Chen K, Wang YL, Identifying targets of human micrornas with
Dewey CN et al (2006) A genome-wide map the LightSwitch luciferase assay system using
of conserved microRNA targets in C. elegans. 30 UTR-reporter constructs and a microRNA
Curr Biol 16(5):460–471 mimic in adherent cells. J Vis Exp (55):e3343
87. Enright AJ, John B, Gaul U, Tuschl T, 101. Ito Y, Inoue A, Seers T, Hato Y, Igarashi A,
Sander C, Marks DS (2003) MicroRNA tar- Toyama T et al (2017) Identification of tar-
gets in Drosophila. Genome Biol 5(1):R1 gets of tumor suppressor microRNA-34a
88. Gaidatzis D, van Nimwegen E, Hausser J, using a reporter library system. Proc Natl
Zavolan M (2007) Inference of miRNA tar- Acad Sci U S A 114(15):3927–3932
gets using evolutionary conservation and 102. Kuhn DE, Martin MM, Feldman DS, Terry
pathway analysis. BMC Bioinformatics 8 AV Jr, Nuovo GJ, Elton TS (2008) Experi-
(1):69 mental validation of miRNA targets. Methods
89. Ekimler S, Sahin K (2014) Computational 44(1):47–54
methods for microRNA target prediction. 103. Ebner OA, Selbach M (2011) Whole cell pro-
Genes 5(3):671–683 teome regulation by microRNAs captured in a
90. Didiano D, Hobert O (2006) Perfect seed pulsed SILAC mass spectrometry approach.
pairing is not a generally reliable predictor In: Argonaute proteins. Humana Press, pp
for miRNA-target interactions. Nat Struct 315–331
Mol Biol 13(9):849–851
104. Fridrich A, Hazan Y, Moran Y (2019) Too bioinformatics. Humana Press, New York,
many false targets for microRNAs: challenges NY, pp 99–116
and pitfalls in prediction of miRNA targets 117. Gunaratne PH, Coarfa C, Soibam B, Tandon
and their gene ontology in model and A (2012) miRNA data analysis: next-gen
non-model organisms. BioEssays 41 sequencing. In: Next-generation microRNA
(4):1800169 expression profiling technology. Humana
105. Henry VJ, Bandrowski AE, Pepin AS, Gonza- Press, Totowa, NJ, pp 273–288
lez BJ, Desfeux A (2014) OMICtools: an 118. Riffo-Campos ÁL, Riquelme I, Brebi-Mieville
informative directory for multi-omic data P (2016) Tools for sequence-based miRNA
analysis. Database 2014:bau069 target prediction: what to choose? Int J Mol
106. Demirci MDS, Baumbach J, Allmer J (2017) Sci 17(12):1987
On the performance of pre-microRNA detec- 119. Yousef M, Showe L, Showe M (2009) A study
tion algorithms. Nat Commun 8(1):1–9 of microRNAs in silico and in vivo: bioinfor-
107. Kozomara A, Birgaoanu M, Griffiths-Jones S matics approaches to microRNA discovery
(2019) miRBase: from microRNA sequences and target identification. FEBS J 276
to function. Nucleic Acids Res 47(D1): (8):2150–2156
D155–D162 120. Mendes ND, Freitas AT, Sagot MF (2009)
108. Brown J, Phillips AR, Lewis DA, Mans MA, Current tools for the identification of
Chang Y, Tanguay RL et al (2019) Bioinfor- miRNA genes and their targets. Nucleic
matics resource manager: a systems biology Acids Res 37(8):2419–2433
web tool for microRNA and omics data inte- 121. Griffiths-Jones S, Saini HK, van Dongen S,
gration. BMC Bioinformatics 20(1):255 Enright AJ (2007) miRBase: tools for micro-
109. Chen X, Xie D, Wang L, Zhao Q, You ZH, RNA genomics. Nucleic Acids Res 36
Liu H (2018) BNPMDA: bipartite network (suppl_1):D154–D158
projection for MiRNA–disease association 122. Samur MK, Yan Z, Wang X, Cao Q, Munshi
prediction. Bioinformatics 34 NC, Li C, Shah PK (2013) canEvolve: a web
(18):3178–3186 portal for integrative oncogenomics. PLoS
110. Huang TH, Fan B, Rothschild MF, Hu ZL, One 8(2):e56228
Li K, Zhao SH (2007) MiRFinder: an 123. Gennarino VA, d’Angelo G,
improved approach and software implemen- Dharmalingam G, Fernandez S,
tation for genome-wide fast microRNA pre- Russolillo G, Sanges R et al (2012) Identifica-
cursor scans. BMC Bioinformatics 8(1):341 tion of microRNA-regulated gene networks
111. Peace RJ, Hassani MS, Green JR (2019) by expression analysis of target genes.
miPIE: NGS-based prediction of miRNA Genome Res 22(6):1163–1172
using integrated evidence. Sci Rep 9(1):1–10 124. Coronnello C, Benos PV (2013) ComiR:
112. Renaud L, Simpson J, Glen JW, Hazard ES, combinatorial microRNA target prediction
Chung D, Hardiman G (2018) miRmapper: a tool. Nucleic Acids Res 41(W1):
tool for interpretation of miRNA mRNA W159–W164
interaction networks. Genes 9(9):458 125. Chorostecki U, Palatnik JF (2014) comTAR:
113. Shukla V, Varghese VK, Kabekkodu SP, a web tool for the prediction and characteri-
Mallya S, Satyamoorthy K (2017) A compila- zation of conserved microRNA targets in
tion of web-based research tools for miRNA plants. Bioinformatics 30(14):2066–2067
analysis. Brief Funct Genomics 16 126. Rasmussen SH, Jacobsen A, Krogh A (2013)
(5):249–273 cWords-systematic microRNA regulatory
114. Lukasik A, Zielenkiewicz P (2019) An over- motif discovery from mRNA expression
view of miRNA and miRNA target analysis data. Silence 4(1):2
tools. In: Plant microRNAs. Humana Press, 127. Dai E, Yu X, Zhang Y, Meng F, Wang S, Liu X
New York, NY, pp 65–87 et al (2014) EpimiR: a database of curated
115. Hertel J, Langenberger D, Stadler PF (2014) mutual regulation between miRNAs and epi-
Computational prediction of microRNA genetic modifications. Database 2014:
genes. In: RNA sequence, structure, and bau023
function: computational and bioinformatic 128. Davis JA, Saunders SJ, Mann M, Backofen R
methods. Humana Press, Totowa, NJ, pp (2017) Combinatorial ensemble miRNA tar-
437–456 get prediction of co-regulation networks with
116. Passetti F, Jorge NAN, Durham A (2014) non-prediction data. Nucleic Acids Res 45
Using bioinformatics tools to study the role (15):8745–8757
of microRNA in cancer. In: Clinical
miRNA Targeting 129
129. Betel D, Wilson M, Gabow A, Marks DS, 142. Ahadi A, Sablok G, Hutvagner G (2016)
Sander C (2008) The microRNA. org miRTar2GO: a novel rule-based model
resource: targets and expression. Nucleic learning method for cell line specific micro-
Acids Res 36(suppl_1):D149–D153 RNA target prediction that integrates Ago2
130. Xie B, Ding Q, Han H, Wu D (2013) miR- CLIP-Seq and validated microRNA–target
Cancer: a microRNA–cancer association data- interaction data. Nucleic Acids Res 45(6):
base constructed by text mining on literature. e42–e42
Bioinformatics 29(5):638–644 143. Sticht C, De La Torre C, Parveen A, Gretz N
131. Jeggari A, Marks DS, Larsson E (2012) miR- (2018) miRWalk: an online resource for pre-
code: a map of putative microRNA target sites diction of microRNA binding sites. PLoS
in the long non-coding transcriptome. Bioin- One 13(10):e0206239
formatics 28(15):2062–2063 144. Wang D, Gu J, Wang T, Ding Z (2014)
132. Wong N, Wang X (2015) miRDB: an online OncomiRDB: a database for the experimen-
resource for microRNA target prediction and tally verified oncogenic and tumor-
functional annotations. Nucleic Acids Res 43 suppressive microRNAs. Bioinformatics 30
(D1):D146–D152 (15):2237–2238
133. Tokar T, Pastrello C, Rossos AE, Abovsky M, 145. Rukov JL, Wilentzik R, Jaffe I, Vinther J,
Hauschild AC, Tsay M et al (2017) mirDIP Shomron N (2013) Pharmaco-miR: linking
4.1—integrative database of human micro- microRNAs and drug effects. Brief Bioinform
RNA target predictions. Nucleic Acids Res 15(4):648–659
46(D1):D360–D370 146. Bhattacharya A, Cui Y (2015) miR2GO:
134. Xiao F, Zuo Z, Cai G, Kang S, Gao X, Li T comparative functional analysis for micro-
(2008) miRecords: an integrated resource for RNAs. Bioinformatics 31(14):2403–2405
microRNA–target interactions. Nucleic Acids 147. Krek A, Grün D, Poy MN, Wolf R,
Res 37(suppl_1):D105–D110 Rosenberg L, Epstein EJ et al (2005) Combi-
135. Vejnar CE, Zdobnov EM (2012) MiRmap: natorial microRNA target predictions. Nat
comprehensive prediction of microRNA tar- Genet 37(5):495
get repression strength. Nucleic Acids Res 40 148. Kertesz M, Iovino N, Unnerstall U, Gaul U,
(22):11673–11683 Segal E (2007) The role of site accessibility in
136. Artzi S, Kiezun A, Shomron N (2008) miR- microRNA target recognition. Nat Genet 39
NAminer: a tool for homologous microRNA (10):1278
gene search. BMC Bioinformatics 9(1):39 149. Dai X, Zhuang Z, Zhao PX (2018) psRNA-
137. Barta T, Peskova L, Hampl A (2016) miRNA- Target: a plant small RNA target analysis
song: a web-based tool for generation and server (2017 release). Nucleic Acids Res 46
testing of miRNA sponge constructs in silico. (W1):W49–W54
Sci Rep 6:36625 150. Miranda KC, Huynh T, Tay Y, Ang YS, Tam
138. Fan Y, Siklenka K, Arora SK, Ribeiro P, WL, Thomson AM et al (2006) A pattern-
Kimmins S, Xia J (2016) miRNet-dissecting based method for the identification of Micro-
miRNA-target interactions and functional RNA binding sites and their corresponding
associations through network-based visual heteroduplexes. Cell 126(6):1203–1217
analysis. Nucleic Acids Res 44(W1): 151. Rehmsmeier M, Steffen P, Höchsmann M,
W135–W141 Giegerich R (2004) Fast and effective predic-
139. Friedman Y, Naamati G, Linial M (2010) tion of microRNA/target duplexes. RNA 10
MiRror: a combinatorial analysis web tool (10):1507–1517
for ensembles of microRNAs and their tar- 152. Muñoz-Mérida A, Perkins JR, Viguera E,
gets. Bioinformatics 26(15):1920–1921 Thode G, Bejarano ER, Pérez-Pulido AJ
140. Lu TP, Lee CY, Tsai MH, Chiu YC, Hsiao (2012) Semirna: searching for plant miRNAs
CK, Lai LC, Chuang EY (2012) miRSystem: using target sequences. Omics 16
an integrated system for characterizing (4):168–177
enriched functions and pathways of micro- 153. Bhattacharya A, Cui Y (2015) SomamiR 2.0:
RNA targets. PLoS One 7(8):e42390 a database of cancer somatic mutations alter-
141. Hsu JBK, Chiu CM, Hsu SD, Huang WY, ing microRNA–ceRNA interactions. Nucleic
Chien CH, Lee TY, Huang HD (2011) miR- Acids Res 44(D1):D1005–D1010
Tar: an integrated system for identifying 154. Bonnet E, He Y, Billiau K, Van de Peer Y
miRNA-target interactions in human. BMC (2010) TAPIR, a web server for the predic-
Bioinformatics 12(1):300 tion of plant microRNA targets, including
target mimics. Bioinformatics 26 158. Gentleman RC, Carey VJ, Bates DM,
(12):1566–1568 Bolstad B, Dettling M, Dudoit S et al
155. Agarwal V, Bell GW, Nam JW, Bartel DP (2004) Bioconductor: open software devel-
(2015) Predicting effective microRNA target opment for computational biology and bioin-
sites in mammalian mRNAs. elife 4:e05005 formatics. Genome Biol 5(10):R80
156. Huang KY, Lee TY, Teng YC, Chang TH 159. Tutar Y (2014) Editorial (thematic issue:“-
(2015) ViralmiR: a support-vector-machine- miRNA and cancer; computational and exper-
based method for predicting viral microRNA imental approaches”). Curr Pharm
precursors. BMC Bioinformatics 16(1):S9 Biotechnol 15(5):429–429
157. Qureshi A, Thakur N, Monga I, Thakur A, 160. Tutar L, Tutar E, Tutar Y (2014) MicroRNAs
Kumar M (2014) VIRmiRNA: a comprehen- and cancer; an overview. Curr Pharm Biotech-
sive resource for experimentally validated viral nol 15(5):430–437
miRNAs and their targets. Database 2014:
bau103
Chapter 7
MicroRNA Databases and Tools

Tharcı́sio Soares de Amorim, Daniel Longhi Fernandes Pedro,
and Alexandre Rossi Paschoal
Abstract
In this era of big data, sets of methodologies and strategies are designed to extract knowledge from huge
volumes of data. However, the cost of where and how to get this information accurately and quickly is
extremely important, given the diversity of genomes and the different ways of representing that informa-
tion. Among the huge set of information and relationships that the genome carries, there are sequences
called miRNAs (microRNAs). These sequences were described in the 1990s and are mainly involved in
mechanisms of regulation and gene expression. Having this in mind, this chapter focuses on exploring the
available literature and providing useful and practical guidance on the miRNA database and tools topic. For
that, we organized and present this text in two ways: (a) the update reviews and articles, which best
summarize and discuss the theme; and (b) our update investigation on miRNA literature and portals
about databases and tools. Finally, we present the main challenge and a possible solution to improve
resources and tools.
Key words miRNAs, Mirtrons, Databases, Repository, Resources, Tools, Prediction, Target gene,
Bioinformatics, Data analysis
1 Introduction
MicroRNA (miRNA) is the most studied small noncoding RNA

(ncRNA) [1]. This small RNA was first described in 1993 [2] and
now constitutes more than 38,589 entries according to miRBase
release 22.1 [3]. As a consequence, miRNA is the noncoding RNA
(ncRNA) having a large number of available databases and tools
[1, 4]. Studies conducted on miRNAs occur in larger proportion
relative to other ncRNAs, mainly because of their mechanisms of
control at the transcriptional and posttranscriptional level of
mRNAs [5]. The biology of miRNA has some differences between
animals and plants, which influence biological assay and
Electronic supplementary material: The online version of this chapter (https://doi.org/10.1007/978-1-0716-

1170-8_7) contains supplementary material, which is available to authorized users.
131
132 Tharcı́sio Soares de Amorim et al.
computational approaches for these groups [6, 7]. To aim to over-

come problems involved in organizing and identifying miRNAs,
bioinformatics databases and tools are a powerful strategy for dis-
covery, to investigate novel hypotheses and to perform analysis of
miRNA in several levels (e.g., miRNA itself, target functional,
network motifs) [8]. In this chapter, we provide three contribu-
tions: (a) we highlighted the most updated reviews and papers on
this topic or related to it; (b) we detailed the information about
each database and tool found among the literature up to March
2019; and (c) we described how to easily search for information
about databases and tools in two web portals that catalog this
information. Finally, for item (b), we also performed an exploratory
data analysis, grouped the resources and tools in similar categories,
and summarized the main information among each of these in the
Supplementary Material.
2 A Comprehensive Overview of miRNA Databases and Tools in the Literature
We first looked for recent publications (papers/chapters, reviews)

having an overview of databases and tools. We summarize this
information in Table 1. For example, Kozomara et al. (2019)
described the data in about 25 databases and presented how these
databases are integrated (see Fig. 2 in [5]). Also, these report
miRNAs data associated with single nucleotide polymorphisms
(SNP), disease, etc., and discuss these against miRNA data
extracted from the literature. Another publication reported in
depth bioinformatics tools for miRNA prediction application
[6]. A relationship between miRNA research, data, and tools was
presented in several graphical views that provide a fascinating way to
display literature information (see Figs. 2 and 3 in [6]). More about
the machine learning techniques for prediction miRNA is discussed
in another review [9]. The authors also compared supervised and
unsupervised methods for pre-miRNA prediction. In another pub-
lication, the discovery of transfer RNA (tRNA) and ribosome RNA
(rRNA) in the 1950s up to 2017, where a timeline of ncRNA
research is shown, was discussed [10]. The review also explored
the data analysis in databases and tools for ncRNAs context with
some examples for miRNAs.
Finally, we also found the main portal that could help to search
quickly by miRNA databases and tools. The Non-coding RNA
Databases Resource (NRDR) is the first portal that catalogs more
than 200 databases about noncoding RNAs wherein 50% are spe-
cific for the miRNA class [1, 4]. The other portal is the Tools4miRs
[11], a manually curated platform with more than 170 methods for
miRNA analysis (tools and databases). We explore these portals
more fully later.
MicroRNA DBs and Tools 133
Table 1
Summary of recent literature on miRNA databases and tools
Article Pubmed
type ID Main subject Summary Year
Review 27021491 miRNA databases Description about 25 miRNA databases and some 2017
graphics that summarize their information.
Review 29982332 miRNA tools This review deeply shows and discuss about 2018
miRNA bioinformatics tools since 2003.
Review 29800232 miRNA machine Focus on machine learning approaches for the 2018
learning methods prediction novel miRNAs
Review 29939215 ncRNAs databases Overview on the ncRNAs repositories and 2018
and tools ncRNAs interaction data in eukaryotes and
tools.
Chapter 30635896 miRNA prediction, This chapter tries to review miRNA-centered 2019
target tools and databases, algorithms, and tools to predict
resources miRNA and their targets.
Chapter/ 30635897 ncRNAs and miRNA The first central repository ncRNAs database 2019
paper 22336709 databases information. The novel version contains more 2012
than 200 resources described and sequence data
from 30 databases.
3 Exploratory Data Analysis on miRNA Databases and Tools from the Literature
We argue now how is the information found on publications about

database and tools. For that, we first mined the literature in
PubMed by using two query search strings.
l For databases “(miRNA[Title] OR miRNAs[Title] OR micro-
RNA[Title] OR microRNAs[Title]) AND (databases[Title] OR
resource[Title] OR resources[Title] OR repository[Title] OR
repositories[Title]).”
l For tools “(miRNA[Title/Abstract] OR microRNA[Title/
Abstract]) AND (tool[Title] OR software[Title] OR “web
server” [Title]).”
In this case, we considered publications to March 2019. Next,
we manually collected those entries that are only about tools or
databases. We extracted the XML file from PubMed database and
performed an exploratory analysis (graphics) with the Scientific
Miner tool.1 At the end, a total of 130 databases and 67 tools
were selected. Finally, we compared the databases and tools
1
Victor Scapin, Katia Romero Felizardo Scannavino, and Alexandre Rossi Paschoal. SCIENTIFIC MINER: FE
RRAMENTA PARA APOIO À REVISÃO SISTEMÁTICA E ANÁLISE BIBLIOMÉTRICA. Master of Infor-
matics. Dissertation, 2017.
Fig. 1 The number (blue) and cumulative (red) distribution of publications in databases (a) and tools (b) over
the years
information collected against the NRDR portal [1, 4] and Tools4-

miRs website [11]. Although the number of publications has been
growing each year (cumulative publication in Fig. 1: red bars),
there is a fluctuation in the number of publications of each year
(Fig. 1: blue bars). miRBase continues to be the database state-of-
the-art in miRNAs data. However, its published databases appear as
novel notices. One case is the first repository dedicated to mirtrons
(mirtronDB–12). Also, a previous database for studying the
Fig. 2 Analysis with cloud words and journals articles in databases (a) and tools (b)
Fig. 3 Databases bubble plot among all journals and numbers of each publication by year
relationship between miRNA and transposable elements [PlanTE-

MIR DB–13] is now covering other noncoding RNA classes. We
also plot an exploratory analysis on the words among the abstract
by using cloud words plot (Fig. 2a, b), a bubble chart of the
publications among the years and journals (Figs. 3 and 4), and
the top 10 journal-published tools/databases (Fig. 5).
Fig. 4 The tools bubble plot among all journals and numbers of each publication by year
4 Tools
To understand ncRNAs (noncoding RNA), its derivative types, and

the current increase of sequenced data, exploration for understand-
ing such a class of sequences and their transcriptional interactions
has become an interesting universe to be exploited in any organisms
[14–16]. This knowledge has boosted the development of strate-
gies and tools to detect miRNA sequences, the most studied type of
noncoding RNA [17, 18]. Applying a set of those techniques also
improves the ability to relate the interaction between miRNAs; it
also allows researchers to build sources of data for a wide analysis
[19, 20]. In addition, the decreasing cost of sequencing has sup-
ported the appearance of models for the large-scale analysis
[21, 22] of data sequencing originated by NGS (next-generation
sequencing) technologies.
An exploratory research for identification of miRNA detection
tools was performed at the NCBI. A total of 67 tools are listed in
Table 2: many tools are miRNA related and perform a variety of
functions.
There are tools for the identification of miRNAs Precursors,
Mature, Interaction, for Target Functional Analysis, miRNA asso-
ciation, and others. Sixty-seven tools were reported in Table 2,
related to species, year tool was released, platform, and link access.
Claiming to be responsible by controlling and regulation of
gene expression within cells, we divided miRNAs nine different
ways (Table 3):
Fig. 5 The top 10 journals that most published databases (a) and tools (b)
(a) Interaction: miRNAs that may be related to mRNAs or clus-

ters of them.
(b) Mature: A short sequence mostly with ~22 nucleotides (nt), it
is ready to inhibit the target (mRNA) by interaction.
Table 2
138
Tools to identify miRNA and its covariance related to functional role, species, and platform available
First Pubmed
Name publication Organism Available Plataform Access/download ID
C-mii 2012 Plants Yes Standalone http://www.biotec.or.th/isl/c-mii 23281648
CMTCN 2018 Homo sapiens Yes Web http://www.cbportal.org/CMTCN 30473937
ComiR 2013 Miscellaneous Yes Standalone http://www.benoslab.pitt.edu/comir/ 23703208
comTAR 2014 Animals and plants Yes Web http://rnabiology.ibr-conicet.gov.ar/comtar 24632500
DIANA-microT 2011 Miscellaneous Yes Web http://www.microrna.gr/webServer 21551220
Infinity 2016 Homo sapiens Yes Both https://github.com/bio-devel/infinity 27082112
Tharcı́sio Soares de Amorim et al.
isomiR-SEA 2016 Homo sapiens Yes Standalone http://eda.polito.it/isomir-sea/ 27036505

MAGIA 2010 Homo sapiens No Web http://gencomp.bio.unipd.it/magia 20484379
miARma-Seq 2016 Homo sapiens Yes Standalone https://sourceforge.net/projects/miarma/ 27167008
MicroInspector 2005 Miscellaneous No Web http://www.imbb.forth.gr/en/ 15980566
microinspector
MicroRNA 2017 Homo sapiens Yes Standalone https://www.ncbi.nlm.nih.gov/pmc/ 29657285
MultiTool articles/PMC5832007/bin/ncrna-03-
00013-s001.rar
MicroSNiPer 2010 Homo sapiens No Web http://cbdb.nimh.nih.gov/microsniper 20809528
Mi-DISCOVERER 2010 Homo sapiens No Standalone – 21364831
mirAct 2011 Homo sapiens No Web http://sysbio.ustc.edu.cn/software/mirAct 21596785
miRanalyzer 2011 Homo sapiens, Rattus norvegicus, No Web https://bioinfo2.ugr.es/ceUGR/ 21515631
Caenorhabditis elegans miranalyzer/
MiRAuto 2013 Plants No Standalone http://nature.snu.ac.kr/software/miRAuto. 23625170
htm
miRClassify 2014 Miscellaneous No Web http://datamining.xmu.edu.cn/software/ 24480175
MIR/home.html
miRDeep 2013 Homo sapiens Yes Standalone http://www.australianprostatecentre.org/ 23221645
research/software/mirdeep-star
miRDeep2 2012 Animal Yes Standalone https://www.mdc-berlin.de/content/ 21911355
mirdeep2-documentation
miRDeepFinder 2012 Arabidopsis thaliana No Standalone http://www.leonxie.com/DeepFinder.php 22290409
miRDeep-P2 2018 Plants Yes Standalone https://sourceforge.net/projects/mirdp2/ 30521000
miRDis 2018 Animals and plants Yes Web http://sbbi.unl.edu/miRDis/index.php 28073746
miREval 2.0 2013 31 species Yes Web http://mimirna.centenary.org.au/mireval/ 24048357
MiRFinder 2007 Animal Yes Standalone http://www.bioinformatics.org/mirfinder/ 17868480
miRiaD 2016 Homo sapiens Yes Web http://biotm.cis.udel.edu/miRiaD 27216254
Mirinho 2015 Animals and plants Yes Standalone http://mirinho.gforge.inria.fr/ 26022464
miR-isomiRExp 2016 Homo sapiens No Web http://mirisomirexp.aliapp.com 27009551
miRLocator 2019 Miscellaneous Yes Standalone https://github.com/cma2015/miRLocator 30701493
miRmapper 2018 Homo sapiens Yes Standalone https://github.com/MUSC-CGM/ 30223528
miRmapper
miRMOD 2015 Homo sapiens Yes Standalone http://bioinfo.icgeb.res.in/miRMOD 26623179
MiRNA-BD 2018 Homo sapiens Yes Standalone http://sysbio.suda.edu.cn/MiRNA-BD/ 30081733
miRNACancerMAP 2018 Homo sapiens Yes Web http://cis.hku.hk/miRNACancerMAP 29897412
miRNAfe 2015 Homo sapiens, Arabidopsis thaliana Yes Web http://fich.unl.edu.ar/sinc/web-demo/ 26499212
mirnafe-full/
miRNAFold 2016 Miscellaneous Yes Both https://evryrna.ibisc.univ-evry.fr/ 27242364
MicroRNA DBs and Tools
miRNAFold
miRNAkey 2010 Miscellaneous Yes Standalone http://ibis.tau.ac.il/miRNAkey/ 20801911
139
(continued)
140
Table 2
(continued)
First Pubmed
Name publication Organism Available Plataform Access/download ID
miRnalyze 2017 Homo sapiens Yes Web http://www.mirnalyze.in 28365733
miRNAminer 2008 Homo sapiens Yes Web http://pag.csail.mit.edu/mirnaminer 18215311
miRNAmotif-A 2018 Animals and plants Yes Standalone http://www.github.com/martynaut/ 30562930
mirnamotif
miRNAsong 2016 219 species Yes Web http://www.med.muni.cz/histology/ 27857164
miRNAsong/
MiRPara 2011 Metazoa, Viridiplantae, viruses Yes Standalone http://www.whiov.ac.cn/bioinformatics/ 21504621

mirpara
miRPlant 2014 Plants Yes Standalone http://www.australianprostatecentre.org/ 25117656
research/software/mirplant
MirPlex 2013 Miscellaneous Yes Standalone https://sourceforge.net/projects/mirplex/ 23184675
miRpower 2016 Homo sapiens Yes Web www.kmplot.com/mirpower 27744485
miR-PREFeR 2014 Plants Yes Standalone https://github.com/hangelwen/miR- 24930140
PREFeR
miRQuest 2016 30 species Yes Web http://mirquest. 27050998
integrativebioinformatics.me
MiRror 2010 Homo sapiens, Mus musculus, Yes Web http://www.proto.cs.huji.ac.il/mirror/ 20529892
Drosophila melanogaster
miRspring 2013 Homo sapiens Yes Standalone http://mirspring.victorchang.edu.au 23775795
miRTarVis+ 2015 Miscellaneous Yes Web http://hcil.snu.ac.kr/research/mirtarvisplus 28600227
miRTour 2011 Plants No Web http://bio2server.bioinfo.uni-plovdiv.bg/ 21887016
miRTour/
miRVaS 2016 Homo sapiens Yes Both http://mirvas.bioinf.be/ 26384425
miTRATA 2016 Animals and plants No Web https://wasabi.dbi.udel.edu/apps/ta/ 26454275
mrSNP 2014 Homo sapiens No Standalone http://mrsnp.osu.edu/ 24629096
MTide 2015 Arabidopsis thaliana, Oryza sativa, Yes Standalone http://bis.zju.edu.cn/MTide/ 25256573
Zea mays, Glycine max, Sorghum
bicolor
myMIR 2011 Homo sapiens, Mus musculus No Standalone https://www.itb.cnr.it/micro/ 22021901
PGS 2014 Homo sapiens Yes Standalone https://github.com/feizhe/PGS 24947752
PROGmiR 2012 Homo sapiens Yes Web http://www.compbio.iupui.edu/progmir 23272654
SimiRa 2015 Homo sapiens Yes Web http://vsicb-simira.helmholtz-muenchen.de 26383775
SoMART 2012 Plants No Web http://somart.ist.berkeley.edu 22268718
sPARTA 2014 Plants No Web https://mpss.udel.edu/tools/mirna_apps/ 25120269
comPARE.php
STarMir 2014 Human, mouse, worm Yes Both http://sfold.wadsworth.org/starmir.html 24803672
StarSeeker 2018 Animals and plants Yes Standalone https://github.com/pnatsi/StarSeeker 29946534
SylArray 2010 Miscellaneous Yes Web http://www.ebi.ac.uk/enright/sylarray 20871108
TAM 2.0 2018 Homo sapiens Yes Web http://www.lirmed.com/tam2/ 29878154
TAPIR 2010 Plants Yes Web http://bioinformatics.psb.ugent.be/ 20430753
webtools/tapir
Target-align 2010 Animals and plants No Both http://www.leonxie.com/targetAlign.php 20934992
TarHunter 2018 Plants Yes Web http://tarhunter.genetics.ac.cn/ 29236948
TFmiR 2015 Homo sapiens Yes Web http://service.bioinformatik.uni-saarland. 25943543

de/tfmir
141
Table 3
Tools related to their function of identification for miRNAs in sequences or genomes
Type Tool
Interaction miRNAmotif-A, SimiRa
Mature StarSeeker, miRNAFold, miRQuest
Mature, precursor miRLocator, miRClassify
Mature, precursor, target DIANA-microT
functional analysis
Network CMTCN, PROGmiR, TFmiR
Precursor MicroRNA MultiTool, Infinity, miRVaS, miTRATA, miRNAfe,
miRMOD, miR-PREFeR, miRPlant, miREval 2.0, miRspring,
MiRAuto, MirPlex, miRDeep, miRDeep-P2, MiRPara, mi-DISCOV
ERER, miRNAkey, miRanalyzer, miRNAminer, MiRFinder,
miRDeep
Precursor, target gene miARma-Seq, Mtide, miRDeepFinder, C-mii, miRTour, MiRror,
MAGIA, miRnalyze
IsomiRs miRDis, isomiR-SEA, miR-isomiRExp
Target functional analysis TAM 2.0, mirAct, SylArray
Target gene miRTarVis+, sPARTA, comTAR, ComiR, myMIR, Target-align,
MicroInspector, miRNACancerMAP
Target gene (biomarker) miRpower
Target gene, network miRmapper, MiRNA-BD
Target gene, ceRNA TarHunter
CeRNA miRNAsong
MiRNA association miRiaD, PGS, mrSNP, MicroSNiPer, STarMir, SoMART, TAPIR,
Mirinho
(c) Precursor: It is the result from the pi-miRNA cleavage used in

the miRNA gene prediction.
(d) IsomiRs: It refers to sequences that have some variation with
respect to the references of the miRNA sequence.
(e) Target Functional Analysis: Tools that contribute to an explor-
atory analysis of miRNA functional data and regulation.
(f) Network: To build an illustration of networking inference
linking miRNA to putative mRNA targets.
(g) Target Gene: mRNA or set of target mRNAs that are regu-
lated by miRNA.
(h) CeRNA (competing endogenous RNA): Sequences that reg-
ulate other RNAs; compete for miRNA activity.
(i) miRNA Association: miRNAs that are associated to others

functions, sequences, or even sequences that potentially regu-
late genes of interest; for example, diseases.
Although the detection of miRNAs by mathematical models
and different strategies is important, it is necessary to identify the
targets [23, 24], network of action, and related sequences [25–27],
which may be beyond our understanding yet, once they have the
capacity to influence the biogenesis of the cell.
Predictive tools of miRNAs are methodologies developed for
the purpose of identification in various forms of data: complete
genomes, incomplete genomes, or data provided by RNA-seq.
4.1 Interaction Taking into account that miRNAs are sequences that have as role to
bind other target sequences, the identification made by protein
complexes of RNA is a fundamental point to improve the under-
standing of how this posttranscriptional regulation happens.
miRNAmotif-A [28] software makes use of the technique to iden-
tify and recognize motifs in several RNA-seq data; although the
motif relationship supports the understanding of the interaction of
miRNAs and mRNAs, it is important to note that other types of
noncoding RNAs can benefit from the use of this strategy.
The SimiRa [29] tool does not use unique strategies for identi-
fication between miRNAs and mRNAs: there is a combination of
approaches that focus on guaranteeing the result of the analysis. In
addition to the contextual and protein sequence analysis, a compar-
ison is made with nuclear and cytoplasmic interactions that are not
explained by binding targets in mRNAs. There is a set of validations
with KEGG pathways and GO-terms for the functional categoriza-
tion of the biological context between mRNAs and RNA-binding
protein gene target sets.
4.2 Mature Applied research and studies in computational biology are relatively
new: tools that support the understanding of biological mechan-
isms are being widely used. The StarSeeker [30] software has
strategies and an automated pipeline for the duplex identification
of miRNAs based on secondary structures modeled by the precur-
sor sequence. It requires as input two FASTA files, one for precur-
sor sequences and one for sequences containing mature sequences;
the tool searches for each mature miRNA sequence combinations
within the sequence data set precursor sequences.
In contrast, miRNAFold [31] uses mechanisms to identify
hairpin structures to predict pre-miRNAs in the genomes. For a
knowledge base, the tool uses miRbase for different observations of
pre-miRNAs to modeling a set of requested criteria. By identifying
their hairpins and a set of constraints, the candidates for
pre-miRNAs are selected.
For identification of miRNAs, miRQuest [32] is shown to be

easy to use even for researchers with no bioinformation background
on specific aspects. The search for miRNAs is based on two main
functions: (a) integration of tools for the prediction of miRNAs in
an interactive environment; and (b) comparison of results of these
prediction methods. For prediction, the software requires an orga-
nization of the sequences in FASTA file standards for such analysis
and comparison.
4.3 Precursor A large collection and several strategies of precursor miRNA pre-
diction tools are highlighted in the survey we have done. Among a
greater number of tools, predicting sequences that give rise to
mature miRNAs [33, 34] are very valuable, given their capacity
for expression and transcriptional regulation, even for the different
mechanisms by which animal and plant miRNAs are related and
involved. Tools such as Infinity [35] describe the ability to search
for patterns that are upstream or downstream in genomic sequences
which are strongly represented in human miRNAs. In addition, the
CCAAT pattern is strongly represented in promoters of genes
encoding proteins that are linked to deregulation in colorectal
cancer, and deregulation of miRNA (noncoding genes) expression
has been found in colorectal cancer.
Although some tools use well-established characteristics that
help in the identification of miRNAs, other methods have proved to
be very powerful; strategies that implement artificial intelligence
[36–38] techniques in the prediction of miRNAs show allied bio-
informatics. miRPlant [39] is a tool that uses 16 data sets of
miRNAs reported in four different species, in addition to not
relying on any third-party software, to develop strategies to identify
regions of excisions and hairpin structures to filter putative
sequences of miRNAs in plants.
4.4 IsomiRs Nowadays the exploration of understanding miRNAs is extensively

studied, deep sequencing strategies that opened up other possible
variations and possibilities to ascertain relationships in different
types of sequences, as well as different forms of gene regulation
and suppression. The tool miRDis [40] claims to be a web-based
software that promotes data analysis for endogenous miRNAs or
exogenous in data from RNA-seq. Following three strategies
defined in this tool, it is able to identify miRNAs with high sensi-
tivity and quantify the expression, and also identifies form isomiRs
and nonhost miRNAs.
Although a collection of software has the ability to identify
variants in target-related isomiRs/miRNAs [41], the isomiR-SEA
[42] tool explores RNA-seq information with expression analyses
of miRNAs, isomiRs, and the interaction of regions of miRNAs-
mRNAs.
4.5 Target Functional The increasing use of high-throughput sequencing strategies has
Analysis increased the need for tools for systemic analyses that can extract
information [15]. The use of tools developed exclusively to treat
the types of regulatory networks and to investigate potential
miRNA activities [43] in these data is stimulated by interest in
predicting possible gene targets in that particular set of data.
Although some of these tools use strategies such as verification
of signatures for different types of ncRNAs [21], other software
offers techniques for identifying miRNA activities by mining such
rich gene expression data [44].
4.6 Network Even though miRNAs [45] may be linked to transcriptional, post-
transcriptional, and complex biological processes, they are also
associated with different types of diseases involving the regulation
of gene expression, transcription factors, and in the miRNAs them-
selves. Tools developed for the exploration [26] of data in cancers
in humans are of extreme importance, because correlating networks
of gene expression involved in diseases can give better understand-
ing about its evolution and possible treatments.
The CMTCN [46] tool uses evidence of the relationship
between miRNAs and transcription factors (TFs). With the aid of
motif analysis, combined with miRNAs and TFs, there is a need for
shared target gene regulation. A collection of information on TF–
genes, TF–miRNA, miRNA–gene, and miRNA–TF includes
another collection of data of gene–miRNA relationships. These
approaches can relate and create systematic networks by combining
all these data inventories. Thus, with the aid of motif analysis there
is a paired combination of the regulatory networks between these
sequences.
4.7 Target Gene Studies for the exploration and correction of miRNAs-mRNAs
[27, 47] gained prominence from studies attesting to a certain
association [48, 49] because this is an underexploited universe
with very high evidence of interactions with other sequences, dis-
eases, and types of cellular regulation in general [17, 50]. Moreover,
quantifying and characterizing levels of expressions of a given set of
data [51], in addition to possible changes or mutations [19, 52], is
essential to outline strategies and promote understanding of these
mechanisms.
Among tools that aim at the construction and inference of
regulatory networks for cancers involving miRNAs, an example is
miRNACancerMAP [25]. This tool defines a pipeline that corre-
lates sequence-based combinations, in addition to analyzing the
specific context of noncorrelation of gene expression. Thus, it
provides an identification that correlates miRNA with target
mRNAs.
Strategies to approach the relationship of miRNAs with the
cellular environment is necessary to promote the understanding
of several mechanisms. However, because using strategies and fos-

tering data in an attractive and easy-to-interact fashion is important,
there are databases that from an analysis provide web tools to
consult possible evidence using biomarkers [53, 54].
4.8 CeRNA Several types of ncRNAs are correlated in the literature, their
mechanisms differing from organism to organism. The main ele-
ments responsible for the system of expression and gene suppres-
sion, in light of this, are data reporting that another type of ncRNA
can exist and can regulate the miRNAs themselves. The competing
endogenous RNAs (ceRNAs) or spontaneous RNAs are targeted as
suppression of miRNAs, which in turn regulate mRNAs.
The miRNAsong tool [55] is shown with the strategy of gen-
erating and testing data on sponge miRNAs. With a wide coverage
of 219 species and approximately 35,000 records of miRNA
sequences, the tool aims at generation and testing to search for
ceRNAs that inhibit certain types of miRNAs. Basically, it makes use
of techniques used in vivo for the in silico modeling of the isolation
to predict a set of these putative ceRNAs.
4.9 miRNA Several types of relationships may involve miRNAs, as well as the
Association possibility of inhibiting mRNAs. Other associations may be asso-
ciated with these regulatory sequences. A set of strategies are used
to circumvent this problem: artificial intelligence techniques
become well applied mainly by the abstraction of characteristics of
these sequences.
Mirinho [22] is a tool that identifies and predicts miRNAs for
small-scale data set (sRNA-Seq–short RNA-seq) and genomic
scales. The model is useful for animal genomes and plants: for its
analysis the strategy uses as a basis the length of the loops, stem
arms, and the free energy of the pre-miRNA hairpin.
Because other approaches provide identification for miRNAs
and implement the ability to analyze gene targets [56, 57] and
mimic targets [23], different approaches contribute to accuracy
and support to create methodologies that can be extended to any
other types of sequences.
5 Databases
For miRNA database search in the literature, we used the

Non-coding RNA Databases Resource (NRDR) [1, 4] as our
main resource. To give a better support to the search, we also
used Tools4Mirs [11] as a secondary resource.
The NRDR repository was first published in 2012 and it pro-
vides a complete list and description for biological databases: it is
also very user friendly and allows filtered search. We found
110 miRNA-related databases on NRDR. The Tools4mirs is a
Table 4
List of databases
First Pubmed
Name publication Organism Website Download Available ID
Antagomirbase 2011 Unspecified http://bioinfopresidencycollegekolkata.edu.in/ No No 21904438
antagomirs.html
ARN 2016 Unspecified http://210.27.80.93/arn/ No Yes 27503118
AtmiRNET 2015 Arabidopsis thaliana http://atmirnet.itps.ncku.edu.tw/ No Yes 25972521
BioM2MetDisease 2017 Animal http://www.bio-bigdata.com/BioM2MetDisease/ Yes Yes 28605773
CCGD 2010 Homo sapiens http://crdd.osdd.net/ No Yes 21045064
ceRDB 2014 Homo sapiens http://www.oncomir.umn.edu/cefinder/ No Yes 23055620
CircuitsDB 2010 Homo sapiens http://biocluster.di.unito.it/circuits/ Yes Yes 20731828
CoGemiR 2008 Homo sapiens, Rattus http://cogemir.tigem.it/ Yes Yes 18837977
norvegicus
ComiRNet 2015 Homo sapiens http://comirnet.di.uniba.it:8080/ No Yes 26051695
comTAR 2015 Plants http://rnabiology.ibr-conicet.gov.ar/comtar/ No Yes 24632500
CREAM 2014 Homo sapiens www.jianglab.cn/CREAM/ Yes Yes 25978092
CSmiRTar 2017 Homo sapiens http://cosbi.ee.ncku.edu.tw/CSmiRTar/ No Yes 28704505
dbDEMC 2010 Homo sapiens http://159.226.118.44/dbDEMC/ Yes No 21143814
dbSMR 2009 Homo sapiens http://miracle.igib.res.in/polyreg/ No No 19371411
DIANA miRPath 2009 Homo sapiens, others http://snf-515788.vm.okeanos.grnet.gr/ No Yes 22649059
DMD 2015 Miscellaneous http://sbbi-panda.unl.edu:5000/dmd/ No Yes 26030752
doRiNA 2011 Miscellaneous https://dorina.mdc-berlin.de/go No Yes 22086949
dPORE 2011 Homo sapiens http://apps.sanbi.ac.za/dpore/ No No 21326606

EpimiR 2014 Miscellaneous http://210.46.85.180:8080/EpimiR/index.jsp Yes Yes 24682734
147
(continued)
148
Table 4
(continued)
First Pubmed
EpimiRBase 2016 Unspecified https://www.epimirbase.eu/ Yes Yes 26748106
ExcellmiRDB 2015 Homo sapiens http://www.excellmirdb.brfjaisalmer.com/ No Yes 25562198
ExprTargetDB 2010 Homo sapiens http://www.scandb.org/apps/microrna/ Yes Yes 20975837
HMDD 2015 Homo sapiens http://202.38.126.151/hmdd/mirna/md/ No No 18923704
HNOCDB 2012 Homo sapiens http://gyanxet.com/hno.html No Yes 22024348
HOCTARdb 2011 Homo sapiens http://hoctar.tigem.it/ No Yes 21435384
IGDB.NSCLC 2011 Homo sapiens http://igdb.nsclc.ibms.sinica.edu.tw/ No Yes 22139933

Immune-miR 2017 Homo sapiens http://biominingbu.org/immunemir/index.html No Yes 28124611
IntmiR 2011 Homo sapiens, others http://rgcb.res.in/intmir/ No No 21423893
Isomirs 2012 Homo sapiens, others http://hood.systemsbiology.net/cgi-bin/isomir/ No Yes 20876832
find.pl
MDTE DB 2016 Miscellaneous http://bioinf.njnu.edu.cn/MDTE/MDTE.php Yes Yes 28055900
mESAdb 2010 Homo sapiens, others http://konulab.fen.bilkent.edu.tr/mirna/ No Yes 21177657
MicroPC 2009 Miscellaneous http://www3a.biotec.or.th/micropc/ No Yes 19660144
microPIR 2012 Homo sapiens http://www4a.biotec.or.th/micropir Yes Yes 22439011
microRNA.org 2008 Miscellaneous http://www.microrna.org/ No No 18158296
microRNAviewer 2011 Miscellaneous http://people.csail.mit.edu/akiezun/ No Yes 22330228
microRNAviewer
microTranspoGene 2008 Miscellaneous http://transpogene.tau.ac.il/microTranspoGene. Yes* Yes 17986453
html
MicroTrout 2016 Oncorhynchus mykiss http://www.mennigen-lab.com/microtrout.html No Yes 27494513
mimiRNA 2010 Homo sapiens http://mimirna.centenary.org.au/ No Yes 19933167
miR-EdiTar 2012 Homo sapiens http://microrna.osumc.edu/mireditar No Yes 23044546
Mir@NT@N 2010 Miscellaneous http://mironton.uni.lu/ No No 21375730
miR2Disease 2008 Homo sapiens http://www.mir2disease.org/ No No 18927107
miRandb 2017 Unspecified http://mirandb.ir/ No Yes 28049134
miRandola 2012 Homo sapiens http://atlas.dmi.unict.it/mirandola/index.html No No 23094086
miRBase 2007 Miscellaneous http://www.mirbase.org/ Yes Yes 14681370
miRBase tracker 2014 Miscellaneous http://www.mirbasetracker.org/ No Yes 25157074
miRCancer 2013 Homo sapiens http://mircancer.ecu.edu/ No No 23325619
mirCoX 2013 Homo sapiens http://210.212.254.116/mircox/pages/index.php No Yes 24267917
miRDB 2014 Miscellaneous http://www.mirdb.org/ Yes Yes 18426918
miRDeathDB 2012 Miscellaneous http://rna-world.org/mirdeathdb/ No No 22743998
mirDIP 2017 Miscellaneous http://ophid.utoronto.ca/mirDIP/ Yes Yes 21364759
miRdSNP 2011 Homo sapiens http://mirdsnp.ccr.buffalo.edu/ Yes Yes 22276777
miRecords 2008 Miscellaneous http://mirecords.biolead.org/ No Yes 18996891
miReg 2016 Homo sapiens http://www.iioab-mireg.webs.com/ No Yes 20693604
miREnvironment 2011 Miscellaneous http://cmbi.bjmu.edu.cn/miren No No 21984757
mirEX 2011 Arabidopsis thaliana http://bioinfo.amu.edu.pl/mirex Yes No 22013167
miRFANs 2011 Arabidopsis http://www.cassava-genome.cn/mirfans No No 22583976
thaliana, others
miRGate 2015 Homo sapiens, others http://mirgate.bioinfo.cnio.es/miRGate/ No Yes 25858286

miRGator 2007 Homo sapiens http://mirgator.kobic.re.kr/ No Yes 17942429
149
(continued)
Table 4
150
(continued)
First Pubmed
mirGen 2006 Homo sapiens, Mus http://www.microrna.gr/mirgen/ Yes Yes 17108354
musculus
MirGeneDB.org 2014 Homo sapiens, others http://mirgenedb.org/ Yes Yes 26473382
miRHrt 2010 Mus musculus, others http://sysbio.suda.edu.cn/mirhrt/ Yes No 20716610
miRmine 2017 Homo sapiens http://guanlab.ccmb.med.umich.edu/mirmine/ No Yes 28108447
miRNA 2009 Vitis vinifera http://www.itb.cnr.it/ptp/grapemirna/ No Yes 19563653
miRNA - target gene 2005 Drosophila http://www.russelllab.org/mirnas/ Yes No 15723116

prediction at EMBL
miRNA body map 2011 Homo sapiens, others http://www.mirnabodymap.org/ No No 21835775
miRNA SNiPer 2011 Homo sapiens, others http://www.integratomics-time.com/miRNA- No No 22303335
SNiPer
miRNA_Targets 2012 Miscellaneous http://mamsap.it.deakin.edu.au/~amitkuma/mirna_ No No 22940442
targetsnew/find_mirna.html
miRnalyze 2017 Homo sapiens http://www.mirnalyze.in/ No Yes 28365733
miRNAMap 2006 Homo sapiens, others http://mirnamap.mbc.nctu.edu.tw/ Yes No 16381831
miRNApath 2007 Homo sapiens, others http://lgmb.fmrp.usp.br/mirnapath/ No Yes 18058708
miRNASNP 2015 Unspecified http://bioinfo.life.hust.edu.cn/miRNASNP2/ Yes Yes 25877638
miRNEST 2011 Animal, plant, virus http://rhesus.amu.edu.pl/mirnest/copy/ Yes Yes 24243848
miRNeye 2010 Mus musculus http://mirneye.tigem.it/ No Yes 21171988
miRò 2009 Homo sapiens http://ferrolab.dmi.unict.it/miro/ No No 20157481
miROrtho 2008 Miscellaneous http://cegg.unige.ch/mirortho/ No Yes 18927110
miRPathDB 2016 Homo sapiens, Mus https://mpd.bioinf.uni-sb.de/overview.html Yes Yes 27742822
musculus
mirPub 2014 Miscellaneous http://diana.imis.athena-innovation.gr/ No Yes 25527833
DianaTools/index.php?r¼mirpub
miRSel 2010 Homo sapiens, others http://services.bio.ifi.lmu.de/mirsel/ No Yes 20233441
MirSNP 2012 Homo sapiens http://bioinfo.bjmu.edu.cn/mirsnp/search/ Yes Yes 23173617
miRSponge 2015 Miscellaneous http://www.bio-bigdata.com/miRSponge/ Yes Yes 26424084
miRStress 2013 Homo sapiens, others https://sourceforge.net/projects/mirstress/ Yes Yes 24244721
miRSystem 2012 Homo sapiens, Mus http://mirsystem.cgm.ntu.edu.tw/ No Yes 22870325
musculus
miRTarBase 2010 Miscellaneous http://mirtarbase.mbc.nctu.edu.tw/ Yes Yes 21071411
mirtronDB 2019 Plants http://mirtrondb.cp.utfpr.edu.br/ Yes Yes 30874795
miRvar 2011 Homo sapiens http://genome.igib.res.in/mirlovd No Yes 21618345
miRWalk 2011 Homo sapiens, others http://www.ma.uni-heidelberg.de/apps/zmf/ No Yes 16381827
mirwalk/
miSolRNA 2010 Solanum http://www.misolrna.org/ No No 21059227
lycopersicum
miTALOS 2011 Homo sapiens, Mus http://mips.helmholtz-muenchen.de/mitalos/ No Yes 21441347
musculus
MtiBase 2015 Homo sapiens, Mus http://mtibase.sysu.edu.cn./ No No 26490638
musculus
multiMiR 2014 Homo sapiens, Mus http://multimir.ucdenver.edu/ Yes Yes 25063298
musculus
NCG 2016 Homo sapiens http://ncg.kcl.ac.uk/index.php Yes Yes 26516186
NetAge 2010 Miscellaneous http://netage-project.org/ No No 20186480

OCDB 2015 Homo sapiens http://ocdb.di.univr.it/ No No 26228432
151
(continued)
152
Table 4
(continued)
First Pubmed
OncomiRDB 2014 Homo sapiens, Mus http://tdb.ccmb.res.in/OncomiRdbB/index.htm No Yes 24651967
musculus
PAmiRDB 2019 Plants, virus http://bioinfo.icgeb.res.in/pamirdb No Yes 30874591
PASmiR 2013 Plants http://pcsb.ahau.edu.cn:8080/PASmiR/ Yes Yes 23448274
Patrocles 2009 Miscellaneous http://www.patrocles.org/ No Yes 19906729
Pharmaco-miR 2013 Unspecified http://www.pharmaco-mir.org/ Yes Yes 23376192
PhenomiR 2009 Homo sapiens http://mips.helmholtz-muenchen.de/phenomir/ Yes Yes 20089154

PlanTE-MIR DB 2016 Plants http://bioinfo-tool.cp.utfpr.edu.br/plantemirdb/ Yes Yes 26887375
PmiRExAt 2016 Arabidopsis thaliana, http://pmirexat.nabi.res.in Yes Yes 27081157
others
PmiRKB 2010 Arabidopsis thaliana, http://bis.zju.edu.cn/pmirkb/ No Yes 20719744
others
PMRD 2009 Plants http://bioinformatics.cau.edu.cn/PMRD/ Yes Yes 19808935
PMTED 2013 Plants http://pmted.agrinome.org/ No No 23725466
PNRD 2014 Plants http://structuralbiology.cau.edu.cn/PNRD/ Yes No 25398903
PolymiRTS 2006 Homo sapiens, Mus http://compbio.uthsc.edu/miRSNP/ Yes Yes 17099235
musculus
Psmir 2016 Homo sapiens http://www.bio-bigdata.com/Psmir/index.jsp No No 26759061
PuTmiR 2010 Homo sapiens http://www.isical.ac.in/~bioinfo_miu/TF-miRNA. Yes Yes 20398296
php
RCGD 2012 Homo sapiens http://www.juit.ac.in/attachments/jsr/rcdb/ No Yes 22608002
homenew.html
RepTar 2010 Miscellaneous http://bioinformatics.ekmd.huji.ac.il/reptar/ Yes No 21149264
RiceATM 2016 Oryza sativa http://syslab3.nchu.edu.tw/rice/ No No 28025342
S-MED 2010 Homo sapiens http://www.oncomir.umn.edu/SMED/ No Yes 20212452
SM2miR 2012 Unspecified http://210.46.85.180:8080/sm2mir/index.jsp Yes Yes 23220571
smiRNAdb 2007 Miscellaneous http://www.mirz.unibas.ch/cloningprofiles/ No No 17604727
SomamiR 2012 Homo sapiens http://compbio.uthsc.edu/SomamiR/ Yes Yes 26578591
SorghumFDB 2016 Sorghum bicolor http://structuralbiology.cau.edu.cn/sorghum/ No No 27352859
index.html
Ssa miRNAs DB 2012 Salmo salar http://www.molgenv.com/ssa_mirnas_db_home. No No 22493538
php
starBase 2010 Miscellaneous http://starbase.sysu.edu.cn/ Yes No 21037263
TarBase 2015 Miscellaneous http://diana.cslab.ece.ntua.gr/tarbase/ Yes Yes 16373484
targetHub 2013 Homo sapiens http://app1.bioinformatics.mdanderson.org/ No Yes 24013925
tarhub/_design/basic/index.html
TargetScanS 2005 Miscellaneous http://genes.mit.edu/tscan/targetscanS2005.html No Yes 15652477
TMREC 2015 Homo sapiens http://210.46.85.180:8080/TMREC/ Yes Yes 25932650
TransmiR 2009 Miscellaneous http://202.38.126.151/hmdd/mirna/tf/ Yes No 19786497
UCbase & miRfunc 2008 Homo sapiens, others http://microrna.osu.edu/.UCbase4/ Yes No 18945703
vHoT 2011 Homo sapiens, others http://best.snu.ac.kr/vhot/ No Yes 22160653
Vir-Mir db 2007 Viruses http://alk.ibms.sinica.edu.tw/cgi-bin/miRNA/ Yes Yes 17702763
miRNA.cgi
VIRmiRNA 2014 Viruses http://crdd.osdd.net/servers/virmirna/ Yes Yes 25380780
ViTa 2006 Homo sapiens, others http://vita.mbc.nctu.edu.tw/ No Yes 17148483

YM500 2012 Homo sapiens, Mus http://ngs.ym.edu.tw/ym500/ Yes Yes 23203880
musculus
153
ZooMir 2010 Miscellaneous http://insr.ibms.sinica.edu.tw/ZooMir/ No Yes 20347954

manually curated repository specifically for miRNAs tools and

databases. We found 68 miRNAs databases on Tools4miRs. With
the two database lists, we eliminated the redundancies and accrued
a final list of 130 databases (Table 4). Finally, we classified the
databases in seven categories: regulation network, disease related,
miRNA-target, pathway analysis, single nucleotide polymorphisms
(SNPs), functional annotation, and specifics. Table 5 shows a data-
base list manually grouped by these categories.
First of all, we give a special highlight to the miRbase [60]
repository. Created in 2006, miRBase was the first online reposi-
tory for miRNAs sequence data and annotation. Since its release, it
has been updated periodically, most recently in 2018. The database
provides an online browse by species or miRNA, and download of
sequences in FASTA format.
5.1 Regulation Assuming that miRNAs do not work alone, some databases aim to
Network Databases construct miRNA regulation networks. In this field, we can high-
light AtmirNET [61], which is a database specific for Arabidopsis
thaliana. The AtmirNET database uses information about tran-
scriptional factors (TFs), miRNA–target interactions, and transcrip-
tional start sites (TSSs) to infer the regulatory networks. The
database provides a search by miRNA name, transcription factor,
or target gene.
Still in the field of networks, miRDeathDB [62] serves to
analyze and provide information about miRNAs that act in a cell
death network. The miRDeathDB also provides five model species
with 2010 associations of miRNAs and programmed cell death.
The ComiRNet [63] database takes advantage of a gap in the
literature, that is the lack of information about functional targeting;
by that, it combines computational approaches such as semi-
supervised ensemble-based classifier and bi-clustering algorithm,
to build a repository about the role of miRNAs in biological
processes.
For the last, we highlight the CircuitsDB [64] database, which
aims to gather information about TFs to provide information about
regulatory networks for miRNAs and their targets. The database
also provides information about particular transcription factors for
mRNA and coding gene regulation.
5.2 Disease-Related Several forms of evidence suggest that miRNAs have crucial roles in
Databases disease regulation, diagnoses, and progression. In this scenario,
there are some manually curated disease-related database we can
highlight. First, we have OCDB [65], a database that studies the
chromosome candidate associated to obsessive compulsive disorder
(OCD), gathering information about genes, miRNAs, and drugs.
In the field of metabolic diseases, we have the BioM2MetDisease
[66] database, which is a manually curated database designed to
join information about biomolecules and drugs for metabolic
diseases.
Table 5
miRNA databases manually grouped by category
Type Database
Regulation ARN, AtmiRNET, CircuitsDB, NetAge, OCDB, Psmir
network
Disease BioM2Disease, CCGD, dbDEMC, EpimiRBase, HMDD, HNOCDB, IGDB.
NSCLC, Immune-miR, IntmiR, miR2Disease, NCG, RCGD, S-MED, SomamiR,
TMREC, OncomiRdbB
miRNA-target ceRDB, ComiRNet, comTAR, ExprTargetDB, CSmiRTar, HOCTARdb, MicroPC,
microPIR, mimiRNA, miR-EdiTar, MI@NT@N, miRDeathDB, TargetScanS,
doRiNA, mirDIP, miRecords, miRFANs, miRGate, targetHub, miRNAMap,
miRNA - target gene, miRNA_Targets, miRNEST, miRSel, mirRDB, miRSponge,
miRTarBase, miRWalk, miSolRNA, miTALOS, Patrocles, VirmiRNA PolymiRTS,
RepTar, Ssa miRNAs DB, MtiBase, TarBase, miRSystem vHoT, ViTa, mirCoX,
PMTED, microPIR, PAmiRDB
Pathway analysis DIANA miRPath, miRHrt, miRnalyze, miRNApath, miRPathDB
SNPS dPORE, miRdSNP, MiRSNP, miRNA SNiPer, u, dbSMR, miRNASNP, CREAM
Functional miRGator, miRNA Body Map, MiRFANS, SorghumFDB
annotation
Specifics Antagomirbase, miRStress, CoGemiR, doRiNA, EpimiR, DMD, ExellmiRDB,
Vir-Mir db, PNRD, Isomirs, MDTE DB, microRNAviewer, MicroTrout,
PmiRexAt, mirPub, miRandb, miReg, miREnvironment, miRBase Tracker,
MultiMir, PASmiR, mESAdb, mirEX, RiceATM, mirGen, miRmine, miRNeye,
miROrtho, miRNA, PlanTE-MIR DB, MirGeneDB.org, mirtronDB, Mirò,
microTranspoGene, DMD, SM2miR, miRvar, Pharmaco-miR, PmiRKB, PMRD,
Starbase PuTmiR, TransmiR, Ucbase & miRfunc, YM500, ZooMir
In the field of cancer-related databases, we highlight dbDEMc

[67], NCG [68], HNOCDB [69], and IGDB.NSCLC [70]. The
dbDEMc repository provides information on 14 types of human
cancer, curated from 48 microarray datasets available in publica-
tions. These databases use the significance analyses of microarray to
retrieve miRNAs that have different expression levels in cancers
when compared to normal tissues. The NCG database focuses on
building networks of cancer genes by curating genes from the
literature. The database stores more than 1500 cancer genes, of
which 518 are known cancer genes. The HNOCDB focuses on
head, neck, and oral cancers. The database tabulates the different
types of cancer separately under appropriate subtype names; it also
provides a chromosomal map their related genes and miRNAs. Last
in the cancer field, we highlight IGDB.NSCLC, which is a reposi-
tory of curated and integrated information of several datasets about
cell lung cancer that provides graphic visualization and tabular
search.
In the field of general disease, the MiR2Disease [71] repository

stores information about miRNA deregulation in human disease.
The authors searched on PubMed for miRNA disease-related
papers and manually curated more than 600 published works.
5.3 miRNA–Target It has been established in the literature that the miRNAs are able to
Databases regulate the expression of mRNAs/genes and that mRNAs can be
regulated by multiple miRNAs. When a miRNA regulates a tran-
script, we call this transcript a target for the miRNA. Many works
about miRNA–targets databases have been published in the litera-
ture, and some of them we highlight in this section.
First, we have the ceRDB [72] repository that uses the Tar-
getScan [73] software to predict miRNA–target interactions. The
authors used the matrix approach to explore the structure of the
dataset; they found more than 50,000 human miRNA–mRNA
interactions. ExprTargetDB [74] is a database of miRNA targets
and expression information gathered from an integration of
computational and experimental approaches that allows a search
by keyword or tag. Similar to ExprTargetDB, the ComTar [75]
database also provides information about miRNA–target pairs,
their function, and evolutionary information in plant genomes.
Also in the field of plants, the MicroPC [76] database runs a
large-scale expressed sequence tag analysis and provides miRNA–
target information in 125 species of plants.
Other important databases for miRNA–target interactions
include microRNA.org, microPIR, miRecords, and miRGate. For
more information, see the Supplementary Material.
5.4 Pathway A few databases focus on the study of biological pathways, which
Analysis are actions in molecules that may lead to products or changes in the
cell. The Diana miRPath [77], for example, is a web application that
performs an enrichment analysis of molecular pathways potentially
altered by the expression of single or multiple miRNAs. The miR-
nalyze [78] database does a similar job by identifying putative
regulation of cell signaling pathways by miRNAs. Further, the
miRPathDB [79] repository contains more than 2595 miRNAs,
reactome pathways, and other information. Last, the miTALOS
[80] database provides a dataset of gene regulation in biological
pathways.
5.5 SNP Databases Single nucleotide polymorphism (SNP) is a condition that affects
1% of the population. Recently, SNPs have been identified in non-
coding regions in genomes, which has increased the number of
databases in this field. In this section we highlight two of these
databases.
The miRdSNP [81] database provides manually curated infor-
mation of dSNPs on Homo sapiens. More than 630 unique dSNPs
and 204 diseases are available in the repository. The second,
miRNA SNiPer [82], is a database for genetic variations in miRNAs

of different vertebrates that provides information related to SNPs
located in seed regions of miRNAs.
5.6 Functional In this section, we highlight two functional annotation-focused

Annotation databases. The miRGator [83] database is a repository of functional
annotation that integrates publicly available expression data of
miRNA and target mRNA/proteins from different tissues, cell
lines, diseases, and other sources. Similar to this is the miRFANs
[84] database, which is the plant repository for functional annota-
tion that includes miRNA–target interactions, transcription factors,
expression profiles, genomic annotations, and pathways. For target
prediction, these use information from miRTarBase [85] and Tar-
Base [86]. Finally, we have the SorghumFDB [87], a database that
joins information about the sorghum gene family classifications,
gene annotations, miRNA and target gene information, and ortho-
logous pairs to construct functional annotation and a general geno-
mic data repository.
5.7 Specific Some databases focus on specific subjects, rather than transcript,
Databases analysis, or organism types. First, we have Antagomirbase [88],
which is a repository for antagomirs, a novel class of chemically
engineered oligonucleotides. Second is a repository of metazoan
evolutionarily conserved miRNAs called CoGemiR [89], which
provides information about miRNA host gene, genomic location,
and expression.
On mutual regulation, we highlight EpimiR [90], which
focuses on regulation between epigenetic modifications and
miRNA, providing more than 1900 relationships between
19 types of epigenetic modifications and 617 miRNAs in 7 different
species.
We highlight mirtronDB [12], a database dedicated for mir-
trons and containing a total of 1407 mirtron precursors and 2426
matures: the information is deposited in a very user-friendly and
graphic website.
The PAmiRDB [91] is a database specific for viral targets of
plant miRNA. The current version contains 2600 plant miRNAs
and their interactions in approximately 500 viral species.
Finally, it seems that new databases have been focusing on
predicting the effects of SNPs in miRNA genes or miRNA-binding
sites [59].
6 Searching Information About Databases and Tools
In the Introduction, we presented two portals (NRDR and Tools4-

miRs) that could help in the search for miRNA databases and tools.
In the NRDR (www.ncrnadatabases.org), two main options are
Fig. 6 Databases for resource in non-coding RNAs
available: (a) the catalog page, which contains the detailed informa-
tion of each database; and (b) the NR2, which is the novel option to
retrieve sequences from 30 databases (Fig. 6). For the purpose of
this chapter, we focus only on the NRDR option. In this website,
the user can find the database information by (a) search page,
(b) browser page by categories (e.g., RNA class), and (c) by the
statistics report. The Search page provides 10 options (see Fig. 7)
from database name to organisms and RNA families. For example,
the user could filter by only the miRNA databases that have infor-
mation for download in Arabidopsis thaliana, which returns
12 results. This can easily help to filter the information according
Fig. 7 Online panel for searching for noncoding RNAs and filter parameters
to the user needs. The second portal is Tools4miRs (https://

tools4mirs.org), which contains both information about tools and
databases (Fig. 8). The database is displayed in a search page where
the tools are grouped in nine categories as search options: Known
miRNA Identification, isomiRs Identification, Novel miRNA/Pre-
cursor Analysis, Differential Expression Analysis, Target Prediction,
Target Functional Analysis, miRNA-SNP Analysis, Other Tools,
and Target Prediction Server. For each of these, there are specific
search pages with some similar filter options such as Organism,
Data Collection, and Tools availability (see Fig. 9 for database
search page example). After the filter options are chosen, the data-
base or tools description table (Name, Description, Organism Spe-
cific, Data Collection, Data Available For Download, and
Reference) is shown. The results page allows the user to easily
find and filter the information according to interest.
Fig. 8 Online tool for miRNA analysis and its correlations
Fig. 9 Search filter for analysis in miRNAs

7 Conclusion, Challenges, and Perspective
Considering all the data analyzed and the ongoing discussion, it is

still possible to extract a few more ideas and try to learn more. For
example, the dataset is extremely important and necessary for sev-
eral reasons: performing data exploratory analysis (e.g., [92, 93]);
developing machine learning methods and tools [8]; building data-
bases [1, 4]; comparing the state-of-the-art [6]; helping to align
and seeing if your sequence was previously described; and guiding
to select a potential candidate for experimental validation in the lab.
Keeping this in mind, only 53 databases (40.76%) make available
any data to download (e.g., GFF, CSV, FASTA), whereas 43 data-
bases (33%) are currently available online and provide any type of
data to download. However, if we look only at sequences for
download without any request or restriction, this number decreases
to just 13 databases (10%): (CircuitsDB [64], CoGemiR [89],
miRbase [94], MirGeneDB.org [95], miRNEST [96], miRSponge
[97], mirtronDB [12], PlanTE-MIR DB [13], PMRD [98], Soma-
miR [99], TarBase [86], Vir-Mir db [100], and VIRmiRNA
[101]). Such results are not satisfactory if we looking for informa-
tion that actually available in the literature. However, outside of
redundancy, it is expected to be close to 100% of sequence data
available, which raises some important questions. Once the authors
have published their work, why cannot data be publicly available to
the scientific community? With these data, everyone can obtain
needed information and advance one step further in their research.
Now, as to numbers about the tools: a total of 16 tools (23.8%)
are not available (we tested several times). From the 61 tools that
remain (76, 2%), only 6 (9.8%) are specific for plants whereas
22 tools (36%) are specific for the human genome, representing
almost four fold more available tools for humans compared to
plants. This bias for human research over plants is not new, which
indicates the scientific community interest in humans or mammals.
However, this approach is starting to change, considering that large
groups are starting to make data for plants more available and
standardized (e.g., Ensembl Plants: plants.ensembl.org). Regard-
ing the disponibility of the tool, miRNAFold [31] is available both
as online version and the standalone, for example. However, as
mentioned in the website, a standalone version of miRNAFold is
available upon request. We tried twice to request the tool but did
not obtain any answer. This failure highlighted another problem:
tools that are available on the web but not for large-scale analysis in
a standalone manner. Another aspect about tools: looking only for
tools published and no longer available in the past 5 years
(2014–2018), we found four tools in this category [2016:
miTRATA [102]; 2014: miRClassify [103], mrSNP [52], sPARTA
[104]]. Even when tools are published in reputable journals (e.g.,
Nucleic Acids Research, Bioinformatics, BMC Bioinformatics), this

problem is still noticed.
All the scientific community (scientists, editors, journals, etc.)
needs to establish a standard request for authors to provide at least
2 or 5 years of availability for their databases and tools. Some
journals already do so, which could be a way for authors to guaran-
tee availability. One suggestion to solve such problems is to always
make available the full tool in a general platform (e.g., Git Hub).
The same for databases: authors could consider uploading the
entire dataset to a data repository such as FigShare or Data Dryad
to ensure data availability. Of course, it is necessary to measure the
database size to do that. But all these problems need to be solved to
make possible reproducing all scientific investigations. Otherwise,
we will continue to have just a small fraction of all the data pub-
lished really available. Even now as we are facing the big data age,
we are far from the big available data age. Again, we need the data
or tool available to improve the next step of scientific research.
References
1. Paschoal AR, Maracaja-Coutinho V, Setubal 8. Lee RC, Feinbaum RL, Ambros V (1993)
JC, Simões ZL, Verjovski-Almeida S, Durham The C. elegans heterochronic gene lin-4
AM (2012) Non-coding transcription charac- encodes small RNAs with antisense comple-
terization and annotation: a guide and web mentarity to lin-14. Cell 75(5):843–854
resource for non-coding RNA databases. 9. Stegmayer G, Di Persia LE, Rubiolo M,
RNA Biol 9(3):274–282 Gerard M, Pividori M, Yones C, Bugnon
2. Kozomara A, Birgaoanu M, Griffiths-Jones S LA, Rodriguez T, Raad J, Milone DH
(2019) miRBase: from microRNA sequences (2018) Predicting novel microRNA: a com-
to function. Nucleic Acids Res 47(D1): prehensive comparison of machine learning
D155–D162 approaches. Brief Bioinform 20
3. Negri TDC, Alves WAL, Bugatti PH, Saito (5):1607–1620
PTM, Domingues DS, Paschoal AR (2018) 10. Chen Q, Meng X, Liao Q, Chen M (2018)
Pattern recognition analysis on long noncod- Versatile interactions and bioinformatics anal-
ing RNAs: a tool for prediction in plants. Brief ysis of noncoding RNAs. Brief Bioinform 20
Bioinform 20(2):682–689 (5):1781–1794
4. Maracaja-Coutinho V, Paschoal AR, Caris- 11. Lukasik A, Wójcikowski M, Zielenkiewicz P
Maldonado JC, Borges PV, Ferreira AJ, Dur- (2016) Tools4miRs: one place to gather all
ham AM (2019) Noncoding RNAs databases: the tools for miRNA analysis. Bioinformatics
current status and trends. Methods Mol Biol 32(17):2722–2724
1912:251–285 12. Da Fonseca BHR, Domingues DS, Paschoal
5. Singh NK (2017) microRNAs databases: AR (2019) mirtronDB: a mirtron knowledge
developmental methodologies, structural and base. Bioinformatics 35(19):3873–3874
functional annotations. Interdiscip Sci 9 13. Lorenzetti APR, De Antonio GYA, Paschoal
(3):357–377 AR, Domingues DS (2016) PlanTE-MIR
6. Chen L, Heikkinen L, Wang C, Yang Y, DB: a database for transposable element-
Sun H, Wong G (2018) Trends in the devel- related microRNAs in plant genomes. Funct
opment of miRNA bioinformatics tools. Brief Integr Genomics 16(3):235–242
Bioinform 20(5):1836–1852 14. Ma X, Liu C, Gu L, Mo B, Cao X, Chen X
7. Szwacka M, Pawełkowicz M, Skarzyńska A, (2018) TarHunter, a tool for predicting con-
Osipowski P, Wojcieszek M, Przybecki Z, Pla˛- served microRNA targets and target mimics in
der W (2018) Biological significance, compu- plants. Bioinformatics 34(9):1574–1576
tational analysis, and applications of plant 15. Li J, Han X, Wan Y, Zhang S, Zhao Y, Fan R,
microRNAs. Acta Physiol Plant 40:146 Cui Q, Zhou Y (2018) TAM 2.0: tool for
MicroRNA set analysis. Nucleic Acids Res 46 28. Urbanek-Trzeciak MO, Jaworska E, Krzyzo-
(W1):W180–W185 siak WJ (2018) miRNAmotif: a tool for the
16. Kaushik A, Saraf S, Mukherjee SK, Gupta D prediction of pre-miRNA–protein interac-
(2015) miRMOD: a tool for identification tions. Int J Mol Sci 19(12):4075
and analysis of 50 and 30 miRNA modifications 29. Preusse M, Marr C, Saunders S, Maticzka D,
in next generation sequencing small RNA Lickert H, Backofen R, Theis F (2015)
data. Peer J 3:e1332 SimiRa: a tool to identify coregulation
17. Corrada D, Viti F, Merelli I, Battaglia C, between microRNAs and RNA-binding pro-
Milanesi L (2011) myMIR: a genome-wide teins. RNA Biol 12(9):998–1009
microRNA targets identification and annota- 30. Natsidis P, Kappas I, Karlowski WM (2018)
tion tool. Brief Bioinform 12(6):588–600 StarSeeker: an automated tool for mature
18. Artzi S, Kiezun A, Shomron N (2008) miR- duplex microRNA sequence identification
NAminer: a tool for homologous microRNA based on secondary structure modeling of
gene search. BMC Bioinformatics 9:39 precursor molecule. J Biol Res (Thessalon)
19. Barenboim M, Zoltick BJ, Guo Y, Weinberger 25:11
DR (2010) MicroSNiPer: a web tool for pre- 31. Tav C, Tempel S, Poligny L, Tahi F (2016)
diction of SNP effects on putative microRNA miRNAFold: a web server for fast miRNA
targets. Hum Mutat 31(11):1223–1232 precursor prediction in genomes. Nucleic
20. Sales G, Coppe A, Bisognin A, Biasiolo M, Acids Res 44(W1):W181–W184
Bortoluzzi S, Romualdi C (2010) MAGIA, a 32. Aguiar RR, Ambrosio LA, Sepúlveda-
web-based tool for miRNA and genes Hermosilla G, Maracaja-Coutinho V,
integrated analysis. Nucleic Acids Res 38: Paschoal AR (2016) miRQuest: integration
W352–W359 of tools on a web server for microRNA
21. Bartonicek N, Enright AJ (2010) SylArray: a research. Genet Mol Res 15(1)
web server for automated detection of 33. Lei J, Sun Y (2014) miR-PREFeR: an accu-
miRNA effects from expression data. Bioin- rate, fast and easy-to-use plant miRNA predic-
formatics 26(22):2900–2901 tion tool using small RNA-Seq data.
22. Higashi S, Fournier C, Gautier C, Gaspin C, Bioinformatics 30(19):2837–2839
Sagot MF (2015) Mirinho: an efficient and 34. Humphreys DT, Suter CM (2013) miR-
general plant and animal pre-miRNA predic- spring: a compact standalone research tool
tor for genomic and deep sequencing data. for analyzing miRNA-seq data. Nucleic Acids
BMC Bioinformatics 16:179 Res 41(15):e147
23. Bonnet E, He Y, Billiau K, Van de Peer Y 35. Falcone E, Grandoni L, Garibaldi F, Manni I,
(2010) TAPIR, a web server for the predic- Filligoi G, Piaggio G, Gurtner A (2016) Infin-
tion of plant microRNA targets, including ity: an in-silico tool for genome-wide predic-
target mimics. Bioinformatics 26 tion of specific DNA matrices in miRNA
(12):1566–1568 genomic loci. PLoS One 11(4):e0153658
24. L’Yi S, Jung D, Oh M, Kim B, Freishtat RJ, 36. An J, Lai J, Lehman ML, Nelson CC (2013)
Giri M, Hoffman E, Seo J (2017) miRTarVis miRDeep*: an integrated application tool for
+: web-based interactive visual analytics tool miRNA identification from RNA sequencing
for microRNA target predictions. Methods data. Nucleic Acids Res 41(2):727–737
124:78–88 37. Hackenberg M, Sturm M, Langenberger D,
25. Tong Y, Ru B, Zhang J (2018) miRNACan- Falcón-Pérez JM, Aransay AM (2009) miRa-
cerMAP: an integrative web server inferring nalyzer: a microRNA detection and analysis
miRNA regulation network for cancer. Bioin- tool for next-generation sequencing experi-
formatics 34(18):3211–3213 ments. Nucleic Acids Res 37:W68–W76
26. Hamed M, Spaniol C, Nazarieh M, Helms V 38. Huang TH, Fan B, Rothschild MF, Hu ZL,
(2015) TFmiR: a web server for constructing Li K, Zhao SH (2007) MiRFinder: an
and analyzing disease-specific transcription improved approach and software implementa-
factor and miRNA co-regulatory networks. tion for genome-wide fast microRNA precur-
Nucleic Acids Res 43(W1):W283–W288 sor scans. BMC Bioinformatics 8:341
27. da Silveira WA, Renaud L, Simpson J, Glen 39. An J, Lai J, Sajjanhar A, Lehman ML, Nelson
WB Jr, Hazard ES, Chung D, Hardiman G CC (2014) miRPlant: an integrated tool for
(2018) miRmapper: a tool for interpretation identification of plant miRNA from RNA
of miRNA–mRNA interaction networks. sequencing data. BMC Bioinformatics 15:275
Genes (Basel) 9(9):458 40. Zhang H, Resende EV, Silva B, Cui J (2018)
miRDis: a web tool for endogenous and
exogenous microRNA discovery based on 52. Deveci M, Catalyürek UV, Toland AE (2014)
deep-sequencing data analysis. Brief Bioin- mrSNP: software to detect SNP effects on
form 19(3):415–424 microRNA binding. BMC Bioinformatics
41. Guo L, Yu J, Liang T, Zou Q (2016) 15:73
miR-isomiRExp: a web-server for the analysis 53. Lin Y, Wu W, Sun Z, Shen L, Shen B (2018)
of expression of miRNA at the miRNA/iso- MiRNA-BD: an evidence-based bioinformat-
miR levels. Sci Rep 6:23700 ics model and software tool for microRNA
42. Urgese G, Paciello G, Acquaviva A, Ficarra E biomarker discovery. RNA Biol 15
(2016) isomiR-SEA: an RNA-Seq analysis (8):1093–1105
tool for miRNAs/isomiRs expression level 54. Subhra Das S, James M, Paul S, Chakravorty
profiling and miRNA–mRNA interaction N (2017) miRnalyze: an interactive database
sites evaluation. BMC Bioinformatics 17:148 linking tool to unlock intuitive microRNA
43. Maragkakis M, Vergoulis T, Alexiou P, regulation of cell signaling pathways. Data-
Reczko M, Plomaritou K, Gousis M, base 2017:bax015
Kourtis K, Koziris N, Dalamagas T, Hatzi- 55. Barta T, Peskova L, Hampl A (2016) miRNA-
georgiou AG (2011) DIANA-microT web song: a web-based tool for generation and
server upgrade supports fly and worm testing of miRNA sponge constructs in silico.
miRNA target prediction and bibliographic Sci Rep 6:36625
miRNA to disease association. Nucleic Acids 56. Rennie W, Liu C, Carmack CS, Wolenc A,
Res 39:W145–W148 Kanoria S, Lu J, Long D, Ding Y (2014)
44. Liang Z, Zhou H, He Z, Zheng H, Wu J STarMir: a web server for prediction of micro-
(2011) mirAct: a web tool for evaluating RNA binding sites. Nucleic Acids Res 42:
microRNA activity based on gene expression W114–W118
data. Nucleic Acids Res 39:W139–W144 57. Li F, Orban R, Baker B (2012) SoMART: a
45. Goswami CP, Nakshatri H (2012) PROG- web server for plant miRNA, tasiRNA and
miR: a tool for identifying prognostic target gene analysis. Plant J 70(5):891–901
miRNA biomarkers in multiple cancers using 58. Pedro DLF, Lorenzetti APR, Domingues DS,
publicly available data. J Clin Bioinforma 2 Paschoal AR (2018) PlaNC-TE: a compre-
(1):23 hensive knowledgebase of non-coding RNAs
46. Li R, Chen H, Jiang S, Li W, Li H, Zhang Z, and transposable elements in plants. Database
Hong H, Huang X, Zhao C, Lu Y, Bo X 2018:bay078. https://doi.org/10.1093/
(2018) CMTCN: a web tool for investigating database/bay078
cancer-specific microRNA and transcription 59. Fehlmann T et al (2019) A review of databases
factor co-regulatory networks. Peer J 6:e5951 predicting the effects of SNPs in miRNA
47. Rusinov V, Baev V, Minkov IN, Tabler M genes or miRNA-binding sites. Brief Bioin-
(2005) MicroInspector: a web tool for detec- form 20(3):1011–1020. https://doi.org/10.
tion of miRNA binding sites in an RNA 1093/bib/bbx155
sequence. Nucleic Acids Res 33:W696–W700 60. Griffiths-Jones S, Grocock RJ, van Dongen S,
48. Xie F, Zhang B (2010) Target-align: a tool for Bateman A, Enright AJ (2006) miRBase:
plant microRNA target identification. Bioin- microRNA sequences, targets and gene
formatics 26(23):3002–3003 nomenclature. Nucleic Acids Res 34:
49. Lánczky A, Nagy Á, Bottai G, Munkácsy G, D140–D144
Szabó A, Santarpia L, Gyorffy B (2016) miR- 61. Chien CH, Chiang-Hsieh YF, Chen YA,
power: a web-tool to validate survival- Chow CN, Wu NY, Hou PF, Chang WC
associated miRNAs utilizing expression data (2015) AtmiRNET: a web-based resource
from 2178 breast cancer patients. Breast Can- for reconstructing regulatory networks of
cer Res Treat 160(3):439–446 Arabidopsis microRNAs. Database (Oxford)
50. Andrés-León E, Núñez-Torres R, Rojas AM 2015:bav042
(2016) miARma-Seq: a comprehensive tool 62. Xu J, Li YH (2012) miRDeathDB: a database
for miRNA, mRNA and circRNA analysis. bridging microRNAs and the programmed
Sci Rep 6:25749 cell death. Cell Death Differ 19(9):1571
51. Zheng Y, Fei Z, Zhang W, Starren JB, Liu L, 63. Pio G, Ceci M, Malerba D, D’Elia D (2015)
Baccarelli AA, Li Y, Hou L (2014) PGS: a tool ComiRNet: a web-based system for the analy-
for association study of high-dimensional sis of miRNA-gene regulatory networks.
microRNA expression data with repeated BMC Bioinformatics 16(suppl 9):S7
measures. Bioinformatics 30(19):2802–2827 64. Friard O, Re A, Taverna D, De Bortoli M,
Corá D (2010) CircuitsDB: a database of
mixed microRNA/transcription factor feed- characterization of conserved microRNA tar-

forward regulatory circuits in human and gets in plants. Bioinformatics 30
mouse. BMC Bioinformatics 11:435 (14):2066–2067
65. Privitera AP, Distefano R, Wefer HA, Ferro A, 76. Mhuantong W, Wichadakul D (2009)
Pulvirenti A, Giugno R (2015) OCDB: a MicroPC (microPC): a comprehensive
database collecting genes, miRNAs and resource for predicting and comparing plant
drugs for obsessive-compulsive disorder. microRNAs. BMC Genomics 10:366
Database (Oxford) 2015:bav069 77. Vlachos IS, Kostoulas N, Vergoulis T,
66. Xu Y, Yang H, Wu T, Dong Q, Sun Z, Georgakilas G, Reczko M, Maragkakis M,
Shang D, Li F, Xu Y, Su F, Liu S, Zhang Y, Paraskevopoulou MD, Prionidis K,
Li X (2017) BioM2MetDisease: a manually Dalamagas T, Hatzigeorgiou AG (2012)
curated database for associations between DIANA miRPath v.2.0: investigating the
microRNAs, metabolites, small molecules combinatorial effect of microRNAs in path-
and metabolic diseases. Database (Oxford) ways. Nucleic Acids Res 40:W498–W504
2017:bax037 78. Subhra Das S, James M, Paul S, Chakravorty
67. Yang Z, Ren F, Liu C, He S, Sun G, Gao Q, N (2017) miRnalyze: an interactive database
Yao L, Zhang Y, Miao R, Cao Y, Zhao Y, linking tool to unlock intuitive microRNA
Zhong Y, Zhao H (2010) dbDEMC: a data- regulation of cell signaling pathways. Data-
base of differentially expressed miRNAs in base (Oxford) 2017(1):bax015
human cancers. BMC Genomics 11(suppl 79. Backes C, Kehl T, Stöckel D, Fehlmann T,
4):S5 Schneider L, Meese E, Lenhof HP, Keller A
68. Repana D, Nulsen J, Dressler L, (2017) miRPathDB: a new dictionary on
Bortolomeazzi M, Venkata SK, Tourna A, microRNAs and target pathways. Nucleic
Yakovleva A, Palmieri T, Ciccarelli FD Acids Res 45(D1):D90–D96
(2019) The Network of Cancer Genes 80. Kowarsch A, Preusse M, Marr C, Theis FJ
(NCG): a comprehensive catalogue of (2011) miTALOS: analyzing the tissue-
known and candidate cancer genes from can- specific regulation of signaling pathways by
cer sequencing screens. Genome Biol 20(1):1 human and mouse microRNAs. RNA 17
69. Mitra S, Das S, Das S, Ghosal S, Chakrabarti J (5):809–819
(2012) HNOCDB: a comprehensive database 81. Bruno AE, Li L, Kalabus JL, Pan Y, Yu A, Hu
of genes and miRNAs relevant to head and Z (2012) miRdSNP: a database of disease-
neck and oral cancer. Oral Oncol 48 associated SNPs and microRNA target sites
(2):117–119 on 30 UTRs of human genes. BMC Genomics
70. Kao S, Shiau CK, Gu DL, Ho CM, Su WH, 13:44
Chen CF, Lin CH, Jou YS (2012) IGDB. 82. Zorc M, Obsteter J, Dovc P, Kunej T (2015)
NSCLC: integrated genomic database of Genetic variability of microRNA genes in
non-small cell lung cancer. Nucleic Acids Res 15 animal species. J Genomics 3:51–56
40:D972–D977 83. Nam S, Kim B, Shin S, Lee S (2008) miRGa-
71. Jiang Q, Wang Y, Hao Y, Juan L, Teng M, tor: an integrated system for functional anno-
Zhang X, Li M, Wang G, Liu Y (2009) miR2- tation of microRNAs. Nucleic Acids Res 36:
Disease: a manually curated database for D159–D164
microRNA deregulation in human disease. 84. Liu H, Jin T, Liao R, Wan L, Xu B, Zhou S,
Nucleic Acids Res 37:D98–D104 Guan J (2012) miRFANs: an integrated data-
72. Sarver AL, Subramanian S (2012) Competing base for Arabidopsis thaliana microRNA
endogenous RNA database. Bioinformation 8 function annotations. BMC Plant Biol 12:68
(15):731–733 85. Hsu SD, Lin FM, Wu WY, Liang C, Huang
73. Agarwal V, Bell GW, Nam JW, Bartel DP WC, Chan WL, Tsai WT, Chen GZ, Lee CJ,
(2015) Predicting effective microRNA target Chiu CM, Chien CH, Wu MC, Huang CY,
sites in mammalian mRNAs. elife 4:e05005 Tsou AP, Huang HD (2011) miRTarBase: a
74. Gamazon ER, Im HK, Duan S, Lussier YA, database curates experimentally validated
Cox NJ, Dolan ME, Zhang W (2010) microRNA–target interactions. Nucleic Acids
Exprtarget: an integrative approach to pre- Res 39:D163–D169
dicting human microRNA targets. PLoS 86. Sethupathy P, Corda B, Hatzigeorgiou AG
One 5(10):e13534 (2006) TarBase: a comprehensive database of
75. Chorostecki U, Palatnik JF (2014) comTAR: experimentally supported animal microRNA
a web tool for the prediction and targets. RNA 12(2):192–197
87. Tian T, You Q, Zhang L, Yi X, Yan H, Xu W, 96. Szczesniak MW, Makalowska I (2014) miRN-
Su Z (2016) SorghumFDB: sorghum func- EST 2.0: a database of plant and animal
tional genomics database with multidimen- microRNAs. Nucleic Acids Res 42:D74–D77
sional network analysis. Database (Oxford) 97. Wang P, Zhi H, Zhang Y, Liu Y, Zhang J,
2016:baw099 Gao Y, Guo M, Ning S, Li X (2015) miR-
88. Ganguli S, Mitra S, Datta A (2011) Antago- Sponge: a manually curated database for
mirbase: a putative antagomir database. experimentally supported miRNA sponges
Bioinformation 7(1):41–43 and ceRNAs. Database (Oxford) 2015:
89. Maselli V, Di Bernardo D, Banfi S (2008) bav098
CoGemiR: a comparative genomics micro- 98. Zhang Z, Yu J, Li D, Zhang Z, Liu F, Zhou X,
RNA database. BMC Genomics 9:457 Wang T, Ling Y, Su Z (2010) PMRD: plant
90. Dai E, Yu X, Zhang Y, Meng F, Wang S, microRNA database. Nucleic Acids Res 38:
Liu X, Liu D, Wang J, Li X, Jiang W (2014) D806–D813
EpimiR: a database of curated mutual regula- 99. Bhattacharya A, Cui Y (2016) SomamiR 2.0:
tion between miRNAs and epigenetic modifi- a database of cancer somatic mutations alter-
cations. Database (Oxford) 2014:bau023 ing microRNA–ceRNA interactions. Nucleic
91. Satish D, Mukherjee SK, Gupta D (2019) Acids Res 44(D1):D1005–D1010
PAmiRDB: a web resource for plant miRNAs 100. Li SC, Shiau CK, Lin WC (2008) Vir-Mir db:
targeting viruses. Sci Rep 9(1):4627 prediction of viral microRNA candidate hair-
92. Rorbach G, Unold O, Konopka BM (2018) pins. Nucleic Acids Res 36:D184–D189
Distinguishing mirtrons from canonical miR- 101. Qureshi A, Thakur N, Monga I, Thakur A,
NAs with data exploration and machine Kumar M (2014) VIRmiRNA: a comprehen-
learning methods. Sci Rep 8(1):7560 sive resource for experimentally validated viral
93. Paschoal AR, Lozada-Chávez I, Domingues miRNAs and their targets. Database (Oxford)
DS, Stadler PF (2018) ceRNAs in plants: 2014:bau103
computational approaches and associated 102. Patel P, Ramachandruni SD, Kakrana A,
challenges for target mimic research. Brief Nakano M, Meyers BC (2016) miTRATA: a
Bioinform 19(6):1273–1289 web-based tool for microRNA truncation and
94. Griffiths-Jones S (2006) miRBase: the micro- tailing analysis. Bioinformatics 32
RNA sequence database. Methods Mol Biol (3):450–452
342:129–138 103. Zou Q, Mao Y, Hu L, Wu Y, Ji Z (2014)
95. Fromm B, Billipp T, Peck LE, Johansen M, miRClassify: an advanced web server for
Tarver JE, King BL, Newcomb JM, Sempere miRNA family classification and annotation.
LF, Flatmark K, Hovig E, Peterson KJ (2015) Comput Biol Med 45:157–160
A uniform system for the annotation of verte- 104. Kakrana A, Hammond R, Patel P, Nakano M,
brate microRNA genes and the evolution of Meyers BC (2014) sPARTA: a parallelized
the human microRNAome. Annu Rev Genet pipeline for integrated analysis of plant
49:213–242 miRNA and cleaved mRNA data sets, includ-
ing new miRNA target-identification soft-
ware. Nucleic Acids Res 42(18):e139
Chapter 8
Computational Detection of Pre-microRNAs

Müşerref Duygu Saçar Demirci
Abstract
MicroRNA (miRNA) studies have been one of the most popular research areas in recent years. Although
thousands of miRNAs have been detected in several species, the majority remains unidentified. Thus,
finding novel miRNAs is a vital element for investigating miRNA mediated posttranscriptional gene
regulation machineries. Furthermore, experimental methods have challenging inadequacies in their capa-
bility to detect rare miRNAs, and are also limited to the state of the organism under examination (e.g.,
tissue type, developmental stage, stress-disease conditions). These issues have initiated the creation of high-
level computational methodologies endeavoring to distinguish potential miRNAs in silico. On the other
hand, most of these tools suffer from high numbers of false positives and/or false negatives and as a result
they do not provide enough confidence for validating all their predictions experimentally. In this chapter,
computational difficulties in detection of pre-miRNAs are discussed and a machine learning based approach
that has been designed to address these issues is reviewed.
Key words miRNA, Ab initio prediction, In silico miRNA prediction, Pre-miRNA datasets
1 Introduction
MicroRNAs (miRNA) are single-stranded and small RNAs that

function as the posttranscriptional regulators of gene expression
by using translational inhibition and/or destabilization of the tar-
get mRNAs [1, 2]. Although the first miRNA was discovered in
Caenorhabditis elegans, as a regulator of the developmental timing
[3], numerous important processes are controlled through miR-
NAs’ action in hundreds of organisms ranging from viruses to
higher eukaryotes. Furthermore, many reports show that miRNAs
are involved in human diseases such as cancer and neurodegenera-
tive diseases [4–6].
There are various experimental and computational methods [7]
designed to explore miRNAs; from standard molecular biology
methods like forward genetic screening to more sophisticated
high technology systems like Next Generation Sequencing
(NGS). While experimental confirmation is the gold standard for
miRNA detection, such methods suffer from certain challenges
167
168 Müşerref Duygu Saçar Demirci
such as: high cost, arduous steps, and limitations to scan all possible
candidates in different systems. Thus, an effective ab initio compu-
tational approach is essential for miRNA prediction studies.
2 Computational miRNA Detection
The most straightforward computational approach is using already

known miRNAs to search for their homologs in other organisms
also known as “homology based search.” Since it is a well-known
and standard method for analyzing biological sequences, the sys-
tem is quite sufficient for widely conserved miRNAs. However, it is
not possible to detect nonconserved and/or species-specific miR-
NAs by using homology-based approaches so ab initio systems are
the only option for an extensive search of novel miRNAs. Addition-
ally, homology detection is a challenging topic as reviewed in [8].
Machine Learning (ML) might be one of the most popular
approaches [9] that has been used to analyze biological data from
genomics, systems biology, evolution, microarray, and proteomics
[10]. Although there are some examples of application of unsuper-
vised methods to perform miRNA target prediction [11], majority
of the tools using ML for miRNA gene prediction is based on
supervised [12, 13] (Table 3).
In general, a supervised ML workflow includes five main parts
[14]:
1. Data acquisition for training and testing.
2. Selection of features.
3. Selection of ML algorithms.
4. Training scheme (e.g., validation method).
5. Adjustment of classifier parameters.
The first and maybe the most important step in building an
effective ML based miRNA prediction system, is obtaining data
sets. However, there are limited numbers of databases providing
hairpin sequences. Table 1 lists some of those sources hosting
hairpin sequences of various species. While there are some specific
databases for miRNA—disease associations [15] and/or validated
targets of known miRNAs [16], such services might provide mature
sequences but not hairpins. The standard repository for miRNA
hairpin sequences is miRBase [17] and even though it is frequently
used there are some concerns regarding its dubious entries [18].
Majority of miRNA predictors are based on a 2—class classifi-
cation system and only a few is built by using one class classification
strategies [19]. However, for building a model through 2-class
learner, not only known miRNAs but also negative samples (similar
Detection of Pre-miRNA 169
Table 1
List of databases storing pre-miRNA sequences. Note that only databases providing hairpin
sequences are listed
Database # Species #miRNA Link

miRBase [21] 271 38,589 www.mirbase.org
MirGeneDB [22] 34 8857 mirgenedb.org
miRNAMap [23] 14 mirnamap.mbc.nctu.edu.tw
miRNest [24] 544 57,165 rhesus.amu.edu.pl/mirnest/copy/home.php
Table 2
List of available negative datasets for pre-miRNA classification. The table is adapted from [13] and
sorted by size
Dataset Size Source

NegHsa 68,046 http://adaa.polsl.pl/agudys/huntmi/huntmi.htm
Zou 14,246 http://datamining.xmu.edu.cn/main/~leyiwei/mirnaDetect.html
Pseudo 8492 http://web.bii.a-star.edu.sg/archive/stanley/Publications/Supp_materials/
06-002-supp.html
Chen 3054 http://bioinformatics.hitsz.edu.cn/iMiRNA-SSF/Material.jsp
NotBestFold 1881 https://data.mendeley.com/datasets/mgh5r9wny7/1
Shuffled 1423 https://data.mendeley.com/datasets/mgh5r9wny7/1
hsaFR 5000 https://data.mendeley.com/datasets/mgh5r9wny7/1
hsaBQ 5000 https://data.mendeley.com/datasets/mgh5r9wny7/1
hsaAM 5000 https://data.mendeley.com/datasets/mgh5r9wny7/1
pseudoFR 5000 https://data.mendeley.com/datasets/mgh5r9wny7/1
pseudoBG 5000 https://data.mendeley.com/datasets/mgh5r9wny7/1
pseudoAM 5000 https://data.mendeley.com/datasets/mgh5r9wny7/1
to miRNAs in some ways but are not true miRNAs) are required.
Unlike known miRNAs, there is no standard dataset for sequences
that can be used as negative. Moreover, the quality and quantity of
negative data for miRNA analysis are not sufficient. Table 2 shows
some of the available negative datasets for miRNA classification
studies.
Most of the time, features describing miRNA hairpins include
sequence, structure, and thermodynamic properties of miRNAs
such as minimum free energy (mfe) required for the secondary
structure formation, number of individual nucleotides and motifs,
length of hairpin sequence (Fig. 1). At the end, there are various
Fig. 1 MiRNA hairpin structure of hsa-mir-21 created by using RNAfold WebServer (http://rna.tbi.univie.ac.at//
cgi-bin/RNAWebSuite/RNAfold.cgi)
components affecting the overall performance of the model gener-

ated, but one of the most crucial one is the efficient selection of
features since hundreds of parameters describing miRNAs have
been proposed so far [13, 14].
Among numerous supervised ML algorithms some of them like
Support Vector Machine (SVM) is often used for miRNA studies.
Classifiers like Naive Bayes (NB), Multi Layered Perceptron
(MLP), Random Forest (RF), Asymmetric Partial Least Squares
Classification (APLSC), and Generalized Gaussian Density Estima-
tor (G2DE) have also been used in miRNA analysis (Table 3).
In any case, there are two foremost challenges with the existing
ML based miRNA prediction systems.
– Class imbalance: Recently, it is estimated that human genome
hosts more than 60 million hairpin structures that could be
possible miRNAs [13]. Furthermore, while building a classifica-
tion model with known miRNAs and negatives, the number of
miRNAs is usually less than negative ones; for example, miRBase
22.1 has 1917 human hairpin sequences, and frequently used
Pseudo negative dataset includes more than 8000 sequences
(Tables 1 and 2). It has been shown that such imbalance
between dataset size might cause significant decrease in the
performance of miRNA prediction process [20].
– Tool selection: Even though there are several software for the
computational prediction of pre-miRNAs (Table 3), majority of
them are either no longer updated and/or working. Moreover,
many researchers tend to choose a tool according to reported
performance measures such as accuracy, specificity, and sensitiv-
ity. However, a direct comparison between diverse tools based
on their published performance scores is not possible since each
of them use different elements to build their system; different
datasets, feature groups, classifiers etc.
To overcome these challenges, an ensemble pre-miRNA pre-
diction approach, izMiR framework was developed [13]. izMiR
allows for detailed comparison of 13 popular tools and presents a
method for miRNA prediction through synergistic combination of
available approaches. The idea behind izMiR framework is based on
generating models for each of these 13 tools by using three differ-
ent classifiers and picking the best models from 3000 trials (Fig. 2).
Table 3
List of various tools for miRNA prediction. The table is adapted and updated from [25] and sorted by
publication year. ML: Machine Learning, Cite: Number of Google Scholar citations. Conservation,
Structure and Sequence columns indicate if these properties are used in the analysis. NGS: Next
Generation Sequencing; tools that can perform predictions from sequences obtained from NGS
experiments
Tool Year Conservation Structure Sequence ML NGS Cite

miRscan [26] 2003 + + 1502
miRAlign [27] 2005 + + 312
ProMiR [28] 2005 + + + + 90
Triplet-SVM [29] 2005 + + + 465
miR-abela [30] 2005 + + + 278
RNAmicro [31] 2006 + + + 214
BayesMiRNAfind [32] 2006 + + + 191
miRFinder [33] 2007 + + + 141
miPred [34] 2007 + + 436
MiRRim [35] 2007 + + + 57
miRDeep [36] 2008 + + 914
miRanalyzer [37] 2009 + + 277
SSCprofiler [38] 2009 + + + + + 59
HHMMiR [39] 2009 + + + 83
MIReNA [40] 2010 + + + + 123
miRPara [41] 2011 + + + 129
miRNAFold [42] 2012 + + 37
miREval 2.0 [43] 2013 + + + + 19
miR-PREFeR [44] 2014 + + 43
miRBoost [45] 2015 + + + 19
iMiRNA-SSF [46] 2016 + + + 40
izMiR [13] 2017 + + + + 15
miRge2.0 [47] 2018 + + + + 1
PmiRDiscVali [48] 2019 + + + +
3D representation [49] 2019 + + +
DeepMir [50] 2019 Uses images for training and application + +
Fig. 2 Simplified version of izMiR workflow. Features describing positive data (known miRNA hairpins) and
negative data sets are calculated. Through 1:1 positive/negative ratio and 1000 fold Monte Carlo Cross
Validation, 70% of data sets are used for learning, while remaining are tested for the performance. Models
with the highest accuracy scores are saved and applied on predictions
These models were applied to all known pre-miRNAs from miR-

Base. Even though during the training steps of izMiR human
pre-miRNA were used, it was able to show high performance on
distinct species’ miRNAs.
3 Conclusion
Over the years with the increasing popularity of both miRNAs and
ML applications, numerous tools attempting to predict
pre-miRNAs have been developed. Some of those tools suffer
from certain limitations due to characteristics of miRNA hairpins
and facts like quality of the datasets. Recently, an easily updatable,
free, and user-friendly approach, izMiR was developed. With vali-
dated and trustable datasets and carefully selected features, an
efficient computational pre-miRNA prediction seems possible.
Regardless of the selection of computational tool or approach
applied for miRNA prediction, in order to be labeled as a true
miRNA, all of the predicted pre-miRNA candidates need to be
validated experimentally. Therefore, we need to have a system,
which would be highly accurate and sensitive with minimum num-
ber of false positives and negatives.
References
1. Piast M, Kustrzeba-Wójcicka I, Matusiewicz M human microRNA–disease associations.
et al (2005) Molecular evolution of enolase. Nucleic Acids Res 47:D1013–D1017
Acta Biochim Pol 52:507–513 16. Chou C-H, Shrestha S, Yang C-D et al (2018)
2. Filipowicz W, Bhattacharyya SN, Sonenberg N miRTarBase update 2018: a resource for exper-
(2008) Mechanisms of post-transcriptional imentally validated microRNA-target interac-
regulation by microRNAs: are the answers in tions. Nucleic Acids Res 46:D296–D302
sight? Nat Rev Genet 9:102–114 17. Griffiths-Jones S (2010) miRBase: microRNA
3. Lee RC, Feinbaum RL, Ambros V (1993) The sequences and annotation. Curr Protoc Bioin-
C. elegans heterochronic gene lin-4 encodes formatics Chapter 12:Unit 12.9.1-10
small RNAs with antisense complementarity 18. Saçar MD, Hamzeiy H, Allmer J (2013) Can
to lin-14. Cell 75:843–854 MiRBase provide positive data for machine
4. Bushati N, Cohen SM (2007) microRNA func- learning for the detection of MiRNA hairpins?
tions. Annu Rev Cell Dev Biol 23:175–205 J Integr Bioinform 10
5. Hébert SS, Horré K, Nicolaı̈ L et al (2009) 19. Yousef M, Allmer J, Khalifa W (2015)
MicroRNA regulation of Alzheimer’s amyloid Sequence motif-based one-class classifiers can
precursor protein expression. Neurobiol Dis achieve comparable accuracy to two-class lear-
33:422–428 ners for plant microRNA detection. J Biomed
6. Wang G, van der Walt JM, Mayhew G et al Sci Eng 08:684–694
(2008) Variation in the miRNA-433 binding 20. Sacar MD, Allmer J (2013) Data mining for
site of FGF20 confers risk for Parkinson disease microrna gene prediction: on the impact of
by overexpression of??-Synuclein. Am J Hum class imbalance and feature number for micro-
Genet 82:283–289 rna gene prediction. In: 2013 8th international
7. Saçar Demirci MD, Yousef M, Allmer J (2019) symposium on health informatics and bioinfor-
Computational prediction of functional micro- matics. IEEE, pp 1–6
RNA–mRNA interactions. Methods Mol Biol 21. Kozomara A, Griffiths-Jones S (2014) miR-
1912:175–196 Base: annotating high confidence microRNAs
8. Velandia-Huerto CA, Yazbeck AM, Schor J, using deep sequencing data. Nucleic Acids Res
Stadle PF (2021) Evolution and phylogeny of 42:D68–D73
microRNAs—protocols, pitfalls, and problems. 22. Fromm B, Domanska D, Hackenberg M et al
In: Allmer J, Yousef M (eds) miRNomics: (2018) MirGeneDB2.0: the curated micro-
microRNA biology and computational analy- RNA Gene Database. bioRxiv 258749
sis. Methods in molecular biology, vol 2257. 23. Hsu S-D, Chu C-H, Tsou A-P et al (2008)
Springer, New York miRNAMap 2.0: genomic maps of microRNAs
9. Saçar MD, Allmer J (2014) Machine learning in metazoan genomes. Nucleic Acids Res 36:
methods for microRNA gene prediction. D165–D169
Methods Mol Biol 1107:177–187 24. Szczesniak MW, Makalowska I (2014) miRN-
10. Larranaga P (2006) Machine learning in bioin- EST 2.0: a database of plant and animal micro-
formatics. Brief Bioinform 7:86–112 RNAs. Nucleic Acids Res 42:D74–D77
11. Heikkinen L, Kolehmainen M, Wong G (2011) 25. Gomes CPC, Cho J-H, Hood L et al (2013) A
Prediction of microRNA targets in Caenorhab- review of computational tools in microRNA
ditis elegans using a self-organizing map. Bio- discovery. Front Genet 4:81
informatics 27:1247–1254 26. Lim LP, Lau NC, Weinstein EG et al (2003)
12. Zhang B-T, Nam J-W (2008) Supervised The microRNAs of Caenorhabditis elegans.
learning methods for MicroRNA studies. In: Genes Dev 17:991–1008
Machine learning in bioinformatics. Wiley, pp 27. Wang X, Zhang J, Li F et al (2005) MicroRNA
339–365 identification based on sequence and structure
13. Saçar Demirci MD, Baumbach J, Allmer J alignment. Bioinformatics 21:3610–3614
(2017) On the performance of pre-microRNA 28. Nam J-W, Kim J, Kim S-K et al (2006) ProMiR
detection algorithms. Nat Commun 8:330 II: a web server for the probabilistic prediction
14. Saçar Demirci MD, Allmer J (2017) Delineat- of clustered, nonclustered, conserved and non-
ing the impact of machine learning elements in conserved microRNAs. Nucleic Acids Res 34:
pre-microRNA detection. PeerJ 5:e3131 W455–W458
15. Huang Z, Shi J, Gao Y et al (2019) HMDD 29. Xue C, Li F, He T et al (2005) Classification of
v3.0: a database for experimentally supported real and pseudo microRNA precursors using
local structure-sequence features and support 40. Mathelier A, Carbone A (2010) MIReNA:
vector machine. BMC Bioinformatics 6:310 finding microRNAs with high accuracy and no
30. Sewer A, Paul N, Landgraf P et al (2005) Iden- learning at genome scale and from deep
tification of clustered microRNAs using an ab sequencing data. Bioinformatics
initio prediction method. BMC Bioinformatics 26:2226–2234
6:267 41. Wu Y, Wei B, Liu H et al (2011) MiRPara: a
31. Hertel J, Stadler PF (2006) Hairpins in a Hay- SVM-based software tool for prediction of
stack: recognizing microRNA precursors in most probable microRNA coding regions in
comparative genomics data. Bioinformatics genome scale sequences. BMC Bioinformatics
22:e197–e202 12:107
32. Yousef M, Nebozhyn M, Shatkay H et al 42. Tempel S, Tahi F (2012) A fast ab-initio
(2006) Combining multi-species genomic method for predicting miRNA precursors in
data for microRNA identification using a genomes. Nucleic Acids Res 40:e80
naive Bayes classifier. Bioinformatics 43. Gao D, Middleton R, Rasko JEJ et al (2013)
22:1325–1334 miREval 2.0: a web tool for simple microRNA
33. Huang T-H, Fan B, Rothschild MF et al prediction in genome sequences. Bioinformat-
(2007) MiRFinder: an improved approach ics 29:3225–3226
and software implementation for genome- 44. Lei J, Sun Y (2014) miR-PREFeR: an accurate,
wide fast microRNA precursor scans. BMC fast and easy-to-use plant miRNA prediction
Bioinformatics 8:341 tool using small RNA-Seq data. Bioinformatics
34. Jiang P, Wu H, Wang W et al (2007) MiPred: 30:2837–2839
classification of real and pseudo microRNA 45. Tran VDT, Tempel S, Zerath B et al (2015)
precursors using random forest prediction miRBoost: boosting support vector machines
model with combined features. Nucleic Acids for microRNA precursor classification. RNA
Res 35:W339–W344 21:775–785
35. Terai G, Komori T, Asai K et al (2007) miR- 46. Chen J, Wang X, Liu B (2016) iMiRNA-SSF:
Rim: A novel system to find conserved miRNAs improving the identification of microRNA pre-
with high sensitivity and specificity. RNA cursors by combining negative sets with differ-
13:2081–2090 ent distributions. Sci Rep 6:19062
36. Friedl€ander MR, Chen W, Adamidi C et al 47. Lu Y, Aras AS, Halushka MK (2018) miRge
(2008) Discovering microRNAs from deep 2.0: an updated tool to comprehensively ana-
sequencing data using miRDeep. Nat Biotech- lyze microRNA sequencing data
nol 26:407–415 48. Yu D, Wan Y, Ito H et al (2019) PmiRDiscVali:
37. Hackenberg M, Sturm M, Langenberger D an integrated pipeline for plant microRNA dis-
et al (2009) miRanalyzer: a microRNA detec- covery and validation. BMC Genomics 20:133
tion and analysis tool for next-generation 49. Saçar Demirci MD (2019) MicroRNA predic-
sequencing experiments. Nucleic Acids Res tion based on 3D graphical representation of
37:W68–W76 RNA secondary structures. Turk J Biol
38. Oulas A, Boutla A, Gkirtzou K et al (2009) 43:274–280
Prediction of novel microRNA genes in 50. Cordero J, Menkovski V, Allmer J (2019)
cancer-associated genomic regions--a com- Detection of pre-microRNAs with convolu-
bined computational and experimental tional neural networks detection of
approach. Nucleic Acids Res 37:3276–3287 pre-microRNAs with convolutional neural net-
39. Kadri S, Hinman V, Benos PV (2009) works. bioRxiv 840579
HHMMiR: efficient de novo prediction of
microRNAs using hierarchical hidden Markov
models. BMC Bioinformatics 10(Suppl 1):S35
Chapter 9
Computational Methods for Predicting Mature microRNAs

Malik Yousef, Alisha Parveen, and Abhishek Kumar
Abstract
Tiny single-stranded noncoding RNAs with size 19–27 nucleotides serve as microRNAs (miRNAs), which
have emerged as key gene regulators in the last two decades. miRNAs serve as one of the hallmarks in
regulatory pathways with critical roles in human diseases. Ever since the discovery of miRNAs, researchers
have focused on how mature miRNAs are produced from precursor mRNAs. Experimental methods are
faced with notorious challenges in terms of experimental design, since it is time consuming and not cost-
effective. Hence, different computational methods have been employed for the identification of miRNA
sequences where most of them labeled as miRNA predictors are in fact pre-miRNA predictors and provide
no information about the putative miRNA location within the pre-miRNA. This chapter provides an update
and the current state of the art in this area covering various methods and 15 software suites used for
prediction of mature miRNA.
Key words Micro RNA, Mature miRNA, Mature miRNA prediction, Machine learning, Support
vector machine, Random forest
1 Introduction
MicroRNAs (miRNAs) are short single-stranded noncoding

mRNA with a size of 19–27 nucleotides (nt), which regulate gene
expression at the posttranscriptional level. miRNAs are crucial for
the development of many species and are frequently involved in a
wide array of human diseases like cancer [1] and cardiovascular
disorders [2]. During the miRNA biogenesis process, the micro-
processor complex (with two major enzymes: Drosha and its cofac-
tor DGCR8/Pasha) cleaves long primary miRNA transcripts
(pri-miRNAs) into double-stranded precursor miRNAs
(pre-miRNAs) [3]. Pre-miRNAs are hairpin intermediates with a
size of ~70 nt with an overhang of two nt on their 30 ends [4] These
intermediates are exported from the nucleus with carrier proteins
such as exportin-5 [4] and the endonuclease DICER cleaves these
intermediates into ~22 bp miRNA–miRNA* duplexes [5]. This
miRNA–miRNA* duplex possesses two stable strands. One strand
of this miRNA–miRNA* duplex remains as the mature miRNA
175
176 Malik Yousef et al.
[5]. This mature miRNA is incorporated into the RNA-induced

silencing complex (RISC) and the RISC will subsequently target
mRNAs (Fig. 1a). using partial sequence complementarity [5]. The
nucleotide sequence of mature miRNA is crucial in the recognition
of targets [5].
Experiment methods for identification of mature miRNA are
cloning and constructing short RNA libraries, but these methods
have their own limitations as all miRNAs are not cloneable and it is
also hindered by problems of low level of expression and tissues-
specific RNA library construction [6]. Rapid changes in whole
genome sequencing approaches primarily because of next-
generation sequencing (NGS), it is possible to decipher tissue-
specific miRNAs on a genome-wide scale by deep sequencing
[6]. But it pushes another challenge with billions of reads as it is a
notorious task to unravel individual small RNA [6]. Hence, there is
urgent requirements of overviewing computational approaches and
software suites for the prediction of mature miRNAs, which will
help dealing full length and mature miRNAs. Herein, we have
summarized computational approaches and resulting software
suites, which are widely used in the prediction of mature miRNAs.
There are two major computational approaches were utilized for
mature miRNA prediction miRNA as (a) rule-based approaches
and machine learning (ML)-based approaches [6]. From 2004
onward, 15 different software suites have become available, which
are widely used for identifying mature miRNAs from their
pre-miRNA transcripts (Fig. 1b).
2 Homology Guided Detection Algorithms
Homology-based approaches are based on scanning a similar

sequence of pre-miRNA from experimentally verified miRNAs
among the species using an evolutionary genomics approach [7],
combining sequence homology and/or secondary structural
homology. There are various homology-based mature miRNA
gene prediction software such as MirScan [8], miRNAminer [9],
ProMiR II [10], and MirFinder [11].
Limitations of this approach is that it become irrelevant for
divergent miRNAs without structural similarities. Additionally,
miRNAs are species-specific novel mature miRNA be missed by
this method. Therefore other strategies need to be used in parallel.
A powerful approach developed for genome-wide screening of
phylogenetically well-conserved pre-miRNAs between closely
related species uses multiple sequence alignment. However, it also
suffers from low sensitivity, especially for more divergent evolution-
ary distances [12].
Bioinformatics Methods for Mature miRNA Prediction 177
Fig. 1 Overview of mature miRNA and their prediction tools. (a) Processing mature miRNA from Pre-miRNA. (b)
Timeline of major computation methods for detection of mature miRNA
3 Ab Initio miRNA Gene Prediction Algorithms
The ab initio method does not require any prior information of

homology for miRNA gene prediction [13]. These approaches
require proper negative and positive datasets; which is conceptually
similar to homology-based approaches. Unlike the homology
method, the ab initio approach may enable the identification of
new miRNAs [13]. The main drawback of ab initio approach is to
design proper parameters in both datasets that allow determining a
given sequence to be a miRNA based on its given feature properties
[13]. For instance, highly weighted features include hairpin struc-
ture and folding free energy for miRNAs prediction [13]. In this
method generating proper parameter is a crucial step, otherwise it
does not provide high specificity, which leads to an increase in false
position results [13]. There are many programs using ab initio
methods with machine learning (ML) approaches including
Triplet-SVM [14] and Random Forest (RF) based MiPred [15].
4 Machine Learning (ML) Method
Machine learning (ML) is a subbranch of artificial intelligence

(AI) which improves the performance of prediction systems for
the discovery of knowledge from novel datasets using statistical
and mathematical approaches. Several mature miRNA predicting
tools are exploiting ML-based approaches after extracting feature

sets like sequence and structural component of miRNA sequences
and/or miRNA duplexes as by miRLocator [6], miRFinder [11],
MiRPara [16], MatureBayes [17], MiRdup [18], and
MatPred [19].
5 Widely Used Feature Sets
Generally, there are thousands of features that were extracted from

the pre-miRNA to be used for pre-miRNA prediction, however not
all features are used at a time and there are features MiRNAfe tool
[20] provides a wide arrays of feature lists which comprises most of
features. There are several features involved in the detection of
mature miRNA that consists of sequence content characteristics,
the stability of the secondary structure, thermodynamic energy,
sequence alignment measure, and folding. Sequence features
include frequency of nucleotide after every two and three alignment
position and the frequency of aggregated dinucleotide present in
the same sequence [21]. Structural features include free energy of
stem and loop regions that consist of diversity, frequency, entropy,
and enthalpy related properties of each structure [6]. There are
several other features involved in preparing training datasets such as
loop length, hairpin length, consecutive base pairs, and the ratio of
loop length to the hairpin length of pre-miRNA secondary struc-
ture. Thermodynamics energy of pre-miRNAs secondary structure
includes the minimal free energy of folding structure of
pre-miRNAs, overall free energy of the structure, and energy for
dissolving property of a secondary structure [6].
A recent study has studied the significant of the features of most
of the algorithms that suggest a tool for mature miRNA prediction
stating that they still predict an impractical amount of false positives
[22]. Most of the proposed features are based on the structure of
precursors of the miRNA only, not considering the important and
relevant information contained in the mature miRNA. In this work,
they propose three new features that take particularly into account
the order in which the nucleotides are presented in the mature
miRNA, which can effectively improve the sequence representa-
tion. The first proposed features is based on the Levenshtein dis-
tance [23]. The second and third proposed features were inspired,
from the point of view of the information theory, considering the
randomness of a sequence that would encode a mature miRNA in
the hairpin.
6 Learning Model for Prediction
For building a learning classifier algorithm for mature miRNA

prediction proper feature-based dataset is needed. There are also
requirements of evaluations of the algorithm by cross-validation
process. To examine the robustness of the trained classifier model,
validated experimental results are applied to the classifier model.
Inequality of positive and negative datasets is a major limitation in
learning machine learning model [7]. There are exponential
increase in the validated experimentally verified datasets, but due
there are problems in verifications of pseudo datasets and this is
known as class imbalance [24]. This causes imbalance in a sense that
the positive datasets are higher in number as compared to the
pseudo datasets. These affect the efficiency and robustness of the
classifier model. There is another reason for the imbalance problem
as total numbers of pre-miRNAs and corresponding mature miR-
NAs are yet not fully knowns from majority of genomes [7]. Under
and oversampling mechanism has been applied to solve the uneven-
ness of the dataset. In the under-sampling process, the majority of
the class label with the even value of instances is removed whereas in
oversampling addition of new data sample or duplication of data
point applies to the existing minor class label [24]. There are several
ML-based microRNA predictor classifier that is used to remove the
class imbalance problem such as triplet-SVM [14] and miPred
[15]. Developers of miRLocator have illustrated effects of class
imbalance on the prediction performance of miRNA predictors.
They improvised the prediction models in miRLocator with an
appropriate RF based-ML approach and the ratio between positive
and negative samples (RPNS, 1:5 [6]).
The other weakness in the learning classifier is that most of the
current research in machine learning in microRNA based on the
assumption of the length of the stem, loop, and thermodynamic
energy of the data. Therefore, sequence outside the hairpin struc-
ture cannot be considered as a base sequence for the production of
a mature miRNA sequence, which reduces the accuracy and perfor-
mance of prediction classifier algorithm [7].
7 Commonly Used Software Suites
We have provided provides summary of known software suites for

predicting mature miRNA (Table 1) in the reverse chronological
order from 2019 to 2004.
7.1 miRFinder The most recent tool for mature miRNA prediction is the tool
named miRFinder [11]. miRFinder is Adaboost-SVM-based prob-
ability algorithm for the prediction of all mature based on
Table 1
Overview of major software suites for predicting mature microRNAs
Software suites Year Speciesa Featuresb Approachesc Availability References

miRFinder [11] 2019 SF ML, SVM https://github.com/ [11]
wangying0128/
miRFinder
matPred 2015 H SVM [19]
miRLocator 2015 P SSF ML, RM https://github.com/ [6]
cma2015/miRLocator
miRdup [18] 2013 MS SSF RM X [18]
MiRRim2 2012 H/CI SSF [25]
mirMat 2012 V FE/SF ML, RM http://mcube.nju.edu.cn/ [26]
jwang/lab/soft/
MiRmat/
MiRPara 2011 H/P SSF ML, SVM https://code.google.com/ [16]
archive/p/mirpara/
wikis
MaturePred 2011 P SSF ML, SVM X [27]
MatureBayes 2010 H/M SSF NB http://mirna.imbb.forth. [17]
gr/MatureBayes.html
Microprocessor 2007 H SSF ML, SVM X [28]
SVM
mirCoS 2007 H/M [29]
BayesMiRNAfind 2006 MS SSF NB https://mirnafind.fbk.eu/ [30]
engine/
BayeSVMmiRNAfind/
ProMir 2005 H SSF HMM https://bi.snu.ac.kr/ [3]
Research/ProMiR/
ProMiR.html
MiRalign 2005 A/P SS AL http://bioinfo.au. [31]
tsinghua.edu.cn/
miralign
MIRcheck 2004 P SSF SOR Application/script [32]
X This tool is not available anymore
a
A animal, CI Ciona intestinalis, H human, M mouse, MS multi-species, P plant, V vertebrate
b
FE free energy, SF structural features, SSF sequence and structural features
c
AL alignment, RM random forest, ML machine learning, NB Naı̈ve Bayes, HMM hidden Markov model, SOR set of
rules
structured-sequence features [11]. The tool is using structure fea-

tures to form hundreds of features, but the study did not make a
specific value of the size of the feature vector [11]. miRFinder [11]
was compared to other similar approaches, MiRPara [16],
MatureBayes [17], MiRdup [18], and MatPred [19] to report that

miRFinder [11] is superior to the other methods in identification
accuracy and average position deviation. We notice that in the study
of miRFinder [11], authors have been inconsistence with respect of
the name of the tool and they have used name as matFinder and
miRFinder [11].
7.2 miRLocator miRLocator is a random forest (RF)-based ML suite implemented

in Python [6] and it can reuse locally compiled datasets. It is used
for predicting mature miRNAs from plant genomes [6]. miRLoca-
tor is based on 440 sequence and structural features extracted from
miRNA duplexes [6]. After a tenfold cross-validation evaluation
using 5854 experimentally validated miRNAs (19 plant species),
Authors claimed demonstrated miRLocator is competitive to the
state-of-the-art miRNA predictor miRdup with similar or slightly
better results [6].
7.3 MatPred MatPred is based on the radial basis function kernel based-SVM
(RBF-SVM) with objective of predicting the start position of the
mature miRNA in a pre-miRNA transcript [19]. This tool uses
94 features that includes the length of the mature and precursor
miRNA transcript, the distances from the stem loop, and the mini-
mum free energy of the RNA complex and integrated features that
describe the nucleotide-specific RNA secondary structure charac-
teristics [19]. MatPred was evaluated by two independent datasets
derived from (a) a set of experimentally validated mature miRNAs
derived from miRBase (version 20) and (b) an additional dataset of
nonoverlapping with used training dataset [19]. Hence, MatPred
illustrated significant improvement on the prediction accuracy
compared to several existing methods [19]. Remarkably, ~35%
predictions accurately reported the known locations of mature
miRNAs, and over 90% of the predicted positions were within
5 bp of known sites [19].
7.4 miRdup miRdup (miRNA duplex) uses Random Forest (RF) algorithm
integrated with adaptive boost (Adaboost) applied on mature
miRNA represented as 100 features [18]. miRdup takes as input a
pre-miRNA sequence and the position of a candidate miRNA, and
returns a score that reflects the likelihood that the candidate is a
true miRNA [18]. This tool was just compared with the tool
matureBayes [17] claiming better performance. Additionally,
miRdup was trained for different clade to be more specific and
yield increased accuracy as there is different between species in
term of sequences and structural features [18]. miRdup is also fast
and auto-trainable user-friendly tool [18].
7.5 MiRRim2 MiRRim2 is designed for predicting human mature miRNA candi-
dates using a conditional random fields (CRFs) method, based on
evolutionary conserved features upstream of Drosha cleavage sites
[25]. miRRim2 is reported as accurately detect miRNA hairpins
and detects the location of a mature miRNA sequence over the
hairpin [25]. This tool has considered the whole human genome to
evaluate its efficiency in capturing the location of the mature miR-
NAs and claiming that it is outperform other similar computational
tools by capturing 40% of the known mature miRNAs [25]. Addi-
tionally, this tool was tested on the Ciona intestinalis genome
capturing 40% of the known mature miRNA [25].
7.6 MiRmat MiRmat is a ML-based method [26] to predict the mature miRNA
sequence that based on the biological fact that microRNAs are
known to be generated from primary transcripts mainly through
the sequential cleavages by two enzymes (Drosha and Dicer). MiR-
mat is combined by two sequential steps [26] as.
(a) The first is for Drosha processing site prediction, which uses
the energy distribution pattern as the feature and Random
Forest (RF) algorithm as classifier [26].
(b) The second step is for Dicer processing site prediction, which
uses RF algorithm to recognize the structural features. By
predicting Drosha and Dicer processing sites, MiRmat can
finally obtain the microRNA mature sequences [26]. The
study of MiRmat [26] was compared to other similar tools
such as proMIR [3]. MatureBayes [17] and MaturePred [27],
demonstrating that MiRmat has better performance. MiRmat
was shown to identify 77.8% of the Drosha processing sites
and 92.8% of the Dicer sites [26].
7.7 miRpara miRpara is a SVM classifier trained to predicts most probable

mature miRNA coding regions from genome scale sequences in a
species specific manner [16]. The algorithm uses 77 features in
order to present the data. The 77 features are related to physical
properties of the pre-miRNA and its miRNAs [16]. The tool
detects top 25 features that enhance the performance of the tool.
The reported accuracy is about 80% tested on validated mature
miRNA [16]. mirPara was trained two SVMs against experimen-
tally verified plant and mammalian sequences from miRBase and a
third SVM against all experimentally verified sequence data from
miRBase data [16]. Authors have compared miRpara with other
three similar tools miRAlign [31], mirEval [33], and BayesMiR-
NAfind [30] and they claimed it superiority [16].
7.8 MaturePred MaturePred is an ML (based on SVM) suite focusing on the pre-

diction of the miRNA positions for the new plant pre-miRNA
candidates [27]. The tool initially extracts the position-specific
features, the energy related features, the structure related features,

and stability related features from real/pseudo miRNA–miRNA*
duplexes [27].
7.9 MatureBayes MatureBayes identifies the starting sites of mature miRNAs for
mice and humans based on Naı̈ve Bayes [17]. As a result, the
method finds that 7, 8, and 9 nt from the starting position have
the typical biological features to distinguish mature miRNAs. The
tool is relying on the sequence and secondary structure information
of their miRNA precursors. A comparison study was done against
ProMiR [3] and BayesMiRNAfind [30] statin that MatureBayes
[17] is getting better results.
7.10 Microprocessor Microprocessor SVM predicts 50 Drosha processing sites in hairpins

SVM that are candidate miRNAs [28]. Microprocessor SVM correctly
predicts the processing site for 50% of known human 50 miRNAs,
and 90% of its predictions are within two nucleotides of the true site
[28]. Microprocessor uses a SVM classifier to identify Drosha
cleavage sites based on 686 sequence- and structure-related
features [28].
7.11 mirCos mirCos constructed a model based on SVM to predict miRNAs

conserved between human and mouse [29].
7.12 Bayes- BayesMiRNAfind [30] is a machine leaner approach that based on

MiRNAfind sequence and structure features for prediction pre-miRNAs and the
exact location of the mature miRNA over it. The tools split the
pre-miRNA secondary structure into 3 parts, where the mature
part is considered as one of it. Features are extracted from each
part to form the set of features for the naı̈ve bays classifier [30]. The
tool combines different miRNA species to create a general model.
However, considering the low number of the discovered micro-
RNA at the time of the tool the combing was a necessary step.
7.13 ProMir ProMir identifies human pre-miRNAs and their mature miRNAs by
combining sequence and structural features in a paired hidden
Markov model [3]. Additional versions of the tool were developed
ProMirii [22] and ProMiriii as incremental improvements to the
first version.
7.14 MiRalign MiRalign [31] is a genome-wide computational approach to detect

miRNAs in animals based on both sequence and structure align-
ment using BLAST. The des advantage of this tool that no new
mature miRNA is predicted. It is a search tool that looking for
known miRNA over other genomes.
7.15 MIRcheck MIRcheck identifies 20-nt regions of a given plant pre-miRNA

using a predetermined set of rules and constraints based on six
features [32]. The six features are plant specific characteristics that
are related to sequence and structure features [32].
8 Conclusion
We have surveyed the state-of-the-art approaches for computa-

tional prediction of mature miRNA over last 15 years based on
the classical homology-based methods, ab initio methods and
machine learning methods. There are 15 different tools developed
from 2004 to 2019 and only a dozen are active at the moment and
large fraction of these tools are using ML-based methods either
support vector machine (SVM) or Random Forest focusing on the
predictions from various animal (including human and plant gen-
omes. Sequence and structural features are used by these tools as a
primary source of feature collections.
The comparisons between those tools was not well established
[34]. We are not convinced about the reported results for some of
that tools that claim superiorities. We would suggest that it would
be useful to create one data set to be as a standard to the commu-
nity to be used in order to evaluate the tools, previous and future
tools. Additionally, we recommend that there is a need to establish
one tool that combine all the features suggested by the 15 tools and
create one system to the community.
However, during the current survey we notice that there are no
studies focusing the differences and the similarity between mature
miRNA sequences from different species. Thus, raising a question
about detecting specific pattern among matures that are unique for
a specific species.
Acknowledgments
AK is a recipient of Ramalingaswami Re-Retry Faculty Fellowship

(Grant; BT/RLF/Re-entry/38/2017) from Department of Bio-
technology (DBT), Government of India (GOI).
References
1. Esquela-Kerscher A, Slack FJ (2006) Oncomirs model of sequence and structure. Nucleic
- microRNAs with a role in cancer. Nat Rev Acids Res 33(11):3570–3581
Cancer 6(4):259–269 4. Tijsterman M, Plasterk RH (2004) Dicers at
2. Small EM, Olson EN (2011) Pervasive roles of RISC; the mechanism of RNAi. Cell 117
microRNAs in cardiovascular biology. Nature (1):1–3
469(7330):336–342 5. Brennecke J et al (2005) Principles of
3. Nam JW et al (2005) Human microRNA pre- microRNA-target recognition. PLoS Biol 3
diction through a probabilistic co-learning (3):e85
6. Cui H, Zhai J, Ma C (2015) miRLocator: 21. Ding J, Zhou S, Guan J (2010) MiRenSVM:
machine learning-based prediction of mature towards better prediction of microRNA pre-
microRNAs within plant pre-miRNA cursors using an ensemble SVM classifier with
sequences. PLoS One 10(11):e0142753 multi-loop features. BMC Bioinformatics 11:
7. Sacar MD, Allmer J (2014) Machine learning S11
methods for microRNA gene prediction. 22. Raad J, Stegmayer G, Milone DH (2020)
Methods Mol Biol 1107:177–187 Complexity measures of the mature miRNA
8. Lim LP et al (2003) The microRNAs of Cae- for improving pre-miRNAs prediction. Bioin-
norhabditis elegans. Genes Dev 17 formatics 36(8):2319–2327
(8):991–1008 23. Levenshtein VI (1966) Binary codes capable of
9. Artzi S, Kiezun A, Shomron N (2008) miR- correcting deletions, insertions, and reversals.
NAminer: a tool for homologous microRNA Sov Phys Dokl 10(8):707–710
gene search. BMC Bioinformatics 9:39 24. Marques YB et al (2016) Mirnacle: machine
10. Nam JW et al (2006) ProMiR II: a web server learning with SMOTE and random forest for
for the probabilistic prediction of clustered, improving selectivity in pre-miRNA ab initio
nonclustered, conserved and nonconserved prediction. BMC Bioinformatics 17(Suppl
microRNAs. Nucleic Acids Res 34(Web Server 18):474
issue):W455–W458 25. Terai G et al (2012) Prediction of conserved
11. Huang TH et al (2007) MiRFinder: an precursors of miRNAs and their mature forms
improved approach and software implementa- by integrating position-specific structural fea-
tion for genome-wide fast microRNA precur- tures. PLoS One 7(9):e44314
sor scans. BMC Bioinformatics 8:341 26. He C et al (2012) MiRmat: mature microRNA
12. Zhang T et al (2019) miRLocator: a Python sequence prediction. PLoS One 7(12):e51673
implementation and web server for predicting 27. Xuan P et al (2011) MaturePred: efficient iden-
miRNAs from pre-miRNA sequences. Meth- tification of microRNAs within novel plant
ods Mol Biol 1932:89–97 pre-miRNAs. PLoS One 6(11):e27422
13. Tempel S, Tahi F (2012) A fast ab-initio 28. Hu X, Ma C, Zhou Y (2013) A novel two-layer
method for predicting miRNA precursors in SVM model in miRNA Drosha processing site
genomes. Nucleic Acids Res 40(11):e80 detection. BMC Syst Biol 7(Suppl 4):S4
14. Xue C et al (2005) Classification of real and 29. Sheng Y, Engstrom PG, Lenhard B (2007)
pseudo microRNA precursors using local Mammalian microRNA prediction through a
structure-sequence features and support vector support vector machine model of sequence
machine. BMC Bioinformatics 6:310 and structure. PLoS One 2(9):e946
15. Jiang P et al (2007) MiPred: classification of 30. Yousef M et al (2006) Combining multi-
real and pseudo microRNA precursors using species genomic data for microRNA identifica-
random forest prediction model with com- tion using a naive Bayes classifier. Bioinformat-
bined features. Nucleic Acids Res 35(Web ics 22(11):1325–1334
Server issue):W339–W344 31. Wang X et al (2005) MicroRNA identification
16. Wu Y et al (2011) MiRPara: a SVM-based soft- based on sequence and structure alignment.
ware tool for prediction of most probable Bioinformatics 21(18):3610–3614
microRNA coding regions in genome scale 32. Jones-Rhoades MW, Bartel DP (2004)
sequences. BMC Bioinformatics 12:107 Computational identification of plant micro-
17. Gkirtzou K et al (2010) MatureBayes: a proba- RNAs and their targets, including a stress-
bilistic algorithm for identifying the mature induced miRNA. Mol Cell 14(6):787–799
miRNA within novel precursors. PLoS One 5 33. Gao D et al (2013) miREval 2.0: a web tool for
(8):e11843 simple microRNA prediction in genome
18. Leclercq M, Diallo AB, Blanchette M (2013) sequences. Bioinformatics 29(24):3225–3226
Computational prediction of the localization of 34. Akgül B, Stadler PF, Hawkins LJ, Hadj-
microRNAs within their pre-miRNA. Nucleic Moussa H, Storey KB, Ergin K, Çetinkaya R,
Acids Res 41(15):7200–7211 Paschoal AR, Nachtigall PG, Tutar Y, Yousef
19. Li J et al (2015) MatPred: computational iden- M, Allmer J (2021) 44 Current challenges in
tification of mature microRNAs within novel miRNomics. In: Allmer J, Yousef M (eds) miR-
pre-MicroRNAs. Biomed Res Int Nomics: microRNA biology and computa-
2015:546763 tional analysis. Methods in molecular biology,
20. Yones CA et al (2015) miRNAfe: a comprehen- vol 2257. Springer, New York
sive tool for feature extraction in microRNA
prediction. Biosystems 138:1–5
Chapter 10
Computational Detection of MicroRNA Targets

Abstract
MicroRNAs (miRNAs) are small noncoding RNAs that are recognized as posttranscriptional regulators of
gene expression. These molecules have been shown to play important roles in several cellular processes.
MiRNAs act on their target by guiding the RISC complex and binding to the mRNA molecule. Thus, it is
recognized that the function of a miRNA is determined by the function of its target (s). By using high-
throughput methodologies, novel miRNAs are being identified, but their functions remain uncharted.
Target validation is crucial to properly understand the specific role of a miRNA in a cellular pathway.
However, molecular techniques for experimental validation of miRNA–target interaction are expensive,
time-consuming, laborious, and can be not accurate in inferring true interactions. Thus, accurate miRNA
target predictions are helpful to understand the functions of miRNAs. There are several algorithms
proposed for target prediction and databases containing miRNA-target information. However, these
available computational tools for prediction still generate a large number of false positives and fail to detect
a considerable number of true targets, which indicates the necessity of highly confident approaches to
identify bona fide miRNA–target interactions. This chapter focuses on tools and strategies used for miRNA
target prediction, by providing practical insights and outlooks.
Key words Noncoding RNA, miRNA recognition element, Target prediction tools, Bioinformatics,
Computational biology
1 Introduction
MicroRNAs (miRNAs) were first described in Caenorhabditis ele-

gans more than two decades ago [1, 2]. In 2000, the second
miRNA was described as a regulator of development in C. elegans,
and being spread across vertebrate species [3, 4]. Today, it is known
that miRNAs are small molecules of noncoding RNA (approxi-
mately 21 nts) that are loaded into the guide channel of Argonaute
proteins to form a silencing complex responsible for the regulation
of gene expression [5]. The miRNA genes are integrated as impor-
tant regulators to most of the biological processes in animals and
plants and are present in eukaryotic viruses [6–9]. MiRNAs have a
big impact by controlling the development, homeostasis of organs
and tissues, and response to environmental stimuli [10–
187
188 Pedro Gabriel Nachtigall and Luiz Augusto Bovolenta
12]. Moreover, it is well documented that many miRNAs show cell-

specific expression signatures [13], and any dysregulation at the
miRNA expression level can lead to disorders and diseases, such as
cancer [14, 15].
The combined advances in deep-sequencing approaches and
bioinformatics tools have enabled researchers to detect and quan-
tify the expression of miRNAs in several species, developmental
stages, diseases, and disease treatment responses (reviewed in
[16]). These efforts led to the 38,589 miRNA entries in miRBase
(release 22; [17]) related to the mature sequence of miRNAs from
more than one hundred species of animals, plants, and viruses.
Despite the high amount of data related to annotated miRNAs,
there are still thousands of miRNAs with targets and functions
uncharted, which reveals the necessity of precise and accurate
computational approaches for target prediction of miRNAs.
The function of a miRNA is defined by its target genes and the
effective regulation of their expression [18]. Computational pre-
dictions estimate that in some cases one miRNA regulates hundreds
of distinct targets and that commonly one target can be regulated
by one or more miRNAs, suggesting that a large number of genes
(around 60% in humans) are under the control of the miRNA
regulatory mechanism [19]. In animals, miRNAs recognize their
targets by binding preferentially at the 30 UTR sequence of mRNAs
[20], but functional binding sites were also found in the coding
domain sequence (CDS) and 50 UTR [21–23]. The core of the
silencing process via miRNAs is based on protein abundance, which
can be indirect, by degradation of the mRNA, or direct, by repres-
sion of the translation, or a combinatorial effect of both processes
[24–26]. Four distinct ways of translational repression have been
proposed so far: (a) inhibition of translation initiation;
(b) inhibition of translation elongation; (c) co-translational protein
degradation; and (d) premature termination of translation
(reviewed by [27]). Independently, the silencing mechanism relies
on the interaction between miRNA and its target, which includes
miRNAs in the precise fine-tuning of gene expression, and inte-
grates them into the functionality of cell biology. In this sense,
the identification of miRNA–mRNA interactions is the key to
define the functions of miRNAs in the context of complex
biological regulatory networks orchestrated by these small
molecules.
1.1 Target MicroRNAs interact with their target mRNAs through Watson–
Recognition Features Crick base pairing involving A:U, G:C, and also G:U pairs. The
target recognition is different between plants and animals (Fig. 1).
In plants, many gene targets contain high complementarity with
the miRNA [28, 29], however, the simple pairing process could be
considered a suboptimal parameter in the prediction (reviewed in
[30]), once some studies showed that partial complementarity can
Predicting MicroRNA Targets 189
Fig. 1 MiRNA target recognition in animals (top) and plants (bottom)
be also functional [29, 31], and other aspects could add relevance
in the determination of the target interaction such as mismatches,
position-specific effects, G-U wobbles, and bulges which offer
stronger effects at some positions than others [32]. In animals,
most of the interactions are based on an imperfect complementarity
between the mature sequence and the miRNA recognition element
(MRE) at the mRNA sequence. The essential region of the miRNA
for target recognition and binding relies on the “seed” sequence,
which is characterized by seven nucleotides (from 2 to 8) at the 50
end of the miRNA [19, 33]. This pairing between the seed and
MRE can be classified into five types, named as 8-mer, 7-mer-m8,
7-mer-A1, 6-mer, and offset-6-mer ( [34]; Fig. 2). Although not
usually considered in target prediction, a recent report showed that
the pairing type 5-mer can directly influence the target gene abun-
dance in specific cases [35]. Added to the seed pairing, several
studies reported effectiveness regulation of miRNAs through non-
perfect seed interactions (also known as noncanonical types), which
showed that the 30 region of the miRNA should be relevant for
confident target recognition ( [36–39]; Fig. 3).
In addition to the essential Watson–Crick pairings between the
miRNA and the MRE, there is a set of miRNA recognition features
considered for the identification of bona fide miRNA–target inter-
actions that are based on three main parameters: duplex features,
local context features, and global context features [40].
The duplex features, which evaluate the binding properties
between the miRNA and its target, include seed match, seed pairing
Fig. 2 Pairing types between miRNA and binding site at mRNA target. (a) offset-6-mer and (b) 6-mer are
perfect matches for nts 3–8 and 2–7, respectively; (c) 7-mer-A1 is a perfect match for nts 2–7, with an
stability (SPS), 30 contribution, hetero-duplex free energy, and

p-value [40, 41]. Seed match assesses the type of seed pairing,
whereas the SPS evaluates the nucleotide composition of the seed
[42]. The 30 contribution evaluates the binding property of the
30 UTR region containing the MRE of the miRNA [43], which can
improve the binding strength between miRNA and target by com-
pensatory pairing [19, 20]. The hetero-duplex free energy mea-
sures if the minimum free energy between the MRE and the
miRNA is relatively strong for a positive hybridization [44]. The
p-value indicates the probability of the predicted interaction to be a
random event.
The local context features consider properties of the mRNA
sequence that has a direct influence on the binding efficacy with the
miRNA, which encompasses site accessibility and the existence of a
flanking region rich in AU. The site accessibility evaluates the self-
Fig. 3 Noncanonical pairing types. The 30 compensatory type (top) that presents a strong miRNA–target
complementarity at the 30 region of the miRNA, which compensates an imperfect seed match (Friedman et al.,
2009). The centered type (bottom) that presents a perfect pairing of 11–12 bases to miRNA positions 4–15
[34]
Fig. 2 (continued) adenine at relative nt 1 in the mRNA; (d) 7-mer-m8 is a perfect match for nts 2–8; (e) 8-mer
is a perfect match for nts 2–8, with an adenine at relative nt 1 in the mRNA; The adenine at relative nt position
1 of the mRNA supports the efficacy of miRNA regulation, even when the opposite nt does not form a
Watson–Crick pairing [25]
folding potential of the region containing the MRE to form a

strong secondary structure that can disrupt the binding feasibility
with the miRNA [45]. The flanking AU measures the richness of A
and U nucleotides around the MRE site. It is known that a flanking
region rich in A and U nucleotides enhance the regulatory effect of
the miRNA [46].
Global context features take into consideration properties that
can act indirectly on the target recognition, such as transcript
length, 30 UTR length, pairing position at 30 UTR, conservation,
and abundance of the target in the transcriptome. The total length
of the transcript analyzed can increase the chance of a false predic-
tion, therefore, the transcript length should be evaluated to
improve the prediction accuracy [41]. Moreover, the 30 UTR
length measures the potential that a miRNA has to bind in the
MRE, once larger 30 UTRs were reported to be under a stringency
regulation than shorter 30 UTRs [47]. The pairing position evalu-
ates the MRE position in a 30 UTR context, which takes into con-
sideration that efficient and strong regulatory miRNA binding sites
are located near to the start and/or end of the 30 UTR [46]. Con-
servation is based on the extent of similarity of MREs (mainly the
seed region) at the 30 UTR among several species [20, 48]. The
transcriptome abundance is related to the number of MREs along
with all transcripts in the transcriptome, which can reduce or
enhance the regulatory function of the miRNA on a specific target
[42, 49]. Actually, not all parameters responsible for the target
recognition are known and new features are being reported
together with the advances of molecular tools and bioinformatics
approaches [50].
In this sense, all of the parameters employed together can
potentially improve and refine the determination of highly confi-
dent miRNA–target interactions in a regulatory context.
1.2 Target Prediction Interestingly, the sequence analysis of the first miRNA target vali-
Tools dated lead to the hypothesis that the miRNA–target interaction
occurs under the pressure of pattern based on pairing complemen-
tarity [1, 2]. Which led to the development of several computa-
tional tools aiming to predict miRNA targets (reviewed in [51]).
These developed methodologies are based on a distinct set of
multiple parameters and strategies to increase target prediction
efficacy, which leads to similarities and singularities of each tool.
Furthermore, these programs commonly use specific scores to
weigh and measure the probability of real miRNA–target interac-
tion. A summary of some of the previously published miRNA target
prediction tools is listed in Table 1.
Many predictors were developed for being used alone as high-
performance programs and became popular along time ago, such as
TargetScan, miRanda, RNA22, RNAhybrid, PicTar, and others.
TargetScan [52] is one of the most popular target prediction
Table 1
A non-comprehensive list of miRNA target prediction tools
mRNA region
coverage
Web Source CD Last

Tool server code Database Speciesa 50 UTR S 30 UTR Version update Website Reference
TargetScan X X X h, c, rh, m, r, d, – – X v7.2 2018 targetscan.org [34, 52]
cw, o, ch, fr,
z, f, w
miRanda X X – h, m, r, f, w – – X v3.3a 2010 microrna.org [53]
PITA X – – h, m, f, w – – X v6.0 2008 genie.weizmann.ac.il [45]
RNA22 X – – h, m, f, w X X X v2.0 2015 cm.jefferson.edu/rna22 [41]
RNAhybrid X X – f, w – – X v2.1.2 2006 bibiserv2.cebitec.uni-bielefeld.de/ [44]
rnahybrid
PicTar X – – h, m, f, w – – X v1.0 2007 pictar.mdc-berlin.de [54]
poptargs X – X h – – X v2.0 2019 poptargs.essex.ac.uk [55]
maTE – X – Any species – – X v1.0 2019 malikyousef.com/matemate- [56]
discovering-expressed-microrna-
target-interactions/
FilTar – X – Animal species – – X v1.2.3 2019 github.com/TBradley27/FilTar [57]
isoTar X – – h – – X v1.1 2019 ncrnaome.osumc.edu/isotar/ [58]
BRM X – – h, rh, m, r, z – – X v2.3 2019 cbb.pnnl.gov/brm/ [59]
dbMTS – – X h – – X v4.0 2019 sites.google.com/site/jpopgen/ [60]
dbNSFP
Predicting MicroRNA Targets
miRDB – – X h, m, r, d, ch – – X v5.0 2019 mirdb.org [61, 62]

miRFA X X – h – – X v1.0 2019 github.com/emmbor/miRFA [63]
193
(continued)
194
Table 1
(continued)
mRNA region
coverage
Web Source CD Last

Tool server code Database Speciesa 50 UTR S 30 UTR Version update Website Reference
SeedVicious X X X h, c, m, r, f, w – – X v1.1 2018 seedvicious.essex.ac.uk/ [64]
miRAW – X – h – – X v1.0 2018 bitbucket.org/account/user/bipous/ [65]
projects/MIRAW
miRTPred – X – h – – X v1.0 2018 bicresources.jcbose.ac.in/zhumur/ [66]
mirtpred
Avishkar – X – h – – X v1.0 2018 bitbucket.org/cellsandmachines/ [67]
avishkar
mirDIP X – – h – – X v4.1 2018 ophid.utoronto.ca/mirDIP/ [68]
miRNet X – – h, m, r, cw, ch, – – X v1.0 2018 mirnet.ca [69]
z, f, w
TarBase – – X Animal species – – X v8.0 2018 carolina.imis.athena-innovation.gr/ [70]

diana_tools/web/index.php?
r¼tarbasev8%2Findex
miRTarBase – – X Animal species – – X v7.0 2017 miRTarBase.mbc.nctu.edu.tw/ [71]
miRTarVis+ X – – h – – X v1.0 2017 hcil.snu.ac.kr/research/mirtarvisplus [72]
metaMIR X X – h – – X v1.1.0 2017 rna.informatik.uni-freiburg.de/ [73]
metaMIR/
miRabel – – X h – – X v1.0 2017 bioinfo.univ-rouen.fr/mirabel/ [74]
miRWalk X – X h, m, r, d, cw X X X v3.0 2017 mirwalk.umm.uni-heidelberg.de/ [75]
TarPmiR – X – h, m – – X v1.0 2016 hulab.ucf.edu/research/projects/ [50]
miRNA/TarPmiR/
miSTAR X – X h – – X v0.66 2016 mi-star.org [76]
miRTarget – – X h, m, r, d, ch – – X v3.0 2016 mirdb.org [77]
MBStar X X – h – – X v1.0 2015 isical.ac.in/~bioinfo_miu/MBStar30. [78]
htm
StarScan X – – h, c, m, r, d, ch, – – X v1.0 2015 mirlab.sysu.edu.cn/starscan/ [79]
rh, cw, o, fr, z,
f, w
Cupid – X – h, c, rh, m, r, d, – – X v1.0 2015 cupidtool.sourceforge.net/ [80]
cw, o, ch, fr,
z, f, w
miRHunt – X – h – – X v0.4 2015 castelli-lab.net/apps/apps_mirhunt. [81]
php
miRPath X – X h, m, r, ch, z, f, X X X v.3.0 2015 snf-515788.vm.okeanos.grnet.gr/ [82]
w
DIANA- X – X h, m, f, w – X X v5.0 2014 diana.imis.athena-innovation.gr/ [83]
microT- DianaTools/
CDS
VIRmiRNA – – X Virus – – X v1.0 2014 crdd.osdd.net/servers/virmirna/ [84]
StarBase – – X h, m, w – – X v2.0 2013 starbase.sysu.edu.cn/ [85, 86]
psRNATarget X – X Plant species X X X v2.0 2017 plantgrn.noble.org/psRNATarget/ [87, 88]
Targetfinder – X – Plant species X X X v1.0 2016 github.com/carringtonlab/ [89]
TargetFinder
Tapir X X X Plant species X X X V1.0 2016 bioinformatics.psb.ugent.be/ [90]
webtools/tapir/
github.com/eb00/tapir
Predicting MicroRNA Targets
a
h human, c chimpanzee, rh rhesus, m mouse, r rat, d dog, cw cow, o opossum, ch chicken, fr frog, z zebrafish, f fly, w worm
195
tools. The last update of TargetScan (v7.2, March 2018) uses

16 features considered relevant for target recognition, such as
seed match, 30 contribution, SPS, SA, flanking AU, TA, pairing
position, 30 UTR length, conservation, and many others [34]. Its
recent score model, named as “context++”, is being considered a
powerful approach to help researchers validate miRNAs placement
in their regulatory networks of vertebrates. In this sense, TargetS-
can has the highest number of species available for the analysis, due
to the conservation analysis, and has noncanonical pairing interac-
tions predicted for human and mouse species (i.e., the 30 compen-
satory and centered types). Furthermore, TargetScan presents the
30 UTR profile with the distinct isoforms produced from multiple
cell lines or tissues generated by the 3P-seq methodology [34, 91,
92] for human and mouse species, which permits the user to
consider specific MREs under certain conditions available in the
TargetScan data. The website is user-friendly and permits rapid
access to the precomputed predictions with scores for miRNAs
and their binding sites in specific genes and/or to check the puta-
tive list of target genes. Moreover, all data used for the analysis and
the precomputed predictions are accessible, and the source code is
freely available, being easily applicable for stand-alone use.
The miRanda algorithm performs the MRE recognition by the
use of features such as seed match, 30 contribution, SA, flanking
AU, pairing position, 30 UTR length, and conservation [53]. Fur-
thermore, the last update empowered the approach by the use of
the mirSVR score (miRanda-mirSVR) to rank potential interac-
tions, which indicates the strength of the regulatory effect
[40]. Moreover, the mirSVR score helped to increase the MREs
identified at noncanonical and nonconserved sites. On the website,
the user can access the list of predicted targets of a specific miRNA
and/or a set of miRNAs binding to a specific gene. Precomputed
predictions are available, but the last update is based on miRBase
release 16 (2010 August). The source code is available for use, but
the mirSVR score is not applicable when performing a stand-alone
prediction.
PITA uses five main parameters for target prediction [45]. It
was last updated in August 2008, and the precomputed results are
related to miRNAs data of miRBase release 11.0 (April 2008). The
online version permits to change the parameter cutoffs and enables
the predictions on specific miRNAs and 30 UTR sequences
uploaded by the user. The source code is available for a stand-
alone application.
RNA22 focuses on the entire sequence of the mRNA (i.e., 50
UTR, CDS, and 30 UTR) to search for matches [41]. Despite the
main features used by this tool, the main differences of targets
predicted by RNA22 can be attributed to the use of CDS and 50
UTR sequences in the prediction [93]. The tool is non-intuitive
and the user can provide the 30 UTR sequence and the miRNA
sequence and/or download the precomputed predictions. Further-

more, the source code can be downloaded for stand-alone use.
RNAhybrid was developed to search for MREs in target
sequences mainly based on the minimum free energy at the
miRNA–mRNA duplexes [44]. In this sense, the program ranks
the result by featuring seed match, heteroduplex free energy, and
p-value. However, taken into consideration solely the minimum
free energy as a parameter can lead to erroneous predictions for
miRNAs and MREs that are rich in AU composition, due to the
high value of minimum free energy resulting from the high amount
of A-U pairings. This tool has no precomputed predictions and
requires the input of the miRNA and target sequences provided by
the user, which can set the parameters to change the stringency of
analysis. It was last updated in 2006 but is used to calculate the
minimum free energy of the binding site to check for the stability
and strength involved in the interaction.
PicTar [54] is a popular program, which performs the predic-
tion task regarding the conservation of MREs across multiple spe-
cies. It also takes into consideration the cooperation of multiple
miRNAs and the spatiotemporal coexpression between miRNA and
their potential targets. The source code is not available, but the
precomputed predictions can be freely accessed. Unfortunately,
PicTar is out of date, due to no updates since the last release in
March 2007, and all the data available for use is based on miRBase
release 9.1 (2007 February). Thus, no miRNAs detected after this
release are included in the set of predictions.
Many other prediction tools are being developed with singular
features that can be applied to complement the target prediction
task by giving new insights into the miRNA–target interactions.
One example is SeedVicious [64]. This tool allows the analysis of
near-target sites [64], which can be helpful for population analysis
aiming to find differences in the miRNA–target interactions among
individuals and can be a valuable resource in evolutionary studies.
Two applications take into consideration the expression of the
target at the tissue/cell being analyzed. The FilTar [57] tool filters
the mRNA targets that are not expressed in the transcriptome of
the cell/tissue of interests and attempts to improve the 30 UTR
annotation for that specific dataset. However, FilTar does not
consider the miRNA expression data. On the other hand, the
maTE algorithm [56] considers the target and miRNA expression
profiles in two conditions, then outputs a list of deregulated genes
clustered by their targeting miRNAs.
Interestingly, some predictors were designed with the use of
machine learning (ML) technologies [94–96]. These programs
take into consideration validated interactions obtained through
published experiments to train the algorithm for the target predic-
tion task. Some of these tools were developed based on supervised
learning approaches such as miRTPred [66], Avishkar [67],
metaMIR [73], MBStar [78], NBmiRTar [97], and many others.

Some research groups developed ML-based algorithms to use deep
learning methodologies, such as miRAW [65], miRTDL [98] and
deepTarget [99], which are based on deep neural networks, con-
volutional neural networks, and recurrent neural networks, respec-
tively. Whereas the TarPmiR algorithm [50] uses four distinct ML
methods together to improve the target prediction based on pub-
lished CLASH (cross-linking ligation and sequencing of hybrids)
data from human and mouse species. Despite the ML approach
used for each tool, these ML-based predictors need a high amount
of validated targets to be improved with high accuracy. However,
considering that validated interactions are being continuously
reported, the ML-based approaches are a valuable resource with
big opportunities for future improvement.
Some applications were developed to attempt to find bona fide
interactions by considering the results from multiple prediction
tools, such as miRSystem [100], MIRDIP [101], ComiR [102],
miRabel [74], miRTarVis+ [72], and many others. These programs
scan the results obtained by distinct predictors and show the puta-
tive interactions identified by one or more tools. These approaches
can give robustness to the binding sites detected and allow the user
to easily combine predictions without programming skills.
In addition to prediction tools, there are diverse open-access
web servers and databases containing a list of putative and validated
targets by in silico prediction and experimental analyses. These web
servers scan published and experimental evidence of the regulatory
effect of miRNAs over their targets and take into consideration the
molecular methodology applied for the target validation. One
example is the DIANA tools web server, which contains the
DIANA-microT-CDS tool focused on target prediction and func-
tional analysis of miRNAs [83]. This web server has a user-friendly
interface and provides manuals and protocols for new users. The
last version (v5.0) is based on miRBase release 18, but considers
only four representative species of animals (Homo sapiens, Mus
musculus, Drosophila melanogaster, and C. elegans). Moreover,
this server integrates the DIANA-TarBase, which contains more
than 3000 miRNA–target interactions retrieved from published
articles based on several experimental techniques commonly used
to validate these interactions [70]. Another example relies on miR-
TarBase, which is a database containing validated miRNA–target
interactions [71]. It is a database with comprehensive annotation of
experimentally validated miRNA–target interactions with more
than 300 thousand such interactions identified by high-throughput
assays (next-generation sequencing experiments) and more than
seven thousand detected with strong evidence by reporter assays
and western blot. The web server is user-friendly and permits the
user to look at the binding sites and check the interactions among
miRNAs and their targets, and the expression profile of miRNAs
and target genes in multiple tissues and cells are available from
Gene Expression Omnibus (GEO) and The Cancer Genome Atlas
(TCGA). It was recently updated (v7.0, released in September
2017 uses data from miRBase release 21) and comprises more
than 20 species. Furthermore, the user can customize the search
by selecting the experimental methodology applied for the valida-
tion and several other parameters. There are other good examples
of web servers and databases:
l miRWalk [75, 103], which stores predicted data obtained
through the use of ML algorithms and includes experimentally
verified interactions miRNA–target interactions.
l The integrative web server StarBase2 [86], which was designed
to detect interaction networks of ncRNAs, including miRNAs,
through retrieved data from 108 experiments from CLIP-seq
technology.
l miRecords [104], which is a useful application to scan miRNA
binding sites in multiple animal species, whereas their results are
composed by the combinatorial analysis of manually curated
data of validated targets and target prediction data of eleven
different algorithms.
l poptargs [55], which is a tool designed for studying the popula-
tion genetics of human miRNA target sites.
l isoTar [58], which is a web-based application focused on
performing consensus miRNA targeting prediction and func-
tional enrichment analysis.
l Many other tools (Table 1).
All of these approaches attempt to find with high accuracy a list
of bona fide interactions. Despite the design and development of
such databases and applications, the constant reduction in the costs
of high-throughput technologies developed to validate miRNA–
target interaction on a large-scale, these programs are still in need of
more frequent updates to help researchers to elucidate the func-
tional roles of miRNAs and implement the knowledge around the
features of miRNA-target recognition.
1.3 Issues of Target Although the predictors are in constant evolution, there are some
Prediction Algorithms issues regarding the predictions, such as the high level of low
precision, low sensitivity, and coexpression of miRNAs and target
genes, that need to be addressed with caution when performing
downstream analysis based on the list of predicted targets.
It is well documented that the estimations of the false-positive
rates of some target predictors are around 50% [20, 52, 54]. A
recent study revealed a high amount of false positives in predictions
[105]. The analysis, from Pinzón and colleagues, considered a list
of 196 conserved targets predicted to miR-223. By using a single
cell type as the model (i.e., mature neutrophils) and microarray
data, they showed that only 40 targets (from 196) presented a

differential expression that can be related to miR-223 modulation.
Moreover, their findings suggested that the majority portion of
targets identified by these programs are not real targets, but they
can act as a modulatory mechanism on the free and bound con-
centrations of miRNA, which can fluctuate the efficiency of miRNA
activity. However, further analysis of additional miRNAs and cell
types is necessary to expand it as a general conclusion.
A comparison study of eight target prediction methodologies
demonstrated that even highly sensitive programs failed to recog-
nize a high amount of validated target genes [106]. Recent research
comparing four commonly used tools showed, by using a high-
quality dataset of validated targets, that around 18% of the true
targets for the miRNAs considered were not detected by any tool
[93]. Despite the continuous updating applied in some tools to
improve the prediction quality, the actual status of predictors still
needs to improve sensitivity to avoid the missing of true targets.
However, the true targets not detected by any predictor could be
explained by the fact that all features for target recognition were not
described yet, which indicates that these true targets that were not
detected by any tool can be on the rules of these features still
uncharted.
A key point to analyze miRNA–target interaction relies on the
biological characteristics of miRNAs related to the temporal
and/or cell type-specific expression [13, 107]. Then, to find poten-
tial miRNA–target interactions, it is necessary to consider the coex-
pression of miRNA and mRNA target with respect to space and
time. In general, the predictors do not attempt to resolve this issue
in their analysis, and checking the coexpression of miRNA and
mRNA by using available data or personal data (e.g., transcriptome,
microarray, or qPCR data) is highly recommended.
2 Practical Insight
Besides the improvement and upgrading of prediction tools avail-

able today, there is no consensus about the best methodology
and/or strategy to search for bona fide targets. Although several
studies commonly use the intersection or single tools whereas few
studies use the union method, none of them compared and
described the best workflow.
A recent study published by Oliveira et al. [93] focused to
compare the performance of widely used prediction target tools in
the literature (TargetScan, miRanda, Pita, and RNA22) by consid-
ering the strategies of single, union, and intersection. The study
used a dataset of 10 miRNAs and 1400 mixed genes of 700 vali-
dated and 700 nonvalidated. The tools analyzed showed a high-
quality performance by returning a few false-positive targets
predicted, but they did not detect several true targets (~18%). By
analyzing the intersection and union, they showed that the inter-
section reduces significantly the performance, whereas the union
improves the performance by increasing significantly the sensitivity
with a low decrease in specificity. In summary, the authors con-
cluded that the union of the considered tools is the best strategy to
improve the performance by increasing the sensitivity, whereas the
use of tools alone is better to improve the specificity.
Taking into account the actual status of prediction strategies
performance, we suggest to researchers to use the union approach
to obtain high-quality predictions for global analysis of miRNA–
target interactions and to use a single tool to find specific miRNA–
target interactions. These suggestions can help to minimize the
efforts and loss of resources and time in downstream functional
experiments for validation of specific miRNA–target interaction by
using in vitro and/or in vivo models.
2.1 30 UTR Sequence To perform target prediction in a specific dataset, next to the
Availability miRNA sequence, the mRNA sequence is necessary, especially the
30 UTR sequence. As discussed above, although there is evidence
for binding sites along the entire mRNA sequence (50 UTR, CDS,
and 30 UTR), most of the programs take into consideration only
the 30 UTR sequences to search for putative miRNA–target inter-
action, due to the high amount of validated targets within the 30
UTR. Then, 30 UTRs can be considered crucial for prediction, since
the absence of 30 UTR region in the analysis must impair the
efficiency of the in silico predictions of a miRNA anchoring on a
specific MRE.
The 30 UTR sequences and coordinates can be obtained for
distinct species from several databases (Ensembl, NCBI, UCSC
Genome browser, etc.). The most user-friendly and commonly
used database for obtaining 30 UTR annotated data is the Ensembl
(release 93) by using the BioMart tool (see Fig. 4 for more details).
Even with the large availability of species, most of them have a poor
annotation of around 15% do not have 30 UTRs annotated. In this
sense, there is a database directed to the UTR annotation, named
UTRdb ( [108], last release in May 2015). This is a manually
curated database of 50 and 30 UTR sequences of eukaryotic
mRNAs with an actual availability of 5,987,587 50 UTR and
6,270,325 30 UTR for 3,661,773 and 3,969,632 genes, respec-
tively, from 395 species. Then, the UTRdb is being used as a tool to
improve and help elucidate the 30 UTR sequences for several pur-
poses, including miRNA target prediction.
When researchers work with model species, such as human,
mice, worm, or fruit fly, we notice the existence of a large amount
of 30 UTR annotation with a high coverage depth of the genes
identified in these species. However, when using nonmodel species
in the research, the poor annotation and difficulty to acquire
30 UTR sequences is a critical problem for the analysis.
Fig. 4 Summary of 30 UTR annotation on Ensembl database (release 93). (a)

Number of protein-coding genes with and without 30 UTR annotated. (b) Length of
30 UTR annotated
Some computational tools were developed to assist in the

prediction and annotation of 30 UTRs, such as 3USS [109], UTRs-
can [110], and GETUTR [111]. Although both depend on high
coverage and well-annotated genomes and use RNA-seq data to
retrieve the 30 UTR sequence, which makes it difficult to study
nonmodel species without previous data or with unknown ancestry.
New applications are emerging to address this issue. One example is
ExUTR [112], which is a pipeline developed for obtaining 30 UTR
through prediction from RNA-seq data by using De novo or
Reference-based assembly. Then, it can be applied to RNA-seq
data independently of the existence of high or poorly assembled
genomes of the relevant species. In this sense, ExUTR covers model
and nonmodel species. However, all methods available for retriev-
ing 30 UTR are dependent on previously available
transcriptome data.
Recently, a molecular methodology was developed for sequenc-
ing the entire 30 UTR present in RNA samples, it is named 3P-seq
(Poly(A)-Position Profiling by Sequencing). It has been used to the
characterization of the 30 UTR profile by discovering the isoforms
of 30 UTR being expressed at a specific tissue and/or developmen-
tal timing [91, 113–115]. Moreover, it is known that 30 UTR iso-
forms can directly affect the regulatory function of miRNAs by
changing the MREs available for the interactions [116, 117],
which indicates that the 30 UTR characterization is crucial for
miRNA-target analysis. The 3P-seq is a reliable methodology for
30 UTR identification and isoforms discoveries, however, it needs
large amounts of RNA, it is technically laborious to perform, and it
is still relatively expensive for such a specific purpose.
Then, a challenging task is to expand the 30 UTR annotation by
the use of high throughput sequencing techniques and the experi-
mental validation of the 30 UTR computationally predicted data.
Recent studies showed that MREs presented in a particular 30 UTR,
may not be available for the miRNA–mRNA interaction. This could
be due to the difference and preference of different sizes of 30 UTRs
expressed among different tissue types and developmental timing
[34, 92, 118], which may explain the absence or presence of action
of miRNAs on specific MREs. The availability of 30 UTR data in
different cell types, developmental stages, and distinct species can
be considered a critical point for the improvement of target predic-
tion tools for model and nonmodel organisms.
3 Discussion and Outlook
The current computational prediction tools are heterogeneous,

both in target recognition features and performance, and all have
to be improved for an accurate and precise target screening.
Although the algorithms are evolving in parallel with increasing
knowledge about miRNAs and their functions, not all tools are
regularly updated. The constant upgrade policy is an important
issue since the rules of target recognition are still being elucidated,
novel miRNAs are reported frequently, and validated targets are
being published every year. Moreover, the actual approaches have
low specificity and low sensitivity and fail to detect all validated
targets, which indicates that high confidence predictions are still on
demand.
Much is known about the features for target recognition, but
the actual knowledge is still limited, but the forthcoming under-
standing from the constant growth of validated targets by reporter
and high-throughput assays will help to elucidate all rules of
miRNA–target interactions, which will improve the performance
for screening bona fide interactions.
The targets predicted by molecular approaches are often used
to infer biological functions for miRNAs. However, the actual
status is concerned with high false-positive rates [105] and low
sensitivity [93]. This implies that functional inferences can suffer
bias through contamination by false and missing targets. Indicating
that the downstream analysis using the predictions need to be
performed with caution to avoid erroneous conclusions.
Target validation, by using molecular approaches, is required to
uncover the functionality and activity of a particular miRNA in a
biological pathway, once predictors only generate a putative targets
list. Despite the fact that large-scale methodologies (microarray,
degradome-seq, transcriptome, etc.) are decreasing in cost, they
still have some limitations on validating targets when compared to
low-throughput experiments (e.g., reporter assay). However, the
low-throughput validation techniques are very expensive, labori-
ous, and time-consuming. In this sense, computational predictions
are very helpful for researchers.
To summarize, computational target prediction programs are
widely used in literature by considering several distinct programs
and strategies, but their performance and output are quite different
from each other. Thus, the list of potential targets generated by
computational tools should be used with wisdom.
4 Conclusion
Since the discovery of miRNAs and target recognition in 1993,

there was an explosion in the number of validated targets by molec-
ular approaches that were supported by the efforts in the constant
development of computational target prediction algorithms. Albeit
the predictors are still lacking performance, they are valuable
resources for the investigation of miRNA functional roles by help-
ing to elucidate the regulatory activity of miRNAs on the protein
output of new targets. Considering that miRNAs and their binding
sites are the collaborative keys for the fine-tuning of the protein
abundance, any putative target site should be carefully taken into
consideration to infer biological functionality or by validation
experiments. Despite the improvement of prediction algorithms
and strategies in the last decades, we can conclude that there is
still a lot of work to be done to reach the gold standard of the
computational miRNA target prediction.
Acknowledgments
This work was supported by grants from São Paulo Research Foun-
dation (FAPESP - processes numbers: 2013/06864-7; 2018/
26520-4), CAPES (process number 88887.177457/2018-00),
and The Brazilian National Council for Scientific and Technologi-
cal Development (CNPq—process number 167444/2017-4).
References
1. Lee RC, Feinbaum RL, Ambros V (1993) development and disease. Dev Cell 11
The C. elegans heterochronic gene lin-4 (4):441–450
encodes small RNAs with antisense comple- 11. Qiu C, Chen G, Cui Q (2012) Towards the
mentarity to lin-14. Cell 75(5):843–854 understanding of microRNA and environ-
2. Wightman B, Ha I, Ruvkun G (1993) Post mental factor interactions and their relation-
transcriptional regulation of the heterochro- ships to human diseases. Sci Rep 2:318
nic gene lin-14 by lin-4 mediates temporal 12. Nachtigall PG, Dias MC, Carvalho RF et al
pattern formation in C. elegans. Cell 75 (2015) MicroRNA-499 expression distinc-
(5):855–862 tively correlates to target genes sox6 and
3. Reinhart BJ, Slack FJ, Basson M et al (2000) rod1 pro-files to resolve the skeletal muscle
The 21-nucleotide let-7 RNA regulates devel- phenotype in nile tilapia. PLoS One 10(3):
opmental timing in Caenorhabditis elegans. e0119804
Nature 403(6772):901–906 13. Sood P, Krek A, Zavolan M et al (2006) Cell-
4. Pasquinelli AE, Reinhart BJ, Slack F et al type-specific signatures of microRNAs on tar-
(2000) Conservation of the sequence and get mRNA expression. Proc Natl Acad Sci U S
temporal expression of let-7 heterochronic A 103(8):2746–2751
regulatory RNA. Nature 408(6808):86–89 14. Im H-I, Kenny PJ (2012) MicroRNAs in neu-
5. Bartel DP (2004) MicroRNAs: genomics, ronal function and dysfunction. Trends Neu-
biogenesis, mechanism, and function. Cell rosci 35(5):325–334
116(2):281–297 15. Lujambio A, Lowe SW (2012) The microcos-
6. Sunkar R, Girke T, Jain PK et al (2005) Clon- mos of cancer. Nature 482(7385):347–355
ing and characterization of microRNAs from 16. Berezikov E (2011) Evolution of microRNA
rice. Plant Cell 17(5):1397–1411; diversity and regulation in animals. Nat Rev
ISSN1040-4651 Genet 12(12):846–860
7. Jia W, Li Z, Lun Z (2008) Discoveries and 17. Kozomara A, Griffiths-Jones S (2014) miR-
functions of virus-encoded microRNAs. Chin Base: annotating high confidence microRNAs
Sci Bull 53(2):169–177; ISSN1001-6538 using deep sequencing data. Nucleic Acids
8. Taylor RS, Tarver JE, Hiscock SJ et al (2014) Res 42(Database issue):D68–D73
Evolutionary history of plant microRNAs. 18. Jonas S, Izaurralde E (2015) Towards a
Trends Plant Sci 19(3):175–182 molecular understanding of microRNA-
9. Bartel DP (2018) Metazoan microRNAs. Cell mediated gene silencing. Nat Rev Genet 16
173(1):20–51; ISSN00928674 (7):421–433
10. Kloosterman WP, Plasterk RHA (2006) The
diverse functions of microRNAs in animal
19. Bartel DP (2009) MicroRNAs: target recog- heterochromatic siRNAs. Curr Opin Plant
nition and regulatory functions. Cell 136 Biol 27:118–124
(2):215–233 33. Lewis BP, Shih I-h, Jones-Rhoades MW et al
20. Friedman RC, Farh KK-H, Burge CB et al (2003) Prediction of mammalian microRNA
(2009) Most mammalian mRNAs are con- targets. Cell 115(7):787–798; ISSN 0092-
served targets of microRNAs. Genome Res 8674
19(1):92–105 34. Agarwal V, Bell GW, Nam J-W et al (2015)
21. Lytle JR, Yario TA, Steitz JA (2007) Target Predicting effective microRNA target sites in
mRNAs are repressed as efficiently by mammalian mRNAs. elife 4:e05005
microRNA-binding sites in the 50 UTR as in 35. Hart M, Kern F, Backes C et al (2018) The
the 30 UTR. Proc Natl Acad Sci U S A 104 deterministic role of 5-mers in microRNA-
(23):9667–9672 gene targeting. RNA Biol 15(6):819–825
22. Forman JJ, Legesse-Miller A, Coller HA 36. Tay Y, Zhang J, Thomson AM et al (2008)
(2008) A search for conserved sequences in MicroRNAs to nanog, oct4 and sox2 coding
coding regions reveals that the let-7 micro- regions modulate embryonic stem cell differ-
RNA targets dicer within its coding sequence. entiation. Nature 455(7216):1124–1128;
Proc Natl Acad Sci U S A 105 ISSN 1476-4687
(39):14879–14884 37. Nelson PT, Wang W-X, Mao G et al (2011)
23. Brümmer A, Hausser J (2014) MicroRNA Specific sequence determinants of miR-15/
binding sites in the coding region of 107 microRNA gene group targets. Nucleic
mRNAs: extending the repertoire of post- Acids Res 39(18):8163–8172
transcriptionalgene regulation. Bioessays 36 38. Chi SW, Hannon GJ, Darnell RB (2012) An
(6):617–626 alternative mode of microRNA target recog-
24. Lim LP, Lau NC, Garrett-Engele P et al nition. Nat Struct Mol Biol 19(3):321–327
(2005) Microarray analysis shows that some 39. Broughton JP, Lovci MT, Huang JL et al
microRNAs downregulate large numbers of (2016) Pairing beyond the seed supports
target mRNAs. Nature 433(7027):769–773; MicroRNA targeting specificity. Mol Cell 64
ISSN 1476-4687 (2):320–333
25. Baek D, Vill én J, Shin C et al (2008) The 40. Betel D, Koppal A, Agius P et al (2010) Com-
impact of microRNAs on protein output. prehensive modeling of microRNA targets
Nature 455(7209):64–71; ISSN 1476-4687 predicts functional non-conserved and
26. Djuranovic S, Nahvi A, Green R (2012) non-canonical sites. Genome Biol 11(8):R90
miRNA-mediated gene silencing by transla- 41. Miranda KC, Huynh T, Tay Y et al (2006) A
tional repression followed by mRNA deade- pattern-based method for the identification of
nylation and decay. Science 336 MicroRNA binding sites and their
(6078):237–240 corresponding hetero duplexes. Cell 126
27. Huntzinger E, Izaurralde E (2011) Gene (6):1203–1217
silencing by microRNAs:contributions of 42. Garcia DM, Baek D, Shin C et al (2011) Weak
translational repression and mRNA decay. seed-pairing stability and high target-site
Nat Rev Genet 12(2):99–110 abundance decrease the proficiency of lsy-6
28. Rhoades MW, Reinhart BJ, Lim LP et al and other microRNAs. Nat Struct Mol Biol
(2002) Prediction of plant microRNA targets. 18(10):1139–1146
Cell 110(4):513–520; ISSN 0092-8674 43. Witkos TM, Koscianska E, Krzyzosiak WJ
29. Ding J, Zhou S, Guan J (2012) Finding (2011) Practical aspects of microRNA target
microRNA targets in plants: current status prediction. Curr Mol Med 11(2):93–109;
and perspectives. Genomics Proteomics Bio- ISSN 1566-5240
informatics 10(5):264–275 44. Rehmsmeier M, Steffen P, Hochsmann M
30. Axtell MJ, Meyers BC (2018) Revisiting cri- et al (2004) Fast and effective prediction of
teria for plant microRNA annotation in the microRNA/target duplexes. RNA 10
era of big data. Plant Cell 30(2):272–284 (10):1507–1517; ISSN 1355-8382
31. Brousse C, Liu Q, Beauclair L et al (2014) A 45. Kertesz M, Iovino N, Unnerstall U et al
non-canonical plant microRNA target site. (2007) The role of site accessibility in micro-
Nucleic Acids Res 42(8):5270–5279 RNA target recognition. Nat Genet 39
32. Wang F, Polydore S, Axtell MJ (2015) More (10):1278–1284
than meets the eye? Factors that affect target 46. Grimson A, Farh KK-H, Johnston WK et al
selection by plant miRNAs and (2007) MicroRNA targeting specificity in
mammals: determinants beyond seed pairing. 61. Wong N, Wang X (2015) miRDB: an online
Mol Cell 27(1):91–105 resource for microRNA target prediction and
47. Sandberg R, Neilson JR, Sarma A et al (2008) functional annotations. Nucleic Acids Res 43
Proliferating cells express mRNAs with short- (Database issue):D146–D152
ened 30 untranslated regions and fewer micro- 62. Liu W, Wang X (2019) Prediction of func-
RNA target sites. Science 320 tional microRNA targets by integrative mod-
(5883):1643–1647; ISSN 1095-9203 eling of microRNA binding and target
48. Nielsen CB, Shomron N, Sandberg R et al expression data. Genome Biol 20(1):18
(2007) Determinants of targeting by endoge- 63. Borgm¨ astars E, de Weerd HA, Lubovac-Pilav
nous and exogenous microRNAs and siRNAs. Z et al (2019) miRFA: an automated pipeline
RNA 13(11):1894–1910 for microRNA functional analysis with corre-
49. Arvey A, Larsson E, Sander C et al (2010) lation support from TCGA and TCPA expres-
Target mRNA abundance dilutes microRNA sion data in pancreatic cancer. BMC
and siRNA activity. Mol Syst Biol 6:363 Bioinformatics 20(1):393
50. Ding J, Li X, Hu H (2016) TarPmiR: a new 64. Marco A (2018) SeedVicious: analysis of
approach for microRNA target site predic- microRNA target and near-target sites. PLoS
tion. Bioinformatics 32(18):2768–2775 One 13(4):e0195532
51. Hamzeiy H, Allmer J, Yousef M (2014) 65. Pla A, Zhong X, Rayner S (2018) miRAW: a
Computational methods for microRNA tar- deep learning-based approach to predict
get prediction. Methods Mol Biol microRNA targets by analyzing whole micro-
1107:207–221 RNA transcripts. PLoS Comput Biol 14(7):
52. Lewis BP, Burge CB, Bartel DP (2005) Con- e1006185; ISSN 1553-7358
served seed pairing, often flanked by adeno- 66. Maji RK, Khatua S, Ghosh Z (2020) A super-
sines, indicates that thousands of human vised ensemble approach for sensitive micro-
genes are microRNA targets. Cell 120 RNA target prediction. IEEE/ACM Trans
(1):15–20; ISSN 0092-8674 Comput Biol Bioinform 17(1):37–46
53. Enright AJ, John B, Gaul U et al (2003) 67. Ghoshal A, Zhang J, Roth MA et al (2018) A
MicroRNA targets in drosophila. Genome distributed classifier for MicroRNA target
Biol 5(1):R1 prediction with validation through TCGA
54. Krek A, Grün D, Poy MN et al (2005) Com- expression data. IEEE/ACM Trans Comput
binatorial microRNA target predictions. Nat Biol Bioinform 15(4):1037–1051
Genet 37(5):495–500; ISSN 1061-4036 68. Tokar T, Pastrello C, Rossos AEM et al
55. Hatlen A, Helmy M, Marco A (2019) Pop- (2018) mirDIP 4.1 - integrative database of
Targs: a database for studying population evo- human microRNA target predictions. Nucleic
lutionary genetics of human microRNA target Acids Res 46(D1):D360–D370
sites. Database (Oxford) 2019 69. Fan Y, Siklenka K, Arora SK et al (2016)
56. Yousef M, Abdallah L, Allmer J (2019) maTE: miRNet - dissecting miRNA-target interac-
discovering ex-pressed interactions between tions and functional associations through
microRNAs and their targets. Bioinformatics network-based visual analysis. Nucleic Acids
35(20):4020–4028 Res 44(W1):W135–W141
57. Bradley T, Moxon, S. (2019) FilTar: using 70. Karagkouni D, Paraskevopoulou MD, Chat-
RNA-seq data to improve microRNA target zopoulos S et al (2018) DIANA-TarBase v8: a
prediction accuracy in animals. BioRxiv decade-long collection of experimentally sup-
ported miRNA-gene interactions. Nucleic
58. Distefano R, Nigita G, Veneziano D et al Acids Res 46(D1):D239–D245
(2019) isoTar: consensus target prediction
with enrichment analysis for MicroRNAs har- 71. Chou C-H, Shrestha S, Yang C-D et al (2018)
boring editing sites and other variations. miRTarBase update 2018: a resource for
Methods Mol Biol 1970:211–235 experimentally validated microRNA-target
interactions. Nucleic Acids Res 46(D1):
59. Brown J, Phillips AR, Lewis DA et al (2019) D296–D302
Bioinformatics resource manager: a systems
biology web tool for microRNA and omics 72. L’Yi S, Jung D, Oh M et al (2017) miRTarVis
data integration. BMC Bioinformatics 20 +: web-based interactive visual analytics tool
(1):255; ISSN 1471-2105 for microRNA target predictions. Methods
124:78–88
60. Li C, Swartz MD, Yu B, et al (2019) dbMTS:
a comprehensive database of putative human 73. Davis JA, Saunders SJ, Mann M et al (2017)
microRNA target site SNVs and their func- Combinatorial ensemble miRNA target pre-
tional predictions. BioRxiv diction of co-regulation networks with
non-prediction data. Nucleic Acids Res 45 86. Li J-H, Liu S, Zhou H et al (2014) starBase
(15):8745–8757 v2.0: decoding miRNA-ceRNA, miRNA-
74. Quillet A, Saad C, Ferry G et al (2019) ncRNA and protein-RNA interaction net-
Improving bioinformatics prediction of works from large-scale CLIP-seq data.
microRNA targets by ranks aggregation. Nucleic Acids Res 42(Database issue):
Front Genet 10:1330 D92–D97
75. Dweep H, Gretz N (2015) miRWalk2.0: a 87. Dai X, Zhao PX (2011) psRNATarget: a plant
comprehensive atlas of microRNA-target small RNA target analysis server. Nucleic
interactions. Nat Methods 12(8):697 Acids Res 39(Web Server issue):W155–W159
76. Van Peer G, De Paepe A, Stock M et al (2017) 88. Dai X, Zhuang Z, Zhao PX (2018) psRNA-
miSTAR: miRNA target prediction through Target: a plant small RNA target analysis
modeling quantitative and qualitative server (2017 release). Nucleic Acids Res 46
miRNA binding site information in a stacked (W1):W49–W54
model structure. Nucleic Acids Res 45(7):e51 89. Fahlgren N, Carrington JC (2010) miRNA
77. Wang X (2016) Improving microRNA target target prediction in plants. Methods Mol
prediction by modeling with unambiguously Biol 592:51–57
identified microRNA-target pairs from CLIP- 90. Bonnet E, He Y, Billiau K et al (2010) TAPIR,
ligation studies. Bioinformatics 32 a web server for the prediction of plant micro-
(9):1316–1322 RNA targets, including target mimics. Bioin-
78. Bandyopadhyay S, Ghosh D, Mitra R et al formatics 26(12):1566–1568
(2015) MBSTAR: multiple instance learning 91. Jan CH, Friedman RC, Ruby JG et al (2011)
for predicting specific functional binding sites Formation, regulation and evolution of Cae-
in microRNA targets. Sci Rep 5:8004 norhabditis elegans 30 UTRs. Nature 469
79. Liu S, Li J-H, Wu J et al (2015) StarScan: a (7328):97–101
web server for scanning small RNA targets 92. Nam J-W, Rissland OS, Koppstein D et al
from degradome sequencing data. Nucleic (2014) Global analyses of the effect of differ-
Acids Res 43(W1):W480–W486 ent cellular contexts on microRNA targeting.
80. Chiu H-S, Llobet-Navas D, Yang X et al Mol Cell 53(6):1031–1043
(2015) Cupid: simultaneous reconstruction 93. Oliveira AC, Bovolenta LA, Nachtigall PG
of microRNA-target and ceRNA networks. et al (2017) Combining results from distinct
Genome Res 25(2):257–267 MicroRNA target prediction tools enhances
81. Porto IOP, Mendes-Junior CT, Fel ´ıcio LP the performance of analyses. Front Genet
et al (2015) MicroRNAs targeting the immu- 8:59
nomodulatory HLA-g gene: a new survey 94. Yousef M, Najami N, Khalifav W et al (2010)
searching for microRNAs with potential to A comparison study between one-class and
regulate HLA-g. Mol Immunol 65 two-class machine learning for microRNA tar-
(2):230–241 get detection. J Biomed Sci Eng 3(03):247
82. Vlachos IS, Zagganas K, Paraskevopoulou 95. Yousef M, Allmer J, Khalifa W (2016) Feature
MD et al (2015) Diana-mirpath v3. 0: deci- selection for MicroRNA target prediction -
phering microRNA function with experimen- comparison of one-class feature selection
tal support. Nucleic Acids Res 43(W1): methodologies. In: Proceedings of the 9th
W460–W466 international joint conference on biomedical
83. Paraskevopoulou MD, Georgakilas G, Kos- engineering systems and technologies. SCI-
toulas N et al (2013) DIANA-microT web TEPRESS - Science and Technology Publica-
server v5.0: service integration into miRNA tions, pp 216–225; ISBN978-989-758-170-
functional analysis workflows. Nucleic Acids 0
Res 41(Web Server issue):W169–W173 96. Yousef M, Khalifa W, Acar IE, et al (2017)
84. Qureshi A, Thakur N, Monga I et al (2014) Distinguishing between microrna targets
VIRmiRNA: a comprehensive resource for from diverse species using sequence motifs
experimentally validated viral miRNAs and and k-mers. In: Proceedings of the 10th inter-
their targets. Database (Oxford) 2014: national joint conference on biomedical engi-
bau103 neering systems and technologies - Volume 3:
85. Yang J-H, Li J-H, Shao P et al (2011) star- Bioinformatics, (BIOSTEC 2017).
Base: a database for exploring microRNA- INSTICC, SciTePress, pp 133–139; ISBN
mRNA interaction maps from argonaute 978-989-758-214-1
CLIP-seq and degradome-seq data. Nucleic 97. Yousef M, Jung S, Kossenkov AV et al (2007)
Acids Res 39(Database issue):D202–D209 Naı̈ve Bayes for microRNA target
predictions–machine learning for microRNA 108. Pesole G, Liuni S, Grillo G et al (2002)

targets. Bioinformatics 23(22):2987–2992 UTRdb and UTRsite:specialized databases
98. Cheng S, Guo M, Wang C et al (2016) of sequences and functional elements of 50
MiRTDL: a deep learning approach for and 30 untranslated regions of eukaryotic
miRNA target prediction. IEEE/ACM Trans mRNAs. Update 2002. Nucleic Acids Res
Comput Biol Bioinform 13(6):1161–1169 30(1):335–340
99. Lee, B., Baek, J., Park, S., et al. (2016) Deep- 109. Le Pera L, Mazzapioda M, Tramontano A
target: End-to-end learning framework for (2015) 3USS: a webserver for detecting alter-
microrna target prediction using deep recur- native 30 UTRs from RNA-seq experiments.
rent neural networks. In: Proceedings of the Bioinformatics 31(11):1845–1847
7th ACM international conference on bioin- 110. Grillo G, Turi A, Licciulli F et al (2010)
formatics, computational biology, and health UTRdb and UTRsite(RELEASE 2010): a
informatics, BCB’16, New York, NY, USA. collection of sequences and regulatory motifs
Association for Computing Machinery, pp of the untranslated regions of eukaryotic
434–442; ISBN 9781450342254 mRNAs. Nucleic Acids Res 38(Database
100. Lu T-P, Lee C-Y, Tsai M-H et al (2012) miR- issue):D75–D80
System: an integrated system for characteriz- 111. Kim M, You B-H, Nam J-W (2015) Global
ing enriched functions and pathways of estimation of the 30 untranslated region land-
microRNA targets. PLoS One 7(8):e42390 scape using RNA sequencing. Methods
101. Shirdel EA, Xie W, Mak TW et al (2011) 83:111–117
NAViGaTing themicronome–using multiple 112. Huang Z, Teeling EC (2017) ExUTR: a
microRNA prediction databases to identify novel pipeline for large-scale prediction of
signalling pathway-associated microRNAs. 30 -UTR sequences from NGS data. BMC
PLoS One 6(2):e17429 Genomics 18(1):847
102. Coronnello C, Benos PV (2013) ComiR: 113. Ulitsky I, Shkumatava A, Jan CH et al (2012)
combinatorial microRNA target prediction Extensive alternative polyadenylation during
tool. Nucleic Acids Res 41(Web Serverissue): zebrafish development. Genome Res 22
W159–W164 (10):2054–2066
103. Dweep H, Sticht C, Pandey P et al (2011) 114. Jambor H, Surendranath V, Kalinka AT et al
miRWalk–database: prediction of possible (2015) Systematic imaging reveals features
miRNA binding sites by “walking” the genes and changing localization of mRNAs in dro-
of three genomes. J Biomed Inform 44 sophila development. elife:4
(5):839–847 115. Lakshmanan V, Bansal D, Kulkarni J et al
104. Xiao F, Zuo Z, Cai G et al (2009) miRecords: (2016) Genome-wide analysis of polyadenyla-
an integrated resource for microRNA-target tion events in Schmidtea mediterranea. G3
interactions. Nucleic Acids Res 37(Database): (Bethesda) 6(10):3035–3048
D105–D110 116. Ji Z, Lee JY, Pan Z et al (2009) Progressive
105. Pinzón N, Li B, Martinez L et al (2017) lengthening of 30 untranslated regions of
microRNA target prediction programs pre- mRNAs by alternative polyadenylation during
dict many false positives. Genome Res 27 mouse embryonic development. Proc Natl
(2):234–245 Acad Sci U S A 106(17):7028–7033
106. Alexiou P, Maragkakis M, Papadopoulos GL 117. Hoffman Y, Bublik DR, Ugalde AP et al
et al (2009) Lost in translation: an assessment (2016) 30 utr shortening potentiates
and perspective for computational microRNA microrna-based repression of
target identification. Bioinformatics 25 pro-differentiation genes in proliferating
(23):3049–3055 human cells. PLoS Genet 12(2)
107. Wienholds E, Kloosterman WP, Miska E et al 118. Spies N, Burge CB, Bartel DP (2013) 30
(2005) MicroRNA expression in zebrafish UTR-isoform choice has limited influence
embryonic development. Science 309 on the stability and translational efficiency of
(5732):310–311; ISSN 1095-9203 most mRNAs in mouse fibroblasts. Genome
Res 23(12):2078–2090
Chapter 11
Evolution and Phylogeny of MicroRNAs — Protocols,

Pitfalls, and Problems
Cristian A. Velandia-Huerto, Ali M. Yazbeck, Jana Schor,
and Peter F. Stadler
Abstract
MicroRNAs are important regulators in many eukaryotic lineages. Typical miRNAs have a length of about
22nt and are processed from precursors that form a characteristic hairpin structure. Once they appear in a
genome, miRNAs are among the best-conserved elements in both animal and plant genomes. Functionally,
they play an important role in particular in development. In contrast to protein-coding genes, miRNAs
frequently emerge de novo. The genomes of animals and plants harbor hundreds of mutually unrelated
families of homologous miRNAs that tend to be persistent throughout evolution. The evolution of their
genomic miRNA complement closely correlates with important morphological innovation. In addition,
miRNAs have been used as valuable characters in phylogenetic studies. An accurate and comprehensive
annotation of miRNAs is required as a basis to understand their impact on phenotypic evolution. Since
experimental data on miRNA expression are limited to relatively few species and are subject to unavoidable
ascertainment biases, it is inevitable to complement miRNA sequencing by homology based annotation
methods. This chapter reviews the state of the art workflows for homology based miRNA annotation, with
an emphasis on their limitations and open problems.
Key words RNA secondary structure; gene duplication; homology search; data curation; evolution
1 Introduction
Since their discovery at the turn of the millennium, microRNAs

(miRNAs) have been described in almost all animals and plants.
Most recently, miRNAs have also been discovered in an increasingly
diverse collection of unicellular Eukaryotes. Their function in post-
transcriptional gene silencing [1] depends on the presence of the
RNA interference pathways [2, 3], which presumably date back to
the Eukaryote ancestor.
Canonical miRNAs are produced from primary precursor tran-
scripts, the pri-miRNAs, which are transcribed by pol-II. These
carry hairpin-shaped precursors, the pre-miRNAs, which are
excised in the nucleus. The details of processing steps that extract
211
212 Cristian A. Velandia-Huerto et al.
a miRNA/miRNA* duplex from the pre-miRNA hairpin differs

substantially between plants and animals, see e.g. [4]. Single-
stranded mature miRNAs (miR) have a length of about 22nt and
are one of many types of small RNAs that can direct the Argonaut
complex to its target. As summarized in [5], the processing implies
a list of criteria for miRNA annotation:
(a) Evidence for the expression of both the mature miR and its
miR* from the opposite side of the precursor hairpin.
(b) A sequence distinct from the processing products of other
structured RNAs such rRNAs, tRNAs, snRNAs, or snoRNAs.
(c) A sequence that is non-repetitive in the genome of origin
(d) Consistent 5’ processing of both the miR and the miR*
sequence
In addition, different structural constraints apply to the precursor
hairpins of animals and plants. The conditions (a)–(d) pertain to
canonical microRNAs. We shall briefly return to “non-canonical
miRNAs” that do no follow the canonical processing pathway
below. For the purpose of this contribution, we single out (canoni-
cal) microRNAs from a much broader class of small RNAs that
include e.g. the processing products of structured “housekeeping
RNAs excluded by condition (b) above, endogenous siRNAs [6],
or miRNA-like are small RNAs released in to cytosol by intracellular
pathogens [7].
A thorough re-analysis of miRNAs reported for various protist
lineages [5], shows that only a fraction of them conforms to these
annotation criteria. Nevertheless there is now good evidence for
the convergent emergence of miRNAs in several independent
events in dinoflagellates, brown algae, some lineages of Green
algae, land plants, excavates, slime molds, porifera and higher
metazoa [8]. A recent report of microRNAs in Ichthyosporea [9]
suggests that the history in the animal lineage may be even more
complex.
Most miRNAs are among the most highly conserved genetic
elements, at least in animals and plants both the mature sequence
and the entire precursor is usually under strong stabilizing selection
[10]. This makes it possible to investigate the evolutionary history
of miRNAs despite the limited amount of sequence information
contained in the short (< 100 nt) precursor sequences
[11, 12]. Like other genes, miRNAs often form families of para-
logs, evolving by gene duplication and loss [13–17]. These families
form the basis of the miRBase nomenclature [18]. The presence of
miRNA families has been advocated as excellent phylogenetic mar-
kers [19–22] since miRNA families are rarely lost entirely, although
individual paralogs are much more flexible in this respect. The
massive restructuring of the miRNA complement of tunicates is
an important exception to this rule [23].
Evolution and Phylogeny of MicroRNAs — Protocols, Pitfalls, and Problems 213
In contrast to protein-coding genes, miRNAs are readily inno-

vated de novo. The miRBase [24] thus reports a large number of
evolutionarily very young and species-specific miRNAs
[25, 26]. For fruit flies, an innovation rate of 12 new miRNA
genes per million years has been estimated [27]. This is consistent
with theoretical and simulation results suggesting that hairpins that
can be processed as miRNA precursors are frequently formed by
random RNA sequences [13, 28, 29]. Given that a major fraction
of (the non-coding part of) eukaryotic genomes is transcribed,
many such candidate precursors become available to be evaluated
by selection. Repetitive elements are a major source of raw material
for miRNA innovation [30–34]. In plants, inverted repeats of the
target genes are also an important contributor to miRNA innova-
tion, see e.g. [35] and the references therein.
Of course, most of the newly evolved hairpins that enter the
miRNA processing pathways are never placed under stabilizing
selection and even fewer become functionally important enough
to be retained over long evolutionary time scales. There are positive
correlations between miRNA expression level, their rate of
sequence evolution, and the evolutionary age of miRNAs
[36, 37]. This can be explained by a gradual increasing of both
the required copy number and the selection pressure due to the
acquisition of additional targets [38]. Once the number of distinct
targets is large enough, stabilizing selection becomes sufficiently
strong to prevent the loss of the miRNA. The net growth rate of the
set of these permanent miRNAs can be estimated by comparing the
miRNA complements between phyla. The resulting general trend
of expanding the miRNA repertoire in most lineages appears to
correlate with increasing morphological complexity [11, 19, 20,
38–41], while massive morphological simplification, as in the case
of tunicates, seems to be accompanied by loss of miRNA families
[23, 42–44].
A comprehensive understanding of the evolution and func-
tional adaptations of miRNAs requires a comprehensive annota-
tion. In this chapter we summarize homology-based technique
available for this task. We emphasize on the pitfalls and open
problems that so far have hampered the collection of data sets
that are sufficiently accurate and complete to be used as data basis
of a detailed quantitative analysis that would settle the ongoing
controversies in the field.
2 Homology Search for miRNAs
Despite their short sequence length, it is usually not too difficult to

identify homologs of known microRNAs. Their mature sequences
are nearly perfectly conserved and thus are convenient anchors for
sequence-based search methods such as blastn. The sequences of
the precursor hairpins are usually also quite well conserved, evol-
ving at rates comparable to coding sequences. The selective con-
straints on the mature sequences stronger by up to an order of
magnitude [45]. The structural constraints on the precursor hair-
pin, finally, imply an approximate complementarity between the
miR and miR* sequences even if the miR* sequence is not func-
tional in its own right. As a consequence, simple, sequence-based
methods are usually successful at least at phylum level. A blastn
search with the human mir-10a precursor, for instance, readily
yields top hits with E-values < 10 40 in mammals, < 10 20 in
sauropsids, and < 10 8 in the duplicated genomes of teleost fishes.
The methods quickly loses power outside the vertebrates, however,
significant hits E < 10 5. . .10 3 are found in some echinoderms and
protostomes. Sensitivity and specificity can be increased by optimiz-
ing blast parameters and by comparing “hit lists” obtained with
different parameter settings [43].
MicroRNA precursors also feature well-conserved secondary
structures. The phylogenetic scope of homology searches can
therefore be expanded by employing Covariance Models (CMs)
[46]. CMs are a generalization of Hidden Markov Models that
incorporates the co-variation of paired bases. Thus, the specificity
of CM-bases homology search with infernal [47, 48] is consid-
erably increased compared to sequence-only methods such as
blast, full dynamic programming alignments [49], and HMMs.
CMs are trained from sequence alignments annotated by a consen-
sus structure for the aligned sequences. The Rfam database
[50, 51] provides such alignments for a wide array of RNA families,
including some miRNAs. It is not comprehensive in its coverage of
miRNAs, however. MirBase [24, 52], on the other hand, provides
a much more complete coverage of the miRNA precursor sequence.
In addition, miRBase also provides alignments for miRNA
families, however, at present, these often have to be manually
curated or extended by additional members. The general workflow
for a de novo construction of a miRNA family alignment is as
follows:
(1) Obtain a set of seed sequences covering as evenly as possible
the phylogenetic range of family members that are known
already.
(2) Construct a multiple sequence alignment (MSA) of the seed
sequences, using one of the many tools such as muscle,
mafft, clustalw, t-coffee, or dialign.
(3) Compute the consensus structure e.g. with
RNAalifold [53].
(4) In general, a curation of the 3’ and 5’ ends is necessary. Ideally,
this step includes the evaluation annotated ends of the precur-
sor relative to the location of the mature sequence and in
relation to secondary structure. If sequences are extended or
trimmed, steps (2) and (3) should be repeated.
MSAs are sufficient to construct HMMs, which provide an alterna-

tive to searching homologs of the seed sequences individually. A
convenient and efficient implementation is nhmmer [54]. The
infernal package provides the necessary tools to convert an
MSA that is annotated by a consensus structure into a CM, to
identify score thresholds for significant matches, and to search a
nucleotide sequence for approximate matches.
Instead of starting from sequence alignments, it is also possible
to first predict secondary structure for each sequence with RNA-
fold [55] and to use as a second step a structure-based alignment
tool such as LocARNA [56, 57] or MARNA [58] to obtain an align-
ment together with a consensus structure. The resulting multiple
alignment will in general require some level of user intervention to
identify any obvious alignment errors, incorrect or corrupted
sequences, and other errors.
In many cases, only a single query sequence is available at the
outset. This is in particular often the case when miRNAs are iden-
tified from small RNA-seq data. The natural first step is then to
approximate the precursor hairpin. This can be achieved by extract-
ing about 100 nt of genomic flanking sequence on both sides,
which is sufficient to ensure that the precursor hairpin is completely
contained in the extracted sequence. The precursor hairpin, if it
exists, can then be identified by computing the secondary struc-
tures. An elegant method for this purpose is RNAplfold [59],
which allows the extraction of locally stable structures by restricting
the base-pair span to 80 or 100 nt. Since miRNAs fold into much
more stable structure compared to random RNA structures with
the same sequence composition, see e.g. [60–62], the most stable
local stem-loop structure identifies the pre-miRNA hairpin. In
[63], instead a simple rule is used to estimate the precursor
sequence depending on whether the miR is assumed to be located
on the 5’- or 3’-side of the hairpin; both are then folded and the
more stable hairpin is retained. Once a plausible precursor has been
found, closely related genomes can be conveniently searched with
blast for an initial set of homologous examples, from these a
structure-annotated alignment can then be computed as above.
Figure 1 summarizes the outline of the workflows, starting
from a single candidate query. Targets for the search are typically
genomic sequences (at various stages of assembly) retrieved from
Ensembl or NCBI . Typically, the iterative searches use relaxed
parameters optimized for sensitivity and necessarily produce large
numbers of false positive candidates. These require extensive cura-
tion steps, which in part are performed by automatic filtering
procedures (considering quantitative measures such as E-values,
sequence identity, the presence of highly conserved miR sequences,
etc.), and in part rely on user intervention and manual inspection.
Most studies use the scheme in Figure 1 as a guideline rather than
as a fully integrated pipeline. A notable exception is the recent
single "miR" seed alignment
extract precursor candidate

Covariance model
blast search sequence−based

in close relatives search
infernal search
curation
realign
sequence alignment
add new hits to
curation
alignment
curation
consensus structure compute consensus update consensus

structure
structure
Fig. 1 General workflow for homology search for miRNAs and other structured RNAs. In the initial phase (l.h.s.)
the goal is to obtain a seed set of trusted homologs starting from a single small RNA (obtained e.g. by
sequencing) or a predicted precursor structure. This seed set is then expanded iteratively. Often a sequence-
based search can efficiently expand the phylogenetic scope considerably, leading to a collection of homologs
with sufficient diversity to allow validation of the consensus structure by patterns of sequence co-variation.
Sequence-based searches may be performed e.g. using blast, full dynamic programming alignment tools
such as ssearch or gotohscan [49], or Hidden Markov Models (inferred from the sequence alignment).
Alignments annotated with a consensus structure allow the construction of co-variance models. Often these
are more sensitive than purely sequence-based models. Importantly, putative homologs need to be curated
either manually or with the help of automatic means to avoid the inclusion of false positives into the next
iteration of the search.
survey of tunicate ncRNAs [43], which combined blastn-based

searches with several different combinations of parameters with
sequence-based searches using Hidden Markov Models to generate
candidate sets. These are then filtered using co-variance models for
the known miRNA families as a first automatic curation step.
Homology searches on large phylogenetic groups are usually
performed in an iterative way, i.e., using new candidates as a means
of expanding and refining the search. Purely blast-based
approaches usually use the new candidates as additional queries.
In HMM or CM based approaches, the alignments are expanded
with the new candidates. This can either be done by re-aligning the
entire set of homologs, or by aligning the new sequence(s) to the
HMM or CM using specialized tools. The augmented alignments
are then used to re-train the model. Typically this process is iterated
until no further candidate homologs can be found.
Iterative homology searches require the evaluation of the can-
didate hits (which will be the topic of the following section), and
typically involve an update to the MSA and its consensus structure.

For the latter task, there are at least two distinct strategies:
(a) If the search started from a trusted seed alignment, new hits
are often added individually. Many sequence alignment tools
offer an option of align individual sequences to a given MSA.
Similarly, individual sequences can be aligned to HMMs and
CMs. Since there is a correspondence between the positions in
the model and the columns of the seed alignment, this also
determines how the candidate fits to the MSA.
(b) If no trusted seed is available it may be preferable to
completely realign the union of query and candidate
sequences. As for the seed alignment, this can be done either
purely sequence-based alignment methods or with the help of
a structure-based alignment method.
The update of the alignment in general also prompts an update of
the consensus structure.
Recently, pipelines implementing this workflow have become
available for general use. RNAlien [64] and GLASSgo [65] are
primarily intended for the small RNAs of procaryotes. To our
knowledge it has not been applied to animal miRNAs so far,
although conceptually it should be suitable for this task as well. A
partial solution for miRNA families in which at least one represen-
tative has a documented miR and miR* sequence has become
available in [66].
The particular structure of miRNA precursors can also be
utilized to devise a miRNA-specific approach. In the first step,
near exact matches of the mature miR are retrieved. In the second
step, the flanking sequences of the initial candidates are retrieved
and investigated for the presence of a significantly stable hairpin
structure. The pre-miRNA candidates that satisfy the filtering cri-
teria can then be treated as above. A quite general workflow, MIR-
fix, combining these structure-based criteria with annotation of
miR and miR* sequence is described in [63]. It has been used for
the annotation of miRNAs in tunicates [67].
3 Curation of Candidate miRNA Homologs
In this contribution we focus on the evaluation of miRNA candi-

dates obtained by homology search, although the initial identifica-
tion of bona fide miRNAs from sRNA-seq data shares many of the
problems. We will not assume in the following, however, that we
have physical evidence for the existence of the mature miR or miR*
sequences and consider two related questions:
(a) Does a new candidate belong to a family represented by an
alignment?
(b) Does a multiple sequence alignment really represent a miRNA

family?
To answer the first question, several tests are available. First very
low E-values reported by the homology search tool, and/or a
sequence that is very similar to known family members is a near
perfect indication for membership in a given family.
In most cases of genome-wide surveys, good phylogenetic trees
cover the target genomes as well as at least a subset of the taxa
included in the seed alignment. It is appealing then to use a
position-wise maximum likelihood model trained on the seed
sequences to predict the most likely sequence of a homolog in the
target taxon and its position-wise uncertainty. The maxAlike soft-
ware [68] implements the prediction of a position-specific proba-
bility matrix from a multiple sequence alignment and a
phylogenetic tree. The miRNA candidate can then be compared
to sequences predicted as maxAlike as the most likely one for the
target sequence; this can yield an increase in both sensitivity and
specificity of the homology assessment. A less systematic way to
implement the same idea is to check whether the mature miRNA
sequence is conserved at least as well as the rest of the precursor.
False positive candidates can also be identified by their unex-
pected position in a phylogenetic tree computed from the final
multiple sequence alignment. Of course, a tree computed from
sequences as short as microRNA precursors will be noisy. Never-
theless, one expects the new sequence(s) to fall at least approxi-
mately into the position expected from the known phylogenetic
affiliation of the taxon in which they were identified. To our knowl-
edge, this idea has not been explored systematically, although it has
been used informally at the level of manual curation e.g. in our own
work of miRNA evolution [11].
Many machine learning applications have been developed to
recognize miRNA precursor hairpins, although most of these tools
are designed for animal miRNAs (recently reviewed in [69]), similar
methods have also been use for plant miRNAs [70]. Such programs
can be used to identify likely false positives. It may also be useful to
check the final alignment using a classifier that evaluates the substi-
tution patterns, such as a RNAmicro [71] (focusing on animal
miRNAs).
Additional confidence can be gained by considering mature and
precursor sequences at the same time. This is particularly appealing
as small-RNA-seq data are becoming available also for many less-
studied organism. The combination of independent information
for mature and precursor sequences also helps to identify errors in
public data sources. We observed that database entries are listed
occasionally in the wrong orientation, and sometimes putative
precursor hairpins are shifted. These issues are addressed in the
MIRfix pipeline for the curation of miRNA data [63, 66]. It aims
at constructing not only a multiple sequence alignment of the

precursor hairpin but also to determine the location of the miR
and miR* products and to determine consistent 5’ and 3’ ends for
all included sequences. While this is fairly straightforward when
both miR and miR* sequences are known, the situation is more
difficult if only one putative mature product is available. In this case
the mature product may be located in either the 5’ or the 3’ flank of
the hairpin, implying two approximate positions of the precursor.
Since miRNA precursors are extremely stable [60–62], these two
alternatives can be distinguished based on their difference in fold-
ing energies (as predicted e.g. by RNAfold [55]) and the position
of the mature product in the hairpin.
For canonical miRNA, the mature product is expected to be
located on either side of the helical part of the hairpin. Its counter-
part on the opposite side of the helix serves as prediction for the
miR*. In many cases, miR and miR* sequences can be inferred
from known homologs in other taxa. In the absence of experimen-
tal evidence both in the focal taxon and in all related taxa, their
position can be inferred at least approximately as the best-
conserved sequences. Known and predicted miR and miR*
sequences then serve as anchors for constructing a new seed align-
ment using dialign [73]. An example of a well-studied miRNA
family whose precursors are most likely incorrectly annotated is
shown in Figure 2, see also [63].
4 MiRNA Paralogs
Like most protein coding gene families, miRNAs are subject to

gene duplication and loss. Both, local tandem duplications and
remnants of chromosome or genome-wide duplications are fre-
quently observed. Local duplication often lead to the formation
of miRNA clusters that are transcribed from the same primary
precursor transcript. Most clusters are small and contain only two
or three precursor hairpins [11, 74], which may or may not be
homologs.
It is important to note that the miRNA nomenclature
employed in miRBase does not necessarily reflect orthology within
a miRNA family. The detailed reconstruction of the gene family
history of larger families requires tedious manual work. Several of
the more complex gene families have been studied in detail, includ-
ing the mir-17 clusters [13, 75, 76], the let-7/mir-100/mir-125
clusters [15, 77], the mir-302/mir-367 cluster [78], the mir-8/
mir-200 [79], and the miR199/mir-214c [80] families. The short
miRNA precursor sequences often do not convey enough informa-
tion to distinguish paralog groups unambiguously. This is an issue,
in particular in the case of lineage-specific losses of some members
of a miRNA family. In these cases synteny, that is, conservation of
Fig. 2 The miR* sequence for mir-30e of the medaka (Oryzias latipes) is
predicted to overlap the loop region of the precursor hairpin in miRBase
entry MI0019479 (left). A re-evaluation of the precursor location MIRfix
[63] (right, see text for details) places the miR at the expected position in the
3’ side of the hairpin. This prediction is supported by basal UG enhancer motif
[72], which now appears immediately upstream. The revised precursor also fits
very well into an alignment of the entire mir-30 family (not shown). Adapted
from [63].
relative gene order, can be used to support paralog assignments.

Within a single miRNA cluster, gene order is typically conserved.
For paralogs located on different chromosomes one can use the
syntenic arrangement of flanking genes in different species to sup-
port correspondences. Synteny has been used systematically to
investigate orthology of tRNA genes [81]. The SMORE pipeline
[82] implements this idea into an automatic workflow that should
also be applicable to miRNAs. At the level of manual inspection,
synteny has been used fairly systematically to trace the evolution of
the let-7 clusters in vertebrates [15].
As an example we consider the mir-17 clusters. The human
genome harbors multiple paralogous copies that in turn are com-
posed of at least three mutually unrelated miRNA families: mir-17,
mir-19 and mir-25. While the mir-17 and mir-19 families are
specific to vertebrates, the mir-25 family is among the oldest in
animals [13, 76, 83]. These three groups are subdivided into
several families according to the miRBase annotation. Figure 3
shows the sequence patterns for six of them. The conservation
pattern can be used to identify the location of the mature
sequences: it is located on the 5’ arm for mir-17/-106 (A),
mir-18 (B) and mir-20 (D), and on the 3’ arm for mir-19a
(C) and mir-19b (E) and mir-92/-25 (F), consistent with data
from deep sequencing. The reduced sequence conservation in the
loop region is clearly visible.
A detailed comparison between the miRNA families involved in
the mir-17 cluster provides strong evidence for homology within
the three unrelated groups: first, synteny among the paralogous
clusters strongly suggests homology among the (sub)families in the
top panel of Fig. 3. This is corroborated by the alignment of the
family-specific sequence logos in the lower part of Fig. 3. A more
quantitative comparison is summarized in Fig. 4. CMCompare [84]
shows the expected significant similarities within the mir-17/mir-
18/mir-20/mir-93/mir-109 group (see also [13, 76]). Surpris-
ingly, this method recovers only a moderate similarity between
the two mir-19 subgroups and instead shows an affinity of
mir-19a to the entire mir-17 super family, which is not visible at
all in the hierarchical clustering of the z-score similarities [13]. We
suspect that this may be driven by structural features. The consen-
sus structures are shown in Fig. 5. The lack of a well-formed hairpin
for the mir-25/-92 family is at least in part due to the substantial
divergence between vertebrate and invertebrates sequences. It is
not observed for the other families, which are confined to
vertebrates.
5 Evolution of miRNA Families
The origin of microRNA families is pinpointed by mapping the

members of the family onto a phylogenetic tree. The parsimony
principle then suggests that the family originated on the branch
preceding the last common ancestor of the known homologs, see
e.g. [11, 12, 16, 19]. This approach has readily extended the
number of family members [12] with the help of a slightly modified
version of Sankoff’s parsimony algorithm [86] that accommodates
Dollo’s principle that a miRNA family may arise only once. This
approach is also amenable to a probabilistic version [87]. A very
detailed account of the evolution of microRNAs in the genus
Drosophila can be found in [88]. It is based on a comprehensive
set of small-RNA sequencing data.
Given the large number of miRNAs that have been reported for
many genomes, a quantitative analysis of miRNAs evolution is of
interest. Large-scale comparative analysis of animal miRNA evolu-
tion have revealed several bursts of innovation of novel miRNA
A B C D E F
a mir-17 mir-18 mir-19a mir-20 mir-19b mir-92
mir-106/ mir-18/ mir-19/ mir-20/ mir-19b/ mir-92/

-106a/ -18-1/ -19a/ -20a/ -19b-1/ -92-1/
-106a-1/ -18-2/ -19a-1/ -20a-1/ -19b-2/ -92-2/
-106a-2/ -18a/ -19a-2/ -20a-2/ -19c/ -92a/
-106b -18b/ -19a-3/ -20a-3/ -19c-1/ -92a-1/
-18c -20b/ -19c-2/ -92a-2/
mir-17/ -20b-1/ -19c-3/ -92a-3/
-17-1/ -20b-2 -19c-4/ -92a-4/
-17-2/ -19d -92b/
-17-3/ mir-93/ -92c/
-17a/ -93a/ -92d/
-17a-1/ -93a-1/
-17a-2/ -93a-2/ mir-25/
-17b -93b/ -25-1/
-25-2/
-25-3/
-25b
b mir-17
2.0
1.5
mir-106/-17
1.0
0.5
0.0
100
101
102
103
104
105
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
1
2
3
4
5
6
7
8
9
2.0
1.5
mir-18
1.0
0.5
0.0
100
101
102
103
104
105
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
1
2
3
4
5
6
7
8
9
2.0
mir-20
1.5
1.0
0.5
0.0
100
101
102
103
104
105
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
1
2
3
4
5
6
7
8
9
** * * * * **** * * * * * ** * ** * ** * * * *
mir-19
2.0
mir-19a
1.5
1.0
0.5
0.0 100
101
102
103
104
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
1
2
3
4
5
6
7
8
9
2.0
mir-19b
1.5
1.0
0.5
0.0
100
101
102
103
104
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
1
2
3
4
5
6
7
8
9
* ** * * ** * * * * * * ** * * * * ** * ** ** * *** * * ** *** * * * **
mir-25/92
2.0
mir-25/92
1.5
1.0
0.5
0.0
100
101
102
103
104
105
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
6
7
8
9
Fig. 3 Organization of consensus sequences of the miRNAs forming the mir-17/mir-19/mir-25 clusters.
a Schematic overview of the cluster organization and miRBase nomenclature of the pertinent miRNA
families. Colors indicate the paralogy relationship between those miRNA families, blue: mir-17, red: mir-19
and green: mir-25/-92. b Sequence logos representing the alignments of the three unrelated groups and their
major sub-groupings. Conserved positions along alignments of sequence Logos are marked with ‘*’, while
non-informative positions are shaded in gray. The miR and miR* sequences are framed in black. The
dominating functional products are the 5’ mature sequences for mir-17, and the 3’ mature products for
mir-19 and mir-25/-92.
0.0
2.0
4.0
6.0
0
12 8.0
25
10.0
-4
mir-92
mir-19b
mir-25
14 12.0
mir-20
Dm-92
mir-17
mir-93
12 23
mir-19b
mir-106b
14.0
mir-18
49
mir-19a
-5 16.0
11
18.0
z-score
Ce-92
Fig. 4 Distant relationships of the miRNA families of the mir-17 cluster. L.h.s.: Comparison of co-variance
models created with CMCompare [84] from vertebrate annotated (miRBase (v.21)) pre-miRNAs of the
miRNA families that form the mir-17 cluster. Line width scales with CM similarity, from high (thick) to low
(thin, dotted). R.h.s.: Hierarchical cluster tree using pairwise z-scores as similarity measure. Adapted from
[13]. The two leaves for mir-19b refer to different paralogous clusters. Dm and Ce indicate the sequences of
Drosophila melanogaster and Caenorhabditis elegans homologs, all other nodes are averages over vertebrate
sequences. The higher diversity of the mir-25/-92 groups is reflected here by smaller z-score of the top
branches compared the mir-19 and mir-17 groups.
families, most notably one associated with the origin of the verte-
brates and another one at the root of the placental mammals
[11, 20, 40]. These findings have subsequently been corroborated
at least semi-qualitatively, see e.g. [12, 16, 75]. Not all develop-
mental innovations are associated with burst of microRNA innova-
tion, however. For example, the advent of complete metamorphosis
in holometabolous insects was accompanied only by the acquisition
of three and the loss of one families relative to a core set of
65 miRNA families in insects [89].
A recent evaluation of well-annotated miRNA complements
shows that miRNAs are lost infrequently at the family level [90]
and confirms that losses are concentrated in a few lineages with
major phenotypic changes, in particular simplifications. Report of
wide-spread loss of miRNA families can be attributed largely to
ascertainment biases in the available data sources, artifacts of data
curation, and occasional false positives in homology searches
[12, 89, 90]. False positive hits in homology searches lead to
sometimes massive over-predictions of loss events based on parsi-
mony approaches. A case in point is mir-7880, erroneously
detected in the squirrel in [12], as noted in [90].
Fig. 5 Structural consensus from all mir-17 paralog groups. Conservation along positions could be observed
with the symbols along miRNA folding. Consensus structures were computed with RNAalifold from
ClustalW alignments of all family members available in miRBase v.21. The graphical representations
were generated with R2R [85], which provides annotations for sequence and structure conservation for
nucleotide identity (measured position-wise in the underlying multiple sequence alignment), base pairings
(using the criteria for covariation implemented in infernal), and presence of nucleotides (indicating the
fraction on non-gaps in an alignment column). In contrast to the typical behavior of most other structured RNA
families, stem regions are more conserved that loops. The mir-25/-92 group is more divergent than the other
groups due its larger phylogenetic depth.
A related source of errors are unrecognized homologies, such

as tunicate mir-1473, which is homologous to the ancient mir-100
family [15]. Such cases lead to the erroneous prediction of both an
innovation and possibly many loss events. Distant homologies are
not trivial to recognize due to limited size and thus information
content of miRNA precursors. So far, two computational
approaches have been explored. A comparison of similarity scores
of pairwise alignments with the randomly shuffled sequences was
proposed in [13] as a means to identify mir-25/mir-92 and
mir-17/mir-18/mir-20/mir-93/mir-106, respectively, as ancient
homologs. A more sophisticated approach is CMCompare, which
directly compares the CMs representing two miRNA families
[91]. The method, which is also available as a web service [84],
identified e.g. the plant families MIR806, MIR811, MIR812,

MIR821, MIR1023, and MIR1151 as likely homologs. A system-
atic analysis of distant homologies among miRNA families has not
been conducted in recent years.
Quantitative studies are not only hampered by the limits of
homology search, but also suffer from other problems with the
available data. First, there is a massive ascertainment bias of empiri-
cal studies, which concentrate on a small number of model species
whose miRNA complements are very well studied, in particular
human, mouse, fruitfly and C. elegans. On the other hand, there
is a substantial ambiguity what exactly constitutes a miRNA as
opposed to another family of small non-coding RNAs. Using a
stringent view requiring canonical processing by Drosha and
Dicer as criteria, a significant fraction of the entries in recent
releases of the miRBase are presumably “false positives”
[90]. This is difficult to decide, however, since miRBase is not
very explicit about where exactly it draws the boundary of “miR-
Base-miRNAs”. In contrast, a quite specific set of rules to identify
canonical miRNAs has been adopted by the MirGeneDB [92],
based on the analysis small-RNA-seq data from a selection of
from metazoan species.
6 Non-Canonical miRNAs
A clear outlier in miRBase is mir-451, whose mature product is

processed by Argonaute2 from the loop region of the precursor
[93]. It is usually treated as miRNA, even though it is not recog-
nized by many automatic annotation tools. Recent studies show
that there is a large number of alternative biogenesis pathways that
overlap more or less with the canonical one eventually producing
small RNAs that incorporate into the Argonaut complex and per-
form miRNA-like functions in gene silencing [94–96].
Mir-451 is the prototypical representative of a larger class of
Argonaut2-processed loop-miRNAs whose mature product, the
“miR-loop”, is excised from the loop rather than the precursor
stem. MiR-loop RNAs are also produced from the precursors of
some canonical miRNAs including mir-33a, mir-34a, mir-192, mir
219-2 [97, 98]. It does not seem too difficult to adapt e.g. the
workflow of [63] to loop-miRNAs; so far, however, no such tool
seems to be available.
A large class of non-canonical miRNAs are mirtrons, whose
precursors are produced by splicing rather than by Drosha [99–
101]. They appear to be much less conserved than canonical miR-
NAs. A recent study indicates major differences between the selec-
tion pressures acting on canonical miRNAs and mirtrons,
respectively, strongly suggesting that they should better be handled
as distinct RNA classes [102]. In Drosophila, for instance, mirtron
hairpins are specifically by the TUTase Tailor, which in this manner

efficiently suppresses the mirtron biogenesis [103]. A computa-
tional analysis shows, furthermore, that mirtrons can be robustly
distinguished from canonical microRNAs [104]. Recently, the
specialized mirtronDB became available, making it much easier
to study mirtron systematically [105]. In some cases,
e.g. mir-1017, only the 5’-end of the precursor is generated by
the splicing machinery, while the 3’-end is the result of exosome-
mediated trimming [106]. There are also examples, such as
mir-3103, where only the 3’-end is splicing-related, while the
5’-end is the result of trimming [100]. Very little is know about
the latter type.
Small RNAs derived from transposable elements (TE) have
been reported to be significantly less conserved than non-TE
derived ones [31], with observed selective pressures differing
between major TE classes. Mariner derived elements (MADE1)
and other miniature inverted-repeat transposable elements
(MITEs) are a major source of reported human-specific micro-
RNAs [32], while those originating from L2 and MIR elements
seem to be better conserved [30]. It is a matter of ongoing discus-
sion to what extent these small RNAs should be considered as
miRNAs, see e.g. [90].
Small RNAs are also produced by processing tRNA precursors,
snoRNAs, and other structured RNAs, some with and some with-
out the help of Dicer, see e.g. [107–109] and the references
therein. Entries of this type (e.g., mir-333) have been systematically
removed from miRBase, other exceptional cases remain firmly
included in the list of miRNA families. Over the last years, func-
tional small RNAs have been described for most of the “classical”
classes of structured RNAs. This includes the sdRNAs deriving
from snoRNAs [110, 111]. A few H/ACA snoRNAs in fact seem
to have acquired primarily miRNA like functions [112]. TFs are an
analogous class of small RNAs produced from tRNAs or their
precursors [113, 114]. Even ribosomal RNA operons are the
source of functional miRNA-like products [115].
In this contribution we focused on the computational methods and

workflows that are available to trace and analyze the evolution of
miRNAs. We deliberately did not discuss experimental methods, in
particular high-throughput methods, to identify miRNAs and their
precursors. These are described in other chapters of this book.
Instead, we assumed that at least one representative of a miRNA
family is known. Although much effort has been expended to
catalog and analyze miRNA families, we still lack a comprehensive,
quantitative picture and many open questions remain.
First, there is still no universal consensus of what exactly dis-

tinguishes a bona fide miRNA from one of the many types of other
RNAs of comparable size. This is not merely a question of nomen-
clature or the decision defining the scope of a data repository such
as miRBase. The distinction between (canonical) miRNAs and
other small RNAs, as well as a classification of miRNAs in canonical
miRNAs and several types of non-canonical ones is of practical
importance for the construction of computational methods.
Machine learning approaches, in particular, need clearly defined
test and training data. This issue is particularly important in the
context of lineage-specific miRNAs and miRNAs associated with or
derived from repetitive elements.
The most pressing open problems, however, concern practical
aspects of data analysis: What is the level of completeness of miRNA
annotation that can be achieved by homology search? In other
words, how good are the available tools in identifying the last
common ancestors of miRNA families, and how good are our
estimates of the gain and loss of individual ortholog groups and
entire families of homologous miRNAs? It would appear that
quantitative statements on patterns and regularities of microRNA
evolution are at present limited by technical (computational) and
biological (definitional) issues. The available evidence seems to
support clear differences between (sub)types of miRNAs, implying
that quantitative studies required clear distinctions of miRNA
types. Additional data will certainly help to clarify these issues.
High coverage small-RNA-seq data would be of particular value
for species that do not have close relatives for which such data are
already available.
The wealth of miRNA data available in miRBase and Rfam, in
particular when augmented by (semi)automatic pipelines to curate
and complete the data by homology search set the stage for inves-
tigating a wide array of questions. While we seem to have a decent
understanding of the evolution of miRNAs at the family level,
much less is known about the histories of the individual paralogs.
It seems natural to ask, therefore, whether it is possible to devise
automatic methods to distinguish orthologs reliably and to phylo-
genetically map duplication and loss events within miRNA families.
So far, only changes in the number of family members have been
investigated systematically (e.g. [12]).
Several aspects of miRNA evolution remain poorly understood.
How prevalent are anti-sense miRNAs, i.e., those produced from a
common locus in both reading directions such as iab-4/iab-
8 [116]? Are evolutionary transitions from miR to miR* common?
Examples of this kind of “arm-switching” have been reported
e.g. in the mir-10 and mir-100 families [117]. Are there cases in
which the position of the precursor hairpin shifts relative to the
mature product, i.e., new miR or miR* are introduced? Finally,
there is of course the broad topic of conservation of miRNA
function and thus of their target sites. Answers to all these topics
require a reliable, accurate, and (reasonably) complete miRNA
annotation. Hence it is more than worth while to address the
many technical issues addressed in this contributions in future
research and to invest in tools and pipelines to process the ever
increasing wealth of miRNA data with much less user intervention
and expert curation.
Acknowledgements
This work was funded in part by the German Federal Ministery for
Education and Research (BMBF 031A538A, de.NBI/RBC),
CAVH was funded by the German Academic Exchange Service
(DAAD) (Forschungsstipendien-Promotionen in Deutschland,
2018/19 (Bewerbung 57299294), Ali M. Yazbeck was funded by
a doctoral stipend of the National Council for Scientific Research of
Lebanon (CNRS-L).
References
1. Ameres SL, Zamore PD (2013) Diversifying 9. Bråte J, Neumann RS, Fromm B, Haraldsen
microRNA sequence and function. Nature AAB, Tarver JE, Suga H, Donoghue PCJ,
Rev Mol Cell Biol 14:475–488 Peterson KJ, Ruiz-Trillo I, Grini PE, Shal-
2. Cerutti H, Casas-Mollano JA (2006) On the chian-Tabrizi K (2018) Unicellular origin of
origin and functions of RNA-mediated silenc- the animal microRNA machinery. Curr Biol
ing: from protists to man. Curr Genet 28:3288-3295, DOI 10.1016/j.
50:81–99 cub.2018.08.018
3. Shabalina SA, Koonin EV (2008) Origins and 10. Price N, Cartwright RA, Sabath N, Graur D,
evolution of eukaryotic RNA interference. Azevedo RB (2011) Neutral evolution of
Trends Ecol Evol 23:578–587 robustness in drosophila microRNA precur-
4. Bologna NG, Schapire AL, Palatnik JF (2013) sors. Mol Biol Evol 28:2115–2123
Processing of plant microRNA precursors. 11. Hertel J, Lindemeyer M, Missal K, Fried C,
Brief Funct Genomics 12:37–45 Tanzer A, Flamm C, Hofacker IL, Stadler PF,
5. Tarver JE, Donoghue PCJ, Peterson KJ The Students of Bioinformatics Computer
(2012) Do miRNAs have a deep evolutionary Labs 2004 and 2005 (2006) The expansion
history? Bioessays 34:857–866 of the metazoan microRNA repertoire. BMC
Genomics 7:15
6. Piatek MJ, Werner A (2014) Endogenous
siRNAs, regulators of internal affairs. Bio- 12. Hertel J, Stadler PF (2015) The expansion of
chem Soc Trans 42:1174–1179, DOI animal microRNA families revisited. Life
10.1042/BST20140068 5:905–920, DOI 10.3390/life5010905
7. Saçar Demirci MD, Bağcı C, Allmer J (2016) 13. Tanzer A, Stadler PF (2004) Molecular evo-
Differential expression of toxoplasma gondii lution of a microRNA cluster. J Mol Biol
microRNAs in murine and human hosts. In: 339:327–335
Leitão A, Enguita F (eds) Non-coding RNAs 14. Li SC, Chan WC, Hu LY, Lai CH, Hsu CN,
and Inter-kingdom Communication, Lin Wc (2010) Identification of homologous
Springer, Cham, pp 143–159, DOI microRNAs in 56 animal genomes. Genomics
10.1007/978-3-319-39496-1_9 96:1–9, DOI 10.1016/j.ygeno.2010.03.009
8. Moran Y, Agron M, Praher D, Technau U 15. Hertel J, Bartschat S, Wintsche A, Otto C,
(2017) The evolutionary origin of plant and The Students of the Bioinformatics Computer
animal microRNAs. Nature Eco Evol 1:0027, Lab 2011, Stadler PF (2012) Evolution of the
DOI 10.1038/s41559-016-0027 let-7 microRNA family. RNA Biology
9:231–241
16. Guerra-Assunção JA, Enright AJ (2012) 28. Campo-Paysaa F, Sémon M, Cameron RA,
Large-scale analysis of microRNA evolution. Peterson KJ, Schubert M (2011) MicroRNA
BMC Genomics 13:218, DOI 10.1186/ complements in deuterostomes: origin and
1471-2164-13-218 evolution of microRNAs. Evol Dev 13:15–27
17. McCreight JC, Schneider SE, Wilburn DB, 29. Marco A, Ninova M, Ronshaugen M,
Swanson WJ (2017) Evolution of microRNA Griffiths-Jones S (2013) Clusters of micro-
in primates. PLoS ONE 12:e0176,596, DOI RNAs emerge by new hairpins in existing
10.1371/journal.pone.0176596 transcripts. Nucleic Acids Res 41:7745–7752
18. Ambros V, Bartel B, Bartel DP, Burge CB, 30. Smalheiser NR, Torvik VI (2005) Mammalian
Carrington JC, Chen X, Dreyfuss G, Eddy microRNAs derived from genomic repeats.
SR, Griffiths-Jones S, Marshall M, Matzke Trends Genet 21:322–326
M, Ruvkun G, Tuschl T (2003) A uniform 31. Piriyapongsa J, Mariño Ramı́rez L, Jordan IK
system for microRNA annotation. RNA (2007) Origin and evolution of human micro-
9:277–279 RNAs from transposable elements. Genetics
19. Sempere LF, Cole CN, McPeek MA, Peterson 176:1323–1337
KJ (2006) The phylogenetic distribution of 32. Zhou X, Ruan J, Wang G, Zhang W (2007)
metazoan microRNAs: insights into evolu- Characterization and identification of micro-
tionary complexity and constraint. J Exp RNA core promoters in four model species.
Zoolog B Mol Dev Evol 306B:575–588 PLoS Comput Biol 3:e37
20. Heimberg AM, Sempere LF, Moy VN, Dono- 33. Roberts JT, Cooper EA, Favreau CJ, Howell
ghue PCJ, Peterson K (2007) MicroRNAs JS, Lane LG, Mills JE, Newman DC, Perry
and the advent of vertebrate morphological TJ, Russell ME, Wallace BM, Borchert GM
complexity. Proc Natl Acad Sci USA (2013) Continuing analysis of microRNA ori-
105:2946–2950 gins: Formation from transposable element
21. Heimberg AM, Cowper-Sallari R, Sémon M, insertions and noncoding RNA mutations.
Donoghue PC, Peterson KJ (2010) Micro- Mob Genet Elements 3:e27,755
RNAs reveal the interrelationships of hagfish, 34. Gim JA, Ha HS, Ahn KA, Kim DS, Kim HS
lampreys, and gnathostomes and the nature of (2014) Genome-wide identification and clas-
the ancestral vertebrate. Proc Natl Acad Sci sification of microRNAs derived from repeti-
USA 107:19,379–19,383 tive elements. Genomics Inform 12:261–267,
22. Wheeler BM, Heimberg AM, Moy VN, Sper- DOI 10.5808/GI.2014.12.4.261
ling EA, Holstein TW, Heber S, Peterson KJ 35. Cui J, You C, Chen X (2017) The evolution of
(2009) The deep evolution of metazoan microRNAs in plants. Curr Opin Plant Biol
microRNAs. Evol Dev 11:50–68 35:61–67, DOI 10.1016/j.pbi.2016.11.006
23. Fu X, Adamski M, Thompson EM (2008) 36. Liang H, Li W (2009) Lowly expressed
Altered miRNA repertoire in the simplified human microrna genes evolve rapidly. Mol
chordate, Oikopleura dioica. Mol Biol Evol Biol Evol 26:1195–1198
25:1067–1080 37. Meunier J, Lemoine F, Soumillon M,
24. Kozomara A, Birgaoanu M, Griffiths-Jones S Liechti A, Weier M, Guschanski K, Hu H,
(2019) miRBase: from microRNA sequences Khaitovich P, Kaessmann H (2012) Birth
to function. Nucleic Acids Res 47: and expression evolution of mammalian
D155–D162, DOI 10.1093/nar/gky1141 microrna genes. Genome Res 23:34–45
25. Bentwich I, Avniel AA, Karov Y, Aharonov R, 38. Lee CT, Risom T, Strauss WM (2007) Evolu-
Gilad S, Barad O, Barzilai A, Einat P, Einav U, tionary conservation of microRNA regulatory
Meiri E, Sharon E, Spector Y, Bentwich Z circuits: an examination of microrna gene
(2005) Identification of hundreds of con- complexity and conserved microRNA-target
served and nonconserved human microRNAs. interactions through metazoan phylogeny.
Nat Genet 37:766–770 DNA Cell Biol 26:209–218
26. Berezikov E, Thuemmler F, van Laake LW, 39. Niwa R, Slack FJ (2007) The evolution of
Kondova I, Bontrop R, Cuppen E, Plasterk animal microRNA function. Curr Opin
RH (2006) Diversity of microRNAs in human Genet Dev 17:145–150
and chimpanzee brain. Nat Genet 40. Prochnik SE, Rokhsar DS, Aboobaker AA
38:1375–1377 (2007) Evidence for a microRNA expansion
27. Lu J, Shen Y, Wu Q, Kumar S, He B, Shi S, in the bilaterian ancestor. Dev Genes Evol
Carthew RW, Wang SM, Wu CI (2008) The 217:73–77
birth and death of microRNA genes in Dro-
sophila. Nat Genet 40:351–355
41. Berezikov E (2011) Evolution of microRNA 52. Kozomara A, Griffiths-Jones S (2014) MiR-
diversity and regulation in animals. Nat Rev Base: Annotating high confidence micro-
Genet 12:846–860 RNAs using deep sequencing data. Nucleic
42. Dai Z, Chen Z, Ye H, Zhou L, Cao L, Acids Res 42:D68–D73, DOI 10.1093/
Wang Y, Peng S, Chen L (2009) Characteri- nar/gkt1181
zation of microRNAs in cephalochordates 53. Bernhart SH, Hofacker IL, Will S, Gruber
reveals a correlation between microRNA rep- AR, Stadler PF (2008) RNAalifold:
ertoire homology and morphological similar- improved consensus structure prediction for
ity in chordate evolution. Evol Dev 11:41–49 RNA alignments. BMC Bioinformatics 9:474
43. Velandia-Huerto CA, Gittenberger A, Brown 54. Wheeler TJ, Eddy SR (2013) nhmmer: DNA
FD, Stadler PF, Bermúdez-Santana CI (2016) homology search with profile HMMs. Bioin-
Automated detection of ncRNAs in the draft formatics 29:2487–2489
genome sequence of a basal chordate: The 55. Lorenz R, Bernhart SH, Höner zu
carpet sea squirt Didemnum vexillum. BMC Siederdissen C, Tafer H, Flamm C, Stadler
Genomics 17:591, DOI 10.1186/s12864- PF, Hofacker IL (2011) ViennaRNA Package
016-2934-5 2.0. Alg Mol Biol 6:26
44. Wang K, Dantec C, Lemaire P, Onuma TA, 56. Reiche K, Stadler PF (2007) RNAstrand:
Nishida H (2017) Genome-wide survey of Reading direction of structured RNAs in mul-
miRNAs and their evolutionary history in tiple sequence alignments. Alg Mol Biol 1:6
the ascidian, Halocynthia roretzi. BMC Geno- 57. Will S, Joshi T, Hofacker IL, Stadler PF, Back-
mics 18:314, DOI 10.1186/s12864-017- ofen R (2012) LocARNA-P: Accurate bound-
3707-5 ary prediction and improved detection of
45. Nozawa M, Miura S, Nei M (2010) Origins structured RNAs for genome-wide screens.
and evolution of MicroRNA genes in Drosoph- RNA 18:900–914
ila species. Genome Biol Evol 2:180–189, 58. Siebert S, Backofen R (2005) MARNA: multi-
DOI 10.1093/gbe/evq009 ple alignment and consensus structure predic-
46. Eddy SR, Durbin R (1994) RNA sequence tion of RNAs based on sequence structure
analysis using covariance models. Nucleic comparisons. Bioinformatics 21:3352–3359
Acids Res 22:2079–2088 59. Bernhart S, Hofacker IL, Stadler PF (2006)
47. Nawrocki EP, Eddy SR (2007) Query- Local RNA base pairing probabilities in large
dependent banding (QDB) for faster RNA sequences. Bioinformatics 22:614–615
similarity searches. PLoS Comput Biol 3:e56 60. Freyhult E, Gardner PP, Moulton V (2005) A
48. Nawrocki EP, Eddy SR (2013) Infernal 1.1: comparison of RNA folding measures. BMC
100-fold faster RNA homology searches. Bio- Bioinformatics 6:241, DOI 10.1186/1471-
informatics 29:2933–2035, DOI 10.1093/ 2105-6-241
bioinformatics/btt509 61. Clote P, Ferré F, Kranakis E, Krizanc D
49. Hertel J, de Jong D, Marz M, Rose D, (2005) Structural RNA has lower folding
Tafer H, Tanzer A, Schierwater B, Stadler PF energy than random RNA of the same dinu-
(2009) Non-coding RNA annotation of the cleotide frequency. RNA 11:578–591
genome of Trichoplax adhaerens. Nucleic 62. Zhang BH, Pan XP, Cox SB, Cobb GP,
Acids Res 37:1602–1615 Anderson TA (2006) Evidence that miRNAs
50. Kalvari I, Argasinska J, Quinones-Olvera N, are different from other RNAs. Cell and
Nawrocki EP, Rivas E, Eddy SR, Bateman A, Molec Life Sci 63:246–254
Finn RD, Petrov AI (2018) Rfam 13.0: shift- 63. Yazbeck AM, Stadler PF, Tout K, Fallmann J
ing to a genome-centric resource for (2019) Automatic curation of large compara-
non-coding RNA families. Nucleic Acids Res tive animal microRNA data sets. Bioinformat-
46:D335–D342, DOI 10.1093/nar/ ics 35:4553–4559, DOI 10.1093/
gkx1038 bioinformatics/btz271
51. Kalvari I, Nawrocki EP, Ontiveros-Palacios N, 64. Eggenhofer F, Hofacker IL, Höner zu Sieder-
Argasinska J, Lamkiewicz K, Marz M, Grif- dissen C (2016) RNAlien – unsupervised
fiths-Jones S, Toffano-Nioche C, Gautheret RNA family model construction. Nucleic
D, Weinberg Z, Rivas E, Eddy SR, Finn RD, Acids Res 44:8433–8441, DOI 10.1093/
Bateman A, Petrov AI (2020) Rfam 14: nar/gkw558
expanded coverage of metagenomic, viral
and microRNA families. Nucleic Acids Res 65. Lott SC, Sch€afer RA, Mann M, Backofen R,
49:D192-D200, DOI 10.1093/nar/ Hess WR, Voß B, Georg J (2018) GLASSgo –
gkaa1047 automated and reliable detection of sRNA
homologs from a single input sequence.
Front Genet 9:124, DOI 10.3389/ 77. Zhao BW, Zhou LF, Liu YL, Wan SM, Gao
fgene.2018.00124 ZXG (2017) Evolution of fish let-7 micro-
66. Yazbeck AM, Tout KR, Stadler PF, Hertel J RNAs and their expression correlated to
(2017) Towards a consistent, quantitative growth development in blunt snout bream.
evaluation of microRNA evolution. J Integra- Int J Mol Sci 18:646, DOI 10.3390/
tive Bioinf 14:20160,013, DOI 10.1515/jib- ijms18030646
2016-0013 78. Chen L, Heikkinen L, Emily Knott K,
67. Parra-Rincón E, Velandia-Huerto CA, Liang Y, Wong G (2015) Evolutionary con-
Fallmann J, Gittenberger AA, Brown Almeida servation and function of the human embry-
FD, Stadler PF, Bermúdez-Santana CI (2020) onic stem cell specific miR-302/367 cluster.
The genome of the “sea vomit” Didemnum Comp Biochem Physiol D 16:83–98, DOI
vexillum Submitted 10.1016/j.cbd.2015.08.002
68. Menzel P, Stadler PF, Gorodkin J (2011) 79. Trümbach D, Prakash N (2015) The con-
maxAlike: Maximum-likelihood based served miR-8/miR-200 microRNA family
sequence reconstruction with application to and their role in invertebrate and vertebrate
improved primer design for unknown neurogenesis. Cell Tissue Res 359:161–177
sequences. Bioinformatics 27:317–325 80. Desvignes T, Contreras A, Postlethwait JH
69. Demirci MDS, Baumbach J, Allmer J (2017) (2014) Evolution of the miR199-214 cluster
On the performance of pre-microRNA detec- and vertebrate skeletal development. RNA
tion algorithms. Nature Comm 8:300 Biol 11:281–294, DOI 10.4161/rna.28141
70. Williams PH, Eyles R, Weiller G (2012) Plant 81. Velandia-Huerto CA, Berkemer SJ,
MicroRNA prediction by supervised machine Hoffmann A, Retzlaff N, Romero Marroquı́n
learning using C5.0 decision trees. J Nucleic LC, Hernández Rosales M, Stadler PF, Ber-
Acids 2012:652,979, DOI 10.1155/2012/ múdez-Santana CI (2016) Orthologs, turn-
652979 over, and remolding of tRNAs in primates and
71. Hertel J, Stadler PF (2006) Hairpins in a fruit flies. BMC Genomics 17:617, DOI
haystack: Recognizing microRNA precursors 10.1186/s12864-016-2927-4
in comparative genomics data. Bioinformatics 82. Berkemer SJ, Hoffmann A, Murray CR, Sta-
22:e197–e202 dler PF (2017) SMORE: Synyeny MOdulator
72. Auyeung VC, Ulitsky I, McGeary SE, Bartel of repetitive elements. LIFE 7:42, DOI
DP (2013) Beyond secondary structure: pri- 10.3390/life7040042
mary-sequence determinants license pri- 83. Wang Y, Luo J, Zhang H, Lu J (2016) Micro-
miRNA hairpins for processing. Cell 152 RNAs in the same clusters evolve to coordi-
(4):844–858 nately regulate functionally related genes. Mol
73. Al Ait L, Yamak Z, Morgenstern B (2013) Biol Evol 33:2232–2247, DOI 10.1093/
DIALIGN at GOBICS—multiple sequence molbev/msw089
alignment using various sources of external 84. Eggenhofer F, Hofacker IL, Höner zu Sieder-
information. Nucleic Acids Research 41 dissen C (2013) CMCompare Webserver:
(W1):W3–W7, DOI 10.1093/nar/gkt283 Comparing RNA families via covariance mod-
74. Jia Y, Wang F, Yang GH, Wang FL, Ma YN, els. Nucleic Acids Res 41:W499–W503, DOI
Du ZW, Zhang JW (2006) Human micro- 10.1093/nar/gkt329
RNA clusters: Genomic organization and 85. Weinberg Z, Breaker RR (2011) R2R – soft-
expression profile in leukemia cell lines. Bio- ware to speed the depiction of aesthetic con-
chem Biophys Res Comm 349:59–68, DOI sensus RNA secondary structures. BMC
10.1016/j.bbrc.2006.07.207 Bioinformatics 12:3, DOI 10.1186/1471-
75. Tanzer A, Riester M, Hertel J, Bermudez- 2105-12-3
Santana CI, Gorodkin J, Hofacker IL, Stadler 86. Sankoff D (1975) Minimal mutation trees of
PF (2010) Evolutionary genomics of micro- sequences. SIAM J Appl Math 28:35–42
RNAs and their relatives. In: Caetano-Anolles 87. Balogh G, Bernhart SH, Stadler PF, Schor J
G (ed) Evolutionary Genomics and Systems (2020) A probabilistic version of Sankoff’s
Biology, Wiley-Blackwell, Hoboken, NJ, pp maximum parsimony algorithm. J Bioinf
295–327 Comput Biol 18:2050004, DOI 10.1142/
76. Guo L, Yang S, Zhao Y, Wu Q, Chen F S0219720020500043, in press
(2013) Dynamic evolution of mir-17-92 88. Mohammed J, Flynt AS, Panzarino AM, Hos-
gene cluster and related miRNA gene families sein Mondal MM, DeCruz M, Siepel A, Lai
in vertebrates. Mol Biol Rep 40:3147–3153, EC (2018) Deep experimental profiling of
DOI 10.1007/s11033-012-2388-z microRNA diversity, deployment, and
evolution across the Drosophila genus. 100. Westholm JO, Lai EC (2011) Mirtrons:
Genome Res 28:52–65, DOI 10.1101/ microRNA biogenesis via splicing. Biochimie
gr.226068.117 93:1897–904, DOI 10.1016/j.
89. Ylla G, Fromm B, Piulachs MD, Belles X biochi.2011.06.017
(2016) The microRNA toolkit of insects. Sci 101. Wen J, Ladewig E, Shenker S, Mohammed J,
Rep 6:37,736, DOI 10.1038/srep37736 Lai EC (2015) Analysis of nearly one thou-
90. Tarver JE, Taylor RS, Puttick MN, Lloyd GT, sand mammalian mirtrons reveals novel fea-
Pett W, Fromm B, Schirrmeister BE, Pisani D, tures of dicer substrates. PLoS Comput Biol
Peterson KJ, Donoghue PCJ (2018) Well- 11:e1004,441, DOI 10.1371/journal.
annotated microRNAomes do not evidence pcbi.1004441
pervasive miRNA loss. Genome Biol Evol 102. Titov II, Vorozheykin PS (2018) Comparing
10:1457–1470, DOI 10.1093/gbe/evy096 miRNA structure of mirtrons and
91. Höner zu Siederdissen C, Hofacker IL (2010) non-mirtrons. BMC Genomics 19 S3:114,
Discriminatory power of RNA family models. DOI 10.1186/s12864-018-4473-8
Bioinformatics 26:i453–i459, DOI 10.1093/ 103. Bortolamiol-Becet D, Hu F, Jee D, Wen J,
bioinformatics/btq370 Okamura K, Lin CJ, Ameres SL, Lai EC
92. Fromm B, Domanska D, Hye E, (2015) Selective suppression of the splicing-
Ovchinnikov V, Kang W, Aparicio-Puerta E, mediated MicroRNA pathway by the terminal
Johansen M, Flatmark K, Mathelier A, uridyltransferase Tailor. Mol Cell
Hovig E, Hackenberg M, Friedl€ander MR, 59:217–228, DOI 10.1016/j.
Peterson KJ (2019) MirGeneDB 2.0: the molcel.2015.05.034
metazoan microRNA complement. Nucleic 104. Rorbach G, Unold O, Konopka BM (2018)
Acids Res 48:D132–D141, DOI 10.1093/ Distinguishing mirtrons from canonical miR-
nar/gkz885 NAs with data exploration and machine
93. Cifuentes D, Xue H, Taylor DW, Patnode H, learning methods. Sci Rep 8:7560, DOI
Mishima Y, Cheloufi S, Ma E, Mane S, Han- 10.1038/s41598-018-25578-3
non GJ, Lawson ND, Wolfe SA, Giraldez AJ 105. Da Fonseca BHR, Domingues DS, Paschoal
(2010) A novel miRNA processing pathway AR (2019) mirtronDB: a mirtron knowledge
independent of Dicer requires Argonaute2 base. Bioinformatics 35:3873–3874, DOI
catalytic activity. Science 328:1694–1698 10.1093/bioinformatics/btz153
94. Kim VN, Han J, Siomi MC (2009) Biogenesis 106. Flynt AS, Greimann JC, Chung WJ, Lima
of small RNAs in animals. Nat Rev Mol Cell CD, Lai EC (2010) MicroRNA biogenesis
Biol 10:126–139, DOI 10.1038/nrm2632 via splicing and exosome-mediated trimming
95. Li L, Liu Y (2011) Diverse small non-coding in Drosophila. Mol Cell 38:900–907, DOI
RNAs in RNA interference pathways. Meth- 10.1016/j.molcel.2010.06.014
ods Mol Biol 764:169–182, DOI 10.1007/ 107. Langenberger D, Bartschat S, Hertel J,
978-1-61779-188-8_11 Hoffmann S, Tafer H, Stadler PF (2011)
96. Okamura K (2012) Diversity of animal small MicroRNA or not MicroRNA? In: de Souza
RNA pathways and their biological utility. ON, Telles GP, Palakal MJ (eds) Advances in
Wiley Interdiscip Rev RNA 3:351–368, DOI Bioinformatics and Computational Biology,
100.1002/wrna.113 6th Brazilian Symposium on Bioinformatics,
97. Winter J, Link S, Witzigmann D, BSB 2011, Springer, Berlin, Heidelberg, Lec-
Hildenbrand C, Previti C, Diederichs S ture Notes in Computer Science, vol 6832, pp
(2013) Loop-miRs: active microRNAs gener- 1–9
ated from single-stranded loop regions. 108. Langenberger D, Çakir MV, Hoffmann S,
Nucleic Acids Res 41:5503–5512, DOI Stadler PF (2012) Dicer-processed small
10.1093/nar/gkt251 RNAs: Rules and exceptions. J Exp Zool:
98. Okamura K, Ladewig E, Zhou L, Lai EC Mol Dev Evol 320:35–46
(2013) Functional small RNAs are generated 109. Roberts JT, Cardin SE, Borchert GM (2014)
from select miRNA hairpin loops in flies and Burgeoning evidence indicates that micro-
mammals. Genes Dev 27:778–792, DOI RNAs were initially formed from transposable
10.1101/gad.211698.112 element sequences. Mob Genet Elements 4:
99. Okamura K, Hagen JW, Duan H, Tyler DM, e29,255
Lai EC (2007) The mirtron pathway gener- 110. Taft RJ, Glazov EA, Lassmann T,
ates microRNA-class regulatory RNAs in Dro- Hayashizaki Y, Carninci P, Mattick JS (2009)
sophila. Cell 130:89–100, DOI 10.1016/j. Small RNAs derived from snoRNAs. RNA
cell.2007.06.028 15:1233–1240, DOI 10.1261/rna.1528909
111. Falaleeva M, Stamm S (2013) Processing of halves: The new players in cancers. Cancer
snoRNAs as a new source of regulatory Letters 452:31–37, DOI 10.1016/j.
non-coding RNAs: snoRNA fragments form canlet.2019.03.012
a new class of functional RNAs. Bioessays 115. Chak LL, Mohammed J, Lai EC, Tucker-
35:46–54, DOI 10.1002/bies.201200117 Kellogg G, Okamura K (2015) A deeply con-
112. Scott MS, Ono M (2011) From snoRNA to served, noncanonical miRNA hosted by ribo-
miRNA: Dual function regulatory somal DNA. RNA 21:375–384, DOI
non-coding RNAs. Biochimie 10.1261/rna.049098.114
93:1987–1992, DOI 10.1016/j. 116. Hui JH, Marco A, Hunt S, Melling J,
biochi.2011.05.026 Griffiths-Jones S, Ronshaugen M (2013)
113. Maute RL, Schneider C, Sumazin P, Structure, evolution and function of the
Holmes A, Califano A, Basso K, Dalla-Favera bi-directionally transcribed iab-4/iab-
R (2013) tRNA-derived microRNA modu- 8 microRNA locus in arthropods. Nucleic
lates proliferation and the DNA damage Acids Res 41:3352–3361, DOI 10.1093/
response and is down-regulated in B cell lym- nar/gks1445
phoma. Proc Natl Acad Sci USA 117. Griffiths-Jones S, Hui JH, Marco A, Ron-
110:1404–1409, DOI 10.1073/ shaugen M (2011) MicroRNA evolution by
pnas.1206761110 arm switching. EMBO Rep 12:172–177,
114. Zhu L, Ge J, Li T, Shen J Yijing anf Guo DOI 10.1038/embor.2010.191
(2019) tRNA-derived fragments and tRNA
Chapter 12
Ensemble Classifiers for Multiclass MicroRNA Classification

Luise Odenthal, Jens Allmer, and Malik Yousef
Abstract
Gene regulation is of utmost importance to cell homeostasis; thus, any dysregulation in it often leads to
disease. MicroRNAs (miRNAs) are involved in posttranscriptional gene regulation and consequently, their
dysregulation has been associated with many diseases.
MiRBase version 21 contains microRNAs from about 200 species organized into about 70 clades. It has
been shown that not all miRNAs collected in the database are likely to be real and, therefore, novel routes to
delineate between correct and false miRNAs should be explored. We introduce a novel approach based on
k-mer frequencies and machine learning that assigns an unknown/unlabeled miRNA to its most likely
clade/species of origin. A simple way to filter new data would be to ensure that the novel miRNA
categorizes closely to the species it is said to originate from. For that, an ensemble classifier of multiple
two-class random forest classifiers was designed, where each random forest was trained on one species–clade
pair. The approach was tested with different sampling methods on a dataset that was taken from miRBase
version 21 and it was evaluated using a hierarchical F-measure. The approach predicted 81% to 94% of the
test data correctly, depending on the sampling method. This is the first classifier that can classify miRNAs to
their species of origin. This method will aid in the evaluation of miRNA database integrity and analysis of
noisy miRNA samples.
Key words miRNA, Categorization, Machine learning
1 Introduction
MicroRNAs (miRNAs) play a key role in posttranscriptional gene

regulation and therefore are involved in controlling cell homeosta-
sis and cellular processes [1]. Mature miRNAs are short (18–24 nt),
single-stranded, noncoding RNA molecules. At least one-third of
all human genes [2] and all human regulatory pathways [3] are
affected by them and they have been implicated in many diseases
such as cancer [4]. MicroRNAs, like any other RNA, need to be
transcribed, however, they are not translated into protein, but are
involved in the regulation of protein expression. A characteristic of
a miRNA is that it forms a hairpin-like structure by folding. After
being processed by different enzymes, the hairpins are transformed
into single-stranded mature miRNAs incorporated in the RISC
235
236 Luise Odenthal et al.
complex. There, the mature miRNA is involved in providing the

recognition of its targets and thereby enables posttranscriptional
gene regulation.
MicroRNAs exist in many different species ranging from
sponges to mammals. They can also be classified into miRNA
families based on function or sequence [5], This implies that the
presence of a specific miRNA provides taxonomic information
[6]. For example, the mir-17 family arises from duplication and
loss of individual members as well as duplications of entire clusters
[7]. MicroRNA evolution is beyond the scope of this chapter, but
more information on miRNA families can be found in [8] of this
volume.
Many miRNAs have been confirmed experimentally and there
exist many miRNA families so that it is essential to keep track of all
miRNAs including their meta information. The current reference
database for this is miRBase [7]. However, some studies lead to the
question if there might be contaminants registered in miRBase
[9, 10]. For RNA-seq measurements, Bağci and Allmer [11]
showed that the discovery of plant miRNAs in human body fluids
was clearly due to the contamination of the samples. This was
further supported by the high correlation of the set of shared
transcripts (inter- and intraspecies) in the particular samples.
To our knowledge, there exists no tool, which automates the
detection of contaminants in databases like miRBase or in
biological samples. As it appears to be difficult to find them with
traditional methods, there is a clear need for such a tool. We
therefore propose a machine learning approach to engage this
problem.
For this approach, we make use of the fact that miRNA can be
distinguished by their sequence features such as k-mers [12–15],
despite its highly conserved structure [16] even between species
[12]. One approach that could distinguish two species by training a
random forest classifier on miRNA k-mer frequencies was devel-
oped by [13]. Further experiments confirmed that miRNA precur-
sors from different species can be distinguished using sequence
features and most prominently k-mers [14, 15, 17]. One important
part of a miRNA is the mature sequence which is directly tied to its
targets. Therefore, it also seems to be possible to distinguish species
based on their target sequences [18, 19] which also extends to
ncRNA predictions [20]. Whether it is possible to distinguish
between large numbers of species when extending the aforemen-
tioned approaches allowing, for example, the automated detection
of contaminants, is a natural question emerging from the findings.
If a miRNA is classified to a species that differs in taxonomy above a
threshold from the species it is thought to originate from, by an
automated system, could help to determine contaminants.
Categorizing MicroRNAs to their Species of Origin 237
Here, we propose a machine learning approach that is able to

perform a multiclass classification. Our novel approach was devel-
oped using and applied to the 126 different species/clades avail-
able. It is an ensemble classifier that consists of multiple random
forest models (RF) [21]. Where each RF is trained on one unique
species–clade pair, resulting in 7875 random forest models. To
classify a miRNA, it is scored by all RFs models. The resulting
scores are combined to obtain the probability of the miRNA to
belong to either species. As the approach can distinguish 126 spe-
cies/clades the final species/clade is computed from those
probabilities.
In this way, our approach can assign a miRNA to its clade of
origin. To rule out contamination the target and predicted clade
can be compared using various distance metrics. If the two clades
are too different in taxonomy, the miRNA is probably the result of
contamination. One key advantage of this approach over most
miRNA detection algorithms is its independence of arbitrary
pseudo negative data [22].
We achieved a hierarchical F-measure of 0.81 to 0.94, depend-
ing on the sampling method with our approach. This shows that it
is possible to classify miRNA to its species of origin and that
approach is a valuable contribution to find contaminants and to
understand the evolution of miRNA a bit better.
2 Methods
2.1 Data We retrieved the data set from miRBase version 21. Initially, all
28,645 hairpins sequences were downloaded. 3553 hairpins were
later removed during the cleaning process. The final dataset
contained 25,092 miRNA sequences of 126 clades that illustrated
in Fig. 1.
2.2 Data Cleaning MicroRNAs can be grouped into families [8] and that already
indicates that there are intraspecies and interspecies homologies
between miRNAs. Having highly similar sequences in the training
dataset biases the resulting classifiers toward these sequences.
Therefore, it is important to clean the data from such biases. Here
we used USEARCH’s [23] UCLUST algorithm to cluster
sequences by their similarity. All hairpin sequences were merged
into one FASTA file that was subjected to the UCLUST algorithm,
which clusters the sequences so that one cluster contains all
sequences that are similar. We set the similarity threshold to 0.9,
where 1.0 refers to complete equality and zero to no similarity.
From each cluster, we kept only one sequence. This procedure
resulted in 25,092 different miRNAs.
Fig. 1 Phylogenetic tree. Phylogenetic structure of the data downloaded from miRBase after filtering for a
minimum of 100 hairpins per species. 0.30 under the tree visualizes the branch length metric according to
phylo.io [36] which was used to create the figure. Gray lines lead to leaves (here species) and blue lines lead to
higher-order phyla such as families
Fig. 2 Schematic overview of our ensemble classification approach. First, a preprocessing step is performed
to clean and group the data, afterward, the k-mer frequencies are computed. Then the data is split into
random training and test sets over tenfold cross-validation. For every unique clade pair in the training set, one
random forest classifier is trained. Then the whole test set is classified by all trained classifiers which results
in a score vector. As the score vector contains multiple scores for the same clade, this vector is compressed to
an MP-Vector that contains one probability for each clade. A maximum selector was used for result integration
before evaluation
2.3 Machine In order to categorize species into their clade of origin, a formida-
Learning Approach ble multiclass classification problem, including hierarchical rela-
tionships among the data, a classification strategy was devised
(Fig. 2).
The species categorization system consists of multiple compo-
nents, with the first one being preprocessing:
1. Group the data hierarchically into their species/clades accord-
ing to the taxonomic tree, so that each clade of the data is made
up of one clade of the taxonomic tree. This should result in a
hierarchical dataset, embedding each clade within its ancestor
clade (Fig. 1).
2. Clean the data of every group using UCLUST with a similarity
threshold of 0.9.
3. Remove all clades with less than 100 members.
The second component transforms the data sets into the k-mer
feature space (k ≦ 3). The k-mer frequency describes how often a
specific k-mer appears in a given sequence relative to the length of

that sequence length, more precisely relative to length-(k1). To
calculate the frequencies of these k-mers in a sequence, one simply
counts how often the particular k-mer appears in a sequence and
divides it by the possible number of k-mers for that sequence.
1. Create a k-mer feature vector of 1, 2, and 3-mers. This results
in an 84-dimensional feature vector including A—U up to
AAA—UUU.
2. Represent each sequence by the frequencies of the k-mer fea-
ture vector [24].
The k-mer features could be directly used with a multiclass
classifier. However, transforming the data to a different space
may increase the performance. The transformation first gener-
ates a classifier for all pairwise clades in the training dataset. For
each a random forest classifier was trained on one species–clade
pair:
3. Generate a set of all pairs of the 126 clades. S ¼ {(ci, cj),
i ∈ {1, 2, . . ., 126}}. It follows that 7875 pairs (126 125/
2 ¼ 7875) needed to be generated.
4. For each member in S:
(a) Train one RF on the two species/clades, where one clade
functions as the positive and the other as the negative
class.
(b) Store the model for downstream processes.
We have used different approaches for balancing the
unbalanced data typically existing among pairs of clades
and species and settled on using the Synthetic Minority
Oversampling Technique (SMOTE) [25]. We used the
default parameters of KNIME’s [26] WEKA random for-
est classifier except for the number of trees, which we set
to 50. The last component is the scoring stage in which
the following steps are applied to a given miRNA
sequence that is classified to one of the 126 clades:
5. Let the vector v be the representation of the given sequence in
the k-mer feature space according to the second component of
the system.
6. Apply all trained models to v. The output of each RF is the
probability of v to belong to the positive class (species/clade)
and the probability to belong to the negative class.
7. Combine all scores that represent the probability of a specific
species/clade in one score vector (125 dimensions).
8. Repeat step a through b for every score vector:
(a) Fit a probability density function (pdf) to the score vector.
(b) Choose the most probable value according to the pdf and
store/recall its species/clade. This value represents the
membership probability of the miRNA to belong to the
species/clade.
9. Concatenate all values from (b) to a membership probability
vector (MP-Vector), so that the dimension represents the spe-
cies/clade and the associated membership probability.
10. Use a weak maximum selector to get the final result from the
MP-Vector:
(a) Select the biggest membership probability maxmp in the
MP-vector.
(b) Choose all membership probabilities in the MP-vector
that are bigger than the maxmp- 0.05 and store their
species/clades.
(c) Choose the species/clade on the lowest taxonomic layer.
11. Compare the predicted species/clade to the target species/
clade with an appropriate distance measure.
12. If the difference is bigger than some threshold, mark it as a
possible contaminant.
In the following, the parts of the categorization system will be
discussed further.
2.3.1 Sampling Methods Imbalanced data such as the data in this study can cause problems
when applying machine learning algorithms. In effect, classification
could be biased in favor of the majority class. Undersampling the
majority class or oversampling the minority class are typical reme-
dies, producing balanced classes. Here we used SMOTE sampling
to balance the classes. Additionally, a custom hierarchically
informed random sampling strategy was developed for this study.
This custom hierarchically informed random sampling method is an
undersampling strategy where the size of the major class is reduced
by choosing n random samples from it, where n is the number of
samples in the minor class. As multiple classifiers for one species
were trained, it might outbalance the information loss due to
undersampling. The special part of this random sampling technique
is that it can balance classes that are made up of several subclasses
(e.g., species in a genus). Therefore, the sampling strategy ensures
that every subclass is represented equally in the drawn set as follows
(Figure 3): If one wants to draw n random samples from the dataset
that is made of m subclasses, one draws l ¼ mn random samples from
each subclass. If a subclass contains only h samples and h < l, then
the difference l h is divided among the remaining subclasses.
Basically this method tries to take an equal random portion of
each subclass contained in the clade. Since some subclades contain
fewer samples, only the maximum number of the subclasses can be
Fig. 3 Balancing multiclass data. Example of how the method drawer balances two clades that are composed
of subclades using undersampling. The first clade (upper) is the smaller one with 17 members from
2 subgroups (purple and orange) so it is kept how it is. The second clade containing 22 members from
three subgroups (red, green, and blue) is undersampled to 17 samples to make it equally sized to the first one.
As there are three subgroups (SG1, SG2, and SG3) one will get six samples from each of them (17/3 ¼ 5.67).
Since SG1 only consists of five samples and SG3 of three samples, both of them are used completely and three
more samples are used from SG1
used. Therefore, it is still possible to produce an imbalance between

the different subclasses. This procedure was realized in java as a
custom KNIME node called Drawer.
In oversampling, artificial data is generated out of the existing
data. A very common tool to generate such data is Synthetic
Minority Oversampling Technique [25]. To generate new samples
for the minor class, SMOTE looks at every sample in the minor
class and computes its k-nearest neighbors. Smote’s default for k is
5. According to the difference between the major and minor clas-
ses, a suitable number of the minor class samples are randomly
selected. The k-nearest neighbors of these points are used to inter-
polate between them and generate a new sample. In this way, the
problem of overfitting is avoided.
2.3.2 Model Parameters For the random forests, we used the default values set by KNIME
[27] for all parameters except for the number of trees, which we set
to 50. We elected to use 50 trees as a compromise between runtime
and performance because the performance increased with the num-
ber of trees but so did the run time needed to train the RF models.
It took 69 s to train the random forest model using 50 trees and
143 s for 100 trees (Lenovo ThinkPad with an Intel Core i7 and
3 GB RAM). As we had to train 7875 such trees, we chose a

runtime of 151 h over 320. Additionally, the overall process had
to be repeated several times so that the longer runtime would be a
limiting factor.
2.3.3 Generating the During the training stage we have generated 7875 RF models that
Score Vector (SV Vector) were trained on each distinct pair of clades and species (126 125/
2 ¼ 7875). The models will be used for the task of categorization.
After training the RF models they can be used to obtain the
score vector (SV). Every dimension of this vector represents the
output of one RF and the value is the probability of a miRNA to
belong to the classifiers positive or negative class (So every RF has
two dimensions in the score vector). That is a transformation from a
k-mer frequency vector to a score vector. Recall that each RF is
trained to distinguish between two clades/species (speciesA, spe-
ciesB). So when a k-mer feature vector of a sample s is fed into a
classifier, it outputs two scores p(speciesA) and p(speciesB). These
scores represent the probability according to the classifier that the
sample s is in species A, respectively species B. A more visual
explanation can be seen in Figure 4.
Since p(speciesA) ¼ 1 p (speciesB) a value of zero for p(spe-
ciesA) means that it is species B. Therefore a value of 0.5 represents
absolute uncertainty.
Scoring an input k-mer vector with all RFs in this way results in
a 15,750-dimensional vector. As one RF was trained for every
unique pair of the 126 clades, there are 7875 classifiers in total in
this study, which will produce a two-dimensional vector of 7875 SV
which means 2 7875 ¼ 15,750 score vectors.
The problem with the obtained 15,750 scores is that they
include 126 scores for each species/clade. Therefore, they need to
be combined to obtain a final score for the corresponding species/
clade. To combine these scores, a probability density function is
fitted to the values, and the most likely value according to that
function is returned (Figure 5). This value is the membership
probability for a miRNA to belong to a specific species/clade.
After doing this for every clade a membership probability vec-
tor (MP-vector) is obtained. In this MP-vector every dimension
represents one species and the value whether the specific sample
belongs to that species.
2.3.4 Selecting the Final Different methods were used to determine the final clade from the
Result from the MP-Vector MP-vector. “Maximum Selector” and “Weak Maximum Selector”
are the two approaches that were selected in this study because they
achieved the best results. These approaches will be detailed in the
following.
Fig. 4 Feature generation. The transformation from the k-mer frequency feature vector (top) to a score feature
vector (bottom). The frequency feature vector is transformed into a score feature vector by the random forest
classifiers. Each RF classifier (middle) gives a probability for the frequency feature vector to belong to either of
the clades it was trained on. All the scores from all classifiers are concatenated to form one score feature
vector
2.3.5 Maximum Selector The Maximum Selector predicts the final clade from the MP-vector
simply by choosing the maximum score in the MP-vector. The
corresponding dimension will be the final clade (see Figure 6).
The mathematical operation is defined as:
Pred ðpm Þ ¼ Pred ðpm 1 , . . . , pmn Þ ¼ max ðpm1 , . . . , pmn Þ
where pm ¼ pm1, . . , pmn is an MP-vector with n dimensions and
max() is the maximum function. The maximum selector performed
best when using random sampling, however, results obtained with
SMOTE sampling were not as successful. The target clade did not
have the highest value in all cases because its ancestor clades had a
similarly high or even higher value the target clade (Figure 6). That
could be expected considering every clade is a subclass of its ances-
tor clade.
Fig. 5 Feature transformation. Application of the probability density function (pdf) to the classification scores to
create the membership probability vector (MP-vector, bottom). The classifiers (above tables) are aggregated
per species using the pdf and the result is represented in the MP-vector. Here n samples lead to n MP-vectors
displayed in one table
Fig. 6 Sample classification result. (a) Tree representing the average MP-value for the target clade (dark gray
node) and taxonomic relatives. The tree structure and labels of the nodes represent the taxonomic relation-
ship. In this case, random sampling was used; (b) Tree representing the average MP-value for the target clade
(dark gray) and taxonomic relatives. In this case SMOTE sampling was used
2.3.6 Weak Maximum A variant of the maximum selector, the weak maximum selector,
Selector was used, which showed better results with SMOTE sampling. This
version of the maximum selector is not as strict and identifies all
clades that have a score that is at least as large as a threshold. For this
study, the threshold was chosen to be the maximum score 0.05:
Threshold ¼ max ðpmÞ 0:05

From all the clades that have a higher score than the threshold
the one on the lowest taxonomic layer is chosen as the result
(Figure 6).
2.4 Testing the To ensure that the model was trained correctly we tested its perfor-
Ensemble Classifier mance. If the RFs were trained on an unsuitable dataset or if their
parameters were chosen incorrectly or the sampling method was
not suitable for the training set, wrong results might be obtained.
To check the performance, we used 10% of the data as a test set
and the other 90% as a training set (Figure 2). Then the training set
was used to train the RFs (Machine Learning Approach steps 1–6)
and for each sample in the test set steps, 7–12 were performed.
Then the species/clade annotated in miRBase (target species/
clade) was compared to the predicted species/clade using the hier-
archical F-measure (hF-measure) [27] to establish performance. We
used tenfold Monte Carlo cross-validation [28] to train and test the
ensemble classifier. In every fold 10% of the dataset was selected as
the test set. The test and training sets were selected by a custom
made stratified random selection method that makes sure all clades
are represented in the same ratio as in the original dataset (see
Subheading 2.3.1). In order to ensure that the sampling methods
are appropriate, the correlation between annotated and predicted
species/clades can be evaluated. Training sets were grouped by
species/clade and for each group the hierarchical F-measure was
computed. If the group size is positively correlated with the group’s
F-measure, it indicates that the sampling method is not suitable and
that small species/clade groups are underrepresented in the train-
ing set. This is because the hF-measure should be independent of
the group size.
2.5 Evaluation There are various evaluation measures in machine learning such as
accuracy, precision, recall, and F-measure. However, the hierarchi-
cal structure and the imbalanced data in this study require adapta-
tions to those measures. The average true positive rate can be used
for imbalanced classes [29]. It calculates the true positive rate per
class and uses the average as the final evaluation measure:
1X
n
AvgTPR ¼ TPR i
n i¼0
where n is the number of classes and TPRi is the true positive rate of
class i. Thereby every class is weighted equally in the computation
of the overall accuracy measurement.
To obtain an accurate evaluation measure that is applicable to
hierarchical data, we adapted common measures such as the true
positive rate. Otherwise we would miss many results in higher layers
of the taxonomic tree. For instance, if a gorilla is classified correctly,
it is also correctly assigned to all ancestor clades (Hominidae,

Primates, ...).
To adapt these measures, for each clade in every hierarchical
layer TP, FP, TN, and FN were computed with a one-vs-rest
scheme. Then those values were used to compute the hierarchical
TPR. A sample was classified as TP if it was predicted correctly and
TN if it was predicted into the clade that was not the annotated
clade and not part of the taxonomic hierarchy. However, we mainly
used the hierarchical F-measure (hF) proposed by [27], which takes
into account that predictions which are taxonomically closer related
to the target, are better than predictions that are distantly related to
the target. For example if a gorilla was identified as a Hominidae
but not as a primate, it would be worse than if it was identified as a
primate. To calculate the hF measure, first the hierarchical precision
and recall need to be computed. The precision and recall are defined
as follows:
for any sample di that belongs to class Ci and is predicted as
class Di, the sets Ci and Di are extended with their corresponding
ancestors:
C 0i ¼ U c k ∈C i Ancestors ð c k Þ, D 0i ¼ U d l ∈Di Ancestors ð d l Þ:
Then the (micro-average) hP (hierarchical precision) and hR
(hierarchical recall) are calculated as:
P
j C 0 \ D0 j
hP ¼ iP i 0 i
i j Di j
P
j C 0 \ D0 j
hR ¼ iP i 0 i
i j Ci j
Those measures are then used to calculate the hF-measure

(hierarchical F-measure).
ð β2 þ 1 Þ hP hR
hF β ¼ , β∈½ 0, þ1Þ
β² hP þ hR
where β was set to 1.
2.6 Implementation The training and testing pipeline was implemented in KNIME [26]
which is an open-source data analytics platform. KNIME includes
many machine learning methods and provides access to WEKA
[30] and other tools. For our ensemble classifier we used KNIME’s
SMOTE implementation and the Knime WEKA random forest
classifier package. The other sampling methods as well as all other
parts of the ensemble classifier and the evaluation were custom
implementations leveraging KNIME’s python node. The project
was realized with three KNIME workflows, one that trains the RF
models, one that performs steps one through 12 in the Subheading
2, and the third one which evaluates the results.
3 Results
All experiments were run on a Lenovo ThinkPad with an Intel Core

i7 and 3 GB RAM. The input data used in this study is imbalanced
(see Subheading 2.1). We tested three balancing approaches and
settled for SMOTE sampling which led to the best results (see
Figure 6 for an example). Two datasets were considered. The first
dataset was downloaded from miRBase version 21 and cleaned as
described above. The second dataset consists of all samples that
were newly introduced in miRBase version 22 and that remained
after the cleaning process. The latter dataset was used for testing,
only. In short, the miRBase v21 data was divided into 90% training
and 10% testing, and the miRNAs new in miRBase v22 were used as
an additional test set. An overview of all results in more detail can be
seen in Table 1.
To deal with the imbalanced datasets we considered two differ-
ent sampling strategies to balance the different clade/species, a
random undersampling method, and SMOTE. The better results
were obtained with SMOTE, it shows better performance on all
datasets, except for the AvgTPR measurement on the newMirBase
dataset. The lower hF value of random sampling is a result of the
undersampling, which leads to an underrepresentation of small
clades/species. This was confirmed by calculating the correlation
between clade/species size and hF measure (r ¼ 0.7). Therefore
SMOTE was the better alternative and was used throughout the
study. We tried different methods to obtain the final results from
the MP-vectors. The best results were achieved with the maximum
and the weak maximum selector. Where the maximum selector was
used with random sampling and the weak maximum selector with
SMOTE. The maximum selector achieved hF scores between 0.54
and 0.81 for the first and second datasets, whereas the weak maxi-
mum selector achieved hF scores between 0.67 and 0.94. Therefore
the best alternative was the weak maximum selector with SMOTE.
In order to assess the classification quality, the rank assigned by
the maximum selector was analyzed. The rank of a species/clade
can be obtained, by sorting the dimensions of the MP-vector by
their values. So the most likely species will get rank one and the least
likely rank 126. The rank of the correct species is then the rank of
the species in the sorted MP-vector. That means, that the lower the
rank, the less likely it is falsely classified. The results show that
almost 100% of the correct clades of the samples had a rank of at
least 10 using SMOTE sampling (Figure 7). The correct classifica-
tions were assigned low ranks and the majority is ranked below five
(Figure 7).
To further analyze how closely related the predicted clade/
species and the true clade/species are taxonomically, a taxonomic
similarly was calculated. To calculate the taxonomic similarity, the
Table 1
Summary of the classification results. The different measures used are the average true positive rate
(avg TPR), the hierarchical precision (hP), recall (hR), and F-measure (hF) as well as the average
hierarchical precision (avg hP), recall (avg hR), and F-measure (avg hF) that are better suited for
imbalanced classes. The comments column indicates which sampling and methods to select the final
result were used. Also BLAST was used to compare our results to traditional methods
Dataset/sampling Avg TPR hP hR hF Avg hP Avg hR Avg hF

Dataset1- random sampling 0.72 0.84 0.78 0.81 0.86 0.84 0.87
Dataset1- SMOTE 0.83 0.97 0.9 0.94 0.96 0.91 0.94
Dataset2- random sampling 0.31 0.53 0.56 0.54 0.63 0.61 0.63
Dataset2- SMOTE 0.3 0.74 0.61 0.67 0.77 0.63 0.7
BLAST 0.39 0.83 0.83 0.83 0.82 0.83 0.83
Fig. 7 Ranks of correct classifications. Ranks established by the maximum selector for dataset 1 using SMOTE
sampling
set of common ancestors of the target and predicted species C was

computed, as well as the ancestors of the predicted class Pb that are
not shared with the target class and the ancestors of the target class
b that are not shared with the predicted class. The similarity is then
T
calculated as follows:
jC j
similarity ¼
b j þj c
jC jþjP Cj
While the majority was correctly classified and hence had a

similarity of one, most of the wrong classifications had at least a
similarity of 60% with the true species/clade (Figure 8).
Since this measure is biased towards the layer on which the
target and predicted class are located. For example if the target class
was a gorilla and the predicted class Homo sapiens, their relationship
according to the taxonomy tree would be the same as if the target
species was Deuterosoma and the predicted species Ecdysozoa. How-
ever, the similarity measure for Gorilla vs. Homo sapiens would be
bigger than Deuterosoma versus Ecdysozoa. Therefore the phyloge-
netic distance by layers in the taxonomy tree was computed in
addition as:

layer distance ¼ max j Pj, b jT
bj
The results for the layer_distance show that only 9% of the test
set when using SMOTE are more than two taxonomic layers apart
(Figure 9).
While the overall achievements of the method are very good,
some miss classifications occur and we were wondering whether
that had anything to do with the big size differences in the same
clades. For instance, sampling may have introduced a bias leading
to a difference between the larger and smaller clades. It turns out
that larger clades tend to produce lower hF scores when using
random sampling (Figure 10). For SMOTE sampling, such a rela-
tionship was not observed (Figure 10).
There are two ad hoc explanations for this outcome: (1) the
sampling methodology introduced a bias or (2) larger species/
clades contain more examples that may not be true. The latter
was also found by Sacar Demirci and colleagues [16]. Perhaps a
combination of both could explain the result.
4 Conclusion
MicroRNAs are involved in gene expression and, therefore, play an

important role in cellular regulation and diseases [31–35]. Known
miRNAs are stored in databases such as miRBase, TarBase, and
others. However, those databases are thought to contain contami-
nants [11]. To the best of our knowledge, there is no automatic
tool that can detect such contaminants. Therefore, we developed a
machine learning approach that successfully classified miRNAs to
their species/clades of origin employing only k-mer frequency
features of the miRNA hairpins. The average performance, accord-
ing to the hierarchical F-measure, is 0.94 for SMOTE sampling
considering dataset1. This outcome is very good (Table 1) and the
results for dataset2 (0.70) are still acceptable. The ranks for the
correct assignments are typically very good (Figure 7) and the
Fig. 8 Taxonomic similarity. Taxonomic similarity using the similarity measure for dataset1 with SMOTE
sampling. Most wrong classified samples (similarity <1) have a similarity of at least 60%
Fig. 9 Hierarchical distance. Layer_distance results for dataset 1 using SMOTE sampling. Only 9% of the
samples are more than two taxonomic layers apart
Fig. 10 hF scores versus class size. hF scores of the species used as a function to the associated class size
(number of hairpin examples). Class size to hF score distribution for random sampling is provided in orange
and the distribution for SMOTE sampling in blue. Note that the x-axis uses a log scale. A functional relationship
between score and class size appears to only exist when employing random sampling
taxonomic distance to the correct clade for wrong classifications is

usually not large (Figure 8), also when considering hierarchical
relationships (Figure 9). Comparing the hF score distribution
against the class size (Figure 10) revealed that the species with a
large number of examples have a lower average hF score for random
sampling, but not for SMOTE sampling. This seems to be due to
the sampling approach but can in part be due to more contamina-
tion of the larger samples.
The approach developed here can be used to detect contami-
nants by comparing the predicted to the species/clade it was
thought to originate from. If the taxonomic distance is above a
threshold, it can be a contaminant. A similar approach is useful for
miRNAs discovered via sRNA-seq where miRNAs can be sample
contamination. Ensuring that the discovered miRNAs are from the
species under investigation or at least fall into the same class for
species without examples in miRBase will add confidence to the
miRNA identifications. Besides that, our approach can contribute
to a better understanding of miRNA evolution [8]. Since a catego-
rization via k-mer frequency is possible, there appears to be some
amount of conservation of these k-mers which may deserve further

investigation.
In the future, we aim to make the tool available as a web service
so that it can be used in miRNA analysis pipelines.
References
1. Bartel D (2004) MicroRNAsGenomics, bio- 12. Zhang B, Pan X, Cannon CH, Cobb GP,
genesis, mechanism, and function. Cell Anderson TA (2006) Conservation and diver-
116:281–297 gence of plant microRNA genes. Plant J
2. Hammond SM (2015) An overview of micro- 46:243–259
RNAs. Adv Drug Deliv Rev 87:3–14 13. Yousef M, Khalifa W, Acar İE, Allmer J (2017)
3. Hamzeiy H, Suluyayla R, Brinkrolf C, Janowski MicroRNA categorization using sequence
SJ, Hofestaedt R, Allmer J (2017) Visualiza- motifs and k-mers. BMC Bioinformatics
tion and analysis of MicroRNAs within KEGG 18:170
pathways using VANESA. J Integr Bioinform 14. Yousef M, Nigatu D, Levy D, Allmer J, Henkel
14:20160004 W (2017) Categorization of species based on
4. Takamizawa J, Konishi H, Yanagisawa K, their microRNAs employing sequence motifs,
Tomida S, Osada H, Endoh H, Harano T, information-theoretic sequence feature extrac-
Yatabe Y, Nagino M, Nimura Y et al (2004) tion, and k-mers. Eurasip J Adv Signal Proc
Reduced expression of the let-7 microRNAs in 2017. https://doi.org/10.1186/s13634-
human lung cancers in association with short- 017-0506-8
ened postoperative survival. Cancer Res 15. Yousef M (2019) Hamming distance and
64:3753–3756 K-mer features for classification of pre-cursor
5. Rodriguez A, Griffiths-Jones S, Ashurst JL, microRNAs from different species. In:
Bradley A (2004) Identification of mammalian Benavente-Peces C, Slama SB, Zafar B (eds)
microRNA host genes and transcription units. Proceedings of the 1st international conference
Genome Res 14:1902–1910 on smart innovation, ergonomics and applied
6. Sempere LF, Cole CN, Mcpeek MA, Peterson human factors (SEAHF). Springer Interna-
KJ (2006) The phylogenetic distribution of tional Publishing, Cham, pp 180–189
metazoan microRNAs: insights into evolution- 16. Demirci MDS, Baumbach J, Allmer J (2017)
ary complexity and constraint. J Exp Zoolog B On the performance of pre-microRNA detec-
Mol Dev Evol 306:575–588 tion algorithms. Nat Commun 8:1–9
7. Kozomara A, Griffiths-Jones S (2010) miR- 17. Yousef M, Allmer J (2019) Classification of
Base: integrating microRNA annotation and pre-cursor microRNAs from different species
deep-sequencing data. Nucleic Acids Res 39: using a new set of features BT—database and
D152–D157 expert systems applications. In: Anderst-
8. Velandia-Huerto CA, Yazbeck AM, Schor J, Kotsis G, Tjoa AM, Khalil I (eds) . Springer
Stadle PF (2021) Evolution and phylogeny of International Publishing, Cham, pp 15–20
microRNAs—protocols, pitfalls, and problems. 18. Yousef M, Khalifa W, Acar İE, Allmer J (2017)
In: Allmer J, Yousef M (eds) miRNomics: Distinguishing between MicroRNA targets
microRNA biology and computational analy- from diverse species using sequence motifs
sis. Methods in molecular biology, vol 2257. and k-mers. In: Proceedings of the 10th inter-
Springer, New York national joint conference on biomedical engi-
9. Meng Y, Shao C, Wang H, Chen M (2012) Are neering systems and technologies.
all the miRBase-registered microRNAs true? A SCITEPRESS—Science and Technology Pub-
structure-and expression-based re-examination lications, Setúbal, pp 133–139
in plants. RNA Biol 9:249–253 19. Yousef M, Levy D, Allmer J (2018) Species
10. Saçar MD, Hamzeiy H, Allmer J (2013) Can categorization via MicroRNAs—based on
MiRBase provide positive data for machine 30 UTR target sites using sequence features:
learning for the detection of MiRNA hairpins? in: proceedings of the 11th international joint
J Integr Bioinform 10:1–11 conference on biomedical engineering systems
and technologies. SCITEPRESS—Science and
11. Bağcı C, Allmer J (2016) One step forward, Technology Publications, Funchal, Madeira,
two steps back; xeno-microRNAs reported in Portugal, pp 112–118
breast milk are artifacts. PLoS One 11:
e0145065
20. Yousef M, Khaleifa W, Onal-Suzek T (2019) In imbalanced data-sets with multiple classes:
silico validation of ncRNA-ncRNA interaction Binarization techniques and ad-hoc
sites with ncRNAs represented by k-mers fea- approaches. Knowl-Based Syst 42:97–110
tures: in: proceedings of the 12th international 30. Hall M, Frank E, Holmes G, Pfahringer B,
joint conference on biomedical engineering Reutemann P, Witten IH (2009) The WEKA
systems and technologies. SCITEPRESS - Sci- data mining software. ACM SIGKDD Explor
ence and Technology Publications, Prague, Newsl 11:10–18. https://doi.org/10.1145/
Czech Republic, pp 168–173 1656274.1656278
21. Ho TK (1995) Random decision forests. In: 31. Yousef M, Abdallah L, Allmer J (2019) maTE:
Proceedings of 3rd international conference discovering expressed interactions between
on document analysis and recognition. IEEE, microRNAs and their targets. Bioinformatics
Piscataway, New Jersey, pp 278–282 35:4020–4028. https://doi.org/10.1093/bio
22. Saçar MD, Allmer J (2014) Machine learning informatics/btz204
methods for microRNA gene prediction. In: 32. Olcum M, Tufekci KU, Genc S (2021) Micro-
miRNomics: MicroRNA biology and compu- RNAs in genetic etiology of human diseases.
tational analysis. Springer, New York, pp In: Allmer J, Yousef M (eds) miRNomics:
177–187 microRNA biology and computational analy-
23. Edgar RC (2010) Search and clustering orders sis. Methods in molecular biology, vol 2257.
of magnitude faster than BLAST. Bioinformat- Springer, New York
ics 26:2460–2461 33. Yildiz MT, Tutar L, Giritlioğlu NI, Bayram B,
24. Kurtz S, Narechania A, Stein JC, Ware D Tutar Y (2021) MicroRNAs and heat shock
(2008) A new method to compute K-mer fre- proteins in breast cancer biology. In: Allmer J,
quencies and its application to annotate large Yousef M (eds) miRNomics: microRNA biol-
repetitive plant genomes. BMC Genomics ogy and computational analysis. Methods in
9:517 molecular biology, vol 2257. Springer, New
25. Chawla NV, Bowyer KW, Hall LO, Kegelmeyer York
WP (2002) SMOTE: synthetic minority over- 34. Karagur ER, Akgun S, Akca H (2021) Compu-
sampling technique. J Artif Intell Res tational and bioinformatics methods for micro-
16:321–357 RNA gene prediction. In: Allmer J, Yousef M
26. Berthold MR, Cebron N, Dill F, Gabriel TR, (eds) miRNomics: microRNA biology and
Kötter T, Meinl T, Ohl P, Thiel K, Wiswedel B computational analysis. Methods in molecular
(2009) KNIME-the Konstanz information biology, vol 2257. Springer, New York
miner: version 2.0 and beyond. AcM SIGKDD 35. Uzuner E, Ulu GT, Gürler SB, Baran Y (2021)
Explor Newsl 11:26–31 The role of MiRNA in cancer: pathogenesis,
27. Kiritchenko S, Matwin S, Nock R, Famili AF diagnosis, and treatment. In: Allmer J, Yousef
(2006) Learning and evaluation in the presence M (eds) miRNomics: microRNA biology and
of class hierarchies: application to text catego- computational analysis. Methods in molecular
rization. In: Conference of the Canadian Soci- biology, vol 2257. Springer, New York
ety for Computational Studies of intelligence. 36. Robinson O, Dylus D, Dessimoz C (2016)
Springer, New York, pp 395–406 Phylo.io : interactive viewing and comparison
28. Xu Q-S, Liang Y-Z (2001) Monte Carlo cross of large phylogenetic trees on the web. Mol
validation. Chemom Intell Lab Syst 56:1–11 Biol Evol 33:2163–2166. https://doi.org/
29. FernáNdez A, LóPez V, Galar M, Del Jesus MJ, 10.1093/molbev/msw080
Herrera F (2013) Analysing the classification of
Chapter 13
MicroRNAs in Genetic Etiology of Human Diseases

Melis Olcum, Kemal Ugur Tufekci, and Sermin Genc
Abstract
Since their first discovery more than 20 years ago, miRNAs have been subject to deliberate research and
analysis for revealing their physiological or pathological involvement. Regulatory roles of miRNAs in signal
transduction, gene expression, and cellular processes in development, differentiation, proliferation, apo-
ptosis, and homeostasis also imply their critical role in disease pathogenesis. Their roles in cancer, neurode-
generative diseases, and other systemic diseases have been studied broadly. In these regulatory pathways,
their mutations and target sequence variations play critical roles to determine their functional repertoire. In
this chapter, we summarize studies that investigated the role of mutations, polymorphisms, and other
variations of miRNAs in respect to pathological processes.
Key words Human disease, Mutation, Single-nucleotide polymorphism, Copy number variation
1 Introduction
Gene expression and regulation are complex processes that include

many regulatory steps. MicroRNAs (miRNAs) present another
layer of complexity in gene regulation, via modulation of gene
expression at the posttranscriptional level [1]. Current estimates
are that 10–30% of genes are regulated at the posttranscriptional
level by miRNAs while, 3–4% of the genes encodes miRNAs
[2]. They mostly target signaling proteins, enzymes and transcrip-
tion factors [3], and all human pathways have miRNA interactions
[4]. This regulatory activity of miRNAs makes them critical players
of development, differentiation, apoptosis, and proliferation
[5, 6]. Studies revealed that miRNAs negatively regulate their
target genes expression [7]. Recently, a number of studies also
showed, activation of pathways is also possible by miRNAs, but
that is rather rare [8, 9]. miRNAs may target more than one mRNA
and an mRNA may have multiple miRNA regulators [10]. Longer
30 UTRs in genes, are more prone to miRNA binding and these
genes are mostly involved in development [11–14]. MiRNAs may
have mutations that alter their function and or dysregulated gene
255
256 Melis Olcum et al.
expression both resulting in disease pathogenesis. Mutations may

occur in miRNA coding, binding, and biogenesis sites [15]. There
may be deletions, insertions, duplications, and single nucleotide
polymorphisms (SNPs) [16]. Therefore, miRNA associated muta-
tions may have roles in disease pathogenesis. In this chapter, path-
ogenic roles of miRNAs in human diseases are summarized.
2 CNV Type Large-Scale Mutations
Copy number variations (CNV) of miRNAs and their target genes

are relevant to several types of human diseases. However, biological
relevance of these CNVs in disease pathogenesis needs more clinical
studies.
Large-scale mutations like deletions, duplications or insertions
may modify miRNA genes. The very first study reported that
miR-338, miR-657, miR-204, miR-383, and miR-589 contain
CNV type mutation in a Han Chinese population [17]. A bioinfor-
matics analysis revealed 209 CNVs in miRNA genes and 11 CNVs
of these were highly polymorphic and highly frequent
[18]. Another bioinformatics study revealed 193 pre-miRNAs
were located in 385 CNV marker regions deposited in the Database
of Genomic Variants [19].
To date, a number of human Mendelian disorders, such as
nonsyndromic autosomal dominant progressive hearing loss [20],
and Feingold syndrome [21] have been associated with CNV such
as mutations of miRNA related genes. In a case study with 10 index
patients, a germline hemizygous deletion in the miR-17–92 cluster
was reported as relevant to patient’s microcephaly, short stature,
and digital abnormalities [22]. Another deletion found in the
12q24.31, which has 31 protein-coding and a miRNA coding
genes, as relevant to intellectual disability and facial dysmorphism
of an 11-year-old female patient [23].
Studies in cancer suggest causative relationships with CNV type
mutations. In chronic lymphocytic leukemia, the 13q deletion is
the most common aberration [24]. Authors further reported that
downregulated miR-143 and overexpressed miR-155 decreased
apoptosis and increased proliferation in patients. Another target
of the 13q deletions is the cluster miR-15a/miR-16-1 (miR-15/
16) in chronic lymphocytic leukemia (CLL) [25]. The miR-15/16
locus encodes tumor suppressor miRNAs, which are downregu-
lated in CLL and several other cancer types such as colorectal
cancer, melanoma, bladder cancer. The cluster allegedly targets
the overexpressed BCL2 in CLL [25]. The cluster miR-15/16
deletions was found in other types of cancer suggesting that the
cluster could be a tumor suppressor [26].
Large-scale mutations in miRNA genes are linked with leuke-
mogenesis. In 56 cases of diffuse large B-cell lymphoma (DLBCL),
MicroRNAs in Genetic Etiology 257
11 cases with the miR-142 mutation were reported. Among these

11 cases, eight mutations affected mature form, and three muta-
tions found in the precursor form of the miRNA. The two cases
with mutations in the seed sequence of miR-142-3p directed the
miRNA to the mRNAs of ZEB1 and ZEB2. The researchers
reported the mutations in mature miRNA, did not have an effect
on ZEB1 and ZEB2 30 UTRs. But the seed sequence mutations led
to the loss of responsiveness in 30 UTRs of RAC1 and ADCY9
which are two of known miR-142-3p targets [27]. Gaballa et al.
reported a deletion in 5q that affects 5q31 and 5q33 which leads to
p53 activation and impaired erythropoiesis in myelodysplastic syn-
drome patients which is a disease that later transforms into acute
myelogenous leukemia (AML). These mutations decreased
miR-145 and miR-146a expressions [28].
Apart from hematological malignancy, other cancer types also
have been associated with CNV type mutations in miRNA- related
genes. Fourteen miRNA genes and two miRNA biogenesis genes
(DICER and DROSHA) showed CNV type mutations in non–
small-cell lung cancer (NSCLC) [29].The most frequently ampli-
fied miRNAs were miR-30d, miR-21, miR-17, and miR-155. The
same group also reported that increased expression of DROSHA
decreased survival while increased expression of DICER increased
survival of NSCLC patients [29]. In thyroid cancer, downregulated
miRNAs from the 14q32 site play roles in the migration, angiogen-
esis, and adhesion processes [30]. MiR-30d is another microRNA
found to be positively correlated with tumor progression due to
CNV amplification. 22.8% of the cases displayed copy number
gains, and 36.4% of the cases showed lymph node metastases with
cutaneous squamous cell carcinoma [31]. Masson et al. identified
142 CNVs related with colorectal cancer risk. As a result of their
meta-analysis in familial adenomatous polyposis patients, CNV
containing 26 regions were targeted by 42 miRNAs, while the
miR-15 family was overrepresented in this hereditary predisposi-
tion of colorectal cancer [32]. miR-650 was reported to be involved
in osteosarcoma via decreasing the expression of inhibitor of
growth family member 4 (ING4) to result in IL-6 increase [33].
Other reports exhibit that CNV type mutations are linked with
other diseases, for instance acute anterior uveitis (AAU). Low copy
number of miR-146a and high copy number of miR-23a were seen
in AUU patients independent from their ankylosing spondylitis
(AS) status. On the other hand, miR-205 was relevant to both
conditions, having low copy numbers in AS+ patients and high
copy numbers in AS patients [34]. CNVs in miR-155 were
reported to be involved in 3 epileptic patients [35]. MiR-155 acts
as a modulator in inflammatory mechanisms in epilepsy. They
speculated that duplication of 21q21 in epileptic patients results
in upregulated miR-155 expression and higher susceptibility to
epileptic seizures.
3 SNP Type Mutations in MicroRNA Genes
SNPs are frequently found polymorphisms in miRNA genes, which

may affect different stages of miRNA biogenesis, namely, matura-
tion, structure, and expression [36]. Furthermore, the single nucle-
otide changes in the mature miRNA can also result in different
pairing possibilities with the target site, thereby affecting miRNA
regulation of gene expression. These functional changes play a role
in disease pathogenesis in cancer and Mendelian disorders. The first
germline mutation in miRNA gene was linked to familial CLL
[37]. An SNP in pri-miR-16-1 was reported to decrease expression
of the mature miRNA in this study. In another study, an SNP in
miR-126 processing was associated with MLL-AF4 [fusion pro-
teins instigated by chromosomal translocation t(4;11)) acute lym-
phoblastic leukemia (ALL)] [38].
Studies for associating cancer risk with miRNA SNPs have been
carried out [39–47]. Papillary thyroid carcinoma was associated
with a polymorphism (rs2910164) in miR-146 [39]. MiR-146
was also found to be associated with breast cancer [41], glioma
[45], and gastric cancer [40]. On the other hand, for some other
cancer types such as; prostate and cervical cancer, and hepatocellu-
lar and esophageal squamous cell carcinoma, miR-146 was found to
be protective [20, 40]. Oral cavity cancers have been associated
with miR-SNPs, some overlapping miR-SNPs in increasing the risk
of head and neck squamous cell carcinoma and laryngeal
cancer [48].
SNP association studies also include cancer types and drug
resistance and are therefore promising stratifiers of high-risk
patients. By the identification, high-risk patients will have the
chance for an early diagnosis by screening programs resulting in
better patient management. Since germline SNPs can also affect
somatic mutations, miR-SNPs gained higher importance
[49]. Fragile sites of chromosomes, cancer susceptibility loci, and
quantitative trait loci are hot spots for SNPs, as the tool SniPer
showed [50]. A germline SNP reported in pri-miR-16-1 decreased
levels of miR-16-1 and disrupted pri/pre-processing [37]. The
tissue in which miRNA is expressed affects cancer prognosis and
progression [51]. Alleles can be associated with an increase in
cancer of one type in one tissue, while they decrease cancer in
another tissue. An SNP in miR-26a-1 reduced the risk of bladder
cancer while it increased the risk of premalignant oral lesions
[43, 52]. Another SNP in miR-423 was associated to an increase
in risk of bladder and ovarian cancer while reducing the esophageal
cancer risk [53] Three CpG island SNPs in miRNA promoters were
identified in breast cancer and associated with cancer susceptibility
[54]. Another SNP in miR-146a (rs2910164) was originally asso-
ciated with an increased risk of papillary thyroid carcinoma, but it is
now known that this SNP affects many more cancer types and
shows minor & major allele switching between populations [55].
Advances in prevention, diagnosis, and treatment were not
sufficient and efficient enough to reduce high mortality and mor-
bidity caused in cardiovascular diseases, such as coronary artery
disease (CAD) and cerebrovascular diseases [56]. In the last decade,
studies have been conducted to elucidate the relationship between
miRNA SNPs and CAD, intensively about polymorphisms in
miR-146, -196 and -499. A pioneering research in Han Chinese
population reported that alteration from G to C allele of miR-146a
(rs2910164) results in higher incidence of CAD and upregulation
of mature miR-146a in peripheral blood mononuclear cells [57].
Mental disorders are also associated with miRNA regulation.
miR-137, which is implicated in Schizophrenia (Sz) pathogenesis,
has been shown to have an important role in embryonic neural stem
cells, neuronal maturation, and synaptic expression. SNPs in
miRNA-137 (namely, rs1198588, rs1625579, rs2660304, and
rs2802535) were located in the noncoding region and associated
with Schizophrenia [58]. In another study, SNPs (rs1625579 and
rs2660304) in miR-13 were found in the pathophysiology of the
disease [59]. One of the SNPs (rs1625579) that was associated
with schizophrenia, was also found associated with bipolar disorder
[60]. Yet, this association is not confirmed in every study done with
Iranian and Asian populations [61, 62]. In a wide association study
including 609 autosomal miRNAs, 98 of the miRNAs were asso-
ciated with bipolar disorder; including miR-499, miR-708 and
miR-1908 [63]. Two SNPs; pre-miR-182 (rs76481776) and pre--
miR-30e (rs178077483) were implicated in major depressive dis-
order [64, 65].
4 SNPs in MicroRNA Target Genes
There are thousands of SNPs in miRNA target genes that regulate

gene expressions [66]. These mutations can be classified as SNPs
that disrupt the miRNA binding site and SNPs that create a miRNA
binding site. From 58,977 SNPs in 90,784 miRNA targets, 20,779
were found to disrupt existing binding sites and 91,711 were found
to create new binding sites [66]. SNPs that create a new binding
site were detected in miR-191 in ovarian cancer [67], miR-140-5p
in bladder cancer [68] and miR-4271 in coronary heart disease
[69]. On the other hand, target sites of miR-96 and miR-182 in
breast cancer [70], miR-199a in pancreatic ductal carcinoma [71],
miR-433 in Parkinson’s disease [72] and miR-27b in bladder can-
cer [73] were found associated with SNPs that impairs with the
target site’s binding efficiency.
The SNPs in miRNA binding sites are associated with several
disorders. MiR-1 and miR-206 were found as the first example in a
sheep model, to bind on an SNP-created target site on 30 UTR of

(growth differentiation factor 8) GDF8 gene and result in miRNA
mediated reduction of myostatin and lead to muscular hypertrophy,
in an F2 population of Texel sheep [74]. In an initial human study,
miR-189’s target gene, SLITRK1 (Slit and Trk like 1), was found
to have an SNP decreasing binding of the miRNA to the target site,
in Tourette’s syndrome (TS) [75].
Most frequently associated diseases for miRNA target site
mutations are cancer. One of genes containing miRNA target site
mutation is Kirsten Rat Sarcoma virus oncogene (KRAS). MiRNA
let-7 binds 30 UTR of KRAS and thereby suppress its expression.
This binding leads to cell growth inhibition and tumor suppression.
One SNP in Let-7 (rs 61764370) was found in almost 20% of
NSCLC patients and was associated to increased cancer risk
[76, 77]. This KRAS variant was also found in 61% of BRCA1
and BRCA2 negative hereditary breast & ovarian cancer cases
[78]. Another SNP in Keratin type II cuticular Hb1 (KRT81) site
was found associated with NSCLC risk [79] and patient
survival [80].
SNPs were suggested to be prognostic and predictive for sev-
eral cancer types. Some of these are breast cancer [67], hepatocel-
lular carcinoma [81, 82], colorectal cancer [69, 83–85], lung
cancer [86, 87], gastric cancer [88–91], bladder cancer [71] and
prostate cancer [72]. SNPs affect drug responses in cancer patients
for the pharmaceuticals imatinib, doxorubicin, daunorubicin, and
cisplatin [92].
MiRNA target site SNPs were also found related to autoim-
mune diseases [93], schizophrenia [94, 95] and epilepsy [96]. An
SNP in the NFKB1A gene (rs11096957) was associated with the
severity of the hip osteoarthritis in Han Chinese Population
[97]. This SNP creates new miRNA binding site and may inhibit
NFKBIA expression.
5 Mutations in MicroRNA Biogenesis Genes
Genes involved in miRNA biogenesis regulate global expression of

miRNAs by their protein products [51]. Mutation in those genes
therefore, affect the expression state and contribute the pathogen-
esis of human diseases (Fig. 1). One of these proteins that is highly
studied is an RNase III enzyme, the Drosha protein. The Drosha
protein transcribes pre-miRNA into mature miRNA, in the
nucleus. Mutations of Drosha was found relevant for several cancers
acting as an oncogene or a tumor suppressor. For more than half of
cervical squamous cell carcinomas, Drosha was an oncogene by
either being overexpressed or having gene copy number increase
mutations [98]. A haplotype in Drosha was found relevant to
survival in lung cancer and adenocarcinoma [99]. In Wilms’
Fig. 1 The effect of polymorphisms on miRNA-mediated regulation of gene expression. SNPs in different steps
of miRNA biogenesis result in diverse effects on the regulation of gene expression
tumors, most of the mutations in Drosha were single missense

mutations that interfere with metal binding and negatively affecting
the RNase domain function [100–102].
A germline mutation in Dicer1 was linked to familial recurrent
liver tumors. 6 of 243 tumors showed somatic Dicer1 mutations
and these tumors showed decreased miRNA expression. Within
these, miR-365b was found as significantly related to Dicer1 muta-
tions [103]. Neointima formation was reported to increase in
conditional deletion of Dicer in SMCs of Apoe / mice due to
decreased proliferation. Dicer in vascular smooth muscle cells
induces smooth muscle cells to a proliferative phenotype rather
than their contractile phenotype, and therefore promotes repair
mechanism to injuries. MiR-27a-3p, for example, suppresses ARH-
GEF26 (Rho guanine nucleotide exchange factor 26) expression.
ARHGEF26 is expressed in a Dicer-dependent manner in medial
and neointimal SMCs. In vitro studies of Zahedi et al. showed that
SMC proliferation is suppressed by miR-27a-3p interaction with
3 of ARHGEF26 mRNAs. They discuss that Dicer prevents vessel
stenosis by generating antiproliferative miRNAs in neointima
formation [104].
Parkinson’s disease is another disease that miRNAs are involved
in. Several miRNAs are linked to, apoptosis and stress response
pathways of Parkinson’s disease. Chmielarz et al. reported a reduc-
tion of Dicer in the ventral midbrain altered miR expression in
(dopaminergic) DA neurons in aged mice. They also reported

stimulation of miRNA biogenesis induced proliferation in DA neu-
rons in vitro. They suggest that Dicer has an essential role in
maintaining adult DA neurons and that stimulation of miR produc-
tion promotes survival [105].
Germline DICER1 mutations were linked to increased risk of
development of rare tumors. These tumors such as ovarian sex cord
stromal tumors, Sertoli-Leydig cell tumor, and embryonal rhabdo-
myosarcoma of the cervix have distinct morphologies that require
genetic evaluation for underlying tumor predisposition [106].
Other than Drosha and Dicer and transcription, transportation
of precursor miRNAs from the nucleus to the cytoplasm by expor-
tin5 (XPO5) are the steps of miRNA regulation. Alterations in
XPO5 affect expression levels of miRNAs and their targets and
associated with cancer susceptibility and tumor formation
[107]. In hepatocellular carcinoma, XPO5 transport ability is
impaired due to aberrant phosphorylation and global miRNA
expressions are downregulated [108]. Gastric, endometrial, and
colon tumors were other types that were linked to somatic or
germline mutations of XPO5 [109].
Yi et al. showed that Dicer1 independent biogenesis of
miR-451 mediates erythrocyte maturation, via an alanine mutation
of the respective binding site Lys56 of the elF1A (eukaryotic tran-
scription initiation factor 1A) transcription factor interaction with
Ago-2 (argonuate-2). They also suggest that eLF1A is a novel
element of Ago-2 mediated RISCs and increases Ago-2 mediated
miRNA biogenesis [110].
6 Epigenetic Regulation of MicroRNA Genes
Expression of microRNA genes is tightly regulated by epigenetic

mechanisms including miRNA gene promoter methylation, histone
acetylation, and methylation [111]. Epigenetic regulation of
miRNA gene expression occurs both in physiological and patho-
logical conditions. Epigenetic control of miRNA genes may be
tissue and disease-specific as well as miRNA and inducer dependent.
Alteration in epigenetic regulation of miRNA genes has been
observed in several diseases such as cancer, autoimmune, cardiovas-
cular, and neurodegenerative disorders.
More than 50% miRNA genes are located in CpG islands
suggesting that their expression can be regulated by methylation.
For instance, miR-203 is silenced in a CpG island, by a methylation
close to its promoter site. Epigenetic downregulation of miR-203
expression has been observed in hematopoietic malignancies,
including chronic myelogenous leukemia [112]. Hypermethylation
of miRNA gene promoters also downregulates miRNA expression
and increase its target genes. The methylation status of miRNA
genes could be used as a predictive factor for cancer prognosis.

miR-137 promoter methylation is associated with poor prognosis
in a great number of cancer including carcinoma of the head and
neck, gastric carcinoma, and non-small-cell lung cancer [113]. Dif-
ferent miRNA hypermethylations including miR-148a, miR-34b/
c, miR-9, and miR-199a-3p are associated with cancer metastasis
and drug resistance [114, 115].
Posttranslational modification of histones in the amino termi-
nal coils via methylation, acetylation, phosphorylation, ubiquitina-
tion, and SUMOylation affects miRNA expression and contributes
pathogenesis of various diseases. Histone modification may result in
either upregulation or downregulation of miRNA expression. For
instance, miR-21 expression is altered by histone methylation and
acetylation in chronic myeloid leukemia, non–small-cell lung can-
cer, and cardiac hypertrophy [116–118].
7 Conclusion
Growing evidence shows critical roles of miRNA regulation in

human diseases and pathologies. In normal physiology, miRNAs
regulate gene expression via mRNA binding and suppression of
translation, while they are also regulated by various factors. Muta-
tions and variations in miRNA genes or target genes change expres-
sion and diversify functional outcomes. The aberrant expression
and regulation caused by mutations form the basis for disease
development and pathogenesis. Identification of these changes is
a promising solution for early prognosis and diagnosis in many
human diseases.
References
1. Catalanotto C, Cogoni C, Zardo G (2016) 5. Othman N, Nagoor NH (2014) The role of

MicroRNA in control of gene expression: an microRNAs in the regulation of apoptosis in
overview of nuclear functions. Int J Mol Sci lung cancer and its application in cancer treat-
17:1712 ment. Biomed Res Int 2014:318030
2. Valinezhad Orang A, Safaralizadeh R, 6. Frixa T, Donzelli S, Blandino G (2015)
Kazemzadeh-Bavili M (2014) Mechanisms Oncogenic MicroRNAs: key players in malig-
of miRNA-mediated gene regulation from nant transformation. Cancers (Basel)
common downregulation to mRNA-specific 7:2466–2485
upregulation. Int J Genomics 2014:970607 7. Galatenko VV, Galatenko AV, Samatov TR
3. Nazarov PV, Reinsbach SE, Muller A et al et al (2018) Comprehensive network of
(2013) Interplay of microRNAs, transcription miRNA-induced intergenic interactions and
factors and target genes: linking dynamic a biological role of its core in cancer. Sci Rep
expression changes to function. Nucleic 8:2418
Acids Res 41:2817–2831 8. Couzigou JM, Lauressergues D, Andre O
4. Hamzeiy H, Suluyayla R, Brinkrolf C et al et al (2017) Positive gene regulation by a
(2017) Visualization and analysis of Micro- natural protective miRNA enables arbuscular
RNAs within KEGG pathways using mycorrhizal Symbiosis. Cell Host Microbe
VANESA. J Integr Bioinform 14:20160004 21:106–112
9. Orom UA, Nielsen FC, Lund AH (2008) 23. Palumbo O, Palumbo P, Delvecchio M et al
MicroRNA-10a binds the 5’UTR of ribo- (2015) Microdeletion of 12q24.31: report of
somal protein mRNAs and enhances their a girl with intellectual disability, stereotypies,
translation. Mol Cell 30:460–471 seizures and facial dysmorphisms. Am J Med
10. Peter ME (2010) Targeting of mRNAs by Genet A 167A:438–444
multiple miRNAs: the next step. Oncogene 24. Hernandez-Sanchez M, Rodriguez-Vicente
29:2161–2164 AE, Hernandez JA et al (2016) MiRNA
11. Cai Y, Yu X, Hu S et al (2009) A brief review expression profile of chronic lymphocytic leu-
on the mechanisms of miRNA regulation. kemia patients with 13q deletion. Leuk Res
Genomics Proteomics Bioinformatics 46:30–36
7:147–154 25. Pekarsky Y, Croce CM (2015) Role of
12. Cheng C, Bhardwaj N, Gerstein M (2009) miR-15/16 in CLL. Cell Death Differ
The relationship between the evolution of 22:6–11
microRNA targets and the length of their 26. Porkka KP, Ogg EL, Saramaki OR et al
UTRs. BMC Genomics 10:431 (2011) The miR-15a-miR-16-1 locus is
13. Sandberg R, Neilson JR, Sarma A et al (2008) homozygously deleted in a subset of prostate
Proliferating cells express mRNAs with short- cancers. Genes Chromosomes Cancer
ened 30 untranslated regions and fewer micro- 50:499–509
RNA target sites. Science 320:1643–1647 27. Kwanhian W, Lenze D, Alles J et al (2012)
14. Osada H, Takahashi T (2007) MicroRNAs in MicroRNA-142 is mutated in about 20% of
biological processes and carcinogenesis. Car- diffuse large B-cell lymphoma. Cancer Med
cinogenesis 28:2–12 1:141–155
15. Cammaerts S, Strazisar M, De Rijk P et al 28. Gaballa MR, Besa EC (2014) Myelodysplastic
(2015) Genetic variants in microRNA genes: syndromes with 5q deletion: pathophysiology
impact on microRNA expression, function, and role of lenalidomide. Ann Hematol
and disease. Front Genet 6:186 93:723–733
16. Moszynska A, Gebert M, Collawn JF et al 29. Czubak K, Lewandowska MA, Klonowska K
(2017) SNPs in microRNA target sites and et al (2015) High copy number variation of
their potential role in human disease. Open cancer-related microRNA genes and frequent
Biol 7:170019 amplification of DICER1 and DROSHA in
17. Lin CH, Li LH, Ho SF et al (2008) A large- lung cancer. Oncotarget 6:23399–23416
scale survey of genetic copy number variations 30. Geraldo MV, Nakaya HI, Kimura ET (2017)
among Han Chinese residing in Taiwan. BMC Down-regulation of 14q32-encoded miRNAs
Genet 9:92 and tumor suppressor role for miR-654-3p in
18. Marcinkowska M, Szymanski M, Krzyzosiak papillary thyroid cancer. Oncotarget
WJ et al (2011) Copy number variation of 8:9597–9607
microRNA genes in the human genome. 31. Zhou Y, Hao Y, Li Y et al (2017) Amplifica-
BMC Genomics 12:183 tion and up-regulation of MIR30D was asso-
19. Duan S, Mi S, Zhang W et al (2009) Compre- ciated with disease progression of cervical
hensive analysis of the impact of SNPs and squamous cell carcinomas. BMC Cancer
CNVs on human microRNAs and their regu- 17:230
latory genes. RNA Biol 6:412–425 32. Masson AL, Talseth-Palmer BA, Evans TJ et al
20. Mencia A, Modamio-Hoybjor S, Redshaw N (2016) Copy number variants associated with
et al (2009) Mutations in the seed region of 18p11.32, DCC and the promoter 1B region
human miR-96 are responsible for nonsyn- of APC in colorectal polyposis patients. Meta
dromic progressive hearing loss. Nat Genet gene 7:95–104
41:609–613 33. Yun JH, Moon S, Lee HS et al (2015)
21. Feingold M, Hall BD, Lacassie Y et al (1997) MicroRNA-650 in a copy number-variable
Syndrome of microcephaly, facial and hand region regulates the production of interleukin
abnormalities, tracheoesophageal fistula, duo- 6 in human osteosarcoma cells. Oncol Lett
denal atresia, and developmental delay. Am J 10:2603–2609
Med Genet 69:245–249 34. Yang L, Du L, Yue Y et al (2017) miRNA copy
22. de Pontual L, Yao E, Callier P et al (2011) number variants confer susceptibility to acute
Germline deletion of the miR-17 approxi- anterior uveitis with or without ankylosing
mately 92 cluster causes skeletal and growth spondylitis. Invest Ophthalmol Vis Sci
defects in humans. Nat Genet 43:1026–1030 58:1991–2001
35. Tao H, Cui L, Li Y et al (2015) Association of 47. Zhan JF, Chen LH, Chen ZX et al (2011) A
tag SNPs and rare CNVs of the MIR155HG/ functional variant in microRNA-196a2 is
miR-155 gene with epilepsy in the Chinese associated with susceptibility of colorectal
Han population. Biomed Res Int cancer in a Chinese population. Arch Med
2015:837213 Res 42:144–148
36. He S, Ou H, Zhao C et al (2018) Clustering 48. Wilkins OM, Titus AJ, Gui J et al (2017)
pattern and functional effect of SNPs in Genome-scale identification of microRNA-
human miRNA seed regions. Int J Genomics related SNPs associated with risk of head and
2018:2456076 neck squamous cell carcinoma. Carcinogene-
37. Calin GA, Ferracin M, Cimmino A et al sis 38:986–993
(2005) A MicroRNA signature associated 49. Carter H, Marty R, Hofree M et al (2017)
with prognosis and progression in chronic Interaction landscape of inherited poly-
lymphocytic leukemia. N Engl J Med morphisms with somatic events in cancer.
353:1793–1801 Cancer Discov 7:410–423
38. Harnprasopwat R, Ha D, Toyoshima T et al 50. Zorc M, Skok DJ, Godnic I et al (2012) Cat-
(2010) Alteration of processing induced by a alog of microRNA seed polymorphisms in
single nucleotide polymorphism in pri-miR- vertebrates. PLoS One 7:e30737
126. Biochem Biophys Res Commun 51. Sacar Demirci MD, Yousef M, Allmer J
399:117–122 (2019) Computational prediction of func-
39. Jazdzewski K, Murray EL, Franssila K et al tional MicroRNA-mRNA interactions. Meth-
(2008) Common SNP in pre-miR-146a ods Mol Biol 1912:175–196
decreases mature miR expression and predis- 52. Yang H, Dinney CP, Ye Y et al (2008) Evalua-
poses to papillary thyroid carcinoma. Proc tion of genetic variants in microRNA-related
Natl Acad Sci U S A 105:7269–7274 genes and risk of bladder cancer. Cancer Res
40. Ryan BM, Robles AI, Harris CC (2010) 68:2530–2537
Genetic variation in microRNA networks: 53. Chen R, Zheng Y, Zhuo L et al (2017) The
the implications for cancer research. Nat Rev association between miR-423 rs6505162
Cancer 10:389–402 polymorphism and cancer susceptibility: a sys-
41. Shen J, Ambrosone CB, DiCioccio RA et al tematic review and meta-analysis. Oncotarget
(2008) A functional polymorphism in the 8:40204–40213
miR-146a gene and age of familial breast/ 54. Chen J, Jiang Y, Zhou J et al (2018) Evalua-
ovarian cancer diagnosis. Carcinogenesis tion of CpG-SNPs in miRNA promoters and
29:1963–1966 risk of breast cancer. Gene 651:1–8
42. Ye Y, Wang KK, Gu J et al (2008) Genetic 55. Tian T, Wang M, Zhu W et al (2017)
variations in microRNA-related genes are MiR-146a and miR-196a-2 polymorphisms
novel susceptibility loci for esophageal cancer are associated with hepatitis virus-related
risk. Cancer Prev Res (Phila) 1:460–469 hepatocellular cancer risk: a meta-analysis.
43. Clague J, Lippman SM, Yang H et al (2010) Aging (Albany NY) 9:381–392
Genetic variation in MicroRNA genes and risk 56. Alagona P Jr, Ahmad TA (2015) Cardiovas-
of oral premalignant lesions. Mol Carcinog cular disease risk assessment and prevention:
49:183–189 current guidelines and limitations. Med Clin
44. Guo H, Wang K, Xiong G et al (2010) A North Am 99:711–731
functional varient in microRNA-146a is asso- 57. Xiong XD, Cho M, Cai XP et al (2014) A
ciated with risk of esophageal squamous cell common variant in pre-miR-146 is associated
carcinoma in Chinese Han. Familial Cancer with coronary artery disease risk and its
9:599–603 mature miRNA expression. Mutat Res
45. Permuth-Wey J, Thompson RC, Burton 761:15–20
Nabors L et al (2011) A functional polymor- 58. Kuswanto CN, Sum MY, Qiu A et al (2015)
phism in the pre-miR-146a gene is associated The impact of genome wide supported
with risk and prognosis in adult glioma. J microRNA-137 (MIR137) risk variants on
Neuro-Oncol 105:639–646 frontal and striatal white matter integrity, neu-
46. Xu B, Feng NH, Li PC et al (2010) A func- rocognitive functioning, and negative symp-
tional polymorphism in pre-miR-146a gene is toms in schizophrenia. Am J Med Genet B
associated with prostate cancer risk and Neuropsychiatr Genet 168B:317–326
mature miR-146a expression in vivo. Prostate 59. Wright C, Gupta CN, Chen J et al (2016)
70:467–472 Polymorphisms in MIR137HG and micro-
RNA-137-regulated genes influence gray
matter structure in schizophrenia. Transl Psy- 72. Bao BY, Pao JB, Huang CN et al (2011)
chiatry 6:e724 Polymorphisms inside microRNAs and micro-
60. Whalley HC, Papmeyer M, Romaniuk L et al RNA target sites predict clinical outcomes in
(2012) Impact of a microRNA MIR137 sus- prostate cancer patients receiving androgen-
ceptibility variant on brain function in people deprivation therapy. Clin Cancer Res
at high genetic risk of schizophrenia or bipolar 17:928–936
disorder. Neuropsychopharmacology 73. Teo MT, Landi D, Taylor CF et al (2012) The
37:2720–2729 role of microRNA-binding site polymorph-
61. Pu X, Xiao X (2016) No evidence of an asso- isms in DNA repair genes as risk factors for
ciation between MIR137 rs1625579 and bladder cancer and breast cancer and their
schizophrenia in Asians: a meta-analysis in impact on radiotherapy outcomes. Carcino-
30 843 individuals. Psychiatr Genet genesis 33:581–586
26:203–210 74. Clop A, Marcq F, Takeda H et al (2006) A
62. Abtahi A, Samaei NM, Gholipour N, Moradi mutation creating a potential illegitimate
N (2018) No association between the SNP microRNA target site in the myostatin gene
rs1625579 in miR-137 gene and schizophre- affects muscularity in sheep. Nat Genet
nia in Iranian population. Egyptian J Med 38:813–818
Hum Genet 19:247–251 75. Abelson JF, Kwan KY, O’Roak BJ et al (2005)
63. Forstner AJ, Hofmann A, Maaser A et al Sequence variants in SLITRK1 are associated
(2015) Genome-wide analysis implicates with Tourette’s syndrome. Science
microRNAs and their target genes in the 310:317–320
development of bipolar disorder. Transl Psy- 76. Chin LJ, Ratner E, Leng S et al (2008) A SNP
chiatry 5:e678 in a let-7 microRNA complementary site in
64. Saus E, Soria V, Escaramis G et al (2010) A the KRAS 30 untranslated region increases
haplotype of glycogen synthase kinase 3beta is non-small cell lung cancer risk. Cancer Res
associated with early onset of unipolar major 68:8535–8540
depression. Genes Brain Behav 9:799–807 77. Nelson HH, Christensen BC, Plaza SL et al
65. Xu Y, Li F, Zhang B et al (2010) MicroRNAs (2010) KRAS mutation, KRAS-LCS6 poly-
and target site screening reveals a pre-mi- morphism, and non-small cell lung cancer.
croRNA-30e variant associated with schizo- Lung Cancer 69:51–53
phrenia. Schizophr Res 119:219–227 78. Pilarski R, Patel DA, Weitzel J et al (2012)
66. Gong J, Tong Y, Zhang HM et al (2012) The KRAS-variant is associated with risk of
Genome-wide identification of SNPs in developing double primary breast and ovarian
microRNA genes and the SNP effects on cancer. PLoS One 7:e37891
microRNA target binding and biogenesis. 79. Du W, Zhu J, Chen Y et al (2017) Variant
Hum Mutat 33:254–263 SNPs at the microRNA complementary site
67. Pelletier C, Speed WC, Paranjape T et al in the B7H1 3’untranslated region increase
(2011) Rare BRCA1 haplotypes including the risk of nonsmall cell lung cancer. Mol
3’UTR SNPs associated with breast cancer Med Rep 16:2682–2690
risk. Cell Cycle 10:90–99 80. Lee SY, Choi JE, Jeon HS et al (2015) A
68. Landi D, Gemignani F, Naccarati A et al genetic variation in microRNA target site of
(2008) Polymorphisms within micro-RNA- KRT81 gene is associated with survival in
binding sites and risk of sporadic colorectal early-stage non-small-cell lung cancer. Ann
cancer. Carcinogenesis 29:579–584 Oncol 26:1142–1148
69. Naccarati A, Pardini B, Stefano L et al (2012) 81. Chen S, He Y, Ding J et al (2010) An inser-
Polymorphisms in miRNA-binding sites of tion/deletion polymorphism in the 30
nucleotide excision repair genes and colorec- untranslated region of beta-transducin
tal cancer risk. Carcinogenesis 33:1346–1351 repeat-containing protein (betaTrCP) is asso-
70. Wynendaele J, Bohnke A, Leucci E et al ciated with susceptibility for hepatocellular
(2010) An illegitimate microRNA target site carcinoma in Chinese. Biochem Biophys Res
within the 3’ UTR of MDM4 affects ovarian Commun 391:552–556
cancer progression and chemosensitivity. Can- 82. Tan C, Liu S, Tan S et al (2015) Polymorph-
cer Res 70:9641–9649 isms in microRNA target sites of forkhead box
71. Luo J, Cai Q, Wang W et al (2012) A O genes are associated with hepatocellular
microRNA-7 binding site polymorphism in carcinoma. PLoS One 10:e0119210
HOXB5 leads to differential gene expression 83. Kang BW, Jeon HS, Chae YS et al (2016)
in bladder cancer. PLoS One 7:e40127 Impact of genetic variation in MicroRNA-
binding site on susceptibility to colorectal 95. Hou Y, Liang W, Zhang J et al (2018)

cancer. Anticancer Res 36:3353–3361 Schizophrenia-associated rs4702 G allele-
84. Lee AR, Park J, Jung KJ et al (2016) Genetic specific downregulation of FURIN expression
variation rs7930 in the miR-4273-5p target by miR-338-3p reduces BDNF production.
site is associated with a risk of colorectal can- Schizophr Res 199:176–180
cer. Onco Targets Ther 9:6885–6895 96. Li T, Kuang Y, Li B (2016) The genetic var-
85. Shaker OG, Mohammed SR, Mohammed iants in 30 untranslated region of voltage-
AM et al (2018) Impact of microRNA-375 gated sodium channel alpha 1 subunit gene
and its target gene SMAD-7 polymorphism affect the mRNA-microRNA interactions and
on susceptibility of colorectal cancer. J Clin associate with epilepsy. BMC Genet 17:111
Lab Anal 32:e22215 97. Tang H, Cheng Z, Ma W et al (2018) TLR10
86. Yang L, Li Y, Cheng M et al (2012) A func- and NFKBIA contributed to the risk of hip
tional polymorphism at microRNA-629- osteoarthritis: systematic evaluation based on
binding site in the 30 -untranslated region of Han Chinese population. Sci Rep 8:10243
NBS1 gene confers an increased risk of lung 98. Muralidhar B, Winder D, Murray M et al
cancer in southern and eastern Chinese popu- (2011) Functional evidence that Drosha over-
lation. Carcinogenesis 33:338–347 expression in cervical squamous cell carci-
87. Lin J, Zandi R, Shao R et al (2017) A noma affects cell phenotype and microRNA
miR-SNP biomarker linked to an increased profiles. J Pathol 224:496–507
lung cancer survival by miRNA-mediated 99. Rotunno M, Zhao Y, Bergen AW et al (2010)
down-regulation of FZD4 expression and Inherited polymorphisms in the
Wnt signaling. Sci Rep 7:9029 RNA-mediated interference machinery affect
88. Petkevicius V, Salteniene V, Juzenas S et al microRNA expression and lung cancer sur-
(2017) Polymorphisms of microRNA target vival. Br J Cancer 103:1870–1874
genes IL12B, INSR, CCND1 and IL10 in 100. Walz AL, Ooms A, Gadd S et al (2015)
gastric cancer. World J Gastroenterol Recurrent DGCR8, DROSHA, and SIX
23:3480–3487 homeodomain mutations in favorable histol-
89. Wang C, Zhao Y, Ming Y et al (2016) A ogy Wilms tumors. Cancer Cell 27:286–297
polymorphism at the microRNA binding site 101. Wegert J, Ishaque N, Vardapour R et al
in the 30 -untranslated region of C14orf101 is (2015) Mutations in the SIX1/2 pathway
associated with the risk of gastric cancer devel- and the DROSHA/DGCR8 miRNA micro-
opment. Exp Ther Med 12:1867–1872 processor complex underlie high-risk blaste-
90. Song X, Zhong H, Zhou J et al (2015) Asso- mal type Wilms tumors. Cancer Cell
ciation between polymorphisms of 27:298–311
microRNA-binding sites in integrin genes 102. Rakheja D, Chen KS, Liu Y et al (2014)
and gastric cancer in Chinese Han population. Somatic mutations in DROSHA and
Tumour Biol 36:2785–2792 DICER1 impair microRNA biogenesis
91. Wang B, Yang H, Shen L et al (2016) through distinct mechanisms in Wilms
Rs56288038 (C/G) in 3’UTR of IRF-1 tumours. Nat Commun 2:4802
regulated by MiR-502-5p promotes gastric 103. Caruso S, Calderaro J, Letouze E et al (2017)
cancer development. Cell Physiol Biochem Germline and somatic DICER1 mutations in
40:391–399 familial and sporadic liver tumors. J Hepatol
92. Manikandan M, Munirajan AK (2014) Single 66:734–742
nucleotide polymorphisms in microRNA 104. Zahedi F, Nazari-Jahantigh M, Zhou Z et al
binding sites of oncogenes: implications in (2017) Dicer generates a regulatory micro-
cancer and pharmacogenomics. OMICS RNA network in smooth muscle cells that
18:142–154 limits neointima formation during vascular
93. de Almeida RC, Chagas VS, Castro MAA et al repair. Cell Mol Life Sci 74:359–372
(2018) Integrative analysis identifies genetic 105. Chmielarz P, Konovalova J, Najam SS et al
variants associated with autoimmune diseases (2017) Dicer and microRNAs protect adult
affecting putative MicroRNA binding sites. dopamine neurons. Cell Death Dis 8:e2813
Front Genet 9:139 106. Stewart CJ, Charles A, Foulkes WD (2016)
94. Manley W, Moreau MP, Azaro M et al (2018) Gynecologic manifestations of the DICER1
Validation of a microRNA target site poly- syndrome. Surg Pathol Clin 9:227–241
morphism in H3F3B that is potentially asso- 107. Wen J, Gao Q, Wang N et al (2017) Associa-
ciated with a broad schizophrenia phenotype. tion of microRNA-related gene XPO5
PLoS One 13:e0194233 rs11077 polymorphism with susceptibility to
thyroid cancer. Medicine (Baltimore) 96: associated with lymph node metastasis and
e6351 poor clinical outcomes in non-small cell lung
108. Wu K, He J, Pu W et al (2018) The role of cancer. Oncol Lett 15:7744–7750
Exportin-5 in MicroRNA biogenesis and can- 114. Deng Y, Zhao F, Hui L et al (2017) Suppres-
cer. Genomics Proteomics Bioinformatics sing miR-199a-3p by promoter methylation
16:120–126 contributes to tumor aggressiveness and cis-
109. Melo SA, Moutinho C, Ropero S et al (2010) platin resistance of ovarian cancer through
A genetic defect in exportin-5 traps precursor promoting DDR1 expression. J Ovarian Res
microRNAs in the nucleus of cancer cells. 10:50
Cancer Cell 18:303–315 115. Leplae R, Tramontano A (1995) PLANET: a
110. Yi T, Arthanari H, Akabayov B et al (2015) phage library analysis expert tool. Physiol
eIF1A augments Ago2-mediated Dicer-inde- Chem Phys Med NMR 27:331–338
pendent miRNA biogenesis and RNA inter- 116. Shen Y, Pan X, Zhao H (2014) The histone
ference. Nat Commun 6:7194 demethylase PHF8 is an oncogenic protein in
111. Piletic K, Kunej T (2016) MicroRNA epige- human non-small cell lung cancer. Biochem
netic signatures in human disease. Arch Tox- Biophys Res Commun 451:119–125
icol 90:2405–2419 117. Yan M, Chen C, Gong W et al (2015)
112. Bueno MJ, Perez de Castro I, Gomez de miR-21-3p regulates cardiac hypertrophic
Cedron M et al (2008) Genetic and epige- response by targeting histone deacetylase-8.
netic silencing of microRNA-203 enhances Cardiovasc Res 105:340–352
ABL1 and BCR-ABL1 oncogene expression. 118. Zhou M, Zeng J, Wang X et al (2015) His-
Cancer Cell 13:496–506 tone demethylase RBP2 decreases miR-21 in
113. Min L, Wang F, Hu S et al (2018) Aberrant blast crisis of chronic myeloid leukemia.
microRNA-137 promoter methylation is Oncotarget 6:1249–1261
Chapter 14
Role of Exosomal MicroRNAs in Cell-to-Cell Communication

Bora Tastan, Emre Tarakcioglu, Yelda Birinci, Yongsoo Park,
and Sermin Genc
Abstract
Exosomes, a type of extracellular vesicle, are small vesicles (30–100 nm) secreted into extracellular space
from almost all types of cells. Exosomes mediate cell-to-cell communication carrying various biologically
active molecules including microRNAs. Studies have shown that exosomal microRNAs play fundamental
roles in healthy and pathological conditions such as immunity, cancer, and inflammation. In this chapter, we
introduce the current knowledge on exosome biogenesis, techniques used in exosome research, and
exosomal miRNA and their functions in biological and pathological processes.
Key words Exosome, MicroRNA, Cancer, Immune System, Inflammation, Central Nervous System,
Cardiovascular System
1 The Extracellular Vesicles
The term “extracellular vesicle” is used to define cell-derived and

membrane-surrounded structures, which are released into extracel-
lular space from multiple cell types ranging from prokaryotes to
eukaryotes [1, 2]. The size of extracellular vesicles (EVs) ranges
from 30 to 10,000 nm [3]. These structures were first described in
1946 as “pro-coagulant platelet-derived particles” [4]. After that,
in 1967, Peter Wolf identified microvesicles as “Platelets dust”
[5]. Secretion of EVs occurs in healthy and diseased conditions
including injury, cellular stress, or changes in homeostasis
[6, 7]. EVs are composed of a lipid bilayer membrane, which
protects the EVs and their cargo from protease and nuclease activity
[6]. The EVs are differ in their size/shape and origin that EVs are
further divided into mainly two sub-classes: exosomes and micro-
vesicles [6, 8]. The exosome derives from the endosome while
microvesicles form through shedding from plasma membranes
[2]. Also apoptotic bodies, a type of EV are released after apoptosis
[1]. Evidently, the EVs are found almost in all bodily fluids includ-
ing blood, urine, saliva, cerebrospinal fluid, milk, semen [6]. EVs
269
270 Bora Tastan et al.
play an important role in cell-to-cell communication by the cargo

molecules stored in EVs. The secreted EVs could carry a variety of
functional cargo molecules including cytosolic proteins, surface
receptors, lipids, miRNAs, and mRNAs to recipient cells [8]. The
content of the cargo reflects the origin of the secreted EVs and
thereby provides clues on the possible functions in recipient cells
[6]. EVs are involved in both physiological processes, that is, tissue
repair, blood coagulation, immunomodulation, and angiogenesis;
and diseases (e.g., cancer, neurodegenerative diseases, or inflamma-
tion) [3, 9]. These features of EVs enable researchers to use EVs for
biomarker research and production of therapeutics [10]
2 Exosome
Exosomes are a type of EV generally ranging from 30 to 100 nm in

diameter [11], which can be secreted from various types of cells
including stem cells, endothelial cells, and tumor cells
[12, 13]. Besides being secreted from various types of cells, exo-
somes are also present in all types of bodily fluid including plasma,
serum, urine, saliva, milk, and cerebrospinal fluid [11, 13].
Exosomes were first described in 1983 as secreted vesicles in
the process of externalization of transferrin receptors in reticulo-
cytes [14]. For a long time, exosomes were considered as a way to
dispose cellular waste out of the cytoplasm or to get rid of unde-
sired membrane proteins [11, 13]. Later, in 1996, a study showed
that the then so-called waste containing vesicles played a role in
antigen presentation [15]. This finding led to the idea that exo-
somes are not only the waste containing EVs but they have a
fundamental role in physiological processes. In 2007, another
study, a milestone for exosome research, demonstrated that exo-
somes originating from mast cells comprise more than 1200
mRNAs, thereby pointing out the functionality and importance
of the exosomes [16].
Lipids and proteins make up the exosomal membrane. Espe-
cially lipid rafts are abundantly present on the exosomal membrane
[17]. Other lipid molecules such as cholesterol, phospholipids,
phosphatidylserine, and prostaglandins constitute lipid content of
the exosome [11]. As a structural component, exosomes have
proteins on their surface such as tetraspanins (CD9, CD37,
CD53, CD63, CD81), heat shock proteins (Hsp60-70-90), and
cytoskeletal proteins (actin, tubulin). These surface proteins could
be universal or specific depending on the donor cells that released
the exosomes [12]. The universal membrane proteins such as CD9,
CD63, and CD81 have been used to identify and quantify
exosomes [11].
The secretion of exosomes occurs from multivesicular bodies
(MVBs) or late endosomes [18]. The MVBs fuse with their plasma
Exosomal miRNAs in Cell Communication 271
membranes and mediate the secretion of exosomes into extracellu-

lar space [19]. The secretion of exosomes may be induced by the
changes in the environment of the donor cells. For example, change
in pH, temperature, and oxygen/carbon dioxide levels, disturbing
homeostasis, could regulate the secretion of exosomes [12].
Exosomes have an important role in cell-to-cell communica-
tion through the cargo molecules they carry [12]. The content of
cargo molecules may contain all types of molecules being produced
in the donor cells. Not only lipids and proteins, which constitute
the exosomal membrane are found in exosomes, but also nucleic
acids including DNA, mRNA, miRNA and other noncoding RNAs
are also present in exosomes [20]. Cell-to-cell communication
depends on the cargo molecules contained in the exosomes [11]
(Fig. 1). Exosomes share phenotypic features with the donor cells,
thus providing the identification of an exosome’s origin [20]. The
cargo molecules provide evidence for the origin of the exosomes
and the physiological and pathological conditions, in which exo-
somes are released [11].
3 Exosome Biogenesis and Secretion
Exosome biogenesis is highly conserved process and regulated by

multiple mechanisms [21]. In general, exosome biogenesis starts
with the formation of intraluminal vesicles (ILVs) when inward
budding of the membranes of late endosomes occurs [22]. Trans-
portation of MVBs to cell surface cause fusion of MVBs with the
plasma membrane, thereby leading to release of enclosed ILVs
towards extracellular space [22]. Many studies support that exo-
some biogenesis is dependent on two major mechanisms.
The first and the most widely described exosome biogenesis
mechanism is the Endosomal Sorting Complex Required for Trans-
port (ESCRT)-dependent mechanism. This process occurs within
the endosomal system [21] and recognizes ubiquitinated proteins
[23]. The ESCRT machinery is one of the proteins on the exosomal
membrane and carries out the process from sorting of the exosomal
cargo to the secretion of the exosome [22, 24]. This mechanism
relies on various proteins and carbohydrates for exosome biogenesis
during MVB formation [21].
ESCRT machinery consists of five soluble multiprotein
complexes ESCRT-0-I-II-III- and VPS4 [22]. The protein
ESCRT-0 initiates the MVB pathway by recognizing and retaining
ubiquitinated proteins in the late endosomal membrane
[24]. Upon ESCRT-0 activation, ESCRT-I/II mediates the deter-
mination and initiation of the involution process of the limiting
membrane into the MVB lumen [23, 24]. ESCRT-I binds to
ubiquitinated proteins and this leads to the activation of ESCRT-
II [25]. After that, ESCRT-III mediates the final stage of
Fig. 1 Exosome secretion and its role in cell-to-cell communication: Exosomes are formed via MVBs or late
endosomes. Then the exosomes carrying biologically active molecules are released extracellular space by
donor cells and taken by recipient cells. Exosomes modulate biological functions of target cells via their cargo
biogenesis by constricting the budding neck and eventually VPS4,

an ATPase, induces membrane scission [23, 24].
However, studies eliminating the ESCRT machinery demon-
strated that ESCRT proteins are not required for exosome biogen-
esis. This finding suggested that other mechanisms besides the
ESCRT-dependent exosome biogenesis should exist. In 2008,
Trajkovic and colleagues showed that lipid rafts play a fundamental
role in the formation of MVBs [26]. Especially, ceramide and
sphingosine initiate ILVs budding into MVBs by inducing negative
membrane curvature [24]. Furthermore, studies demonstrated
that tetraspanin-enriched domains; that is, CD9 and CD63 are
also involved in the formation of ILVs [25, 27].
These two major mechanisms responsible for exosome biogen-
esis seems that they work independently. However, they might
work synergistically [28]. Exosomes from different origins might
adopt different biogenesis mechanisms. Beside these, biogenesis
pathways depend on the type of donor cells or different conditions
in which donor cells release exosomes.
After biogenesis and transportation of MVBs to the plasma
membrane and their fusion with the plasma membrane, the exo-
somes are released to extracellular space. The release of the exo-
somes is mainly mediated by two different mechanisms; that is, the
constitutive release via the trans-Golgi network and the inducible
release via MVBs directly [19, 29]. The exosome release mechanism
depends on the cell types and activation state of the cells [19].
The process of exosome release is regulated by several proteins,
including the Rab protein family and small GTPases [30]. Rab
proteins regulate intracellular vesicle trafficking in several steps
including transport, docking, and fusion by interacting with spe-
cific effector proteins [30]. The regulation of these mechanisms are
specific to the type of cells and the effector proteins [31]. Rab27,
Rab35, and Rab11 regulate exosome release [29, 32]. Studies
revealed that knocking down of Rab27 or its effector proteins
SYTL4 and EXPH5 ameliorated the release of exosomes in HeLa
cells [17]. Furthermore, Rab35 and Rab11 are involved in regula-
tion of exosome release by interacting with the GTPase-activating
protein TBC1 domain family member 10A-C [29].
Exosomes are transferred to the plasma membrane with assis-
tance of microtubules and actin [30]. The cytoskeleton compo-
nents, regulated by the Rab family and their effector proteins, direct
transportation towards the plasma membrane [30]. Once the exo-
somes are docked to the plasma membrane, fusion occurs through
the SNARE (Soluble N-ethylmaleimide-sensitive-fusion-protein-
attachment-protein-receptor) complex and the VAMP (Vesicle-
associated membrane protein) proteins [18].
4 Selective Loading of miRNAs into Exosome
Before exosome release, their cargo molecules including proteins,

lipids, and nucleic acids are packaged in the exosomes. Exosomes
carry RNAs, including mRNAs and noncoding RNAs such as miR-
NAs, circular RNAs (circRNAs), and long noncoding RNAs
(lncRNAs) [33]. These functional RNA molecules in the exosome
play a fundamental role in cell-to-cell communication by modulat-
ing gene expression and biological functions in healthy and patho-
logical conditions [34]. The presence of a miRNA in exosomes
does not always correlate with the cellular level of this miRNA
[34], suggesting that specific mechanisms direct miRNAs to
exosomes [30].
The mechanisms that direct exosome loading remains poorly
understood. There are several proposed mechanisms orchestrating
selective loading of miRNAs into exosomes. One of the mechan-
isms responsible for loading of miRNAs is the miRNA biogenesis
pathway. miRNAs are initially transcribed in the nucleus by RNA
polymerase II. The transcribed molecules called pri-miRNAs are
converted to pre-miRNAs by the Drosha complex [35]. The
pre-miRNAs are then transported to the cytoplasm by Exportin
5 [35]. The transported pre-miRNA is cleaved by the DICER/
TRBP complex and the RNA-induced silencing complex (RISC)
completes the process of miRNA maturation by incorporating the
mature miRNA. A study using monocyte-derived exosome-like
vesicles revealed that AGO2, a part of RISC, and GW182 proteins
are enriched in exosomes suggesting that these proteins might be
responsible for the selective loading of miRNAs into
exosomes [36].
Post-translational modification, such as ubiquitination and
sumoylation, is another mechanism involved in selective loading
of miRNAs into exosomes [34]. This process includes different

assisting molecules and RNA motifs called EXOmotifs [34]. The
heterogeneous nuclear ribonucleoprotein A2B1 (hnRNPA2B1) is
an RNA-binding protein and binds to groups of miRNAs by recog-
nizing their EXOmotifs and then directs exosome loading with
such miRNAs [37]. Furthermore, the sumoylation of hnRNPA2B1
plays a fundamental role in hnRNPA2B1–miRNA binding
[37]. Besides, the 30 end uridylation of miRNAs increases the
loading of miRNAs to exosomes [38]. 30 -uridylated miRNAs are
sorted into exosomes, whereas 30 -adenylated miRNAs remain
within the cytoplasm [39].
In 2013, Villarroya-Beltri and colleagues identified that miR-
NAs enriched in exosomes have specific sequences in T lympho-
blasts. The uncovered motifs, GGAG and CCCU are now called
EXOmotifs [37]. miRNAs containing EXOmotifs are directed to
exosomes, indicating EXOmotifs act as cis-acting elements for
RNA-binding proteins such as hnRNPA2B1 and SYNCRIP and
direct selective sorting [33].
Lipid rafts, a constitutional part of the exosomal membrane, are
associated with loading of miRNAs to exosomes [33]. Kosaka and
colleagues demonstrated that ceramide enriched lipid rafts regulate
the exosomal loading of miRNAs. In their study, they inhibited the
neutral sphingomyelinase 2 (nSMase2) enzyme, which has a role in
ceramide biosynthesis and this reduced exosomal miRNA levels as
well as the overall production of exosomes [40]. It is also known
that ceramide interacts with hnRNPA2B1, suggesting that these
molecules might collaborate for selective loading of miRNAs into
exosome [33].
Lastly, the cellular levels of miRNAs and their targets might
play a role in miRNA loading into exosomes [41]. Squadritoa and
colleagues reported that the increase or decrease in cellular miRNA
levels affects the exosomal miRNA levels. They also showed that the
overexpression of a miRNA’s target decreased exosomal miRNA
levels [42]. All of these studies showed that exosomal miRNAs are
selectively loaded into exosomes and this specific loading is regu-
lated by several mechanisms.
5 Exosome Isolation
In order to work with exosomes, the first step is the isolation of

exosomes. Exosome studies including functional, biomarker, and
therapeutic studies require the isolation of exosomes. The purity of
exosomes is the key to high quality studies since exosomes and EVs
are found in all bodily fluids and secreted from almost all cell types.
To yield exosomes of high purity, several isolation techniques have
been developed. Exosomes can be isolated using the size distribu-
tion of exosomes, exosomal morphology, their quantity, and the
biochemical composition of exosomes [43]. High efficiency and

applicability of the isolation methodology to different cell types and
tissues are two important features of isolation techniques
[43]. Each isolation method relies on a specific feature of the
exosomes, including density, shape, size, and structural compo-
nents like surface proteins [43]. Therefore, each isolation technique
has inherent advantages like specificity, purity and disadvantages
like time-consumption and cost. That is why the selection of the
appropriate isolation method depends on the type of study and its
main research question.
5.1 Differential The differential ultracentrifugation is the most useful method for
Ultracentrifugation exosome isolation and is accepted as the gold standard for exosome
Method studies [44]. Apart from being widely used, this method offers a
user friendly experience without the need of high expertise
[29]. The differential centrifugation process has several distinct
centrifugation steps. Generally, the process starts with low speed
centrifugation (300–500 g for 10–15 min) for removal of cell
debris, then it continues with medium speed (10,000 to
20,000 g for 20 min) to separate larger vesicles and eventually
ends with ultra-high speed centrifugation (100,000 g for >2 h)
for precipitation of exosomes [29, 44]. Another method for isola-
tion of exosomes is the density gradient. In short, the density
gradient method benefits from different densities of EVs leading
to different flotation densities of exosomes [29]. During the pro-
cess viscous solutions such as sucrose or iodixanol are used to form
flotation densities [3]. Then, the exosome and other EVs float
towards their appropriate gradients based on their size, shape, and
densities [3, 29]. Since each gradient keep different type of EVS,
this method offers contaminant-free isolation for exosome
[29]. During the ultracentrifugation-based isolation processes,
some alterations in the configuration of ultracentrifugation method
might be needed because of the diversity of the samples used for
exosome isolation. The samples might differ in some features such
as density or viscosity [44]. Not only the sample but also the
centrifuge affects centrifugation. The features of the centrifuge
such as maximum angular velocity and radius of motors determine
the efficiency of the ultracentrifugation process [44]. As mentioned
above, ultracentrifugation based methods are considered the gold
standard for exosome isolation because of simplicity, reliability and
speed. Nevertheless, the methods might damage exosomes,
decrease the quality of recovery and might be contaminated by
albumin, lipoproteins, and nucleic acids that obstruct further anal-
ysis [45, 46].
5.2 Size Based The second method for exosome isolation is the size based isolation
Exosome Isolation technique. In general, ultrafiltration followed by chromatography
Methods compromises size based isolation methods [44]. In the ultrafiltra-
tion process, a semipermeable membrane with defined pore size is
used [3]. Vesicles smaller than this defined pore size pass through
the membrane, however larger vesicles fail to pass [3]. Therefore,
the exosomes are selectively isolated based on their size [3]. In size
exclusion chromatography, the method relies on the use of small
porous polymer beads [3]. This polymer beads create a mesh which
allows small sized vesicles to travel fast through the mesh while
larger vesicles are retarded [3]. Both these methods are based on
the size and molecular weight of the exosome and offer fast and
cost effective separation techniques [29, 44]. Although being appli-
cable to large volume samples is an advantage, nonselective isola-
tion (including other EVs) and low concentrated recovery are
limitations of these methods [44]. However, the use of concentra-
tor could be a solution to increase recovery efficiency [29].
5.3 Immunoaffinity The next method for exosome isolation is immunoaffinity capture-
Capture-Based based methods. These methods rely on the selective capture of
Methods exosomes based on surface proteins on the exosomal membrane.
The presence of surface proteins serves as biomarkers for the spe-
cific detection of exosomes [29]. The use of specific antibodies
against surface proteins facilitates the selective capture of exosomes
[44]. The antibodies are coated on magnetic beads and after sepa-
ration is completed, the obtained exosomes are detached from the
magnetic beads [44]. The important step for the immunoaffinity
method is the selection of a convenient exosomal membrane pro-
tein. For instance, selection of universal exosomal surface proteins
such as tetraspanins (CD9, CD63, CD81) increases the efficiency
of the immunoaffinity method [29]. Besides specific antibodies, use
of synthetic peptides and single-stranded oligonucleotides are other
options when performing immunoaffinity capture-based methods
[44] Consequently, immunoaffinity capture-based methods offers
high specificity due to the use specific antibodies. Another advan-
tage of this approach is that the obtained exosomes are compatible
for further analysis including proteomics, transcriptomic, and
imaging with electron microscope [44].
5.4 Exosome The precipitation of exosomes using water-excluding polymers

Precipitation have been recently developed for exosome isolation [43]. This
method includes mixing samples with polymers, such as polyethyl-
ene glycol [24], for sedimentation of exosomes with low speed
centrifugation (10,000–20,000 g) [44]. The water excluding
polymers react with water molecules, thus sedimenting less soluble
molecules out of solution [43]. Commercial kits such as Total
Exosome Isolation Kit from (Invitrogen), ExoQuick (System Bios-
ciences), and Exosome Isolation Kit (Life Technologies) employ
the exosome precipitation method. Although this method is sim-
ple, compatible with various bio-fluids and time efficient (no need
for long centrifuge steps), the recovery could contain some lipo-
proteins or immune complexes that reduce the purity of exosomes
[29, 44, 45].
6 Characterization of Exosomes
After isolation, the recovered exosomes can be characterized. In

general, exosomes are characterized by their size and their lipid and
protein composition [47]. There are several techniques used for the
characterization of exosomes including physical and molecular
techniques. Physical techniques are used to confirm vesicle size
and concentration, whereas molecular techniques analyze the con-
tent of exosomes [2].
The most used physical characterization techniques are elec-
tron microscopy (EM), dynamic light scattering (DLS), and nano-
particle tracking analysis (NTA) [2]. EM is one of the optical
techniques used for physical characterization of exosomes
[31]. This technique is accepted as the gold standard for the
characterization of exosomes [2]. During the process, isolated
exosomes are placed on an EM sample grid [18]. A substance like
uranyl acetate or methylcellulose is used for negative staining,
allowing clear visualization of exosomes under the EM [18]. Exo-
somes are visible as cup-shaped double membrane bound vesicles in
the size range 30 to 100 nm [18]. However, this cup-shaped
morphology is due to the drying caused during sample preparation
[48]. Furthermore, antibodies against exosomal markers could be
used in EM [18]. In this case, the secondary antibody containing
gold nanoparticles is used for enhanced visualization using EM
[18]. Although EM is accepted as the gold standard for exosome
characterization in manner of measurements of size, the process is
time consuming, needs expensive equipment and has limitation for
quantification [49].
DLS and NTA are two other exosome characterization techni-
ques relying on light scattering and Brownian motion of particles
[48]. DLS, also known as photon correlation spectroscopy, pro-
vides the relative size distribution by detecting change in the inten-
sity of scattered light [50]. The intensity of light changes because of
the Brownian motion, which is the random motion of particles due
to collision of particles with solvent molecules in a fluidic condition
[2, 50]. The advantage of DLS is that it allows fast and bulk
characterization of exosomes [2]. However, heterogeneity in the
sample can affect the performance and efficiency of DLS [50].
Like DLS, NTA works based on light scattering and the Brow-
nian motion principles. NTA, employing light scattering and Brow-
nian motion [18], determines the absolute size distribution of
exosomes [50]. Unlike DLS, it tracks exosomes in biofluids directly
and individually [2, 18]. NTA also provides the concentration of
exosomes with the use of spike-in beads of known size and concen-
tration [18, 50]. Furthermore, the use of fluorescently labeled
antibodies in NTA is convenient. In this case, the device tracks
the fluorescent signal and determines its size and concentration,
thereby providing more reliable data [50].
Another physical characterization method is flow cytometry. A

flow cytometer is not applicable for exosomes because of the size of
exosomes. Exosomes fail to pass the size threshold. However, the
development of new flow cytometers and the use of latex beads,
which immobilize exosomes, allows exosome characterization
[18]. The conjugation of fluorescently tagged antibodies with
these latex beads is also possible. For this purpose, exosomal surface
markers including membrane proteins such as tetraspanins (CD9,
CD63 and CD81), TSG101, Alix, flotillin 1, or lipids like phos-
phatidylserine could be used [18, 47].
The physical characterization methods are not able to reveal the
content of exosomes. For this purpose, molecular characterization
methods are applied. To investigate the protein content of exo-
somes, conventional methods like western blotting can be used
[2]. However, western blotting is only convenient for known pro-
teins with available antibodies. However, for the characterization of
the complete proteomic content, methods like liquid chromatog-
raphy coupled to tandem–mass spectrometry (LC–MS/MS) is used
[2]. To analyze the RNA content, conventional reverse transcrip-
tion PCR (RT-PCR) can be used. This method, similarly to blot-
ting depends on knowledge about the expected content and
available primers. For the whole transcriptomic content, RNA-seq
should be employed [2].
Besides these characterization methods, several other methods
are available for exosome characterization, for example, Raman
micro spectroscopy, atomic force microscopy, tunable resistive
pulse sensing, field emission scanning electron microscopy
(FESEM), micro nuclear magnetic resonance, small-angle X-ray
scattering (SAXS), and anomalous SAXS [18, 49].
7 Cell-to-Cell Communication with Exosomal miRNAs in Cardiovascular System
Many cell types including cardiac myocytes, endothelial cells, fibro-

blasts, immune cells, and stem cells need to interact with each other
for proper heart function. Exosomal miRNAs mediate communi-
cation between different cell types in the heart. Physiological stress
and statin therapy enhance the release of miR-143/145 by exo-
somes from HUVEC cells and miR-143/145 are transferred to
smooth muscle cells [51] (Table 1). Downregulation of
miR-143/145 leads to an increase in atherosclerotic lesion area,
suggesting that endothelial cells signaling via exosomal miR-143/
145 exhibits atheroprotective function in smooth muscle cells.
miRNAs in cardiac derived exosomes could also alter endothelial
cell function. For example, exosomal transfer of miR-320 from
cardiomyocytes into endothelial cells shows an antiangiogenic
function in type 2 diabetic rats [52]. Exosomal miRNAs play a
role in signal transduction between cardiac muscle cells.
Table 1
List of miRNAs that are transferred between cardiovascular system, together with their observed
biological functions
Donor cell miRNA in exosome Recipient cell Biological function References

HUVEC miR-143/145 Human aortic Promotion of [51]
smooth atheroprotective function
muscle cell
Cardiomyocytes miR-320 Cardiac Inhibition of endothelial cell [52]
endothelial proliferation, migration
cells and tube formation
Cardiomyoblast miR-30a Cardiomyoblast Inhibition of autophagy [53]
cell line cell line
(H9c2) (H9c2)
T cells miR-142-3p HUVEC Increased endothelial [54]
permeability
Primary human miR-155 and HUVEC Activation of NF-κB and [55]
monocytes miR-223 dysfunction of endothelial
cells
Bone-marrow- Mir-155 Cardiac Suppression of proliferation [56]
derived fibroblasts and induction of
macrophages inflammation
Atherogenic miR-146a, miR-128, Naı̈ve Decrease in migration of [57]
macrophages miR-185, miR-365, macrophages macrophage
and miR-503
Plasma miR-223 Human Decreased IL-6 and NLRP3 [58]
Monocytic expression
cell line
(THP1)
Rat cardiac miR-21* Cardiomyocytes Induction of cellular [59]
fibroblasts hypertrophy
Cardiac miRNA-27a, miRNA- Cardiac Inhibition of Nrf2 signaling [60]
fibroblasts 28a, and miRNA- myocytes pathway, induction of
34a oxidative stress
Human cardiac miR-210 and Cardiomyocytes Inhibition of apoptosis, [61]
progenitor miR-132 improvement of cardiac
cells function
Endothelial cells miR-126 Cardiomyocytes Decrease in cardiomyocytes [62]
size, induction of cardiac
dysfunction
Cardiomyocytes miR-1 Hippocampal Dissolution of microtubules [63]
and cortex
neurons
Cardiomyocytes release miR-30a-containing exosomes under hyp-

oxic conditions and inhibit autophagy in neighboring target
cells [53].
Exosomal miRNAs take part in the crosstalk between immune
and cardiac cells. Exosomal miR-142-3p secreted from T cells is
transferred to endothelial cells in order to enhance vascular perme-
ability [54]. Altered expression of miR-155 and miR-223 in exo-
somes from monocytes causes endothelial cell dysfunction via the
TLR4 and NF-κB pathways [55]. mir-155 in macrophage-derived
exosomes can be transferred into cardiac fibroblasts where it may
inhibit proliferation and promotes inflammation [56]. An athero-
genic stimulus leads to the enrichment of miR-146a, miR-128,
miR-185, miR-365, and miR-503 in atherogenic exosomes and
these exosomes are taken up by recipient macrophages thereby
reducing macrophage migration [57]. Plasma exosomal miR-223
from patients with cardiopulmonary bypass regulates IL-6 and
NLRP3 expression in THP-1 monocytes [58]. Cardiac fibroblasts
may contribute to cardiac remodeling through secreted exosomes.
Cardiac fibroblast-derived exosomal miR-21* is transported to
neighboring cells and induces cellular hypertrophy in cardiomyo-
cytes [59]. Cardiac fibroblast-derived miRNA-27a, miRNA-28a,
and miRNA-34a are taken up by cardiomyocytes and induce oxida-
tive stress via the inhibition of the Nrf2 signaling pathway [60]. Car-
diac progenitor cells contribute to the recovery of cardiac function
by altering exosomal miRNA expression. miR-210 and miR-132 in
exosomes secreted by cardiac progenitor cells are taken into cardi-
omyocyte and inhibit apoptosis [61].
Exosomal miRNA-mediated communication can also be
observed between organs. Ischemic stroke directly induces cardiac
dysfunction through low level of miR-126 in exosomes derived
from endothelial cells [62]. Under hypoxic conditions, miR-1
expression is increased in cardiomyocyte-derived exosomes and
transported to neonatal rat hippocampal and cortex neurons, thus
causing neuronal microtubule damage [63].
8 Cell-to-Cell Communication with Exosomal miRNAs in Cancer
miRNA transfer via exosome between cancer cells or cancer-

fibroblast, cancer-endothelial cells or cancer-immune cells may
also contribute to carcinogenesis during initiation, promotion,
invasion, metastasis, angiogenesis, escape of immunity, and for
drug resistance. M2 macrophages secrete miR-223-containing exo-
somes, which are internalized by breast cancer cells and thereby
promote the invasion capacity of breast cancer cells [64] (Table 2).
miR-200 in exosomes is transferred from high metastatic breast
cancer cells to low metastatic breast cancer cells, thereby increasing
the colonization capacity of breast cancer cells [65]. Angiogenesis is
Table 2
List of miRNAs that are transferred between cancer cells and immune or surrounding endothelial
cells, together with their observed biological functions
miRNA in
Donor cell exosome Recipient cell Biological function References
Monocyte-derived miR-223 Breast cancer cell Promotion of breast [64]
macrophages lines (SKBR3 and cancer cells’ invasion
MDA-MB-231)
Metastatic breast miR-200 Poorly metastatic Transfer of metastatic [65]
cancer cell lines breast cancer cell capability
(4T1, HER2+ lines (4TO7, basal-
MCF10CA1a, B TNBC
basal-A TNBC MDA-MB-231)
BPLER
H1299, MDA435, miR-9 Microvascular Promotion of endothelial [66]
Panc1, SF-539, endothelial cells cell migration and
HM7 matching tumor tumor angiogenesis
type
Mouse breast cancer miR-210 HUVEC Enhancement of [67]
cell line (4T1) angiogenesis
Human lung miR-100-5p Human lung Increased resistance of [68]
adenocarcinoma adenocarcinoma recipient cells against
cell line (A549) cell line (A549) cisplatin
Cancer-associated miR-21, Breast Cancer Cell Promotion of stemness, [70]
fibroblasts miR278e, Line (T47D) epithelial–mesenchymal
miR143 transition, and
anchorage-independent
cell growth
Cancer-associated microRNA- Cancer epithelial cells Increased cell [71]
Fibroblasts 146a proliferation, tumor
growth, and
chemoresistance
Human epithelial miR222-3p Macrophages Transition to tumor- [72]
ovarian cancer cell differentiated from associated macrophage
lines (Skov3, monocytic cell line (TAM)-like phenotype
A2780) U937
a main event for tumor growth and tumor cell-derived exosomes

may be involved in this process by transferring pro-angiogenic
miRNAs. miR-9-containing exosomes stimulate HUVEC endo-
thelial cell migration and angiogenesis [66]. Exosomal miR-210
from mouse breast cancer cells is transferred to recipient HUVEC
cells where it enhances angiogenesis in endothelial cells [67]. Exo-
somes play a role in communication between cancer cells by trans-
ferring chemoresistance. Exosomal miR-100-5p from cisplatin-
resistant lung cancer cells can modify cisplatin sensitivity of other
lung cancer cells [68]. In addition, several exosomal miRNAs take

part in chemoresistance transfer, including miR-222, mir-23a,
miR-24, and mir-452 [69]. Cancer-associated fibroblasts support
tumor microenvironment by transferring exosomal miRNAs and
alter target cells’ functions. For instance, EMT and stemness in
breast cancer cells are modulated by miR-21, miR278e, and
miR143 in cancer-associated fibroblast exosomes [70]. miR-146
containing exosomes, which are secreted from cancer-associated
fibroblasts, modulate the chemoresistance of pancreatic cells
[71]. Bidirectional dialogue between immune and cancer cells
also modulate immune responses. Epithelial ovarian cancer cells
secrete exosomes containing miR222-3p, inducing macrophages
into tumor supportive M2 polarization [72]. Various miRNAs
contribute to cancer- immune cell communication such as
miR-23a, miR-203, miR-21, and miR-29a [69].
9 Cell-to-Cell Communication with Exosomal miRNAs in the Central Nervous

System
miRNAs carried in exosomes within the brain, or between the brain

and the periphery can have various roles in health and disease.
Initial studies have focused on the role of exosomal miRNAs in
the communication between brain tumors and neighboring cells. A
study involving glioblastoma cells has shown that EVs—including
exosomes—from glioblastoma alter the transcriptome of mouse
microglia, monocytes, and macrophages [73] (Table 3). The
study has found that EVs isolated from primary human glioblas-
toma and from the mouse glioma line GL261 are enriched in
miR-21 and miR-451. They also demonstrated that primary
mouse microglia take up this miRNA cargo when exposed to the
human glioblastoma EVs; and that monocytes and macrophages in
GL261-implanted mouse brains take up the miRNA cargo of EVs
released from GL261 cells. This transfer of miR-21 and miR-451
results in the downregulation of the c-Myc mRNA (a common
target of both miRs) in both scenarios. Given that c-Myc promotes
growth and proliferation, it is likely that reduction of c-Myc leads to
poorer antitumor activity of immune cells in the brain. In another
study using the human glioblastoma cell line U251, miR-21 and
miR-451 (among other miRs) were highly abundant in vesicles
released from these cells [74]. This study suggests U251 vesicles
have an effect on the transcriptome of brain endothelial cells
exposed to these exosomes. Notably, in contrast to the microglia
study done by van der Vos et al. in 2016, no significant change in
c-Myc mRNA in endothelial cells was observed [73, 74].
Table 3
List of miRNAs that are transferred between neural cells and within the central nervous system,
together with their observed biological functions
miRNA in
Primary human miR-21, Primary mouse microglia; c-Myc downregulation [73]
glioblastoma; miR-451 monocytes, macrophages
GL261 cells
(mouse glioma)
U251 cells (human miR-21, Primary human brain Various changes in [74]
glioblastoma) miR-451 microvascular endothelial transcriptome
Cells
Primary mouse miR-19a Breast cancer metastases in PTEN downregulation [75]
astrocytes brain
Human breast miR-122 Primary mouse astrocytes GLUT1 [76]
cancer cell lines downregulation
(MCF7, SKBR3
etc.)
Not characterized Let-7i Primary human T cells IGF1R and TGFBR1 [82]
downregulation,
reduction of Treg
frequency
Zebrafish neurons, miR-132 Brain endothelial cells Cdh5 upregulation; [83]
rat neurons (zebrafish), b.End3 cells improvement in
(mouse brain vascular blood brain barrier
endothelial cell line)
Not characterized miR-219 Rat oligodendrocyte Differentiation of [84]
progenitor cells (OPCs) OPCs into
functional
oligodendrocytes
Exchange of miRNA-loaded exosomes between the brain and

metastatic tumors has also been reported. One study has revealed
(both in coculture and in vivo) that astrocytes in the brain promote
a tumorigenic phenotype in brain metastases of breast cancer by
secreting exosomes with a cargo of miR-19a, which targets and
downregulates PTEN, a major tumor suppressor, in the tumor cells
[75]. Another study looking into how cancer cells reprogram the
premetastatic niche has reported that secreted exosomes from
human breast cancer cell lines are enriched in miR-122. miR-122-
enriched exosomes reduces the levels of glucose transporter
GLUT1 by downregulating an upstream protein PKM2 in cultured
mouse astrocytes [76]. Further in vivo experiments show that mice
with miR-122-overexpressing tumor xenografts also have a similar
decrease in GLUT1 level in their brain sections [76].
Besides cancer, there have also been studies on exosomal miR-

NAs in neurodegenerative diseases. To date, exosomes isolated
from the blood and cerebrospinal fluid of Parkinson’s disease and
Alzheimer’s disease patients have been analyzed for differentially
expressed (DE) miRNA [77–80]. Exosomal miRNAs that are con-
sistently DE in these diseases include the miR-29 family (down-
regulated in Alzheimer’s disease) and miR-19b (downregulated in
Parkinson’s disease). The effects of these exosomal miRNAs in
recipient cells are not yet well characterized.
Recent studies of exosomal miRNAs in multiple sclerosis (MS),
an autoimmune disease where unchecked inflammatory T cell activ-
ity destroys myelin sheaths in the central nervous system, identified
several miRNAs upregulated in MS patients. A 2017 comparison
study found three miRNAs significantly upregulated in the serum
exosomes of progressive MS patients compared to healthy controls:
miR-370-3p, miR-409-3p, and miR-432-5p [81]. A study of
plasma exosomes also found exosomal miRNAs upregulated in
MS patients; let-7i, miR-19b, miR-25, and miR-92a are the most
significantly changed miRNAs [82]. The latter study also demon-
strates that let-7i targets IGF1R and TGFBR1 mRNAs in T cells,
that Treg cell of MS patients have reduced IGF1R and TGFBR1
expression, and that let-7i exosomes reduce Treg frequency in
cultures of primary T cells from healthy humans.
More physiological roles of exosome-mediated miRNA transfer
in the brain are also reported. In a recent study on zebrafish,
miR-132 produced in neurons is carried to brain endothelial cells
through exosomes, and in turn contributes to the integrity of the
blood brain barrier by upregulating vascular endothelial cadherin
(Cdh5) [83]. In vitro treatment of a mouse vascular endothelial cell
line with miR-132 containing exosomes isolated from rat cortical
neurons results in miR-132 uptake and Cdh5 increase, implying
that this phenomenon is conserved in mammals [83].
A study in rats has found that serum exosomes isolated from
young rats or rats with higher social/physical activity contain
increased levels of miR-219. Administration of these exosomes to
aged rats shows that miR-219 in the exosomes is taken up by
oligodendrocyte progenitor cells and stimulates their differentia-
tion into functional oligodendrocytes, thus promoting
myelination [84].
Another role of miRNAs trafficking in brain might involve
miR-155. Serum exosomes from mice treated intraperitoneally
with LPS are enriched with miR-155. Brains of mice injected with
these exosomes show an upregulation of miR-155, together with
microglial activation [85].
10 Cell-to-Cell Communication with Exosomal miRNAs in Inflammation and

Immunity
Trafficking of miRNAs, including exosomal miRNAs, is a hallmark

of inflammation. There have been several in vivo and in vitro studies
showing that exosomal miRNAs accelerate or reduce inflammation
in health and disease [86].
Rheumatoid arthritis (RA) is a key chronic inflammatory dis-
ease that occurs in the joints and cell-to-cell interaction play a role
in disease pathogenesis [87]; and expectably exosomes and miR-
NAs have roles in it. Exosomal miRNA exchange between macro-
phages and other myeloid cells [88] (Table 4), between
synoviocytes and osteoblasts [89], and between peripheral blood
mononuclear cells and macrophages [90] are implicated in the
progression or prevention of RA. The miRNA let-7b is highly
expressed, predominantly by macrophages, in the synovial fluid of
RA patients and binds to TLR-7 [88]. Moreover, let-7b packaged
in exosomes drives polarization of myeloid cells to an M1 (inflam-
matory) phenotype in vitro as well as inducing joint swelling in wild
type mice [88]. In a mouse model of RA, miR-221-3p is enriched
in synovia of mice with inflamed joints [89]. The same study has
investigated this miRNA further: exosomes isolated from primary
fibroblast-like synoviocytes treated with TNF have an upregulation
of miR-221-3p, and these exosomes inhibit differentiation of (and
mineralization in) primary osteoblasts. Another study of RA has
found that exosomal miR-548a-3p is significantly downregulated
in sera and peripheral blood mononuclear cells of RA patients
compared to healthy controls, and that miR-548a-3p level is nega-
tively correlated with CRP levels in serum [90]. The same study has
identified TLR4 as a target of miR-548a-3p and overexpression of
miR-548a-3p reduces NF-κB levels in LPS-stimulated
macrophages.
There are multiple reports of exosomal miRNA trafficking in
normal immune function, as well. Regulatory T cells, for example,
suppress inflammatory Th1 cells by secreting pro-apoptotic and
anti-prolifeative miRNAs-containing exosomes which are taken
up by Th1 cells [91]. The corresponding study further shows that
one of the enriched miRNAs in these exosomes, let-7d, can sup-
press inflammation by itself in Th1 cells. An analysis of exosomes
isolated from three different immune cell types (J77 T cells, Raji B
cells, and primary dendritic cells derived from human monocytes)
reveals that the miRNA content of exosomes from different cell
types are similar in general compared to miRNA in host cells
[92]. This implies that the exosomal miRNA content is not simply
a byproduct of basal miRNA synthesis. The same study also demon-
strates that formation of an immune synapse dramatically increases
uptake of exosomes between cells. They further show that miR-335
Table 4
List of miRNAs that are transferred between cells of the immune system or cells that are involved in
inflammatory conditions, together with their observed biological functions
miRNA in
Mouse synovial fluid Let-7b Mouse bone Polarization to M1 [88]
macrophages marrow phenotype
macrophages
Mouse fibroblast-like miR-221- Mouse primary Inhibition of [89]
synoviocytes 3p osteoblasts osteoblast
differentiation
Mouse primary Treg cells Let-7d Mouse Th1 cells Suppression of [91]
inflammation
J77 cells (human T cell line) miR-335 Raji cells (human B SOX4 downregulation [92]
cell line)
Mouse bone marrow-derived miR-155 Mouse BMDCs Upregulation of [95]
dendritic cells (BMDCs) inflammatory genes
Mouse BMDCs miR-146a Mouse BMDCs Downregulation of [95]
inflammatory genes
transfer from T cells to B cells in cell culture downregulates the

expression of the miR-335 target SOX4. SOX4 is a differentiation-
associated transcription factor required for pro-B cell survival [93].
Therefore, the transfer of miR-335 from T cells to B cells might be a
mechanism of immunological tolerance. Exosomal miRNA transfer
also occurs between dendritic cells [94, 95]. Exosomes isolated
from immature/tolerogenic and mature/activated dendritic cells
have different profiles of miRNA cargo stored in exosomes [94],
implying that the difference in cell state determines the content of
exosomes.
miR-155 and miR-146a are central miRNAs in inflammation,
where miR-155 contributes to the induction of the NF-κB pathway
by reducing SHIP1 and IRAK-M expression, and miR-146a sup-
presses the same pathway by reducing IRAK1 and TRAF6 expres-
sion [96, 97]. As expected, exosomal trafficking of miR-155 and
miR-146a also plays a key role in the regulation of inflammation.
For example, miR-155 and miR-146a are transferred between
dendritic cells in cell culture, and dendritic cells taking up exosomes
that contain these miRNAs can alter their response to LPS.
miR-155 promotes and miR-146a dampens expression of inflam-
matory genes upon exposure to LPS [95]. The same group has also
demonstrated the effects in mice. Specifically, mice lacking
miR-155 are rescued from reduced responsiveness to LPS when
injected with wild-type exosomes, and mice lacking miR-146a are
rescued from their heightened responsiveness to LPS when injected
with wild type exosomes [95]. A later study from the same group
has recreated the two phenotypes above in one model. They report
that mice with defective exosome secretion (via a Rab27a/Rab27b
double knockout) have a systemic low-grade inflammation charac-
terized by elevated inflammatory cytokine levels and increased
myeloid cell proliferation, as well as reduced responsiveness to
LPS [97]. Thus, without exosomes bearing miR-146a and
miR-155, both stimulation and suppression of the immune system
are compromised. It is also worth noting that miR-155 has diverse
targets including some proinflammatory ones, and thus can act as
an anti-inflammatory miRNA. An example is an in vitro foam cell
model, where higher miR-155 expression reduces TNF-α secretion
from foam cells [98].
An unusual area in the field of exosomal miRNA is the study of
exosome-like vesicles produced by plants, termed exosome-like
nanoparticles (ELNs). These nanoparticles, which are structurally
similar to mammalian exosomes can carry miRNAs and can be
taken up by cells in the mammalian intestine, including stem cells
and macrophages [99–101]. Thus, cell-to-cell miRNA trafficking
might exist between species, albeit in a more roundabout manner.
A recent study of ELNs has identified 11 miRNAs enriched in
ELNs of 11 different plants and predicted target genes that code
for cytokines, suggesting that these miRNAs might potentially be
involved in inflammatory regulation. Some of these miRNAs
include miR-398b from orange (targeting IL-1A), miR-4996
from soybean (targeting IL-10), and miR-1078 from ginger (tar-
geting IL-6) [102]. While cross-kingdom regulation of cell func-
tions via exosomal miRNAs is a revolutionary finding, risk of
contamination and artifact generation in sequencing studies should
not be ignored. For instance, it has been reported that plant miR-
NAs detected in mammalian breast milk exosomes may arise from
contamination of samples during analysis [103].
11 Conclusion
Exosomes play a fundamental role in the communications between

cells via the cargo molecules they carry. Thus, exosomes derived
from different cell types exhibit pleiotropic biological activities in
target cells, either protective or detrimental, which are primarily
dependent on the cellular origin and current status. Discovery of
extracellular vesicles that play a role in cell-to-cell communication
through their cargos such as noncoding RNA open new views of
cell communication. Exosomes, a subclass of EVs, are found in all
types of bodily fluids and carry a variety of molecules including
lipids, proteins, nucleic acids, and noncoding RNA. Exosomal
miRNAs present good candidates for biomarkers.
Although exosomal miRNA transportation is not well under-

stood yet, two main issues have been identified: selective loading of
miRNAs into exosomes and targeting of exosomes. The studies
that show a difference between cellular miRNAs and exosomal
miRNAs indicate selective loading of miRNAs into exosomes
occurs. The mechanism of selective loading of miRNAs into exo-
somes and targeting of exosomes remain poorly understood.
References
1. Hafiane A, Daskalopoulou SS (2018) Extra- 12. Yuan MJ, Maghsoudi T, Wang T (2016) Exo-
cellular vesicles characteristics and emerging somes mediate the intercellular communica-
roles in atherosclerotic cardiovascular disease. tion after myocardial infarction. Int J Med Sci
Metabolism 85:213–222 13:113–116
2. Lane RE, Korbie D, Hill MM et al (2018) 13. Bang C, Thum T (2012) Exosomes: new
Extracellular vesicles as circulating cancer bio- players in cell-cell communication. Int J Bio-
markers: opportunities and challenges. Clin chem Cell Biol 44:2060–2064
Transl Med 7:14 14. Pan BT, Johnstone RM (1983) Fate of the
3. Willms E, Cabanas C, Mager I et al (2018) transferrin receptor during maturation of
Extracellular vesicle heterogeneity: subpopu- sheep reticulocytes in vitro: selective external-
lations, isolation techniques, and diverse func- ization of the receptor. Cell 33:967–978
tions in cancer progression. Front Immunol 15. Raposo G, Nijman HW, Stoorvogel W et al
9:738 (1996) B lymphocytes secrete antigen-
4. Chargaff E, West R (1946) The biological presenting vesicles. J Exp Med
significance of the thromboplastic protein of 183:1161–1172
blood. J Biol Chem 166:189–197 16. Valadi H, Ekstrom K, Bossios A et al (2007)
5. Wolf P (1967) The nature and significance of Exosome-mediated transfer of mRNAs and
platelet products in human plasma. Br J Hae- microRNAs is a novel mechanism of genetic
matol 13:269–288 exchange between cells. Nat Cell Biol
6. Konoshenko MY, Lekchnov EA, Vlassov AV 9:654–659
et al (2018) Isolation of extracellular vesicles: 17. Zhang J, Li S, Li L et al (2015) Exosome and
general methodologies and latest trends. exosomal microRNA: trafficking, sorting, and
Biomed Res Int 2018:8545347 function. Genomics Proteomics Bioinformat-
7. Xu R, Rai A, Chen M et al (2018) Extracellu- ics 13:17–24
lar vesicles in cancer - implications for future 18. Emanueli C, Shearn AI, Angelini GD et al
improvements in cancer care. Nat Rev Clin (2015) Exosomes and exosomal miRNAs in
Oncol 15(10):617–638. https://doi.org/10. cardiovascular protection and repair. Vasc
1038/s41571-018-0036-9 Pharmacol 71:24–30
8. Paolicelli RC, Bergamini G, Rajendran L 19. Record M, Subra C, Silvente-Poirot S et al
(2018) Cell-to-cell communication by extra- (2011) Exosomes as intercellular signalo-
cellular vesicles: focus on microglia. Neurosci- somes and pharmacological effectors. Bio-
ence 405:148–157. https://doi.org/10. chem Pharmacol 81:1171–1182
1016/j.neuroscience.2018.04.003 20. Samanta S, Rajasingh S, Drosos N et al (2018)
9. Yang J, Li C, Zhang L et al (2018) Extracel- Exosomes: new molecular targets of diseases.
lular vesicles as carriers of non-coding RNAs Acta Pharmacol Sin 39:501–513
in liver diseases. Front Pharmacol 9:415 21. Rajagopal C, Harikumar KB (2018) The ori-
10. Lu M, Xing H, Xun Z et al (2018) Functio- gin and functions of exosomes in cancer.
nalized extracellular vesicles as advanced ther- Front Oncol 8:66
apeutic nanodelivery systems. Eur J Pharm Sci 22. Jones LB, Bell CR, Bibb KE et al (2018)
121:34–46 Pathogens and their effect on exosome bio-
11. Alipoor SD, Mortaz E, Garssen J et al (2016) genesis and composition. Biomedicine 6:79
Exosomes and Exosomal miRNA in respira- 23. Farooqi AA, Desai NN, Qureshi MZ et al
tory diseases. Mediat Inflamm 2016:5628404 (2018) Exosome biogenesis, bioactivities and
functions as new delivery systems of natural composition in cells from exosomes. Cell
compounds. Biotechnol Adv 36:328–334 Rep 8:1649–1658
24. Bebelman MP, Smit MJ, Pegtel DM et al 39. Ferguson SW, Nguyen J (2016) Exosomes as
(2018) Biogenesis and function of extracellu- therapeutics: the implications of molecular
lar vesicles in cancer. Pharmacol Ther composition and exosomal heterogeneity. J
188:1–11 Control Release 228:179–190
25. Abels ER, Breakefield XO (2016) Introduc- 40. Kosaka N, Iguchi H, Yoshioka Y et al (2010)
tion to extracellular vesicles: biogenesis, RNA Secretory mechanisms and intercellular trans-
cargo selection, content, release, and uptake. fer of microRNAs in living cells. J Biol Chem
Cell Mol Neurobiol 36:301–312 285:17442–17452
26. Trajkovic K, Hsu C, Chiantia S et al (2008) 41. Bhome R, Del Vecchio F, Lee GH et al (2018)
Ceramide triggers budding of exosome vesi- Exosomal microRNAs (exomiRs): small
cles into multivesicular endosomes. Science molecules with a big role in cancer. Cancer
319:1244–1247 Lett 420:228–235
27. Brinton LT, Sloane HS, Kester M et al (2015) 42. Squadrito ML, Baer C, Burdet F et al (2014)
Formation and role of exosomes in cancer. Endogenous RNAs modulate microRNA
Cell Mol Life Sci 72:659–671 sorting to exosomes and transfer to acceptor
28. Hessvik NP, Llorente A (2018) Current cells. Cell Rep 8:1432–1446
knowledge on exosome biogenesis and 43. Li P, Kaslan M, Lee SH et al (2017) Progress
release. Cell Mol Life Sci 75:193–208 in exosome isolation techniques. Theranostics
29. Boriachek K, Islam MN, Moller A et al (2018) 7:789–804
Biological functions and current advances in 44. Momen-Heravi F, Getting SJ, Moschos SA
isolation and detection strategies for exosome (2018) Extracellular vesicles and their nucleic
Nanovesicles. Small 14 acids for biomarker discovery. Pharmacol
30. Villarroya-Beltri C, Baixauli F, Gutierrez- Ther 192:170–187. https://doi.org/10.
Vazquez C et al (2014) Sorting it out: regula- 1016/j.pharmthera.2018.08.002
tion of exosome loading. Semin Cancer Biol 45. Li Z, Wang Y, Xiao K et al (2018) Emerging
28:3–13 role of exosomes in the joint diseases. Cell
31. Raposo G, Stoorvogel W (2013) Extracellular Physiol Biochem 47:2008–2017
vesicles: exosomes, microvesicles, and friends. 46. Cheruiyot C, Pataki Z, Ramratnam B et al
J Cell Biol 200:373–383 (2018) Proteomic analysis of exosomes and
32. Beach A, Zhang HG, Ratajczak MZ et al its application in HIV-1 infection. Proteomics
(2014) Exosomes: an overview of biogenesis, Clin Appl 12(5):e1700142. https://doi.org/
composition and role in ovarian cancer. J 10.1002/prca.201700142
Ovarian Res 7:14 47. Batrakova EV, Kim MS (2015) Using exo-
33. O’Loughlin AJ, Woffindale CA, Wood MJ somes, naturally-equipped nanocarriers, for
(2012) Exosomes and the emerging field of drug delivery. J Control Release
exosome-based gene therapy. Curr Gene Ther 219:396–405
12:262–274 48. Xu JY, Chen GH, Yang YJ (2017) Exosomes:
34. Guay C, Regazzi R (2017) Exosomes as new a rising star in falling hearts. Front Physiol
players in metabolic organ cross-talk. Diabe- 8:494
tes Obes Metab 19(Suppl 1):137–146 49. Lawson C, Vicencio JM, Yellon DM et al
35. Thind A, Wilson C (2016) Exosomal miRNAs (2016) Microvesicles and exosomes: new
as cancer biomarkers and therapeutic targets. J players in metabolic and cardiovascular dis-
Extracell Vesicles 5:31292 ease. J Endocrinol 228:R57–R71
36. Zomer A, Vendrig T, Hopmans ES et al 50. van der Pol E, Hoekstra AG, Sturk A et al
(2010) Exosomes: fit to deliver small RNA. (2010) Optical and non-optical methods for
Commun Integr Biol 3:447–450 detection and characterization of microparti-
37. Villarroya-Beltri C, Gutierrez-Vazquez C, cles and exosomes. J Thromb Haemost
Sanchez-Cabo F et al (2013) Sumoylated 8:2596–2607
hnRNPA2B1 controls the sorting of miRNAs 51. Hergenreider E, Heydt S, Treguer K et al
into exosomes through binding to specific (2012) Atheroprotective communication
motifs. Nat Commun 4:2980 between endothelial cells and smooth muscle
38. Koppers-Lalic D, Hackenberg M, Bijnsdorp cells through miRNAs. Nat Cell Biol
IV et al (2014) Nontemplated nucleotide 14:249–256
additions distinguish the small RNA
52. Wang X, Huang W, Liu G et al (2014) Cardi- microRNA-1 in mice. J Mol Cell Cardiol
omyocytes mediate anti-angiogenesis in type 120:12–27
2 diabetic rats through the exosomal transfer 64. Yang M, Chen J, Su F et al (2011) Microve-
of miR-320 into endothelial cells. J Mol Cell sicles secreted by macrophages shuttle
Cardiol 74:139–150 invasion-potentiating microRNAs into breast
53. Yang Y, Li Y, Chen X et al (2016) Exosomal cancer cells. Mol Cancer 10:117
transfer of miR-30a between cardiomyocytes 65. Le MT, Hamar P, Guo C et al (2014)
regulates autophagy after hypoxia. J Mol Med miR-200-containing extracellular vesicles
(Berl) 94:711–724 promote breast cancer cell metastasis. J Clin
54. Sukma Dewi I, Celik S, Karlsson A et al Invest 124:5109–5128
(2017) Exosomal miR-142-3p is increased 66. Zhuang G, Wu X, Jiang Z et al (2012)
during cardiac allograft rejection and aug- Tumour-secreted miR-9 promotes endothe-
ments vascular permeability through down- lial cell migration and angiogenesis by activat-
regulation of endothelial RAB11FIP2 expres- ing the JAK-STAT pathway. EMBO J
sion. Cardiovasc Res 113:440–452 31:3513–3523
55. Tang N, Sun B, Gupta A et al (2016) Mono- 67. Kosaka N, Iguchi H, Hagiwara K et al (2013)
cyte exosomes induce adhesion molecules and Neutral sphingomyelinase 2 (nSMase2)-
cytokines via activation of NF-kappaB in dependent exosomal transfer of angiogenic
endothelial cells. FASEB J 30:3097–3106 microRNAs regulate cancer cell metastasis. J
56. Wang C, Zhang C, Liu L et al (2017) Biol Chem 288:10849–10859
Macrophage-derived mir-155-containing 68. Qin X, Yu S, Zhou L et al (2017) Cisplatin-
exosomes suppress fibroblast proliferation resistant lung cancer cell-derived exosomes
and promote fibroblast inflammation during increase cisplatin resistance of recipient cells
cardiac injury. Mol Ther 25:192–204 in exosomal miR-100-5p-dependent manner.
57. Nguyen MA, Karunakaran D, Geoffrion M Int J Nanomedicine 12:3721–3733
et al (2018) Extracellular vesicles secreted by 69. Maia J, Caja S, Strano Moraes MC et al
Atherogenic macrophages transfer Micro- (2018) Exosome-based cell-cell communica-
RNA to inhibit cell migration. Arterioscler tion in the tumor microenvironment. Front
Thromb Vasc Biol 38:49–63 Cell Dev Biol 6:18
58. Poon KS, Palanisamy K, Chang SS et al 70. Donnarumma E, Fiore D, Nappa M et al
(2017) Plasma exosomal miR-223 expression (2017) Cancer-associated fibroblasts release
regulates inflammatory responses during car- exosomal microRNAs that dictate an aggres-
diac surgery with cardiopulmonary bypass. Sci sive phenotype in breast cancer. Oncotarget
Rep 7:10807 8:19592–19608
59. Bang C, Batkai S, Dangwal S et al (2014) 71. Richards KE, Zeleniak AE, Fishel ML et al
Cardiac fibroblast-derived microRNA passen- (2017) Cancer-associated fibroblast exosomes
ger strand-enriched exosomes mediate cardi- regulate survival and proliferation of pancre-
omyocyte hypertrophy. J Clin Invest atic cancer cells. Oncogene 36:1770–1778
124:2136–2146 72. Ying X, Wu Q, Wu X et al (2016) Epithelial
60. Tian C, Gao L, Zimmerman MC et al (2018) ovarian cancer-secreted exosomal miR-222-
Myocardial infarction-induced microRNA- 3p induces polarization of tumor-associated
enriched exosomes contribute to cardiac macrophages. Oncotarget 7:43076–43087
Nrf2 dysregulation in chronic heart failure. 73. van der Vos KE, Abels ER, Zhang X et al
Am J Physiol Heart Circ Physiol 314: (2016) Directly visualized glioblastoma-
H928–H939 derived extracellular vesicles transfer RNA to
61. Barile L, Lionetti V, Cervio E et al (2014) microglia/macrophages in the brain. Neuro-
Extracellular vesicles from human cardiac pro- Oncology 18:58–69
genitor cells inhibit cardiomyocyte apoptosis 74. Li CC, Eaton SA, Young PE et al (2013)
and improve cardiac function after myocardial Glioma microvesicles carry selectively pack-
infarction. Cardiovasc Res 103:530–541 aged coding and non-coding RNAs which
62. Chen J, Cui C, Yang X et al (2017) MiR-126 alter gene expression in recipient cells. RNA
affects brain-heart interaction after cerebral Biol 10:1333–1344
ischemic stroke. Transl Stroke Res 8:374–385 75. Zhang L, Zhang S, Yao J et al (2015)
63. Sun LL, Duan MJ, Ma JC et al (2018) Myo- Microenvironment-induced PTEN loss by
cardial infarction-induced hippocampal exosomal microRNA primes brain metastasis
microtubule damage by cardiac originating outgrowth. Nature 527:100–104
76. Fong MY, Zhou W, Liu L et al (2015) Breast- 89. Maeda Y, Farina NH, Matzelle MM et al
cancer-secreted miR-122 reprograms glucose (2017) Synovium-derived MicroRNAs regu-
metabolism in premetastatic niche to promote late bone pathways in rheumatoid arthritis. J
metastasis. Nat Cell Biol 17:183–194 Bone Miner Res 32:461–472
77. Gui Y, Liu H, Zhang L et al (2015) Altered 90. Wang Y, Xu D, Yan S et al (2018) MiR-548a-
microRNA profiles in cerebrospinal fluid exo- 3p regulates inflammatory response via
some in Parkinson disease and Alzheimer dis- TLR4/NF-kappaB signaling pathway in
ease. Oncotarget 6:37043–37053 rheumatoid arthritis. J Cell Biochem.
78. Cao XY, Lu JM, Zhao ZQ et al (2017) Micro- https://doi.org/10.1002/jcb.26659
RNA biomarkers of Parkinson’s disease in 91. Okoye IS, Coomes SM, Pelly VS et al (2014)
serum exosome-like microvesicles. Neurosci MicroRNA-containing T-regulatory-cell-
Lett 644:94–99 derived exosomes suppress pathogenic T
79. Cheng L, Doecke JD, Sharples RA et al helper 1 cells. Immunity 41:89–103
(2015) Prognostic serum miRNA biomarkers 92. Mittelbrunn M, Gutierrez-Vazquez C,
associated with Alzheimer’s disease shows Villarroya-Beltri C et al (2011) Unidirectional
concordance with neuropsychological and transfer of microRNA-loaded exosomes from
neuroimaging assessment. Mol Psychiatry T cells to antigen-presenting cells. Nat Com-
20:1188–1196 mun 2:282
80. Lugli G, Cohen AM, Bennett DA et al (2015) 93. Mao K, Chen S, Chen M et al (2013) Nitric
Plasma Exosomal miRNAs in persons with oxide suppresses NLRP3 inflammasome acti-
and without Alzheimer disease: altered vation and protects against LPS-induced sep-
expression and prospects for biomarkers. tic shock. Cell Res 23:201–212
PLoS One 10:e0139233 94. Montecalvo A, Larregina AT, Shufesky WJ
81. Ebrahimkhani S, Vafaee F, Young PE et al et al (2012) Mechanism of transfer of func-
(2017) Exosomal microRNA signatures in tional microRNAs between mouse dendritic
multiple sclerosis reflect disease status. Sci cells via exosomes. Blood 119:756–766
Rep 7:14293 95. Alexander M, Hu R, Runtsch MC et al (2015)
82. Kimura K, Hohjoh H, Fukuoka M et al Exosome-delivered microRNAs modulate the
(2018) Circulating exosomes suppress the inflammatory response to endotoxin. Nat
induction of regulatory T cells via let-7i in Commun 6:7321
multiple sclerosis. Nat Commun 9:17 96. Mann M, Mehta A, Zhao JL et al (2017) An
83. Xu B, Zhang Y, Du XF et al (2017) Neurons NF-kappaB-microRNA regulatory network
secrete miR-132-containing exosomes to reg- tunes macrophage inflammatory responses.
ulate brain vascular integrity. Cell Res Nat Commun 8:851
27:882–897 97. Alexander M, Ramstead AG, Bauer KM et al
84. Pusic AD, Kraig RP (2014) Youth and envi- (2017) Rab27-dependent exosome produc-
ronmental enrichment generate serum exo- tion inhibits chronic inflammation and
somes containing miR-219 that promote enables acute responses to inflammatory sti-
CNS myelination. Glia 62:284–299 muli. J Immunol 199:3559–3570
85. Li JJ, Wang B, Kodali MC et al (2018) In vivo 98. Peng H, Li H, Sheehy A et al (2016)
evidence for the contribution of peripheral Dimethyl fumarate alters microglia phenotype
circulating inflammatory exosomes to neu- and protects neurons against proinflamma-
roinflammation. J Neuroinflammation 15:8 tory toxic microenvironments. J Neuroimmu-
86. Meldolesi J (2019) Extracellular vesicles, nol 299:35–44
news about their role in immune cells: physi- 99. Ju S, Mu J, Dokland T et al (2013) Grape
ology, pathology and diseases. Clin Exp exosome-like nanoparticles induce intestinal
Immunol 196:318–327 stem cells and protect mice from
87. Fox DA, Gizinski A, Morgan R et al (2010) DSS-induced colitis. Mol Ther
Cell-cell interactions in rheumatoid arthritis 21:1345–1357
synovium. Rheum Dis Clin N Am 100. Mu J, Zhuang X, Wang Q et al (2014) Inter-
36:311–323 species communication between plant and
88. Kim SJ, Chen Z, Essani AB et al (2016) Iden- mouse gut host cells through edible plant
tification of a novel toll-like receptor 7 endog- derived exosome-like nanoparticles. Mol
enous ligand in rheumatoid arthritis synovial Nutr Food Res 58:1561–1573
fluid that can provoke arthritic joint inflam- 101. Zhang M, Viennois E, Xu C et al (2016) Plant
mation. Arthritis Rheumatol 68:1099–1110 derived edible nanoparticles as a new
therapeutic approach against diseases. Tissue 103. Bagci C, Allmer J (2016) One step forward,
Barriers 4:e1134415 two steps Back; Xeno-MicroRNAs reported
102. Xiao J, Feng S, Wang X et al (2018) Identifi- in breast Milk are artifacts. PLoS One 11:
cation of exosome-like nanoparticle-derived e0145065
microRNAs from 11 edible fruits and vegeta-
bles. PeerJ 6:e5186
Chapter 15
MicroRNAs and Heat Shock Proteins in Breast Cancer

Biology
Mehmet Taha Yildiz, Lütfi Tutar, Nazlı Irmak Giritlioğlu,
Banu Bayram, and Yusuf Tutar
Abstract
Breast cancer has five major immune types; luminal A, luminal B, HER2, Basal-like, and normal-like. Cells
produce a family of protein called heat shock proteins (Hsps) in response to exposure to thermal and other
proteotoxic stresses play essential roles in cancer metabolism and this large family shows a diverse set of Hsp
involvement in different breast cancer immune types. Recently, Hsp members categorized according to
their immune type roles. Hsp family consists of several subtypes formed by molecular weight; Hsp70,
Hsp90, Hsp100, Hsp40, Hsp60, and small molecule Hsps. Cancer cells employ Hsps as survival factors
since most of these proteins prevent apoptosis. Several studies monitored Hsp roles in breast cancer cells
and reported Hsp27 involvement in drug resistance, Hsp70 in tumor cell transformation-progression, and
interaction with p53. Furthermore, the association of Hsp90 with steroid receptors and signaling proteins
in patients with breast cancer directed research to focus on Hsp-based treatments. miRNAs are known to
play key roles in all types of cancer that are upregulated or downregulated in cancer which respectively
referred to as oncogenes (oncomirs) or tumor suppressors. Expression profiles of miRNAs may be used to
classify, diagnose, and predict different cancer types. It is clear that miRNAs play regulatory roles in gene
expression and this work reveals miRNA correlation to Hsp depending on specific breast cancer immune
types. Deregulation of specific Hsp genes in breast cancer subtypes allows for identification of new targets
for drug design and cancer treatment. Here, we performed miRNA network analysis by recruiting Hsp
genes detected in breast cancer subtypes and reviewed some of the miRNAs related to aforementioned Hsp
genes.
Key words Hsp, miRNA, Breast cancer, miRNA network
1 Breast Cancer
Breast cancer (BC) is the most commonly occurring cancer type in

women and the second most common cancer overall. There were
over 2 million new cases in 2018 [1]. BC is a heterogeneous disease
showing different histopathological and biological characteristics.
Electronic supplementary material: The online version of this chapter (https://doi.org/10.1007/978-1-0716-

1170-8_15) contains supplementary material, which is available to authorized users.
293
294 Mehmet Taha Yildiz et al.
Thus, the management of this cancer type requires different treat-

ment and therapeutic strategies.
Distinct BC subtypes grow at different rates and require differ-
ent treatments and thus display diverse risk factors. The World
Health Organization defined 21 distinct subtypes of BC morpho-
logically [2]. Traditionally immunohistochemistry markers (estro-
gen receptor [ER], progesterone receptor [PR], human epidermal
growth factor 2 [HER2]), and variables such as histological type,
tumor size, tumor grade, and lymph node involvement have been
used to make predictions on prognosis, choose the right treatment
and manage BC [3]. More recently, high throughput proteome-
and genome-profiling methods have gained interest to diagnose
BC, allowing the determination of the expression of multiple genes.
The resulting gene expression levels are used to predict the clinical
profile up to 99% [4]. Gene expression profiling has classified BC
into five biologically distinct intrinsic subtypes; luminal A,
luminal B, HER2-positive, basal-like, and normal-like, all of
which have distinct incidence levels, survival rates and treatment
responses [5]. Molecular characteristics of the cancer subtypes give
more reliable information for cell responses to treatments as com-
pared to anatomical factors, resulting in more accurate and repro-
ducible prognoses [6].
Luminal tumors are the most common subtypes of BC. There
are many subgroups of Luminal type cancer. Basically, depending
on the expression level of proliferation-related genes such as nucle-
ase sensitive element binding protein 1 (NSEP1) and cyclin E1
(CCNE1) they are divided into two subgroups called Luminal A
and Luminal B. The prevalence of luminal A is the highest among
luminal subtypes (50–60%). Generally, they show good prognosis,
responses to hormone therapy and express the genes encoding
proteins of luminal epithelial cells. On the other hand, Luminal-B
tumors have lower prevalence (15–20%), exerting a poor prognosis,
lower survival rates, higher proliferative index and recurrence rate
genes [6–8]. The expression levels of ER, PR, HER2, and Ki67
(a nuclear marker for proliferation) are used to separate luminal A
from luminal B subtype [7]. The Luminal A subtype shows a
positive profile for ER and/or PR and a negative profile for
HER2 with low Ki-67 expression rate (<14%). Furthermore, the
Luminal B subtype is ER (+) and/or PR (+) and HER2 (+). When
it shows a negative profile for HER2, it has high Ki-67 expression
rates (>14%). Both subtypes respond to endocrine therapy. As a
difference, Luminal B is less responsive to endocrine therapy, but it
has a better response to chemotherapy [9].
Basal-like breast cancer is an important subtype among BC
types due to its high frequency and poor prognosis. More often
this type of cancer is observed in young women. Basal-like cancer is
also called triple negative as it lacks the expression of ER, PR, and
HER2. This property makes the therapy of this cancer subtype
miRNAs and Hsps in Breast Cancer 295
difficult due to limited therapeutic options resulting in a lower rate

of survival. For example; anti-estrogen hormone therapies or tras-
tuzumab do not function on basal-like cancer types. On the other
hand, chemotherapy has an effect. About 15% of the BC cases
belong to the basal like subtype. It is commonly characterized by
aggressive and highly recurrent lesions [5, 10].
HER2 positive breast cancer accounts for 15–20% of BC sub-
types characterized by highly proliferative tumor growth and poor
prognosis. HER2 gene encodes a transmembrane tyrosine kinase
receptor that is responsible for cell proliferation and survival. Cyto-
toxic agents such as doxorubicin are effective on this cancer. How-
ever, it is resistant to hormonal agents. HER2 gene and other genes
associated with the HER2 pathway are highly expressed in these
tumor subtypes. On the other hand, it lacks the expression of ER,
and PR [7, 9].
Normal-like cancer refers to approximately 5–10% of all breast
cancers. Studies related to this cancer subtype are scarce and it
needs to be determined clinically. Normal breast-like cancer has
the same profile as fibroadenomas and normal breast samples. They
have similar gene expression profiles like adipose tissue. Neoadju-
vant chemotherapy is not effective on this type. Besides, they lack
the expression of ER, PR, and HER2. Due to that characteristic,
they may be referred to as basal-like cancers. However, they show
negativity for CK5 (a high molecular weight cytokeratin) and
EGFR (namely HER1) genes. Therefore, they differ from basal-
like cancers [7, 8].
1.1 Heat Shock Hsps coordinate and cooperate to perform different functions.
Proteins and Related Biochemical functions of Hsps are nascent substrate protein fold-
miRNAs in Breast ing, solubilization of protein aggregates, protein translocation
Cancer across membranes, preservation and restructuring of the cytoskele-
ton, and targeting damaged proteins to proteasomal degradation.
Hsps along with cochaperones coordinate these cellular processes.
Therefore, several metabolic diseases are associated with Hsps and
they are targets for therapy and immunization. Hsp family main-
tains cellular protein homeostasis under adverse conditions by
maintaining the functional three-dimensional form of their sub-
strate proteins. Since cancer cell metabolism is higher, substrate
proteins (i.e., signaling proteins, transcription factors) may not find
enough time to be correctly folded and this may perturb biochemi-
cal functions and block cascades and drive cells into apoptosis [11].
In a study aiming to identify the profile of the Hsps according
to breast cancer subtypes, Zoppino and coworkers revealed that
25% of the Hsps appeared deregulated and some of them correlated
with subtypes and overall survival [12]. Using the deregulated HSP
gene’s lists from this study we applied a pipeline (Fig. 1) to further
spot shared and specific miRNAs to the BC subtypes related to Hsp
genes. According to this, we used subtype-specific up- and
Fig. 1 Graphical representation of the protocol used to construct breast cancer subtype specific HSP gene-
miRNA-pathway networks
downregulated gene lists to find related miRNAs first. Compres-

sion is made for the obtained miRNAs from the first step to find
shared and specific miRNAs for each subtype of breast cancer.
Obtained miRNA lists were further subjected to pathway analysis
to find possibly affected pathways. According to this analysis sub-
type specific and shared HSP related possible miRNAs are revealed
and given in Fig. 2.
hsa-mir-16-5p (miR-16) is a well-studied microRNA for the
diagnosis and prognosis of cancers, especially in breast cancer. It has
been found that exosomal miR-16 has a higher level in plasma of
breast cancer patients, particularly estrogen-positive patients, than
healthy women [12]. miR-16 has been found to be a potential
biomarker to delineate among subtypes of breast cancer
[12, 13]. Furthermore, it has been reported as a biomarker for
the prediction of overall survival in gastric cancer [14].
In 2018, it has found that hsa-mir-193b-3p (miR-193b) has
relation with progression of gastric cancer. The results of in silico
studies of hsa-mir-193b-3p showed that the expressions of
Non-SMC Condensin I Complex, Subunit G gene (NCAPG) and
KLN1 (also known as CASC5 for Cancer Susceptibility Candidate
5) are regulated by hsa-mir-193b-3p in gastric cancer tissue. While
hsa-mir-193b-3p is downregulated, NACPG and KLN1 are upre-
gulated; and it is thought that the regulation mechanism maybe
contributes gastric cell proliferation. Also, another study showed
that a negative correlation between MORC4 and miR-193b, and
downregulation of miR-193b causes upregulation of MORC4. It
ends up the poor survival in breast cancer tissues [15]. miR-193b
has a tumor suppressor role in T-cells by targeting the MYB onco-
gene [16] and not only play important roles just in cancer cell
regulations, also regulates another mechanism such as chondrogen-
esis, preeclampsia, and Alzheimer’s disease. It also has a potential
role in the regulation of human mesenchymal stem cell chondro-
genesis and metabolism of primary human chondrocytes
[17]. Additionally, this miRNA negatively regulates the matrix
Fig. 2 Venn diagrams of miRNAs potentially related to breast cancer subtypes. (a) Upregulated miRNAs. (b)
Downregulated miRNAs
metalloproteinase 16 gene (MMP16) and promotes chondrocyte

sheet extracellular matrix synthesis [18]. miR-193b has a regulatory
role on the matrix metalloproteinase 19 (MMP19) expression in
the interleukin-1-β-induced human chondrocytes [19]. Also aber-
rantly expressed miR-193b causes preeclampsia [20]. The results of
the transgenic mice study were shown that mir-193b-3p is a non-
invasive biomarker of Alzheimer’s disease and overexpression of
miR-193b may suppress the expression of Amyloid Precursor Pro-
tein (APP) [21].
hsa-mir-124-3p (miR-124) has been extensively studied in
various mechanisms. It has a tumor suppressor role by targeting
CBL which regulates the proliferation and invasion of breast cancer
cells with miR-124 [22]. Furthermore, expression levels of
miR-124 and PDCD6 are negatively correlated in breast tumors.
miR-124 prevents tumor metastasis via inhibiting the expression of
PDCD6 [15]. The expression level of miR-124 was found to be
lower with higher DNA methyltransferase 3B expression level in
bladder cancer cells and tissues compared to normal bladder cells
and tissues [16]. The inhibitory role of miR-124 on the inhibition
of cell growth and metastasis in cervical cancer has been demon-
strated [17]. Kang et al. showed that miR-124 regulates the nuclear
factor of activated T-cells (NFAT) signaling pathway
[18]. miR-124 is also involved in pulmonary hypertension by
decreasing miR-124 expression in fibroblasts isolated from calves
and humans. Migration, proliferation and the inflammatory feature
of hypertensive pulmonary adventitial fibroblasts are caused by
decreased expression of miR-124 [19]. Moreover, the decreased
expression of miR-124 is associated with Alzheimer’s disease [20].
hsa-mir-195-5p (miR-195) is involved in many diseases such
as osteosarcoma [21, 23], melanoma [24], colorectal cancer
[25, 26], endometrial carcinoma [27], cervical carcinoma [28],
renal cell carcinoma [29, 30] oral squamous cell carcinoma [31]
and also gestational diabetes [32], schizophrenia [33], and pre-
eclampsia [34]. Recent studies showed that miR-195 acts as a
tumor suppressor. For example, cancer cells have hTERT activation
by themselves and upregulation of hTERT correlates with the
downregulation of miR-195 in melanoma tissues. Also, it has
been shown that increasing of cancer cell apoptosis, inhibition of
stem cell–like functions decreasing of tumor sphere formation cells,
chemoresistance mechanisms in the colorectal cancer are related
with the expression of miR-195 [25, 26]. Especially, miR-195
targets many genes in the metabolism which are related with glu-
cose intolerance during pregnancy [32].
hsa-mir-15a-5p (miR-15a) has a regulatory role on the pro-
tooncogene MYC [35]. Recently, it was found to be a prognostic
biomarker in non–small cell lung cancer [36] and also negatively
regulates endometrial cancer cell proliferation. It appears to be
inhibiting the Wnt signal pathway [37], cell survival and metastasis
Table 1
Breast cancer immune type specific HSP-miRNA interaction
Computational
miRNA Function Marker Identificationa References
Hsa-mir-16- Tumor suppressor in oral Breast cancer Basal [12–15]
5p squamous cell carcinoma Gastric cancer HER2
(miR-16) Luminal B
Hsa-mir- Tumor suppressor Breast cancer Basal [17–23]
193b-3p Gastric cancer HER2
(miR-193b) Alzheimer’s disease Luminal B
Hsa-mir-124- Tumor suppressor in epithelial- Breast cancer Basal [16–22]
3p mesenchymal transition Bladder cancer HER2
(miR-124) (EMT) and lung Alzheimer’s disease Luminal B
adenocarcinoma
Hsa-mir-195- Tumor suppressor in breast Prostate cancer (PCa) Basal [23–35]
5p cancer HER2
(miR-195)
Hsa-mir-15a- Tumor suppressor in leukemic Non–small-cell lung Basal [36–42]
5p cell proliferation cancer HER2
(miR-15a) Non–small-cell lung
cancer
Hsa-mir-15b- Tumor suppressor in HepG2 cell Fatty liver disease Basal [43–50]
5p proliferation HER2
(miR-15b)
Hsa-mir-183- Oncogene in breast cancer Lung cancer Basal [51, 52]
5p Tongue carcinoma HER2
(miR-183)
Hsa-mir- Protooncogene in prostate Bladder cancer and Basal [53–58]
103a-3p cancer malignant HER2
(miR-103a) mesothelioma Luminal B
Hsa-mir-17- Oncogene in liver, gastric, and Colorectal cancer HER2 [59–63]
5p colorectal cancers. Tumor Breast cancer
(miR-17) suppressor in triple-negative
breast cancer
Hsa-mir-20a- Functions in the pathogenesis of Colorectal cancer HER2 [64–68]
5p multiple myeloma
(miR-20a)
Hsa-mir- Tumor suppressor in several Basal [69–71]
519d-3p tumors HER2
Luminal A
luminal B
Hsa-mir-98- Tumor suppressor in non–small HER2 [72–74]
5p (miR-98 cell Luminal A
luminal B
(continued)
Table 1
(continued)
Computational
miRNA Function Marker Identificationa References
Hsa-mir- Oncogene Alzheimer’s disease Basal, HER2 [75–78]
106b-5p Gastric cancer luminal A
(miR-106b) luminal B
Hsa-mir-30a- Tumor suppressor in many types Basal [79–86]
5p of cancers
(miR-30a)
Hsa-mir-20b- Tumor suppressor in thymoma Basal [87–89]
5p and thymoma-associated HER2
(miR-20b) myasthenia gravis Luminal A
luminal B
Hsa-mir-26a- Plays a regulatory roles in many miR-26a is a well- HER2 [90–94]
5p types of cancers defined prognostic
(miR-26a) marker in several
cancer types
Hsa-mir-34a- Tumor suppressor and plays a HER2 [95–99]
5p regulatory role in acute
(miR-34a) ischemic stroke
Hsa-mir-128- Plays a role in the non–small-cell Lung cancer Basal [100–105]
3p lung cancer drug resistancei
and Alzheimer’s desiase
Hsa-mir-92a- Oncogenic and promotes Luminal B [106–108]
3p cervical cancer cell
(miR-92a) proliferation in colorectal
cancer cells
Hsa-mir-577 Plays a regulatory role in many Basal [109–112]
types of cancers HER2
Luminal A
Luminal B
Hsa-mir-4284 Plays a role in tumor cell growth, Diffuse large B-cell HER2 [113, 114]
(miR-4284) migration and invasion lymphoma Luminal A
Luminal B
See Fig. 2
a
by targeting CXCL10 in chronic myeloid leukemia [38]. It has

been known that Krüppel-like factor 4 (KLF4) has significant
roles such as proliferation and angiogenesis of the cancer cells. A
study in 2013 has shown that KLF4 is being targeted by mir-15a
and consequently miR-15a induces antiproliferative and antiangio-
genic characteristics of KLF4 [39]. Also, mir-15a and mir-16-1
negatively regulate the B cell lymphoma 2 protein at the posttran-
scriptional level and help apoptosis in CLL [40]. Additionally,
miR-15 and miR-16 are downregulated in fibroblasts around the
prostate tumors [41].
hsa-mir-15b-5p (miR-15b) was proposed as a potential bio-

marker in several cancer types, such as hepatocellular carcinoma
[42]. In hepatocellular carcinoma, Rab1A is the target of miR-15b
which prevents the endoplasmic reticulum stress, apoptosis, and
growth by acting as a tumor suppressor [43]. Downregulating of
miR-15b causes the proliferation and Warburg effect in osteosar-
coma [44]. Overexpression of miR-15b and mir-16 which are
normally downregulated in the cancer cells may stimulate the
multi-drug resistance cancer cells [45]. Moreover, overexpression
of miR-15b is the cause of gastric cancer metastasis by regulating
PAQR3 [46]. In colorectal cancer, SIRT1 negatively regulates
miR-15b at the transcription level and suppresses metastasis
[47]. Also, it has been shown that the mir-15 family promotes the
radio-sensitivity of breast cancer cells via targeting the G2/M
checkpoint proteins [48]. In the preeclampsia studies, expression
of mir-15b and Argonaute 2 were found to play significant roles on
placentation [41].
It has been found that overexpression of hsa-mir-183-5p
(miR-183) causes the proliferation and migration of cancer cells
as well as the inhibition of apoptosis by downregulating Ezrin in
endometrial cancer and by targeting PDCD4 in breast cancer
[49, 50].
hsa-mir-103a-3p (miR-103a) is thought to be as a target of
the therapy in glioma [51], a biomarker for bladder cancer [52] and
malignant mesothelioma diagnosis [53]. In addition, miR-103a
plays critical roles in osteogenic differentiation and proliferation
of human adipose tissue-derived stromal cells. It was found that
targets of miR-103a are CDK6 and DICER1 [54]. miR-183a also
showed metabolic roles. In obese mice, silencing of miR-103a
improves glucose homeostasis and insulin sensitivity
[55]. mir-103 also has regulatory roles on cytochrome P450
mixed-function oxidase system genes’ translation such as
CYP2C8 [56].
hsa-mir-17-5p (miR-17) has two different roles in various
cancer types. For example, miR-17 acts like a tumor suppressor in
triple-negative breast cancer [57], while it has an oncogenic feature
in liver, gastric and colorectal cancers [58]. Also, it is a prognostic
biomarker in some cancer such as colorectal cancer [59] and a
predictive factor for breast cancer recurrence [60]. miR-17 is over-
expressed in embryonic stem cells for vital cell functions such as
proliferation, apoptosis and cell cycle regulation. Additionally, it has
been found that the inhibition of miR-17 with an antagomiR has a
positive effect on lung and heart function in a pulmonary hyper-
tension study [61].
Overexpression of hsa-mir-20a-5p (miR-20a) in hepatocellu-
lar carcinoma cells causes cell proliferation and migration by target-
ing the tumor suppressor RUNX3 [62]. Also, it has been found
that overexpression of miR-20a causes tumor migration, invasion,
and growth via targeting RUNX3 in triple-negative breast cancer

cells [63]. By comparison of triple-negative tumors with luminal A,
it was shown that miR-20a is expressed in higher levels in triple
negative-tumors [64]. miR-20a has a regulatory role on the expres-
sion of bone morphogenetic protein receptor type 2 (BMPR2). It
has been shown that inhibition of miR-20a with an antagomiR
improves the functional levels of BMPR2 and avoids the vascular
remodeling in the experimental model of hypoxia-induced pulmo-
nary hypertension [65]. Furthermore, miR-20a is upregulated in
both mouse and human pulmonary artery smooth muscle cells’
(PASMCs) response to hypoxia [66].
hsa-mir-519d-3p acts as a tumor suppressor in many types of
cancers. It inhibits proliferation by targeting TRAF4 in prostate
cancer cells (DU-145) [67], HIF-2α in cervical cancer under hyp-
oxic conditions [68], and B-cell Lymphoma 6 in gastric cancer
cells [69].
Downregulation of hsa-mir-98-5p (miR-98) causes cell pro-
liferation and metastasis. As a result it leads to tumor development
via downregulation of MAP4K4 and inhibition of the downstream
MAPK/ERK signaling in pancreatic ductal adenocarcinoma
[70]. miR-98 also inhibits cell proliferation and induces apoptosis
via targeting IGF2BP1 [71] in hepatocellular carcinoma and
STAT3 in A549 cells [72].
hsa-mir-106b-5p (miR-106b) is an extensively studied
human microRNA. Some target genes of miR-106b have been
found (TP53INP1, APP, PCAF, Stat3, Mapk14, and E2F1).
These involved in various diseases such as Adult T-cell leukemia,
Alzheimer’s disease, malignant mesothelioma, hepatocellular carci-
noma, prostate cancer, gastric cancer, and ovarian cancer
[73]. miR-106b has a negative modulatory role on the angiogene-
sis by inhibiting STAT3 expression [74]. Also, it was predicted that
it has other regulatory roles on vascular endothelial growth factor
(VEGF) and angiogenic factors under hypoxic conditions
[75]. Another role of miR-106b is negative control of p21 expres-
sion. miR-106b is upregulated in gastric cancer [76].
hsa-mir-30a-5p (miR-30a) plays a role as a tumor suppressor
in many types of cancer. miR-30a inhibits proliferation in osteosar-
coma [77], gallbladder cancer [78], oral cancer [79], hepatocellular
cancer [80], urothelial carcinoma [81], and breast cancer
[82, 83]. miR-30a also has a role in the posttranscriptional regula-
tion of SEPT7. A study in 2013 showed that synthesized antisense
oligonucleotides inhibited mir-30a by targeting SEPT7 thereby
suppressing glioma cell growth [84].
It was shown that overexpression of hsa-mir-20b-5p
(miR-20b) has negative control on VEGF expression [75]. Since
Monocyte chemoattractant protein-1 (MCP-1)-induced protein
1 (MCPIP1) has anti-dicer RNase activity, MCPIP1 inhibits the
miR-20b functions. Downregulation of miR-20b gives rise to
angiogenesis [85]. Also, miR-20b has roles in placental angiogene-

sis via posttranscriptional control [86]. Knock-down studies proved
that miR-20b prevents endothelial cell proliferation and contri-
butes to cellular senescence [87].
hsa-mir-26a-5p (miR-26a) is a well-defined prognostic
marker in several cancer types. For example, plasma miR-26a was
found to be increased in bladder cancer patients [88]. It was
evidenced that miR-26a could be a potential biomarker [89], and
involved in hepatocellular carcinoma [90] and bladder cancer
[91]. miR-26a has critical arrangements on the smooth muscle
cell (SMC) regulation by supporting proliferation, migration and
inhibiting apoptosis; altering the Transformation Growth Factor
beta (TGF-β) signaling pathway [92].
hsa-mir-34a-5p (miR-34a) is a significant and clearly upregu-
lated miRNA during DNA damage response in CLL cells in chemo-
immunotherapy [93]. Also, it has a regulatory role in proliferation
of PASMCs by targeting platelet-derived growth factor receptor
alpha. Because of this role, it has been thought that miR-34a could
be used as a potential treatment target in Pulmonary arterial hyper-
tension [94]. Also, it was found that miR-34 is an age-associated
miRNA. Expression of miR-34 is triggered by ageing and prevents
age-relate cell death in cardiomyocyte cells [95]. Overexpression of
miR-34a negatively affects the survival of bone marrow-
mononuclear cells [96] and increases the senescence of endothelial
progenitor cells [97].
hsa-mir-128-3p is a potential blood biomarker for the early
detection of lung cancer [98] and a critical regulator of malignant
lung cancer cells. Hsa-mir-128-3p gives rise to lung cancer cell
migration by regulating the depletion of Drosha [99]. Overex-
pressed hsa-mir-128-3p prevents cell migration of the invasive
breast cancer cells via reducing expressions of hepatocyte growth
factor receptor MET and hepatocyte growth factor [100] and
lymphatic endothelial cell proliferation [101]. hsa-mir-128-3p is
also upregulated in Alzheimer’s disease. miR-128 has a role in the
regulation of co-chaperone BAG2 for Tau degradation
[102]. Moreover, hsa-mir-128-3p and some other miRNAs
(miR-148a, miR-130b, and miR-301b) have regulatory roles on
the expression of important proteins which are involved in
cholesterol-lipoprotein trafficking; for example, low-density lipo-
protein receptor and ATP-binding cassette A1 cholesterol
transporter [103].
hsa-mir-92a-3p (miR-92a) has regulatory roles on the
expression of KLF4 and KLF2 which are protection factors against
atherogenesis in arterial endothelium [104]. Also, it was shown
that miR-92a is overexpressed in colorectal cancer cells (CRC)
which causes decreased expression of PTEN at the translational
level and ectopic expression of miR-92a enhances CRC prolifera-
tion, migration, and invasion. Furthermore, the oncogenic role of
miR-92a has been reported [105]. Overexpression of miR92 reg-

ulates ERbeta1 downregulation which is involved in breast
cancer [106].
hsa-mir-577 has significant roles in many cancer types. In the
cell cycle, miR-577 regulates the transitions between phases. Wang
et al. showed that transfection of mir-577 mimics has a blocking
role in G0/G1 phase in hepatocellular carcinoma [107]. Also, it
was shown that overexpression of hsa-mir-577 and testis specific
10 lead to a more rapid G1/S phase transition in esophageal
squamous cell carcinoma [108]. In a colorectal cancer cell line,
overexpression of hsa-mir-577 is responsible for cell cycle arrest in
the G0/G1 transition [109]. Also, hsa-mir-577 has regulatory
roles in cell proliferation, migration, and invasion [110]. Moreover,
overexpression of hsa-mir-577 inhibits cell proliferation [107, 109,
110].
hsa-mir-4284 (miR-4284) is a diagnostic biomarker for dif-
fuse large B-cell lymphoma. Furthermore, miR-4284 could be the
cause of growth, migration, and invasion in gastric cells by target-
ing ten-eleven translocation-1 [111]. Functional overview of the
[112] most of the miRNAs in the Figure are given in Table 1.
2 Materials and Methods
Computational identification of miRNAs related to breast cancer

subtypes: The up and down regulated Hsp genes detected in breast
cancer subtypes obtained from the study [113] were subjected to
gene-miRNA network analysis via the web-based tools on mir-
net platform [114]. As a result of the analysis, two categories
comprising up, and down regulated miRNAs related to the gene
lists for each subtype of breast cancer were constructed (sup. 1).
Obtained miRNAs were represented as a Venn diagram to underlay
potentially shared and specific miRNAs related to cancer subtypes
(sup. 2, sup. 3). miRNAs were clustered according to their miRNA
families to find the most represented miRNA families on each
cancer subtype according to their representation degrees
( p < 0.005) with hypergeometric Fisher’s test (sup. 4). Identified
miRNAs related to cancer subtypes further subjected to DIANA-
miRPath v3.0 [114] analysis to construct pathway association
(“pathways union”) for mostly dysregulated pathways related to
each subtype. Up and down regulated miRNAs were used together
to perform analysis for each subtype of breast cancer. FDR (False
Discovery Rate) correction option is applied to the resulting signif-
icance levels while modified Fisher’s Exact Test was used to perform
enrichment analysis (sup. 5) (Fig. 1).
3 Conclusion
Hsp genes are critically associated with cancer along with miRNAs.
In this chapter, we computationally analyzed Hsp gene-miRNA
network involved in molecular subtypes of breast cancer
(luminal A, luminal B, HER2, and Basal-like). Furthermore, we
reviewed breast cancer and miRNAs related to Hsps associated with
BC. miRNAs either upregulated or downregulated in different
subtypes. For example, hsa-mir-16-5p was upregulated in all sub-
types except luminal A while hsa-mir-5095 was downregulated in
all subtypes. Hsa-mir-455-3p and hsa-mir-6516-5p upregulated
while 14 miRNAs downregulated in all molecular subtypes
(Fig. 2). However, further analysis should be performed to accept
those miRNAs as potential biomarkers. Furthermore, some KEGG
pathways related to cancer revealed from enrichment analysis such
as Proteoglycans in cancer (hsa05205) and Pathways in cancer
(hsa05200). As two different key molecule groups, Hsps and miR-
NAs have critical roles in cancer molecular biology. These two
groups should be analyzed in the sense of their association in cancer
cells and cell molecular machinery to reveal their potential roles as
biomarkers and therapeutics.
References
1. Bray F, Ferlay J, Soerjomataram I et al (2018) breast carcinoma: a 2015 update. Arch Pathol
Global cancer statistics 2018: GLOBOCAN Lab Med 140:806–814
estimates of incidence and mortality world- 8. Yersal O, Barutca S (2014) Biological sub-
wide for 36 cancers in 185 countries. CA types of breast cancer: prognostic and thera-
Cancer J Clin 68:394–424 peutic implications. World J Clin Oncol
2. Lakhani S, Ellis IO, Schnitt SJ et al (2012) 5:412–424
WHO classification of Tumours of the breast. 9. Krishnamurti U, Silverman JF (2014) HER2
In: En: WHO classification of Tumours. Volu- in breast cancer: a review and update. Adv
men 4. IARC WHO Classification of Anat Pathol 21:100–107
Tumours, vol 4, 4th edn. IARC, Lyon 10. Milioli HH, Tishchenko I, Riveros C et al
cedex, pp 22–23 (2017) Basal-like breast cancer: molecular
3. Prat A, Pineda E, Adamo B et al (2015) Clin- profiles, clinical features and survival out-
ical implications of the intrinsic molecular comes. BMC Med Genet 10:19
subtypes of breast cancer. Breast 24:S26–S35 11. Tutar L, Tutar Y (2010) Heat shock proteins;
4. Pusztai L, Mazouni C, Anderson K et al an overview. Curr Pharm Biotechnol
(2006) Molecular classification of breast can- 11:216–222
cer: limitations and potential. Oncologist 12. Madadi S, Soleimani M (2019) Evaluation of
11:868–877 miR-16 as an internal control in the patients
5. Liu Z, Zhang X-S, Zhang S (2014) Breast with breast cancer. Hum Pathol 85:329
tumor subgroups reveal diverse clinical prog- 13. Ni Q, Stevic I, Pan C et al (2018) Different
nostic power. Sci Rep 4:4002 signatures of miR-16, miR-30b and miR-93
6. Dai X, Li T, Bai Z et al (2015) Breast cancer in exosomes from breast cancer and DCIS
intrinsic subtype classification, clinical use and patients. Sci Rep 8:12974
future trends. Am J Cancer Res 5:2929–2943 14. Ren C, Chen H, Han C et al (2016) High
7. Tang P, Tse GM (2016) Immunohistochemi- expression of miR-16 and miR-451 predicat-
cal surrogates for molecular classification of ing better prognosis in patients with gastric
cancer. J Cancer Res Clin Oncol 27. Kang K, Peng X, Zhang X et al (2013)
142:2489–2496 MicroRNA-124 suppresses the transactiva-
15. Yang Z, Zhuang Q, Hu G et al (2019) tion of nuclear factor of activated T cells by
MORC4 is a novel breast cancer oncogene targeting multiple genes and inhibits the pro-
regulated by miR-193b-3p. J Cell Biochem liferation of pulmonary artery smooth muscle
120:4634–4643 cells. J Biol Chem 288:25414–25427
16. Meng F, Li Z, Zhang Z et al (2018) Micro- 28. Wang D, Zhang H, Li M et al (2014)
RNA-193b-3p regulates chondrogenesis and MicroRNA-124 controls the proliferative,
chondrocyte metabolism by targeting migratory, and inflammatory phenotype of
HDAC3. Theranostics 8:2862–2883 pulmonary vascular fibroblasts. Circ Res
17. Chang ZK, Meng FG, Zhang ZQ et al (2018) 114:67–78
MicroRNA-193b-3p regulates matrix metal- 29. An F, Gong G, Wang Y et al (2017) MiR-124
loproteinase 19 expression in interleukin-1- acts as a target for Alzheimer’s disease by reg-
β-induced human chondrocytes. J Cell ulating BACE1. Oncotarget
Biochem 119:4775–4782 8:114065–114071
18. Roth SA, Hald ØH, Fuchs S, Løkke C et al 30. Yang C, Wu K, Wang S et al (2018) Long
(2018) MicroRNA-193b-3p represses neuro- non-coding RNA XIST promotes osteosar-
blastoma cell growth via downregulation of coma progression by targeting YAP via
Cyclin D1, MCL-1 and MYCN. Oncotarget miR-195-5p. J Cell Biochem 119:5646–5656
9:18160–18179 31. Zhou S, Yu L, Xiong M et al (2018) LncRNA
19. Mets E, Van der Meulen J, Van Peer G et al SNHG12 promotes tumorigenesis and metas-
(2015) MicroRNA-193b-3p acts as a tumor tasis in osteosarcoma by upregulating Notch2
suppressor by targeting the MYB oncogene in by sponging miR-195-5p. Biochem Biophys
T-cell acute lymphoblastic leukemia. Leuke- Res Commun 495:1822–1832
mia 29:798–806 32. Chai L, Kang XJ, Sun ZZ et al (2018)
20. Zhou X, Li Q, Xu J et al (2016) The aber- MiR-497-5p, miR-195-5p and miR-455-3p
rantly expressed miR-193b-3p contributes to function as tumor suppressors by targeting
preeclampsia through regulating transform- hTERT in melanoma A375 cells. Cancer
ing growth factor-β signaling. Sci Rep Manag Res 10:989–1003
6:19910 33. Feng C, Zhang L, Sun Y et al (2018) GDPD5,
21. Liu CG, Song J, Zhang YQ et al (2014) a target of miR-195-5p, is associated with
MicroRNA-193b is a regulator of amyloid metastasis and chemoresistance in colorectal
precursor protein in the blood and cerebro- cancer. Biomed Pharmacother 101:945–952
spinal fluid derived exosomal microRNA- 34. Jin Y, Wang M, Hu H et al (2018) Overcom-
193b is a biomarker of Alzheimer’s disease. ing stemness and chemoresistance in colorec-
Mol Med Rep 10:2395–2400 tal cancer through miR-195-5p-modulated
22. Song B, Du J, Song DF et al (2018) Dysre- inhibition of notch signaling. Int J Biol
gulation of NCAPG, KNL1, miR-148a-3p, Macromol 117:445–453
miR-193b-3p, and miR-1179 may contribute 35. Kong F, Ma J, Yang H et al (2018) Long
to the progression of gastric cancer. Biol Res non-coding RNA PVT1 promotes malig-
51:44 nancy in human endometrial carcinoma cells
23. Wang Y, Chen L, Wu Z et al (2016) miR-124- through negative regulation of miR-195-5p.
3p functions as a tumor suppressor in breast Biochim Biophys Acta, Mol Cell Res. https://
cancer by targeting CBL. BMC Cancer doi.org/10.1016/j.bbamcr.2018.07.008
16:826 36. Li M, Ren CX, Zhang JM et al (2018) The
24. Zhang L, Chen X, Liu B et al (2018) Micro- effects of miR-195-5p/MMP14 on prolifera-
RNA-124-3p directly targets PDCD6 to tion and invasion of cervical carcinoma cells
inhibit metastasis in breast cancer. Oncol through TNF signaling pathway based on
Lett 15:984–990 bioinformatics analysis of microarray
25. Zo RB, Long Z (2018) MiR-124-3p sup- profiling. Cell Physiol Biochem
presses bladder cancer by targeting DNA 50:1398–1413
methyltransferase 3B. J Cell Physiol 37. Chen S, Wang L, Yao X et al (2017) miR-195-
234:464–474 5p is critical in REGγ-mediated regulation of
26. Wang P, Zhang L, Zhang J et al (2018) wnt/β-catenin pathway in renal cell carci-
MicroRNA-124-3p inhibits cell growth and noma. Oncotarget 8:63986–64000
metastasis in cervical cancer by targeting 38. Wang K, Sun Y, Tao W et al (2017) Androgen
IGF2BP1. Exp Ther Med 15:1385–1393 receptor (AR) promotes clear cell renal cell
carcinoma (ccRCC) migration and invasion carcinoma, both in vitro and in vivo, by sup-
via altering the circHIAT1/miR-195-5p/ pressing Rab1A. Oncotarget 6:16227–16238
29a-3p/29c-3p/CDC42 signals. Cancer 51. Weng Y, Shen Y, He Y et al (2018) The
Lett 394:1–12 miR-15b-5p/PDK4 axis regulates osteosar-
39. Wang T, Ren Y, Liu R et al (2017) miR-195- coma proliferation through modulation of
5p suppresses the proliferation, migration, the Warburg effect. Biochem Biophys Res
and invasion of Oral squamous cell carcinoma Commun 503:2749–2757
by targeting TRIM14. Biomed Res Int 52. Xia L, Zhang D, Du R et al (2008) miR-15b
2017:7378148 and miR-16 modulate multidrug resistance by
40. Tagoma A, Alnek K, Kirss A et al (2018) targeting BCL2 in human gastric cancer cells.
MicroRNA profiling of second trimester Int J Cancer 123:372–379
maternal plasma shows upregulation of 53. Zhao C, Li Y, Chen G et al (2017) Overex-
miR-195-5p in patients with gestational dia- pression of miR-15b-5p promotes gastric can-
betes. Gene 672:137–142 cer metastasis by regulating PAQR3. Oncol
41. Musumeci M, Coppola V, Addario A et al Rep 38:352–358
(2011) Control of tumor and microenviron- 54. Sun L-N, Zhi Z, Chen L-Y et al (2017) SIRT1
ment cross-talk by miR-15a and miR-16 in suppresses colorectal cancer metastasis by
prostate cancer. Oncogene 30:4231–4242 transcriptional repression of miR-15b-5p.
42. Sandrim VC, Eleuterio N, Pilan E et al (2016) Cancer Lett 409:104–115
Plasma levels of increased miR-195-5p corre- 55. Mei Z, Su T, Ye J et al (2015) The miR-15
lates with the sFLT-1 levels in preeclampsia. family enhances the radiosensitivity of breast
Hypertens Pregnancy 35:150–158 cancer cells by targeting G2 checkpoints.
43. Kontos CK, Tsiakanikas P, Avgeris M et al Radiat Res 183:196–207
(2017) miR-15a-5p, a novel prognostic bio- 56. Yang M, Chen Y, Chen L et al (2016)
marker, predicting recurrent colorectal ade- miR-15b-AGO2 play a critical role in
nocarcinoma. Mol Diagn Ther 21:453–464 HTR8/SVneo invasion and in a model of
44. Ergun S, Güney S, Temiz E et al (2018) Sig- angiogenesis defects related to inflammation.
nificance of miR-15a-5p and CNKSR3 as Placenta 41:62–73
novel prognostic biomarkers in non-small 57. Cheng Y, Xiang G, Meng Y et al (2016)
cell lung cancer. Anti Cancer Agents Med MiRNA-183-5p promotes cell proliferation
Chem 18:1695–1701 and inhibits apoptosis in human breast cancer
45. Wang ZM, Wan XH, Sang GY et al (2017) by targeting the PDCD4. Reprod Biol
miR-15a-5p suppresses endometrial cancer 16:225–233
cell growth via Wnt/β-catenin signaling path- 58. Yan H, Sun BM, Zhang YY et al (2018) Upre-
way by inhibiting WNT3A. Eur Rev Med gulation of miR-183-5p is responsible for the
Pharmacol Sci 21:4810–4818 promotion of apoptosis and inhibition of the
46. Chen D, Wu D, Shao K et al (2017) MiR-15a- epithelial-mesenchymal transition, prolifera-
5p negatively regulates cell survival and tion, invasion and migration of human endo-
metastasis by targeting CXCL10 in chronic metrial cancer cells by downregulating Ezrin.
myeloid leukemia. Am J Transl Res Int J Mol Med 42:2469–2480
9:4308–4316 59. Yu M, Xue Y, Zheng J et al (2017) Linc00152
47. Zheng X, Li A, Zhao L et al (2013) Key role promotes malignant progression of glioma
of microRNA-15a in the KLF4 suppressions stem cells by regulating miR-103a-3p/
of proliferation and angiogenesis in endothe- FEZF1/CDC25A pathway. Mol Cancer
lial and vascular smooth muscle cells. Biochem 16:110
Biophys Res Commun 437:625–631 60. Zhong Z, Lv M, Chen J (2016) Screening
48. Cimmino A, Calin GA, Fabbri M et al (2005) differential circular RNA expression profiles
miR-15 and miR-16 induce apoptosis by tar- reveals the regulatory role of circTCF25-
geting BCL2. Proc Natl Acad Sci U S A miR-103a-3p/miR-107-CDK6 pathway in
102:13944–13949 bladder carcinoma. Sci Rep 6:30919
49. Chen Y, Chen J, Liu Y et al (2015) Plasma 61. Weber DG, Casjens S, Johnen G et al (2014)
miR-15b-5p, miR-338-5p, and miR-764 as Combination of MiR-103a-3p and mesothe-
biomarkers for hepatocellular carcinoma. lin improves the biomarker performance of
Med Sci Monit 21:1864–1871 malignant mesothelioma diagnosis. PLoS
50. Yang Y, Hou N, Wang X et al (2015) One 9:e114483
miR-15b-5p induces endoplasmic reticulum 62. Kim DS, Lee SY, Lee JH et al (2015) Micro-
stress and apoptosis in human hepatocellular RNA-103a-3p controls proliferation and
osteogenic differentiation of human adipose proliferation of DU-145 prostate cancer cells

tissue-derived stromal cells. Exp Mol Med 47: by reducing TRAF4. Xi Bao Yu Fen Zi Mian
e172 Yi Xue Za Zhi 34:16–21
63. Trajkovski M, Hausser J, Soutschek J et al 76. Jiang L, Shi S, Shi Q et al (2018) MicroRNA-
(2011) MicroRNAs 103 and 107 regulate 519d-3p inhibits proliferation and promotes
insulin sensitivity. Nature 474:649–653 apoptosis by targeting HIF-2α in cervical can-
64. Zhang SY, Surapureddi S, Coulter S et al cer under hypoxic conditions. Oncol Res
(2012) Human CYP2C8 is post- 26:1055–1062
transcriptionally regulated by microRNAs 77. Li Y-Y, Shao J-P, Zhang S-P et al (2018)
103 and 107 in human liver. Mol Pharmacol miR-519d-3p inhibits cell proliferation and
82:529–540 invasion of gastric cancer by downregulating
65. Li J, Lai Y, Ma J et al (2017) miR-17-5p B-cell lymphoma 6. Cytogenet Genome Res
suppresses cell proliferation and invasion by 154:12–19
targeting ETV1 in triple-negative breast can- 78. Fu Y, Liu X, Chen Q et al (2018) Downregu-
cer. BMC Cancer 17:745 lated miR-98-5p promotes PDAC prolifera-
66. Bobbili MR, Mader RM, Grillari J et al (2017) tion and metastasis by reversely regulating
OncomiR-17-5p: alarm signal in cancer? MAP4K4. J Exp Clin Cancer Res 37:130
Oncotarget 8:71206–71222 79. Jiang T, Li M, Li Q et al (2017) MicroRNA-
67. Fu F, Jiang W, Zhou L et al (2018) Circulat- 98-5p inhibits cell proliferation and induces
ing Exosomal miR-17-5p and miR-92a-3p cell apoptosis in hepatocellular carcinoma via
predict pathologic stage and grade of colorec- targeting IGF2BP1. Oncol Res
tal cancer. Transl Oncol 11:221–232 25:1117–1127
68. Wang Y, Li J, Dai L et al (2018) MiR-17-5p 80. Liu H, Wei M, Wang G (2018) miR-98-5p
may serve as a novel predictor for breast can- promotes apoptosis and inhibits migration by
cer recurrence. Cancer Biomark 22:721–726 reducing the level of STAT3 in A549 cells. Xi
69. Pullamsetti SS, Doebele C, Fischer A et al Bao Yu Fen Zi Mian Yi Xue Za Zhi
(2012) Inhibition of microRNA-17 improves 34:522–527
lung and heart function in experimental pul- 81. Wu S, Huang S, Ding J et al (2010) Multiple
monary hypertension. Am J Respir Crit Care microRNAs modulate p21Cip1/Waf1 expres-
Med 185:409–419 sion by directly targeting its 30 untranslated
70. Chen Y, Wang X, Cheng J et al (2016) Micro- region. Oncogene 29:2302–2308
RNA-20a-5p targets RUNX3 to regulate pro- 82. Maimaiti A, Maimaiti A, Yang Y et al (2016)
liferation and migration of human MiR-106b exhibits an anti-angiogenic func-
hepatocellular cancer cells. Oncol Rep tion by inhibiting STAT3 expression in endo-
36:3379–3386 thelial cells. Lipids Health Dis 15:51–51
71. Bai X, Han G, Liu Y et al (2018) MiRNA-20a- 83. Hua Z, Lv Q, Ye W et al (2006) MiRNA-
5p promotes the growth of triple-negative directed regulation of VEGF and other angio-
breast cancer cells through targeting genic factors under hypoxia. PLoS One 1:
RUNX3. Biomed Pharmacother e116
103:1482–1489 84. Kim YK, Yu J, Han TS et al (2009) Functional
72. Calvano Filho CM, Calvano-Mendes DC, links between clustered microRNAs: suppres-
Carvalho KC et al (2014) Triple-negative sion of cell-cycle inhibitors by microRNA
and luminal a breast tumors: differential clusters in gastric cancer. Nucleic Acids Res
expression of miR-18a-5p, miR-17-5p, and 37:1672–1681
miR-20a-5p. Tumour Biol 35:7733–7741 85. Tao J, Cong H, Wang H et al (2018)
73. Brock M, Samillan VJ, Trenkmann M et al MiR-30a-5p inhibits osteosarcoma cell prolif-
(2014) AntagomiR directed against miR-20a eration and migration by targeting FOXD1.
restores functional BMPR2 signalling and Biochem Biophys Res Commun
prevents vascular remodelling in hypoxia- 503:1092–1097
induced pulmonary hypertension. Eur Heart 86. Ye YY, Mei JW, Xiang SS et al (2018) Micro-
J 35:3203–3211 RNA-30a-5p inhibits gallbladder cancer cell
74. Zeng Y, Pan Y, Liu H et al (2014) MiR-20a proliferation, migration and metastasis by tar-
regulates the PRKG1 gene by targeting its geting E2F7. Cell Death Dis 9:410
coding region in pulmonary arterial smooth 87. Ruan P, Tao Z, Tan A (2018) Low expression
muscle cells. FEBS Lett 588:4677–4685 of miR-30a-5p induced the proliferation and
75. Li X, Han X, Yang J et al (2018) Overexpres- invasion of oral cancer via promoting the
sion of miR-519d-3p inhibits the
expression of FAP. Biosci Rep 38: by regulating PLOD2 in bladder cancer. Br J

BSR20171027 Cancer 115:354–363
88. Zhang S, Liu Q, Zhang Q et al (2017) Micro- 100. Leeper NJ, Raiesdana A, Kojima Y et al
RNA-30a-5p suppresses proliferation, inva- (2011) MicroRNA-26a is a novel regulator
sion and tumor growth of hepatocellular of vascular smooth muscle cell function. J
cancer cells via targeting FOXA1. Oncol Lett Cell Physiol 226:1035–1043
14:5018–5026 101. Cerna K, Oppelt J, Chochola V et al (2019)
89. Chung Y-H, Li S-C, Kao Y-H et al (2017) MicroRNA miR-34a downregulates FOXP1
MiR-30a-5p inhibits epithelial-to-mesenchy- during DNA damage response to limit BCR
mal transition and upregulates expression of signalling in chronic lymphocytic leukaemia B
tight junction protein Claudin-5 in human cells. Leukemia 33:403–414
upper tract urothelial carcinoma cells. Int J 102. Wang P, Xu J, Hou Z et al (2016) miRNA-
Mol Sci 18:1826 34a promotes proliferation of human pulmo-
90. Li L, Kang L, Zhao W et al (2017) miR-30a- nary artery smooth muscle cells by targeting
5p suppresses breast tumor growth and PDGFRA. Cell Prolif 49:484–493
metastasis through inhibition of LDHA- 103. Boon RA, Iekushi K, Lechner S et al (2013)
mediated Warburg effect. Cancer Lett MicroRNA-34a regulates cardiac ageing and
400:89–98 function. Nature 495:107–110
91. Li W, Liu C, Zhao C et al (2016) Downregu- 104. Xu Q, Seeger FH, Castillo J et al (2012)
lation of β3 integrin by miR-30a-5p modu- Micro-RNA-34a contributes to the impaired
lates cell adhesion and invasion by function of bone marrow-derived mononu-
interrupting Erk/Ets-1 network in triple- clear cells from patients with cardiovascular
negative breast cancer. Int J Oncol disease. J Am Coll Cardiol 59:2107–2117
48:1155–1164 105. Zhao T, Li J, Chen AF (2010) MicroRNA-
92. Jia Z, Wang K, Wang G et al (2013) MiR-30a- 34a induces endothelial progenitor cell senes-
5p antisense oligonucleotide suppresses gli- cence and impedes its angiogenesis via sup-
oma cell growth by targeting SEPT7. PLoS pressing silent information regulator 1. Am J
One 8:e55008 Physiol Endocrinol Metab 299:E110–E116
93. Roy A, Zhang M, Saad Y et al (2013) Anti- 106. Pan J, Zhou C, Zhao X et al (2018) A
dicer RNAse activity of monocyte chemotac- two-miRNA signature (miR-33a-5p and
tic protein-induced protein-1 is critical for miR-128-3p) in whole blood as potential bio-
inducing angiogenesis. Am J Physiol Cell marker for early diagnosis of lung cancer. Sci
Physiol 305:C1021–C1032 Rep 8:16699
94. Chen DB, Wang W (2013) Human placental 107. Frixa T, Sacconi A, Cioce M et al (2018)
microRNAs and preeclampsia. Biol Reprod MicroRNA-128-3p-mediated depletion of
88:130 Drosha promotes lung cancer cell migration.
95. Wong PF, Jamal J, Tong KL et al (2017) Carcinogenesis 39:293–304
Deregulation of hsa-miR-20b expression in 108. Breunig C, Erdem N, Bott A et al (2018)
TNF-α-induced premature senescence of TGFβ1 regulates HGF-induced cell migra-
human pulmonary microvascular endothelial tion and hepatocyte growth factor receptor
cells. Microvasc Res 114:26–33 MET expression via C-ets-1 and miR-128-
96. Wang H, Hu Z, Chen L (2018) Enhanced 3p in basal-like breast cancer. Mol Oncol
plasma miR-26a-5p promotes the progression 12:1447–1463
of bladder cancer via targeting PTEN. Oncol 109. Zhou J, He Z, Guo L et al (2018) MiR-128-
Lett 16:4223–4228 3p directly targets VEGFC/VEGFR3 to
97. Liang L, Zeng JH, Wang JY et al (2017) modulate the proliferation of lymphatic endo-
Down-regulation of miR-26a-5p in hepato- thelial cells through Ca(2+) signaling. Int J
cellular carcinoma: a qRT-PCR and bioinfor- Biochem Cell Biol 102:51–58
matics study. Pathol Res Pract 110. Carrettiero DC, Hernandez I, Neveu P et al
213:1494–1509 (2009) The cochaperone BAG2 sweeps
98. Chang L, Li K, Guo T (2017) miR-26a-5p paired helical filament- insoluble tau from
suppresses tumor metastasis by regulating the microtubule. J Neurosci 29:2151–2161
EMT and is associated with prognosis in 111. Fang Y, Davies PF (2012) Site-specific micro-
HCC. Clin Transl Onco 19:695–703 RNA-92a regulation of Kruppel-like factors
99. Miyamoto K, Seki N, Matsushita R et al 4 and 2 in atherosusceptible endothelium.
(2016) Tumour-suppressive miRNA-26a-5p Arterioscler Thromb Vasc Biol 32:979–987
and miR-26b-5p inhibit cell aggressiveness
112. Wagschal A, Najafi-Shoushtari SH, Wang L colorectal cancer by targeting PTEN. Dig Dis
et al (2015) Genome-wide identification of Sci 59:98–107
microRNAs regulating cholesterol and tri- 114. Al-Nakhle H, Burns PA, Cummings M et al
glyceride homeostasis. Nat Med (2010) Estrogen receptor {beta}1 expression
21:1290–1297 is regulated by miR-92 in breast cancer. Can-
113. Zhang G, Zhou H, Xiao H et al (2014) cer Res 70:4778–4784
MicroRNA-92a functions as an oncogene in
Chapter 16
Role of MicroRNAs in Extreme Animal Survival Strategies

Hanane Hadj-Moussa, Liam J. Hawkins, and Kenneth B. Storey
Abstract
The critical role microRNAs play in modulating global functions is emerging, both in the maintenance of
homeostatic mechanisms and in the adaptation to diverse environmental stresses. When stressed, cells must
divert metabolic requirements toward immediate survival and eventual recovery and the unique features of
miRNAs, such as their relatively ATP-inexpensive biogenesis costs, and the quick and reversible nature of
their action, renders them excellent “master controllers” for rapid responses. Many animal survival strate-
gies for dealing with extreme environmental pressures involve prolonged retreats into states of suspended
animation to extend the time that they can survive on their limited internal fuel reserves until conditions
improve. The ability to retreat into such hypometabolic states is only possible by coupling the global
suppression of nonessential energy-expensive functions with an activation of prosurvival networks, a process
in which miRNAs are now known to play a major role. In this chapter, we discuss the activation, expression,
biogenesis, and unique attributes of miRNA regulation required to facilitate profound metabolic rate
depression and implement stress-specific metabolic adaptations. We examine the role of miRNA in strate-
gies of biochemical adaptation including mammalian hibernation, freeze tolerance, freeze avoidance, anoxia
and hypoxia survival, estivation, and dehydration tolerance. By comparing these seemingly different
adaptive programs in traditional and exotic animal models, we highlight both unique and conserved
miRNA-meditated mechanisms for survival. Additional topics discussed include transcription factor net-
works, temperature dependent miRNA-targeting, and novel species-specific and stress-specific miRNAs.
Key words miRNA, Secondary structure prediction, Homology-based prediction, Ab initio predic-
tion, miRNA prediction accuracy, Multiple sequence alignment-based prediction
1 Introduction
MicroRNAs are recognized as a principal regulator of posttran-

scriptional control over gene expression, and despite their sequence
brevity, they have been implicated in the regulation of virtually all
biological functions. The ability of miRNAs to perform such a feat,
along with their various unique characteristics, has rendered them
vital modulators of physiological functions, particularly as actors
mediating cellular responses to stress conditions. Many animals
Hanane Hadj-Moussa and Liam J. Hawkins contributed equally to this work.
311
312 Hanane Hadj-Moussa et al.
have evolved exquisitely complex survival strategies that often rely

on profound metabolic reorganization in order to endure unfa-
vourable environmental conditions. Recent research has high-
lighted the importance of miRNAs in facilitating metabolically
taxing adaptive strategies to extreme environmental insults, as
made evident by the emerging prominence of miRNAs in the
field of comparative biochemistry [1]. The unique characteristics
that brand miRNAs as excellent candidate “master controllers” of
global metabolic responses include: (a) rapid action, (b) reversible
outcomes, (c) a relatively ATP-inexpensive cost to biosynthesize,
and (d) an ability to exert broad control over many cellular pro-
cesses [1, 2] (Fig. 1). Indeed, it is well-known that each miRNA
species can bind to and influence the translation of multiple mRNA
transcripts and that each mRNA transcript may be targeted by
multiple miRNA species, resulting in a multifactorial combination
of responses that fine-tune mRNA fate (translation, degradation,
and storage) (Fig. 1). Furthermore, the various influences that
affect the regulation of miRNAs themselves set up a multifaceted
regulatory system that can take into account cell/tissue type, input
from diverse extracellular or environmental stress signals, duration
of stress, and the species-specific nature of the survival strategy to
be activated.
When organisms are unequipped with the adaptations neces-
sary to survive environmental stresses that threaten the source of
one or more metabolic inputs (e.g., biomolecules, oxygen, and
water), they can quickly exhaust internal energy reserves such that
prolonged stress exposure can prove lethal. Indeed, for homeostasis
to be maintained, input molecules must be supplied consistently.
Environments where resource availability is dynamic are plentiful
and many organisms have had to develop survival strategies to
overcome temporal resource limitations. One such strategy is to
establish a new balance between the limited resources available by
transitioning to a new lower rate of resource utilization, a state that
can be achieved by suppressing nonvital metabolic pathways. Regu-
lated metabolic rate depression, leading to entry into a hypometa-
bolic state, can necessitate a fuel switch for ATP production
(aerobic respiration to anaerobic fermentation, carbohydrate catab-
olism to lipid catabolism, etc.). This switch can be achieved by
suppression of energy-expensive processes such as gene transcrip-
tion, protein synthesis, and rates of transmembrane active trans-
port, along with sweeping changes to enzyme/protein
posttranslational modifications, among other metabolic adjust-
ments [4]. To induce such a hypometabolic state requires the
intricate contributions of behavioural, physiological, and molecular
regulatory inputs to promote a retreat into dormancy through the
minimization, and in some cases complete cessation of activities
including feeding, locomotion, heartbeat and breathing rates, and
even neural activity [5]. Animal survival strategies including
Bioinformatics Methods for miRNA Gene Prediction 313
Fig. 1 The regulation of gene expression by miRNA during metabolic rate depression. (Left panel) Only vital
genes are expressed to save energy resources. MiRNA promotes degradation of nonessential mRNA tran-
scripts, thereby reducing the costs of energy expensive protein translation. (Middle panel) MiRNA can also
direct mRNA transcripts into temporary storage for translation in the future when the stress is lifted. (Right
panel) MiRNAs that target prosurvival genes are downregulated or suppressed under stress conditions to
promote translation of mRNAs encoding cytoprotective proteins that are needed for survival under specific
harsh environmental stress conditions. (Adapted with permission from [3])
hibernation, estivation, and freeze tolerance, among others, enact

these strategies that require regulation at all molecular stages of
gene and phenotype expression. Given the pervasive involvement of
miRNAs across biological processes, these small noncoding RNAs
have become important targets to explore in the context of envi-
ronmental stress adaptation and metabolic rate depression (Fig. 2).
Genetic, biochemical, and molecular studies collectively point
toward an evolutionarily conserved “biochemical unity” to the
underlying mechanisms of stress-induced adaptation to extreme
environments. Common themes include strong suppression of
metabolic rate (often to <5% of normal), reliance on internal fuel
reserves (e.g., fat depots, glycogen), upregulation of protectants
(e.g., heat shock and other chaperone proteins, antioxidant
defenses, and low molecular weight stabilizers such as glycerol,
sorbitol, trehalose, urea, and others). This chapter summarizes
and discusses the current state of miRNA regulation as a central
aspect of animal survival strategies and shows that miRNAs are
crucial for facilitating proper metabolic transitions to endure pro-
longed environmental stress. This is accomplished by first examin-
ing the major hypometabolic animal survival strategies that include
hibernation, freeze tolerance, freeze avoidance, hypoxia/anoxia
tolerance, estivation, and dehydration tolerance. For each, we
then highlight the responses and roles of microRNAs in imple-
menting and regulating these survival strategies, along with an
examination of the functional significance of these regulatory net-
works during extreme stress. An analysis of the current status of
stress-responsive regulation of miRNA biogenesis in organisms
employing prolonged hypometabolism as a defense against envi-
ronmental stress is also provided. We also explore the interplay
Fig. 2 Regulation of gene expression by microRNA has been observed in multiple animal species in response
to environmental stress. The miRNA response underlies adaptations to various extreme environments
including freeze tolerance, anoxia and hypoxia tolerance, estivation, and mammalian hibernation indicate
the pervasive role of miRNA in responding to dramatic changes to environmental conditions
between miRNAs and complex transcriptional networks during

stress. Throughout this chapter, new and understudied aspects of
miRNA function are explored with topics such as temperature-
dependent miRNA targeting and the discoveries of novel and
species-specific miRNA.
Before delving into the roles that miRNA play in stress adapta-
tion, below is a brief introduction of miRNA modes of action, for
additional information refer to [6–8]. MicroRNAs are a large group
of highly conserved, short, noncoding RNA transcripts (~22 nt)
that target more than 60% of protein-coding genes in humans, and
influence almost every aspect of biological function [9–11]. Indeed,
depending on the miRNA mode of action for each mRNA target,
miRNAs can effectively mold the composition of the transcriptome
during stress. MicroRNAs function by associating with the
RNA-induced silencing complex (RISC), directing the complex
to the appropriate mRNA transcript(s), binding with partial or
perfect complementarity to mRNA transcripts, and ultimately
resulting in the translational repression or degradation of the
bound mRNA [2]. The 50 seed region of the guide strand, involv-
ing nucleotides two to eight, directs RISC activity toward target
mRNAs with complementary motifs in their 30 UTR [12]. Once
bound, the degree of complementarity between the miRNA and
mRNA transcript dictates the fate of the mRNA transcript,
providing another level of control. Imperfect complementarity
causes mRNA transcripts to be sequestered to cytoplasmic loci
such as P-bodies and stress granules, where they are subject to

translational repression; in contrast, perfect complementarity trig-
gers mRNA degradation by Argonaute endonucleases [2]. A tran-
sient outcome of miRNA targeting is accelerated target mRNA
deadenylation which leads to mRNA destabilization and rapid
mRNA decay [13]. The importance of miRNA regulation can be
observed in the high degree of conservation that miRNA sequences
display between distant vertebrate species. Indeed, this is also
emphasized by the conservation of corresponding target mRNA
30 UTR binding motifs that are 90–100% conserved, despite the
overall low conservation generally observed in 30 UTR regions
[14]. Collectively, the unique attributes highlighted above are
what allow miRNAs to play such key roles in stress adaptations.
2 Mammalian Hibernation and Torpor
Hibernation is an energy-saving survival strategy for many mam-

mals that involves reversible cycling through prolonged bouts of
torpor in order to endure seasonal cold exposure and food scarcity.
Studies have demonstrated that a retreat into cold torpor can save
small hibernators >90% of the energy that would otherwise be
required to get through the winter season using continuous endo-
thermy [15]. This adaptive strategy has been identified in numer-
ous species from eleven different mammalian groups including
monotremes, marsupials, primates, bears, bats, and rodents
[15]. Key characteristics of mammalian hibernation include
(1) depression of metabolic rate, often to <5% of normal resting
rate; (2) a significant drop in body temperature, often to near-
ambient; (3) reversible cycling between torpor and arousal; (4) tor-
por bout durations that can last from 4 days to several weeks; and
(5) the “turning off” of most genes [16]. This complex energy-
saving strategy involves intricate changes in gene expression
through a balance of global gene suppression coupled with the
activation of select prosurvival genes. MiRNAs have been impli-
cated in facilitating successful torpor-arousal cycling and this sec-
tion explores their regulation in well-studied model hibernators
such as ground squirrels and bats and discusses miRNA regulation
in more exotic marsupial and lemur hibernators (Table 1).
2.1 Squirrels The first study to explore how miRNAs responded in extreme
animal survival strategies investigated the expression of nine miR-
NAs in four organs (heart, liver, skeletal muscle, and kidney) of
euthermic versus hibernating 13-lined ground squirrels, Ictidomys
tridecemlineatus [23]. This study set the path for larger and deeper
studies of hibernators, as well as other models of metabolic rate
depression. The small subset of miRNAs analyzed in this study gave
evidence of the global suppression of mRNA translation when
Table 1
The collection of microRNA studies discussed herein examining differential expression levels during
environmentally induced hypometabolic animal survival strategies
Stress Animal Tissue Method References

Hibernation and torpor Microcebus murinus Liver qPCR [17]
Dromiciops gliroides Liver qPCR [18]
Muscle qPCR [18]
Myotis lucifugus Brain RT-PCR [19]
Muscle RT-PCR [20]
Myotis ricketti Brain qPCR, NGS [21]
White adipose qPCR, NGS [21]
Spermophilus paryii Liver NGS [22]
Ictidomys Heart qPCR, RT-PCR [23–27]
tridecemlineatus Liver qPCR, RT-PCR, [23–26, 28,
NGS 29]
Skeletal muscle qPCR, RT-PCR [23–26]
Kidney RT-PCR [23]
White adipose qPCR [30]
Brown adipose qPCR [30]
Brain qPCR, microarray [31]
Freeze tolerance Eurosta solidaginis Whole animal qPCR, NGS, [32–35]
microarray
Littorina littorea Hepatopancreas RT-PCR [36]
Foot muscle RT-PCR [36]
Rana sylvatica Brain qPCR [37]
Heart qPCR [38]
Skeletal muscle qPCR [38]
Liver RT-PCR [39, 40]
Chrysemys picta Heart qPCR [41]
(hatchling) Liver qPCR [41]
Freeze avoidance Epiblema scudderiana Whole animal qPCR, NGS [33, 42]
Leptinotarsa Whole animal NGS [43]
decemlineata
Anoxia and hypoxia Trachemys scripta Liver RT-PCR [40, 44]
tolerance elegans Kidney RT-PCR [44, 45]
Skeletal muscle RT-PCR [40]
White skeletal RT-PCR [45]
muscle
Spleen RT-PCR [45]
Orconectes virilis Hepatopancreas qPCR [46]
Tail muscle qPCR [46]
Littorina littorea Hepatopancreas RT-PCR [36]
Foot muscle RT-PCR [36]
Dosidicus gigas Brain qPCR [47]
Brachial heart qPCR [47]
Mantle muscle qPCR [47]
Estivation Xenopus laevis Liver qPCR [48]
Kidney qPCR [48]
Ventral skin qPCR [48]
Brain qPCR [49]
Apostichopus japonicus Respiratory tree RT-PCR, NGS [50]
Intestine NGS [51]
animals enter torpor. For example, the observed upregulation of

miR-21 in kidney of hibernating squirrels has been linked to the
inhibition of apoptosis, a crucial component of cytoprotection and
long-term viability in multiple animal systems during hypometabo-
lism [52, 53]. Furthermore, the downregulation of miR-24 in
cardiac and skeletal muscle during torpor, suggested that miR-24
may be acting to suppress cell growth in these tissues, another key
requirement of global metabolic rate depression [54]. The conclu-
sion from this study was the discovery that differential miRNA
expression is one of the mechanisms involved in facilitating revers-
ible gene silencing over cycles of torpor/arousal.
Subsequently, the regulation of a subset of six metabolic miR-
NAs (miR-29a, miR-152, miR-195, miR-223, miR-378, and
miR-486) was investigated using qRT-PCR in the liver, heart, and
skeletal muscle of hibernating versus euthermic 13-lined ground
squirrels, where stark tissue-specific expression patterns were
observed [24]. Five of the above miRNAs were upregulated during
torpor in liver, whereas only miR-378 was downregulated in skele-
tal muscle, and no changes were detected in heart miRNA expres-
sions. This study was important in emphasizing the tissue-specific
nature of miRNA regulation required to coordinate the unique
metabolic requirements of each tissue. For example, miR-195
that was upregulated in liver of torpid animals targets fatty acid
synthase (FAS) and levels of this enzyme have been shown to be
suppressed in liver during hibernation [24]. Indeed, the miR-195
mediated suppression of FAS, a key node in the fatty acid synthesis
pathway, is one of the mechanisms in place to regulate fatty acid
homeostasis during torpor. A firm inhibition of fatty acid biosyn-
thesis helps to orchestrate the shift to fatty acid oxidation as the
primary fuel for torpor and halts the use of other fuels (e.g.,
glycogen, glucose) as substrates for energy-expensive fatty acid
synthesis [30, 55].
Wu and colleagues carried out the first large qPCR-based anal-
ysis of the differential expression of miRNAs that target lipid-
specific genes in euthermic versus hibernating 13-lined ground
squirrels focusing on white and brown adipose tissue (WAT and
BAT) [30]. These two adipose tissue types play different yet crucial
roles in hibernation: WAT is the fuel storage tissue that
supplies lipids as the major fuel for hibernation whereas BAT is
responsible for the nonshivering thermogenesis that powers the
arousal of animals from the hypometabolic state. Interestingly,
both WAT and BAT showed reduced expression of six miRNAs,
albeit different ones and while WAT displayed increased levels of
miR-143, miR-200a, and miR-519d, BAT only showed upregu-
lated levels of miR-138. Functional targeting of these different
miRNA expression signatures is associated with adipogenesis and
TGFβ signaling in WAT and mitochondrial β-oxidation and MAPK
signaling in BAT during torpor [30]. These targeted processes
highlight the energetic regulatory role played by miRNAs in WAT

and the signaling required for thermoregulatory action in BAT.
The field of comparative biochemistry has recently begun to
take advantage of larger high-throughput approaches such as
microarrays and next generation sequencing to study miRNA
involvement in extreme animal survival strategies. The first such
study used Illumina sequencing technology to sequence the small
RNA transcriptome of hibernating Arctic ground squirrel livers
(Spermophilus paryii) [22]. The authors identified over 200 con-
served miRNAs and 18 novel squirrel-specific miRNAs. The
miRNA fingerprint characterized in this study was implicated in
various processes that promote cell survival [22]. For example,
downregulated levels of miR-211 can result in inhibition of nones-
sential cell growth during torpor [54]. Additionally, reduced levels
of miR-378 are known to modulate cell cycle arrest through the
lack of inhibition of SuFu and Fus-1, two tumor suppressor pro-
teins that were found to be overexpressed during torpor
[22]. Another study of 13-lined ground squirrels identified addi-
tional novel squirrel-specific and hibernation-regulated miRNAs,
with the expression of 17 novel miRNAs measured in the liver,
heart, and skeletal muscle of torpid versus aroused squirrels
[25]. Targets of these novel miRNAs were enriched in cellular
signaling cascades and ion-transport ATPases, crucial in adaptation
to hibernation.
Since the liver is the body’s metabolic hub, there is much
interest in how it coordinates metabolic rate depression and fuel
reprioritization. To that end, two large-scale studies were carried
out on liver of 13-lined ground squirrels [26, 28]. The expression
of 117 miRNAs was analyzed via qRT-PCR in liver, heart, and
skeletal muscle of squirrels across multiple stages of the torpor-
arousal cycle [26]. Similar to the findings for Arctic ground
squirrels, 13-lined ground squirrels also appear to be utilizing a
miRNA-meditated mechanism for cell growth suppression in liver
during hibernation, as well as in heart and skeletal muscle [22]. A
deeper look at the liver-specific changes showed that miR-181a, a
known marker of hepatic cell insulin resistance, is significantly
upregulated during torpor. This upregulation likely acts to stimu-
late the insulin resistance experienced during the early stages of
torpor and possibly contributes to the storage of excess fat required
to fuel hibernation [26]. MicroRNA data for heart and skeletal
muscle displayed stark tissue and torpor-stage specificity, as exem-
plified by the regulation of miR-208b, a positive regulator of mus-
cle development. In the heart, miR-208b was upregulated during
torpor, whereas the opposite was true in skeletal muscle. The
observed upregulation of miR-208b in the heart may contribute
to adaptive cardiac hypertrophy via the inhibition of the GATA4
transcription factor (TF) and myostatin, a negative regulator of
muscle growth [56]. This target prediction corresponds with
observed reduction in GATA4 total protein levels in hibernating

ground squirrel hearts during late torpor [57]. Adaptive cardiac
hypertrophy is typically experienced during torpor and is required
to pump cold viscous blood through the body. Conversely, skeletal
muscle atrophy is experienced during torpor and the downregula-
tion of miR-208 could be implicated in this response [26].
A third study of 13-lined ground squirrel liver used next-
generation sequencing to characterize the differential regulation
of miRNAs in euthermic versus hibernating animals identifying
28 upregulated and 17 downregulated miRNAs [28]. For example,
miR-145a-3p, miR-22-3p, miR-26b-3p, and miR-25-3p, all play
roles in the regulation of fuel utilization via modulation of the lipid
metabolism and the glucose metabolism. For example, PPARy, a
transcription factor responsible for the upregulation of lipid metab-
olism genes, is overexpressed during torpor and is responsible
(in part) for the observed upregulation of miR-145a-3p that is
under PPARy transcriptional control [58]. The molecular basis
for this is attributed to miR-143a-3p. This miRNA alters hepatic
phosphorylation of Akt and negatively affects insulin signaling
[59]. This potential targeting of insulin signaling via miR-143
could be working to facilitate the previously discussed insulin resis-
tance experienced during torpor [60]. The elevated levels of
miR-15b-3p and miR-25-3p observed in this study have also
been associated with cellular defenses. Mir-15b targets mitochon-
drial sirtuin 4, thereby limiting mitochondrial ROS production,
while miR-25 has been shown to protect against oxidative damage
through the inactivation of the mitochondrial apoptosis by directly
targeting the mitochondrial calcium uniporter [61, 62].
In contrast to large-scale miRNA studies, are studies that mea-
sure a subset of miRNAs targeted to specific pathways and pro-
cesses. One such example is the effect that the anti-proliferatory
TGFβ-Smad transcription factor signaling axis has on miRNA reg-
ulation and biogenesis in hibernating versus euthermic 13-lined
ground squirrels [29]. The TGFβ–Smad–miRNA axis appears to
be activated in the liver of torpid animals, as evidenced by the
elevated levels of phosphorylated Smad3 (active) during torpor,
along with torpor-increased nuclear localization and DNA binding
activity [29]. Once activated and in the nucleus, Smad proteins
serve to mediate the transcription and processing of select miRNAs
[63]. An examination of miRNAs that are known to be posttran-
scriptionally processed by Smad proteins, revealed a significant
increase of miR-21 during torpor stages in ground squirrels. This
increase was linked to the previously reported downregulation of
cell growth, protein synthesis, and transcriptional regulation that
occurs during hibernation [64].
Another study examined the relationship between miRNAs and
Nrf2, a transcription factor that coordinates antioxidant defenses,
in the heart of hibernating squirrels [27]. Levels of oxidative
damage-responsive miRNAs (miR-93, miR-141, miR-144, and

miR-200a) were measured. Reduced levels of miR-200a, a regula-
tor of the Nrf2 antioxidant pathway, correlated with elevated Nrf2
mRNA transcript levels and suggested that changes in Nrf2 expres-
sion are, at least in part, mediated by miR-200a [27]. Indeed, the
upregulation of Nrf2 in hibernation has been observed in other
tissues of torpid animals [65]. Interestingly, comparable studies on
other models of hypometabolism, namely freeze-tolerant painted
turtle hatchlings (Chrysemys picta marginata), also reported eleva-
tion of Nrf2 during freezing. This conserved Nrf2 response sug-
gests a broad and crucial role to protect cells and mitigate oxidative
damage associated with prolonged hypometabolism [65].
While the previous two studies looked at the interplay of miR-
NAs with specific transcription factors, a study by Lee and collea-
gues [31] was the first to report a mechanism by which miRNAs
interact with small proteins, modifying networks during torpor.
Ubiquitination and SUMOylation are posttranslational modifica-
tions that regulate protein trafficking, stability, degradation, and
various other biological processes that are pivotal for cell survival
[67]. Previous studies have shown that global SUMOylation occurs
in hibernating ground squirrels, where it acts to protect against
ischemic damage for oxygen and glucose deprived cells [68]. The
study by Lee and colleagues characterized how miRNAs control
ubiquitin-like modifiers to facilitate the natural tolerance to brain
ischemia during hibernation [31]. Hibernating brain miRNA ana-
lyses identified the significantly downregulated miR-200 and
miR-182 families as regulators of the gene expression of these
ubiquitin-like modifiers. This was coupled with the measured
increases in protein levels of ubiquitin-like modifiers, making cells
more resistant to oxygen and glucose deprived induced cell
death [31].
2.2 Bats Bats are a group of well-characterized small mammalian hibernators

that exhibit long bouts of torpor (as much as 7 weeks), interspersed
with brief arousals into euthermia. Similar to squirrels, bats rely on
internal fat reserves for fuel, coupled with strong reductions of
metabolic rate, breathing, heart rate, and prolonged periods of
immobility to survive through the winter [16]. Studies of bat
hibernation also show the involvement of miRNAs in coordinating
the complex metabolic reprogramming needed during torpor. Skel-
etal muscle from the little brown bat, Myotis lucifugus, showed
significant upregulation of eight muscle-specific “myomiRs” in
torpid animals [20, 69]. These included miR-1a-1, miR-29b,
miR-181b, miR-15a, miR-20a, miR-206, and miR-128-1 that
have been shown to target the translation of muscle-specific tran-
scription factors including MEF2, FOXO, SMAD, and myostatin.
These TFs collectively contribute to the preservation of muscle
mass and protect against disuse atrophy through the prolonged
periods of immobility during torpor [20].
Next, the role that miRNAs play in regulating bat neuronal

functions during hibernation was explored in whole bat brains
[19]. Of the eleven miRNAs studied, eight (miR-21, -29b, -103,
-107, -124a, -132, -183, and -501) displayed elevated expression
during torpor, suggesting the suppression of their mRNA tran-
script targets. These miRNAs are known to be involved in modu-
lating axon guidance and neuronal differentiation, by targeting
genes involved in cytoskeletal reorganization, thereby contributing
to the adaptive neuroprotective mechanisms required for successful
torpor-arousal cycling [19]. To elucidate further the role that the
brain plays in integrating peripheral physiology during extreme
survival strategies, a more in-depth analysis of bat brain miRNAs
was conducted in torpid Rickett’s big-footed bat (Myotis ricketti).
High-throughput sequencing lead to the identification and quanti-
fication of 196 miRNAs (including 77 bat-specific novel miRNAs),
where 33 of the 196 miRNAs were differentially regulated during
torpor [21]. KEGG and GO functional predictions suggested the
involvement of these miRNAs in neurospecific functions including
signaling pathways and neuroactive receptors, as well as protective
functions such as chaperone protein regulation and enhanced DNA
damage repair proteins [21]. In particular, the upregulation of
miR-153 during torpor has been shown to serve a neuronal protec-
tive function through the activation of mTOR and the inhibition of
apoptosis [70]. The authors also performed a parallel investigation
of miRNA regulation in white adipose tissue (WAT) of control and
torpid bats. The 25 differentially expressed miRNAs in WAT
appeared to facilitate a reliance on lipid catabolism via the activation
of oxidative phosphorylation, fatty acid catabolism, and other ener-
getic pathways [21].
2.3 Marsupials A unique and relict hibernator that is only just beginning to be
explored in the context of metabolic rate depression is the South
American marsupial monito del monte (Dromiciops gliroides).
When faced with cold exposure and food scarcity, this cold-adapted
marsupial can switch between long, deep bouts of multiday torpor
with metabolic rate dropping to 1–5% of the euthermic rate, and
shallow daily torpor where metabolism is suppressed to 10–60% of
basal rate [71]. Currently, there is only one study that has explored
the role of miRNAs in facilitating marsupial hibernation. Relative
expression levels of 85 miRNAs were analyzed in liver and skeletal
muscle of control and torpid animals [18]. Stark tissue-specific
differences were observed between liver and muscle profiles,
where 39 miRNAs were significantly downregulated in liver,
whereas only 11 miRNAs were differentially expressed in skeletal
muscle of torpid animals. In liver, the affected miRNAs were impli-
cated in the potential activation of MAPK, PI3K-Akt, and mTOR
signaling pathways, suggesting that they may act in coordinating a
compensatory hepatic thermoregulatory mechanism to facilitate
arousal from hibernation. Marsupials lack brown adipose tissue

and, instead, a liver thermogenesis mechanism seems to be the
replacement [72]. By contrast, in muscle, the changing miRNAs
were linked with the promotion of muscle maintenance mechan-
isms through the regulation of focal adhesion, ErbB, and mTOR
pathways [18]. A follow-up in-depth bioinformatics analysis of
these hibernation-responsive miRNAs in liver found that miR-22-
5p, miR-876-5p, miR-21a-3p, miR-18a-3p, miR-16-3p, and
miR-142b-5p represent the smallest set of miRNAs that can differ-
entiate between arousal and torpor metabolic states [73]. This
subset of hibernation miRNA biomarkers was shown to directly
target key elements in the MAPK signaling pathway as well as to
coordinate a compensatory hepatic thermoregulatory mechanism
to facilitate arousal of hibernating marsupials [73]. The role that
these hibernation miRNA biomarkers serve in other mammalian
hibernators has yet to be explored and will prove interesting.
2.4 Lemurs Recent studies have analyzed the biochemical and molecular adap-
tations that support daily torpor in a primate, the gray mouse lemur
(Microcebus murinus) from Madagascar. These animals allow their
body temperature to cool to near ambient during the resting phase
of their day and gain substantial energy savings by turning off
thermogenesis while they sleep. The facility for metabolic rate
depression in this animal (and other species that use daily torpor)
supports the existence of a “biochemical unity” in the basic princi-
ples of torpor/hibernation control [74]. Indeed, a full appreciation
of the molecular underpinnings of hibernation in this primate
model may ultimately reveal a blueprint for torpor that may still
be hidden in the human genome [75]. A recent study of the gray
mouse lemur analyzed 122 conserved miRNAs and predicted
44 novel species-specific miRNAs (from the gray mouse lemur
genome) [17]. Of those, 26 miRNAs were significantly upregu-
lated in liver with in silico predictions suggesting the involvement
of these miRNAs in promoting cell survival pathways during tor-
por. Conversely, 31 miRNAs were significantly downregulated and
were linked primarily with regulating various immune functions
[17]. Enlisting immune functions such as complement activation
and B-cell mediated immunity suggests a miRNA-centered mecha-
nism to prevent impaired immune responses during periods of
torpor. Understanding immune function regulation during torpor
is crucial since impaired immunity increases the risk of infection, as
demonstrated by the rapid spread of white nose syndrome that has
decimated populations of hibernating bats [76].
3 Freeze Tolerance and Freeze Avoidance
When temperatures drop below 0 C, most animals employ

responses to avoid freezing of their tissues; common strategies
include migrating to a warmer climate or retreating underground
or under water or to other thermally buffered hibernacula. How-
ever, some species have evolved incredible adaptations to face the
winter “head-on” and endure deep subzero temperatures. Two
main survival strategies have been described: [1] freeze avoidance,
where animals supercool and body fluids remain liquid, often to
40 C or lower; and freeze tolerance, where body fluids undergo
controlled freezing in extracellular compartments while the cyto-
plasm is packed with cryoprotectants to keep it from freezing. Both
freeze avoidance and freeze tolerance are common among small
invertebrate species, particularly arthropods [77] and molluscs
[78], whereas notable examples among vertebrate poikilotherms
include some reptiles [79], various frogs, and salamanders [80] as
well as some cold-water marine fishes [81].
Most cold-hardy species employ a freeze avoidance strategy.
This strategy relies on the production of low molecular weight
osmolytes and protein/peptide antifreeze molecules to support
equilibrium freezing point depression and stabilize the supercooled
state. In terrestrial arthropods, the integument also acts as a barrier
between environmental ice and body fluids to block ice nucleation.
By contrast, freeze-tolerant animals allow ice crystallization to
occur in extracellular body fluids but do so in a slow and controlled
manner. Incredibly, species such as the wood frog survive the
freezing of 60–70% of total body water [82]. Multiple mechanisms
are common among freeze-tolerant species including: [1] produc-
tion and distribution of high levels of cryoprotectants (glucose,
glycerol, or other polyols); [2] initiation of ice nucleation at high
subzero temperatures to slow the rate of ice propagation; [4] water
loss from cells into extracellular spaces leading to cell shrinkage and
increased intracellular osmolality and ice crystal formation in extra-
cellular spaces; and, in some cases, [5] production of ice-binding
proteins that limit ice-recrystallization, thereby minimizing ice
crystal size [5].
All freeze-tolerant and some freeze-avoiding species rely on
their ability to depress metabolic rate to facilitate long-term sur-
vival. This is of particular importance in freeze tolerant species,
since locomotion and all vital signs (heartbeat, breathing, measur-
able brain activity) are halted during freezing. Energy resources at
this point are highly limited to glycolytic ATP production alone and
thus a new balance between energy resource availability and use
must be achieved to survive until the animal thaws. As previously
discussed, metabolic rate depression involves strict control over
gene expression to silence nonessential cellular and metabolic
processes while activating prosurvival mechanisms. Regulation by

microRNAs is particularly suited in this context given its power,
flexibility, and efficiency to modulate gene expression (Fig. 1;
Table 1). In this section, we describe how miRNAs regulate survival
in cold environments, in both invertebrates and vertebrates. Addi-
tionally, we explore how miRNAs themselves function differently
over wide temperature ranges.
3.1 Cold-Hardy While overwintering strategies have been studied in a myriad of

Invertebrates species, two of the best-characterized species are insect larvae of the
freeze-tolerant goldenrod gall fly (Eurosta solidaginis) and freeze
avoiding gall moth (Epiblema scudderiana). The pairing of these
species is particularly interesting since they occupy a similar geo-
graphic range and overwinter inside stem galls on the same species
of goldenrod, sometimes even on the same stem. These similarities
allow for close comparisons of the physiological, metabolic, and
molecular mechanisms underlying freeze tolerance versus freeze
avoidance.
Initial studies of the miRNA responses to subzero temperatures
by these species focused on specifically selected miRNAs with tar-
gets that were known to contribute to cold-hardiness [32–
34]. E. solidaginis and E. scudderiana were sampled at 5 C (con-
trol condition), 5 C, and 15 C and two miRNAs, miR-1 and
miR-34 were assessed [33]. MiR-1 levels in E. solidaginis increased
significantly at 15 C, which was the only significantly different
miRNA as compared to control conditions in either species. This
was also the only condition in which the larvae were frozen, because
E. scudderiana supercools to near 40 C whereas E. solidaginis
freezes at about 10 C. This miRNA targets metabolic enzymes
such as glucose-6-phosphate dehydrogenase [83] that is involved in
the synthesis of cryoprotectant polyols (glycerol, sorbitol), contri-
butes to glutathione production for antioxidant defense [84], and
targets regulators of the lipid metabolism [85] that would be
necessary to maintain membrane fluidity [86].
Additional targeted studies on the freeze-tolerant E. solidaginis
identified other differentially expressed miRNAs when exposed to
15 C, including miR-11, miR-34, and miR-92b [33, 34]. While
their specific functions have not been elucidated in the context of
freeze tolerance, these miRNAs have been associated with apoptosis
in other systems [87–89], a process that must be under strict
control in species that use cold-hardiness and metabolic rate
depression for winter survival. Recently, miRNAomes of both spe-
cies were sequenced and compared cold-acclimated (5 C) versus
subzero-exposed (15 C) larvae [35, 42]. Similar to previous
results for freeze-tolerant E. solidaginis, small RNA-seq of freeze
avoiding E. scudderiana showed miR-2b as the most upregulated
miRNA in 15 C exposed larvae.
Results from small RNA-seq of freeze-tolerant E. solidaginis

confirmed previous findings that miR-1 is highly upregulated dur-
ing freezing and additionally that levels of miR-14-3p and
miR-31a-3p are elevated [35]. Like miR-1, miR-14 is involved in
lipid metabolism [90] and may contribute to changes in fatty acid
composition as larvae acclimate to lower temperatures as the
autumn-winter seasons progress [91]. On the other hand,
miR-31 is a positive regulator of HIF-1α by negatively regulating
the HIF regulatory factor [92]. E. solidaginis HIF-1α expression
increases not only in response to anoxia exposure, but also when
larvae freeze [93]. This is because anoxia is an unavoidable conse-
quence of whole body freezing since its delivery to tissues is cut off
and, therefore, miRNA such as miR-31 may be involved in a HIF1α
response to freezing.
Research has also explored the idea of exploiting miRNAs to
modulate insect cold-hardiness as a means of controlling agricul-
tural pest species. The first such study examined the Colorado
potato beetle (Leptinotarsa decemlineata) that has spread across
the world and is the most pervasive potato defoliator [94]. Tradi-
tional control strategies, such as the use of insecticides, have begun
to fail as insect resistance increases. Therefore, alternative strategies
to control these pests are in demand. Multiple groups have
explored the use of RNA interference technologies as a selective
insecticide [95–97] against this beetle. The Colorado potato beetle
employs a freeze avoidance strategy of supercooling over the winter
and it was proposed that characterization of the miRNAs involved
in this process could reveal potential targets for new pest control
strategies. Small RNA-seq analysis of beetles exposed to 15 C
(control) versus 5 C revealed several differentially expressed
miRNAs including miR-9a-3p and miR-210-3p, that are both
linked to HIF activity [98, 99], as well as miR-277-3p and
miR-276-5p that target mitochondrial uncoupling proteins
[43]. This area of research is in its infancy, but the potential exists
that disrupting the beetle’s supercooling ability could lead to
reduced population sizes on a year-to-year basis.
The role of miRNAs in invertebrate cold hardiness has also
been assessed in a freeze-tolerant marine mollusc. The common
periwinkle (Littorina littorea) inhabits intertidal zones where it can
be subjected to whole body freezing in the winter at low tide when
air temperatures are below the equilibrium freezing point
(1.9 C) of the snail’s tissues. Select miRNAs were analyzed in
foot muscle and hepatopancreas of 6 C freezing exposed L. lit-
torea, compared with 10 C controls [36]. In foot muscle, six of
seven miRNAs assessed showed increased expression (miR-1a-1,
miR-34a, miR-133a, miR-125b, miR-29b, and miR-2a) in
response to freezing, whereas in hepatopancreas only three
increased (miR-1a-1, miR-34a, and miR-29b). Since this species
is known to significantly decrease total protein synthesis while
frozen, the authors suggested that the observed increase in

miR-29b, that targets the PI3K/AKT pathway [100], may be an
important contributor to the freeze-tolerant phenotype.
3.2 Cold-Hardy Relative to invertebrates, few vertebrate species have developed

Vertebrates strategies to survive sub-zero body temperatures. Mammals and
birds have the advantages of endothermy, insulation (fur, feathers),
sheltered hibernacula, or migration to deal with seasonal cold.
Most ectothermic vertebrates (fish, amphibians, and reptiles) also
elude the cold by wintering underground below the frost line or
underwater. However, as mentioned earlier, a few amphibian and
reptile species have mastered sub-zero survival while wintering on
land using either freeze tolerance or freeze avoidance strategies
[5]. The best-studied cold-hardy vertebrate is the North American
wood frog (Rana sylvatica), both in the context of freeze tolerance
and the involvement of miRNAs. When exposed to freezing tem-
peratures, frogs immediately activate glycogenolysis in the liver to
produce and distribute large quantities of glucose to all tissues for
use as a cryoprotectant. Emanating from ice nucleation on the skin,
the freezing front moves inward through the body, forming ice in
extracellular spaces and drawing water out of cells so that these
shrink until resistance from high levels of intracellular glucose halts
further volume loss [101]. Metabolic rate depression also lowers
energy demands by tissues so that animals can live in an anoxic
frozen state for days, weeks, or months until thawing occurs [5].
The involvement of miRNAs in the metabolic adjustments
needed for vertebrate freeze tolerance has recently received much
attention. Studies have shown that miRNA expression in wood
frogs changes over the course of a freeze–thaw cycle in a tissue-
dependent manner [39, 40]. In the liver miR-16, -21, -26a, -126,
and -217 all showed increased expression in response to freezing
and it was postulated that the potential functions of these miRNA
during freezing was in cell cycle suppression (one component of
energy-saving global metabolic depression) and control over apo-
ptosis to limit cell death under the restrictive ischemic conditions of
the frozen state. Both of these processes are also known to be under
strict control in other species that undergo metabolic rate depres-
sion [102, 103].
A larger follow-up evaluation of 53 miRNAs in cardiac and
skeletal muscles of wood frogs over the freeze–thaw cycle further
identified 21 cardiac miRNAs that were downregulated during
thawing and 16 skeletal miRNAs that were overexpressed during
freezing. Target prediction for heart miRNAs that decreased during
freezing indicated involvement in processes such as hypertrophy,
cell structure, and cardiomyopathy, all of which could serve cardi-
oprotective roles during freezing and thawing [38]. Freeze-
induced structural strengthening of heart cells could be important
to sustain the heart’s role in cryoprotectant distribution
throughout the body for as long as possible even as blood viscosity

and peripheral resistance to blood flow increases hugely, in parallel
to ice accumulation.
The overall pattern of freeze-responsive miRNA expression in
skeletal muscle showed that the majority of differentially expressed
miRNAs were upregulated which implies that the mRNA tran-
scripts that they regulate were targeted for storage or destruction
in support of a suppression of global mRNA translation in the
frozen state [38]. These miRNAs mapped to functions that could
help the cell conserve energy and transition into a metabolically
suppressed state while the animal is frozen. Taken together, the
miRNA responses by wood frog cardiac and skeletal muscles were
distinctly tissue-specific and promoted activation and suppression
phenotypes, respectively. These newly identified freeze-responsive
wood frog miRNAs join the ranks of previously categorized cryo-
miRs, a subset of cold-responsive miRNAs found to facilitate cold
hardiness, hibernation, freeze avoidance, and freeze tolerance
[104]. Indeed, studies of various natural models of environmentally
induced hypometabolism highlight the importance of miRNAs in
regulating major metabolic cascades [105].
The subsequent quantification of 113 miRNAs in the brain of
the wood frog found 41 miRNAs that were differentially expressed
over the course of the freeze–thaw cycle, with 39 of those miRNAs
being significantly downregulated during freezing [37]. This
prominent downregulation suggests a miRNA-mediated mecha-
nism to fine-tune neuronal functions and enhance translation of
selected proteins that provide cyto- or cryo-protective actions and
facilitate whole brain freezing survival. These findings illustrate a
novel miRNA-mediated neuroprotective mechanism that could
allow the brain to circumvent the extensive translational silencing
being undertaken across the rest of the body. The prosurvival
processes potentially being “activated” by downregulation of miR-
NAs fell into categories including synaptic signaling, intracellular
signal transduction, anoxia/ischemia protective mechanisms, and
chaperone-mediated protein folding [37]. Many of the miRNAs
differentially expressed during freezing displayed similar expression
profiles during thawing, suggesting a potential continuation of
their protective roles in this recovery state [37].
The very different miRNA responses observed in wood frog
liver, heart, skeletal muscle, and brain illustrate the two main types
of miRNA regulatory responses: [1] reduction in miRNA levels to
facilitate enhancement of selected target proteins or pathways, and
[2] increased miRNA levels to suppress and/or inhibit miRNA
targeted proteins or pathways [2]. The contrasting miRNA finger-
prints measured highlight the high degree of tissue-specificity and
unique metabolic needs that miRNA regulation of mRNA transla-
tion can address. To fully assess the biological relevance of these
findings, and the validity of these bioinformatic predictions,
downstream qPCR analyses of mRNA transcripts targeted by the

identified miRNAs, as well as immunoblot analysis of the proteins
will need to be performed. Collectively, these findings support the
overall metabolic nature of the liver, heart, muscle, and brain, and
the roles that they play during freeze–thaw.
Only one study to date has evaluated miRNA responses in
another cold-hardy vertebrate. This is the painted turtle (Chrysemys
picta) that exhibits whole body freeze tolerance in the hatchling
stage of life. These young turtles hatch from their eggs in late
summer but remain in their shallow nests (~10 cm deep) until the
following spring, during which time they are exposed to subzero
winter temperatures. An analysis of heart and skeletal muscle
miRNA responses to 20 h of whole body freezing at 3 C revealed
conserved upregulation of miR-16 in both turtle liver and heart
[41], as also occurred in wood frogs, as well as elevated miRNA-21
in the heart. Two other miRNAs (miR-34a and miR133)
responded oppositely, being strongly suppressed.
Importantly, the study of hatchling turtle miRNA expression
was also the first to tackle the question of temperature effects on
miRNA binding to target mRNA [41]. The action of miRNA in
silencing mRNA or targeting it for storage or degradation depends
on miRNA binding at two regions of the mRNA transcript: (a) a
seed region requiring near-perfect pairing by nucleotides 2–7 on
the 50 end of miRNA to the mRNA and (b) complementary bind-
ing of the 30 end of the miRNA to the gene transcript. Perfect
binding sets up the mRNA for degradation by the argonaute endo-
nuclease whereas imperfect binding directs mRNA into storage
(e.g., in P-bodies or storage granules). Typically, thermodynamic
modeling of miRNA target prediction is performed assuming a
temperature of 37 C (i.e., a mammalian model). However, this
temperature is nonphysiological for invertebrates and lower verte-
brates. RNA-RNA duplexing is highly dependent on temperature,
meaning that target prediction modeling is more biologically rele-
vant if done at a temperature reflecting body temperature of the
animal. Modeling target prediction at 1 C, instead of 37 C,
hugely increases the number of miRNA-mRNA interactions possi-
ble. For example, cpm-miR-16 was predicted to target 2153 sites
within 820 genes at 1 C, compared to 263 sites in 173 genes
predicted at 37 C [41]. This suggests that temperature-dependent
miRNA-mRNA interactions could be a broad and potent mecha-
nism for metabolic rate depression and/or metabolic remodeling at
low temperatures. Temperature-dependent processes identified
included apoptosis signaling, the p53 pathway, MAPK signaling,
and the cell cycle, all of which have been shown to be important in
low temperature winter survival of hatchling painted turtles and in
other freeze-tolerant species [5]. These findings demonstrate the
importance of accounting for biologically relevant temperatures
when performing bioinformatic target predictions.
4 Anoxia and Hypoxia Tolerance
The partial or total restriction of oxygen presents metabolic con-

sequences that must be immediately dealt with in higher organisms.
Anaerobic ATP generating pathways are significantly less efficient
than their aerobic counterparts and produce end products with
negative consequences such as cellular acidification. Multiple verte-
brate and invertebrate species have evolved metabolic adaptations
that allow them to survive in their environments when oxygen is
temporarily limiting, and many do so by relying on strong meta-
bolic rate depression to reduce ATP demand to a level that can be
met by anaerobic ATP-generating pathways alone [106]. This sec-
tion dissects the involvement of miRNA in supporting anoxia or
hypoxia tolerance strategies (Table 1).
4.1 Freshwater Among vertebrate species, freshwater turtles of the Trachemys and
Anoxia Tolerance Chrysemys genera are well-known for their extreme tolerance of
anoxia that aids their survival during diving or prolonged underwa-
ter hibernation. Species including the red-eared slider (Trachemys
scripta elegans) and the painted turtle (Chrysemys picta) can survive
as long as 3–4 months submerged in anoxic water at low tempera-
ture, an adaptation that supports winter survival at the bottom of
ice locked ponds where oxygen is quickly depleted [107]. Long
term anaerobic survival by T. s. elegans has been reviewed exten-
sively [107, 108], but briefly is the general result of [1] a switch to
anaerobic glycogen fermentation via glycolysis for ATP production,
[2] the ability to store high levels of lactate in the shell and ability
to buffer acid by the release of carbonates from the shell, and [4]
the suppression of nonessential energy-consuming processes to
lower ATP demand and extend survival time.
The first study examining the role played by miRNAs in an
anoxia tolerant species was conducted in the context of cell cycle
regulation [44]. It was hypothesized that in response to oxygen
deprivation, cell cycle and cell proliferation would be suppressed,
given how energy expensive these processes are. To investigate this,
cyclin D1, which is associated with the initiation of the cell cycle,
was assessed in tissues of T. s. elegans exposed to 20 h anoxia
submergence [44]. Both total protein and nuclear localized cyclin
D1 decreased in the liver and kidney, while cyclin D1 transcript
levels and phosphorylation cyclin D1 protein remained constant.
These results provided evidence that the cell cycle may be sup-
pressed under anoxia, but a discordance between cyclin D1
mRNA and cyclin D1 protein levels suggested that posttranscrip-
tional regulation was also at play. Two miRNAs, miR-15a and
miR-16-1, that can target cyclin D1 mRNA were measured and
both were shown to increase in response to anoxia in the liver and
kidney. This was consistent with the results showing no change in
mRNA levels but a decrease in total cyclin D1 protein under anoxia.

Interestingly, a comparable analysis of white skeletal muscle in this
study found no difference between control and anoxic conditions
for cyclin D1 or its targeting miRNA, further suggesting a link
between anoxia tolerance, cell cycle suppression, and regulation by
miRNA.
Subsequently, the interacting roles of miRNAs and the p53
protein in supporting turtle anoxia tolerance were analyzed
[109]. Activation of p53 is a common occurrence in animals that
undergo metabolic rate depression [110, 111] since it is a major
node in the control of many processes relevant to hypometabolism
such as cell cycle suppression, apoptosis, autophagy, control of
translation, and modification of various metabolic pathways. Turtle
(T. s. elegans) p53 was shown to be phosphorylated at multiple sites
in response to anoxia exposure [109] which is indicative of p53
stabilization and activation. Downstream targets of p53 were eval-
uated and transcript levels of 14-3-3σ and Gadd45α, encoding two
cell cycle suppressor proteins, were found to be upregulated,
whereas PGM transcripts (encodes phosphoglucomutase) were
downregulated. The p53 protein also targets miR-34a, also
known to negatively regulate the cell cycle, and levels of this
miRNA increased significantly in both liver and muscle under
anoxic conditions.
A third study that focused on the miRNA responses to anoxia
exposure in T. s. elegans also considered the effects of low tempera-
ture on miRNA-mRNA duplexing [45], as discussed above for
hatchling painted turtle freeze tolerance. Anoxia exposures of
adult red-eared sliders (T. s. elegans) were conducted at 5 C to
mimic temperatures typical of winter hibernation underwater and
several miRNA species were quantified in liver, white skeletal mus-
cle, kidney, and spleen. Using common bioinformatics target
enrichment tools, typically set to 37 C as the default, relatively
few mRNA targets were predicted for the miRNAs that were quan-
tified. However, when modeling at a temperature of 5 C, not only
did the total number of predicted mRNA targets increase, but
targets related to biological processes known to be relevant to
anoxia tolerance increased substantially. For example, targets
related to cellular and glycogen metabolism, which are extensively
modulated during anoxia [107], were few or nonexistent when
modeled at 37 C, but at 5 C a high proportion of targets related
to these processes emerged. Likewise, cell cycle regulation, signal-
ing cascades, unfolded protein response, and protein modification
processes were all found to be enriched at 5 C, each shown to be
important to anoxia tolerance in previous studies [44, 112, 113].
Freshwater invertebrates also face comparable oxygen depriva-
tion in ice locked winter waters and/or in stagnant summer water.
Although comparatively less studied than T. s. elegans, the northern
crayfish (Orconectes virilis) is a good example of an anoxia tolerant
invertebrate species. Studies of this crayfish species and others have

included confirmation of anaerobic fermentation of glycogen to
lactate [114] and regulation of the pentose phosphate pathway to
support the NADPH needs by antioxidants [115]. A recent study
examined the miRNA expression in the tail muscle and the hepato-
pancreas of O. virilis exposed to 2 or 20 h of anoxia [46]. In both
tissues, 76 miRNAs were quantified and interestingly there was a
large disparity in the number of differentially expressed miRNAs
between tissues. 22 miRNAs were significantly regulated in hepa-
topancreas (Fig. 3a) but only four in muscle tissue. This was pos-
tulated to be a result of the very different metabolic roles that these
tissues play under anoxic conditions. Although metabolic rate
depression is universal to anoxia tolerant species, hepatic-like tissues
must maintain some activity, thus miRNAs may have more of a role
to play in the metabolic reorganization of the hepatopancreas. The
miRNAs differentially expressed in crayfish hepatopancreas under
anoxia, compared with aerobic conditions, were predicted to regu-
late cell proliferation and apoptotic signaling pathways, a subset of
which were direct targets of HIF1α (Fig. 3b), mRNA levels of
which were upregulated during anoxia in these crayfish [46].
4.2 Marine Anoxia Responses to freezing stress by miRNAs were previously discussed
Tolerance for L. littorea in the context of invertebrate freeze tolerance [36],
but, these snails, like other intertidal species, also experience oxy-
gen deprivation during low tide at any season and accumulate
typical molluscan anaerobic end products. This same study also
explored the tissue-specific miRNA responses to anoxia in the
foot muscle and the hepatopancreas. In the hepatopancreas,
miR-210 increased significantly under anoxia [36] and is known
to be a hypoxia responsive miRNA that negatively regulates mito-
chondrial metabolism [116]. This makes metabolic sense since the
suppression of mitochondrial metabolism is important in the con-
text of anoxia tolerance, because suppression of these pathways act
as a means of minimizing oxidative stress and of shunting fermen-
tative fuels to anaerobic pathways. Also, miR-20b was significantly
regulated, a miRNA that was also differentially expressed during
freezing. As previously discussed, this miRNA regulates the PI3K/
AKT pathway, suggesting a role in control over protein synthesis
and turnover. Suppression of protein synthesis during anoxia is just
as crucial as it is in freeze tolerance given that both are strategies
where the capacity for ATP synthesis is limited.
Research on cephalopods is highlighting some of their remark-
able adaptations, including programmable camouflage, high intel-
ligence, behavioural complexity, and RNA editing capabilities
[117]. Multiple adaptations to extreme environmental conditions
are also present in this group. One of these is the impressive hypoxia
tolerance of the vertically migrating Humboldt squid (Dosidicus
gigas) that rises to surface waters at night to hunt for prey but then
Fig. 3 Anoxia responsive microRNA target HIF1α and HIF1α regulatory proteins. (a) Relative expression of
microRNA in response to short-term (2 h) and long-term (20 h) anoxia in the hepatopancreas of the northern
crayfish (Orconectes virilis). Most microRNA showed decreased expression in both experimental conditions.
(b) Hippo and HIF1α signaling pathway with regulation by microRNA. This pathway was significantly enriched
with significantly downregulated microRNA in the hepatopancreas of the crayfish. (Adapted with permission
from [46])
retreats into the oxygen poor depths during the day, where they rest
in a hypometabolic state [118]. A recent study investigated miRNA
expression in D. gigas in the context of hypoxia exposure [47]. In
the brain, all differentially expressed miRNAs were elevated during
hypoxic exposures. Of note was miR-133, which is known to serve
a neuroprotective role and aid in recovery from ischemic injury
[119]. Likewise, the potential for miRNA-mediated neuroprotec-

tive mechanisms to be enhanced during exposure to environmental
stress has been suggested for other hypometabolic survival strate-
gies, including responses to dehydration in Xenopus laevis [49] and
to freezing in R. sylvatica [37].
MiRNA expression was also assessed in the mantle muscle and
the heart of D. gigas. The mantle muscle is used for jet-propulsion
and powers both prey capture and daily excursion to and from the
ocean’s depths. Several miRNAs were responsive to hypoxia in the
mantle muscle, including miR-100 which is linked to combating
oxidative damage and miR-2001 [47]. MiR-2001 is of particular
interest since it is also expressed in the hydrothermal vent shrimp
Rimicaris exoculata [120] but was thought to be lost in other
crustaceans and chordates [121]. Its presence in D. gigas may link
it to invertebrate survival in harsh marine environments. As in the
brain, all differentially regulated miRNAs identified in the heart
were upregulated under hypoxic conditions. Of note are miR-133,
miR-2a-3p, and miR-2c/d-3p, each of which have been shown to
negatively regulate apoptosis [122, 123]. MiR-133 in particular is
linked to inhibition of maladaptive cardiac hypertrophy
[124]. Given its distinct and protective roles in both neural and
cardiac tissues, miR-133 deserves a broader examination as a possi-
ble central mediator of protective strategies in animals adapted to
environmental stress.
5 Estivation and Dehydration Tolerance
5.1 Dehydration The African clawed frog, X. laevis, has long been a model animal in
Tolerance biology for pragmatic reasons such as its short development time,
in Amphibians large egg size, accessibility at all developmental stages, and hardi-
ness in laboratory conditions. Although widely used in embryology
and developmental research, only a minor amount of research has
been devoted to understanding this frog’s metabolic adaptations to
the stresses that it faces in its natural environment, particularly its
dehydration tolerance. These frogs are native to southern Africa
and are broadly distributed in ponds, lakes, and wetlands that
typically disappear during the dry season. In response to habitat
desiccation, frogs may either burrow into the drying mud or
attempt to move to a new pond. Despite these behaviors,
X. laevis can lose 40% or more of its total body water until rainfall
occurs again [125, 126]. This remarkable ability extends the appli-
cability of Xenopus, an already versatile model organism, in
biological research.
Dehydration tolerance by X. laevis is facilitated by multiple
behavioral, physiological, biochemical, and gene regulatory factors.
The first line of defense is to minimize water loss which has been
shown to occur in X. laevis by increasing body osmolality [127] via
retention of nitrogen waste as urea in blood and tissues instead of

excreting it [127, 128]. Accumulation of this osmolyte helps to
retard water loss from the body but nonetheless X. laevis must also
contend with a reduction of plasma volume and associated increases
in plasma viscosity as dehydration progresses [129]. Increased
blood viscosity ultimately leads to impaired oxygen delivery to
tissues and hypoxia stress [125]. This is partially mitigated by a
linear relationship between dehydration and heart-rate [125], and a
higher hemoglobin affinity for oxygen compared to other anurans
[130]. However, dehydration-induced hypoxia also triggers the
activation of anaerobic glycolysis to supplement ATP
production [125].
The miRNA response to dehydration in X. laevis has been
explored in multiple tissues. When challenged with 30% whole
body dehydration, miRNAs in the liver that target solute carriers
(miR-1, miR-16, and miR-125b) were downregulated [48]. Dehy-
dration is accompanied by strong changes in solute concentrations
[129], undoubtedly requiring rebalancing of various solutes
between intra- and extra-cellular compartments. Hence, increased
production of various solute carriers could be expected under these
conditions. Dehydration responsive miRNAs in the kidney and the
ventral skin were also identified and included targets in the PI3K/
AKT-mTOR pathways and the ERK1/2 signaling cascades
[48]. Regulating the PI3K/AKT-mTOR pathway was interpreted
as a means of reducing protein translation, a process that is gener-
ally suppressed in other systems of hypometabolism. However, the
ERK1/2 pathway has been shown specifically to be differentially
regulated in response to dehydration in X. laevis [126].
A study examining miRNA expression in the Xenopus brain
measured 43 miRNAs and found 12 that were significantly down-
regulated during dehydration [49]. Interestingly, when KEGG
pathway analysis was performed on the targets of the downregu-
lated miRNAs, axon guidance and long-term potentiation path-
ways were identified as targets, suggesting potential
neuroprotective effects mediated by these miRNAs. This result is
consistent with previous findings that showed that unlike other
tissues, blood flow to the brain is prioritized during dehydration
in X. laevis, increasing by approximately 20% compared to the
control condition [131]. This makes sense because to successfully
endure dehydration and rebound and rehydrate after lengthy expo-
sures, active brain networks would be needed to coordinate beha-
vioural, physiological, and metabolic responses. Hence, this
miRNA response may be an integral part of neuroprotective adap-
tation to dehydration.
5.2 Estivation Estivation is a hypometabolic state that is associated with hot

in a Marine Species and/or arid environmental conditions. Most studies of estivation
have focused on terrestrial species (snails and anuran amphibians
have received the most attention) [132] but this form of hypome-
tabolism is also known among marine and freshwater species. The
sea cucumber (Apostichopus japonicus) retreats into a state of estiva-
tion when ocean temperatures rise above 25 C, and as high-value
species raised in aquaculture in Asia [133], many recent studies
have been aimed at understanding the process and regulation of
heat-induced estivation in this species. Entry into the hypometa-
bolic state is accompanied by reduced rates of oxygen consumption
and nitrogen excretion, cessation of feeding, and degeneration of
the intestine into a thin filament over an ~4 month period of
estivation [51]. The ecophysiological, biochemical, and molecular
mechanisms behind estivation in this species have been studied by
multiple groups [134, 135], and this includes recent studies of the
contributions of miRNAs to the estivation phenotype [50, 51,
136].
An initial investigation of miRNAs in the intestine of estivating
A. japonicus revealed 14-20,000 unique sequencing reads that
matched to previously annotated miRNAs [51] and when differen-
tial expression analysis was performed (compared with 15 C non-
estivating animals), the putative targets of differentially expressed
miRNAs matched pathways previously reported to be targets in
other hypometabolic models. For example, miR-22 was upregu-
lated in the intestine of estivating sea cucumbers and targeted the
PI3K/AKT pathway, as also seen in hibernating mammals [18, 28],
and freeze-tolerant wood frogs [37]. Also upregulated during esti-
vation in intestine were miR-92 and miR-92a, which are part of the
miR-17-92 cluster that had been shown to negatively regulate
apoptosis [137]. As with miR-22, miRNA in the miR-17-92 cluster
has been identified as regulated in other animal systems of hypo-
metabolism, as discussed earlier in this chapter [34, 42, 43, 138], as
well as in the respiratory tree of estivating A. japonicus [50]. This
finding is consistent with the body of literature showing that anti-
apoptotic mechanisms are important for maintaining cell and tissue
viability during hypometabolism, particularly when dealing with
persistent environmental stress conditions that would otherwise
favour apoptosis [139–143].
Chen et al. [136] also took a functional approach to confirm a
specific miRNA-mRNA pairing within the context of estivating
A. japonicus. Although lipid catabolism would be an important
fuel source in the estivating, nonfeeding state, the suppression of
gas exchange and therefore oxygen availability could also lead to a
limitation of this aerobic pathway during estivation. In this study,
the authors examined expression of the enoyl-CoA hydratase and
the 3-hydroxyacyl CoA dehydrogenase (EHHADH) genes, which
are integral components of the peroxisomal fatty acid ß-oxidation
pathway, and which are both targeted by miR-200-3p. The results

showed that overexpression of miR-200-3p did indeed decrease the
EHHADH transcript and protein levels during estivation. The
authors suggested that this response may also serve as an antioxi-
dant mechanism to protect against oxidative stress inherent to
estivation.
6 MicroRNA Biogenesis During Environmental Stress Responses
Studying the identity, abundance, and activity of miRNAs does not

provide us with the whole picture, since another important aspect
of miRNA regulation is the capacity and functionality of the pro-
teins involved in miRNA biogenesis. Although many studies have
catalogued miRNA differential expression in response to environ-
mental stress and hypometabolism, less information is available on
the effects on the miRNA biogenesis pathway. Indeed, the miRNA
processing machinery is known to be under tight spatial and tem-
poral control and perturbation of cellular homeostasis can alter
each individual step in the miRNA biogenesis pathway, resulting
in either increased sensitivity to cell stress or to enhanced stress
resistance [144]. Hence, it is important to also consider the miRNA
processing machinery and its influence on the cellular responses to
environmental stress.
Briefly, the biogenesis of miRNAs is a multistep process that
begins with the synthesis of long imperfect dsRNA hairpins in the
nucleus and that ends with a single-stranded mature RNA guide
strand in the cytoplasm. This mature miRNA guide strand is
incorporated into the RISC along with argonaute endonucleases
(AGO1-4) and various other proteins that facilitate the negative
regulation and targeting of mRNA via complementary base-pairing
[6]. For more information on the canonical and noncanonical
miRNA processing pathways, refer to [6, 8]. In this section, we
outline how various environmental stresses and their accompanying
adaptive survival strategies are mediated by stress and species-
specific miRNA biogenesis mechanisms. Since the biogenesis
machinery is differentially regulated, these changes are coupled
with downstream consequences that manifest as distinct and
dynamic shifts in miRNA expression.
6.1 Biogenesis Different environmental stress conditions can have varying effects
and Cellular Stresses on miRNA biogenesis; for example, some can induce the accumu-
lation of pre-miRNAs, whereas others have been shown to facilitate
the processing of select miRNAs [145, 146]. Indeed, increased
abundance and activities of Drosha and Dicer proteins, that regu-
late the two main enzymatic cleavage steps of miRNA maturation,
have been associated with enhanced cellular stress resistance
[147, 148]. Conversely, reduced levels of these proteins accelerated
stress-induced cell death by sensitizing cells to reactive oxygen

species (ROS) exposure [149]. Other miRNA biogenesis enzymes
and RISC components have also been detected in stress-induced
and/or stress-sensitive nuclear and cytoplasmic compartments,
including stress granules and processing bodies [147, 150].
One stress condition that has been particularly well-
characterized with respect to miRNA biogenesis is oxygen depriva-
tion. Indeed, hypoxia-induced regulation has been observed at
each individual step of the miRNA maturation process [145]. For
example, hypoxia-induced pulmonary hypertension in rat lungs
and primary pulmonary fibroblasts exposed to 48 h of hypoxia
both displayed reduced mRNA and protein levels of Drosha and
Dicer [151]. Oxidative stress also led to reduced total pools of
pre-miRNAs and mature miRNAs, a finding in myoblasts that was
attributed to decreased DGCR8 (DiGeorge Syndrome Critical
Region 8) regulation by heme oxygenase-1 [152]. In a more
extreme case, chronic hypoxia exposures led to impaired Dicer
expression and activity which resulted in an accumulation of unpro-
cessed precursor miRNAs in human umbilical vein endothelial cells
[148]. By contrast, a study on human trophoblasts showed that
hypoxia (a state that can occur during both normal development
and placental hypoperfusion) did not affect the expression of Ago2,
DP103, Dicer, Drosha, or Exportin-5 [153]. This maintenance of
homeostatic balance during oxygen limitation suggested that vari-
ous checks might be in place to limit altered expression of the
miRNA biogenesis machinery in order to minimize dysregulation
of downstream miRNA expression. However, it should be noted
that the use of primary trophoblast cells and placental tissue
restricts these findings to embryonic developmental stages, limiting
their potential for broad functional relevance [153].
6.2 Biogenesis MiRNA biogenesis is under regulatory control by several of the

and Transcription major cellular transcription factors [145, 154]. For example, the
Factor Networks stimulation of select receptor-specific SMADs via phosphorylation
led to the translocation of SMADs into the nucleus where they
associated with the microprocessor complex, DROSHA and
DGCR8, and enhanced DROSHA cleavage efficiency [154]. Simi-
larly, once activated, the p53 transcription factor was also able to
enhance DROSHA-mediated processing of several miRNA families
[145]. Both SMADs and p53 are known to be differentially regu-
lated during environmental stress-induced metabolic rate depres-
sion, including in response to freezing, hibernation, and anoxia
[29, 109, 110, 155], but their direct relationship with miRNA
biogenesis has yet to be explored. Other miRNA regulators include
hypoxia-sensitive transcription factors that can be broadly categor-
ized as either HIF-dependent or HIF-independent, such as NFκB
[156]. It should be noted that transcriptional control of miRNA
expression is only part of the narrative of hypoxia-centered miRNA
regulation.
Other aspects of control over the miRNA biogenesis axis

include alternative splicing, subcellular localization, posttransla-
tional modifications, and targeting efficiency [147, 157, 158]
with each of these potentially affected directly or indirectly in
response to environmental stress. Indeed, all proteins involved in
the miRNA biogenesis are known to be regulated by posttransla-
tional modifications that can alter protein functionality or target
recruitment [144]. For example, hypoxia was shown to induce the
phosphorylation, hydroxylation, and polyADP-ribosylation of
Ago2, which in turn directed Ago2 to stress granules and reduced
the binding affinity of Ago2 to Dicer [158]. Temperature and
osmotic stresses on cells induced p38 MAPK-mediated phosphor-
ylation of DROSHA, leading to DROSHA destabilization and
reduced DGCR8 binding [149]. Moreover, hypoxia has been
shown to increase the expression of type I collagen prolyl-4-
hydroxylase, a protein that promotes prolyl-hydroxylation and
resulting accumulation and activation of Ago2 [157].
6.3 Biogenesis Very little is currently known about the regulation of the miRNA
in Systems biogenesis in most animal models of extreme survival strategies.
of Stress-Induced Among mammalian hibernators, only Dicer has been assessed.
Metabolic Rate Dicer protein levels were measured in the brain and the skeletal
Depression muscles of hibernating versus aroused little brown bats,
M. lucifigus, but were unchanged by torpor [19, 20]. Similarly,
protein levels of Dicer were unchanged during hibernation in the
white adipose, the brown adipose, and the skeletal muscle of hiber-
nating 13-lined ground squirrels, I. tridecemlineatus [23, 30]. On
the other hand, protein levels of Dicer were upregulated in hearts
but downregulated in kidneys of hibernating squirrels compared
with euthermic controls [23]. This suggested an increase in miRNA
processing in squirrel heart during torpor but a follow-up study
showed that the majority of differentially expressed miRNAs, out of
117 analyzed, were downregulated during hibernation [26]. This
emphasizes the complexity of miRNA expression and highlights
that whereas Dicer is a key processing enzyme, it is not the sole
regulator; indeed, other studies have also reported that the Dicer
protein levels do not always correlate with the miRNA silencing
activity [159]. The only investigation of miRNA biogenesis in an
invertebrate model of stress-induced hypometabolism was an anal-
ysis of the intertidal marine snail, L. littorea. Dicer protein levels
were upregulated in both the foot muscle and the hepatopancreas
of snails exposed to 24 h freezing and 24 h anoxia [36].
The first in-depth investigation of multiple proteins involved in
the miRNA biogenesis during adaptation to extreme environmen-
tal stress explored how the brain of the wood frog, R. sylvatica,
responded to 24 h freezing exposures (Fig. 4). This study revealed a
reduction in miRNA processing proteins during freezing that cor-
responded with the measured decrease of the majority of
Fig. 4 MicroRNA biogenesis pathway and changes in relative protein levels of its protein components in brains
of frozen wood frogs. The microRNA biogenesis pathway is a multistep process involving sequential nuclease
processing steps. For more detailed information see [6]. Inset: Analysis of protein responses of members of
the miRNA biogenesis pathway in wood frog brain over a freeze–thaw cycle, as assessed by immunoblotting.
The histogram shows protein levels, relative to control, of DROSHA, DGCR8, XPO5, RAN, DICER, TRBP, PACT,
AGO1, AGO2, and p-AGO2Tyr393 under control, 24 h frozen, and 8 h thawed conditions. (Adapted with
permission from [37])
differentially expressed miRNAs [37]. The protein levels of

DROSHA, a crucial enzymatic component of the microprocessor
complex, displayed a significant freeze-induced suppression that
was hypothesized to be related to reduced energy availability in
the frozen state [37, 160]. In addition, reduced levels of
EXPORTIN-5, the protein responsible for transporting nascent
pre-miRNAs into the cytoplasm, as well as the DICER binding
proteins TRBP and PACT, all indicated an overall reduced capacity
for miRNA biogenesis. Indeed, previous work has shown that
reduced levels of TRBP and PACT lower the efficiency of posttran-
scriptional gene regulation, promote the destabilization of DICER,
and result in miRNA downregulation [144, 161]. This is also
supported by studies showing that reduced Ago2 expression leads
to lower miRNA synthesis; this argonaute protein is known to
interact with ~60% of all miRNAs and is the sole argonaute with
slicer activity [161]. Contrary to these global trends of suppression

of miRNA biogenesis, the levels of phosphorylated AGO2
(p-AGO2Tyr393) were reduced during thawing, showing that post-
translational control of protein activity is also associated with
reduced binding to Dicer and inhibited miRNA processing [37].
Overall, these results suggest a reduced capacity for miRNA
synthesis in frozen vertebrate brains, particularly during the nuclear
maturation steps, a suggestion that is confirmed with the associated
predominant downregulation of differentially expressed miRNAs.
Indeed, the reduced capacity for miRNA biogenesis was reflected in
the reported decrease of 41 freeze-downregulated and thaw-
downregulated miRNAs in freeze-tolerant wood frog brains
[37]. The reduced levels of select miRNAs during freezing suggest
that these miRNAs are working in concert with the reduced
miRNA biogenesis capacity to potentially allow for the activation,
or more accurately lack of inhibition, of various neuroprotective
processes such as synaptic functions, intracellular signal transduc-
tion, and anoxia/ischemia injury protection [37]. Regulating the
regulators (i.e., regulating the proteins that synthesize miRNAs)
affects the cell’s ability to cope with stress. Taken together, these
findings highlight the stress- and tissue-specific regulation that is in
place to facilitate dynamic miRNA biogenesis and survive periods of
environmentally induced stress.
7 Conclusions
MiRNAs offer an efficient, flexible, and powerful means to regulate

a myriad of biological processes. Taken together, the large number
of miRNA species, their complementary mRNA targets, and the
temperature-dependent effects on binding, all act to generate an
expansive regulatory RNA network capable of enabling broad and
flexible control over mRNA expression during stress. The energetic
cost efficiency of miRNAs for terminating translation relative to the
synthesis of a nonessential protein, or relative to the synthesis of a
negative regulator to inhibit said nonessential protein, makes miR-
NAs a fitting tool for stressed and resource-limited cells. These
features give cells the ability to quickly modify phenotype in
response to changes in the environment, and thus it is no surprise
that miRNAs have shown significant involvement in adaptations to
various environmental extremes. The miRNA response to the rapid
and unpredictable onset of some environmental stresses, such as
freezing and anoxia, bear out this hypothesis in multiple lineages.
However, regulation by miRNA is not merely used as an emergency
response to sudden environmental changes as seen in animals that
recurrently enter a hypometabolic state such as the cyclic nature of
torpor-arousal bouts and the diving habits of the Humboldt squid.
The general hypometabolic miRNA fingerprints, unique stress, and
tissue-specific expression profiles discussed in this chapter highlight

the conserved nature of miRNA regulation for dealing effectively
with stress across the phylogeny. Studying these miRNA responses
is crucial to elucidating the vast and substantial role of miRNAs in
actualizing profound and reversible shifts in cellular metabolic
state.
Acknowledgments
We thank J. M. Storey for editorial review of the manuscript.

Research in the Storey lab is funded by a Discovery grant from
the Natural Sciences and Engineering Council of Canada
(NSERC). KBS holds the Canada Research Chair in Molecular
Physiology, HH holds an NSERC doctoral scholarship, and LJH
holds an Ontario Graduate Scholarship.
References
1. Biggar KK, Storey KB (2018) Functional computational analysis. Methods in molecular
impact of microRNA regulation in models of biology, vol 2257. Springer, New York
extreme stress adaptation. J Mol Cell Biol 9. Bartel DP (2009) MicroRNAs: target recog-
10:93–101 nition and regulatory functions. Cell
2. Bartel DP (2004) MicroRNAs: genomics, 136:215–233
biogenesis, mechanism, and function. Cell 10. Ebert MS, Sharp PA (2012) Roles for micro-
116:281–297 RNAs in conferring robustness to biological
3. Hadj-Moussa H, Storey KB (2020) The Oxy- processes. Cell 149:515–524
miR response to oxygen limitation: a compar- 11. Hamzeiy H, Suluyayla R, Brinkrolf C et al
ative microRNA perspective. J Exp Biol 223 (2017) Visualization and analysis of Micro-
(Pt 10):jeb204594 RNAs within KEGG pathways using
4. Storey KB, Storey JM (1990) Metabolic rate VANESA. J Integr Bioinform 14:1–9
depression and biochemical adaptation in 12. Huang Y, Shen XJ, Zou Q et al (2011)
anaerobiosis, hibernation and estivation. Q Biological functions of microRNAs: a review.
Rev Biol 65:145–174 J Physiol Biochem 67:129–139
5. Storey KB, Storey JM (2017) Molecular phys- 13. Wu L, Fan J, Belasco JG (2006) MicroRNAs
iology of freeze tolerance in vertebrates. direct rapid deadenylation of mRNA. Proc
Physiol Rev 97:623–665 Natl Acad Sci U S A 103:4034–4039
6. Ergin K, Çetinkaya R (2021) Regulation of 14. Friedman RC, Farh KK, Burge CB et al
microRNAs. In: Allmer J, Yousef M (eds) (2009) Most mammalian mRNAs are con-
miRNomics: microRNA biology and compu- served targets of microRNAs. Genome Res
tational analysis. Methods in molecular biol- 19:92–105
ogy, vol 2257. Springer, New York 15. Ruf T, Geiser F (2015) Daily torpor and
7. Yaylak B, Akgül B (2021) Experimental hibernation in birds and mammals. Biol Rev
microRNA detection methods. In: Allmer J, Camb Philos Soc 90:891–926
Yousef M (eds) miRNomics: microRNA biol- 16. Storey KB, Storey JM (2010) Metabolic rate
ogy and computational analysis. Methods in depression: the biochemistry of mammalian
molecular biology, vol 2257. Springer, New hibernation. Adv Clin Chem 52:77–108
York
17. Biggar KK, Luu BE, Wu CW et al (2018)
8. Dweep H, Showe LC, Kossenkov AV (2021) Identification of novel and conserved micro-
Functional annotation of microRNAs using RNA and their expression in the gray mouse
existing resources. In: Allmer J, Yousef M lemur, Microcebus murinus, a primate capa-
(eds) miRNomics: microRNA biology and ble of daily torpor. Gene 677:332–339
18. Hadj-Moussa H, Moggridge JA, Luu BE et al response in ground squirrels during hiberna-
(2016) The hibernating south American mar- tion. Mol Cell Biochem 439:151–161
supial, Dromiciops gliroides, displays torpor- 30. Wu C-W, Biggar KK, Storey KB (2014)
sensitive microRNA expression patterns. Sci Expression profiling and structural characteri-
Rep 6:24627 zation of microRNAs in adipose tissues of
19. Biggar KK, Storey KB (2014) Identification hibernating ground squirrels. Genomics Pro-
and expression of microRNA in the brain of teomics Bioinformatics 12:284–291
hibernating bats, Myotis lucifugus. Gene 31. Lee Y-j, Johnson KR, Hallenbeck JM (2012)
544:67–74 Global protein conjugation by ubiquitin-like-
20. Kornfeld SF, Biggar KK, Storey KB (2012) modifiers during ischemic stress is regulated
Differential expression of mature microRNAs by microRNAs and confers robust tolerance
involved in muscle maintenance of hibernat- to ischemias. PLoS One 7:e47787
ing little brown bats, Myotis lucifugus: a 32. Courteau LA, Storey KB, Morin PJ (2012)
model of muscle atrophy resistance. Geno- Differential expression of microRNA species
mics Proteomics Bioinformatics 10:295–301 in a freeze tolerant insect, Eurosta solidaginis.
21. Yuan L, Geiser F, Lin B et al (2015) Down Cryobiology 65:210–214
but not out: the role of MicroRNAs in hiber- 33. Lyons PJ, Poitras JJ, Courteau LA et al
nating bats. PLoS One 10:1–19 (2013) Identification of differentially regu-
22. Liu Y, Hu W, Wang H et al (2010) Genomic lated micrornas in cold-hardy insects. Cryo-
analysis of miRNAs in an extreme mammalian Letters 34:83–89
hibernator, the Arctic ground squirrel. 34. Lyons PJ, Storey KB, Morin PJ (2015)
Physiol Genomics 42A:39–51 Expression of miRNAs in response to freezing
23. Morin PJ, Dubuc A, Storey KB (2008) Dif- and anoxia stresses in the freeze tolerant fly
ferential expression of microRNA species in Eurosta solidaginis. Cryobiology 71:97–102
organs of hibernating ground squirrels: a 35. Lyons PJ, Govaere L, Crapoulet N et al
role in translational suppression during tor- (2016) Characterization of cold-associated
por. Biochim Biophys Acta 1779:628–633 microRNAs in the freeze-tolerant gall fly
24. Lang-Ouellette D, Morin P (2014) Differen- Eurosta solidaginis using high-throughput
tial expression of miRNAs with metabolic sequencing. Comp Biochem Physiol Part D
implications in hibernating thirteen-lined Genomics Proteomics 20:95–100
ground squirrels, Ictidomys tridecemlineatus. 36. Biggar KK, Kornfeld SF, Maistrovski Y et al
Mol Cell Biochem 394:291–298 (2012) MicroRNA regulation in extreme
25. Luu BE, Biggar KK, Wu C-WW et al (2016) environments: differential expression of
Torpor-responsive expression of novel micro- microRNAs in the intertidal snail Littorina
RNA regulating metabolism and other cellu- littorea during extended periods of freezing
lar pathways in the thirteen-lined ground and anoxia. Genomics Proteomics Bioinfor-
squirrel, Ictidomys tridecemlineatus. FEBS matics 10:302–309
Lett 590:3574–3582 37. Hadj-Moussa H, Storey KB (2018) Microma-
26. Wu C-W, Biggar KK, Luu BE et al (2016) naging freeze tolerance: the biogenesis and
Analysis of microRNA expression during the regulation of neuroprotective microRNAs in
torpor-arousal cycle of a mammalian hiberna- frozen brains. Cell Mol Life Sci 75:1–13
tor, the 13-lined ground squirrel. Physiol 38. Bansal S, Luu BE, Storey KB (2016) Micro-
Genomics 48:388–396 RNA regulation in heart and skeletal muscle
27. Frigault JJ, Gaudet JD, Morin P (2018) over the freeze–thaw cycle in the freeze toler-
Investigating Nrf2-associated non-coding ant wood frog. J Comp Physiol B
RNAs in the hibernating ground squirrel, 186:229–241
Ictidomys tridecemlineatus. J Therm Biol 39. Biggar KK, Dubuc A, Storey KB (2009)
75:38–44 MicroRNA regulation below zero: differential
28. Morin MD, Lang-Ouellette D, Lyons PJ et al expression of miRNA-21 and miRNA-16 dur-
(2017) Characterization of miRNAs modu- ing freezing in wood frogs. Cryobiology
lated by torpor in the hibernating ground 59:317–321
squirrel Ictidomys tridecemlineatus liver by 40. Zhang J, Storey KB (2013) Akt signaling and
next-generation sequencing. Cryo Letters freezing survival in the wood frog, Rana syl-
38:269–277 vatica. Biochim Biophys Acta
29. Wu C-W, Storey KB (2018) Regulation of 1830:4828–4837
Smad mediated microRNA transcriptional 41. Biggar KK, Storey KB (2015)
Low-temperature microRNA expression in
the painted turtle, Chrysemys picta during miRNA in cell growth and apoptosis. Nucleic
freezing stress. FEBS Lett 589:3665–3670 Acids Res 33:1290–1297
42. Lyons PJ, Crapoulet N, Storey KB et al 55. Rose JC, Epperson LE, Carey HV et al (2011)
(2015) Identification and profiling of miR- Seasonal liver protein differences in a hiberna-
NAs in the freeze-avoiding gall moth Epi- tor revealed by quantitative proteomics using
blema scudderiana via next-generation whole animal isotopic labeling. Comp Bio-
sequencing. Mol Cell Biochem 410:155–163 chem Physiol Part D Genomics Proteomics
43. Morin MD, Frigault JJ, Lyons PJ et al (2017) 6:163–170
Amplification and quantification of cold- 56. Callis TE, Pandya K, Seok HY et al (2009)
associated microRNAs in the Colorado potato MicroRNA-208a is a regulator of cardiac
beetle (Leptinotarsa decemlineata) agricul- hypertrophy and conduction in mice. J Clin
tural pest. Insect Mol Biol 26:574–583 Invest 119:2772–2786
44. Biggar KK, Storey KB (2012) Evidence for 57. Luu BE, Tessier SN, Duford DL et al (2015)
cell cycle suppression and microRNA regula- The regulation of troponins I, C and ANP by
tion of cyclin D1 during anoxia exposure in GATA4 and Nkx2-5 in heart of hibernating
turtles. Cell Cycle 11:1705–1713 thirteen-lined ground squirrels, Ictidomys tri-
45. Biggar KK, Storey KB (2017) Exploration of decemlineatus. PLoS One 10:e0117747
low temperature microRNA function in an 58. Eddy SF, Morin P, Storey KB (2005) Cloning
anoxia tolerant vertebrate ectotherm, the red and expression of PPAR-gamma and
eared slider turtle (Trachemys scripta ele- PGC-1alpha from the hibernating ground
gans). J Therm Biol 68:139–146 squirrel, Spermophilus tridecemlineatus. Mol
46. English SG, Hadj-Moussa H, Storey KB Cell Biochem 269:175–182
(2018) MicroRNAs regulate survival in 59. Lorenzen J, Kumarswamy R, Dangwal S et al
oxygen-deprived environments. J Exp Biol (2012) MicroRNAs in diabetes and diabetes-
221:jeb.190579 associated complications. RNA Biol
47. Hadj-Moussa H, Logan SM, Seibel BA et al 9:820–827
(2018) Potential role for microRNA in regu- 60. Wu CW, Biggar KK, Storey KB (2013) Bio-
lating hypoxia-induced metabolic suppression chemical adaptations of mammalian hiberna-
in jumbo squids. Biochim Biophys Acta Gene tion: exploring squirrels as a perspective
Regul Mech 1861:586–593 model for naturally induced reversible insulin
48. Wu CW, Biggar KK, Storey KB (2013) Dehy- resistance. Braz J Med Biol Res 46:1–13
dration mediated microRNA response in the 61. Pan L, Huang BJ, Ma XE et al (2015) MiR-25
African clawed frog Xenopus laevis. Gene protects cardiomyocytes against oxidative
529:269–275 damage by targeting the mitochondrial cal-
49. Luu BE, Storey KB (2015) Dehydration trig- cium uniporter. Int J Mol Sci 16:5420–5433
gers differential microRNA expression in 62. Lang A, Grether-Beck S, Singh M et al (2016)
Xenopus laevis brain. Gene 573:64–69 MicroRNA-15b regulates mitochondrial ROS
50. Chen M, Storey KB (2014) Large-scale iden- production and the senescence-associated
tification and comparative analysis of miRNA secretory phenotype through sirtuin
expression profile in the respiratory tree of the 4/SIRT4. Aging (Albany NY) 8:484–505
sea cucumber Apostichopus japonicus during 63. Davis BN, Hilyard AC, Nguyen PH et al
aestivation. Mar Genomics 13:39–44 (2010) Smad proteins bind a conserved
51. Chen M, Zhang X, Liu J et al (2013) High- RNA sequence to promote MicroRNA matu-
throughput sequencing reveals differential ration by Drosha. Mol Cell 39:373–384
expression of miRNAs in intestine from sea 64. Frerichs KU, Smith CB, Brenner M et al
cucumber during aestivation. PLoS One (1998) Suppression of protein synthesis in
8:1–8 brain during hibernation involves inhibition
52. Chan JA, Krichevsky AM, Kosik KS (2005) of protein initiation and elongation. Proc
MicroRNA-21 is an antiapoptotic factor in Natl Acad Sci U S A 95:14511–14516
human glioblastoma cells. Cancer Res 65. Morin PJ, Ni Z, McMullen DC et al (2008)
65:6029–6033 Expression of Nrf2 and its downstream gene
53. Logan SM, Storey KB (2017) Avoiding apo- targets in hibernating 13-lined ground
ptosis during mammalian hibernation. Tem- squirrels, Spermophilus tridecemlineatus.
perature (Austin) 4:15–17 Mol Cell Biochem 312:121–129
54. Cheng AM, Byrom MW, Shelton J et al 66. Krivoruchko A, Storey KB (2010) Activation
(2005) Antisense inhibition of human miR- of antioxidant defenses in response to freezing
NAs and indications for an involvement of
in freeze-tolerant painted turtle hatchlings. 80. Berman DI, Meshcheryakova EN (2012) Is

Biochim Biophys Acta 1800:662–668 the western boundary of the Siberian sala-
67. Wei W, Lin H-K (2012) The key role of ubi- mander (Salamandrella keyserlingii,
quitination and sumoylation in signaling and Amphibia, Caudata, Hynobiidae) range deter-
cancer: a research topic. Front Oncol 2:187 mined by the specific features of its wintering?
68. Lee Y, Miyake S, Wakita H et al (2007) Pro- Dokl Biol Sci 443:97–100
tein SUMOylation is massively increased in 81. Marshall CB, Fletcher GL, Davies PL (2004)
hibernation torpor and is critical for the cyto- Hyperactive antifreeze protein in a fish.
protection provided by ischemic precondi- Nature 429:153
tioning and hypothermia in SHSY5Y cells. J 82. Costanzo JP, do Amaral MCF, Rosendale AJ
Cereb Blood Flow Metab 27:950–962 et al (2013) Hibernation physiology, freezing
69. McCarthy JJ (2011) The MyomiR network in adaptation and extreme freeze tolerance in a
skeletal muscle plasticity. Exerc Sport Sci Rev northern population of the wood frog. J Exp
39:150–154 Biol 216:3461–3473
70. Fragkouli A, Doxakis E (2014) miR-7 and 83. Cacchiarelli D, Martone J, Girardi E et al
miR-153 protect neurons against MPP+- (2010) MicroRNAs involved in molecular cir-
induced cell death via upregulation of cuitries relevant for the Duchenne muscular
mTOR pathway. Front Cell Neurosci 8:1–14 dystrophy pathogenesis are controlled by the
71. Nespolo RF, Verdugo C, Cortés PA et al dystrophin/nNOS pathway. Cell Metab
(2010) Bioenergetics of torpor in the micro- 12:341–351
biotherid marsupial, monito del Monte (Dro- 84. Joanisse DR, Storey KB (1996) Oxidative
miciops gliroides): the role of temperature stress and antioxidants in overwintering larvae
and food availability. J Comp Physiol B of cold-hardy goldenrod gall insects. J Exp
180:767–773 Biol 199:1483–1491
72. Villarin JJ, Schaeffer PJ, Markle RA et al 85. Zhong D, Huang G, Zhang Y et al (2013)
(2003) Chronic cold exposure increases liver MicroRNA-1 and microRNA-206 suppress
oxidative capacity in the marsupial Monodel- LXRα-induced lipogenesis in hepatocytes.
phis domestica. Comp Biochem Physiol A Cell Signal 25:1429–1437
Mol Integr Physiol 136:621–630 86. Joanisse DR, Storey KB (1996) Fatty acid
73. Zhang J, Hadj-Moussa H, Storey KB (2016) content and enzymes of fatty acid metabolism
Current progress of high-throughput micro- in overwintering cold-hardy gall insects.
RNA differential expression analysis and ran- Physiol Zool 69:1079–1095
dom forest gene selection for model and 87. Song H, Zhang Y, Liu N et al (2016)
non-model systems: an R implementation. J miR-92b regulates glioma cells proliferation,
Integr Bioinform 13:306 migration, invasion, and apoptosis via PTEN/
74. Storey KB (2015) The gray mouse lemur: a Akt signaling pathway. J Physiol Biochem
model for studies of primate metabolic rate 72:201–211
depression. Genomics Proteomics Bioinfor- 88. Ge W, Chen YW, Weng R et al (2012) Over-
matics 13:77–80 lapping functions of microRNAs in control of
75. Hadj-Moussa H, Storey KB (2019) Bringing apoptosis during drosophila embryogenesis.
nature Back: using hibernation to reboot Cell Death Differ 19:839–846
organ preservation. FEBS J 286 89. Hermeking H (2010) The miR-34 family in
(6):1094–1100 cancer and apoptosis. Cell Death Differ
76. Field KA, Sewall BJ, Prokkola JM et al (2018) 17:193–199
Effect of torpor on host transcriptomic 90. Xu P, Vernooy SY, Guo M et al (2003) The
responses to a fungal pathogen in hibernating drosophila microRNA Mir-14 suppresses cell
bats. Mol Ecol 27:3727–3743 death and is required for normal fat metabo-
77. Block W (1990) Cold tolerance of insects and lism. Curr Biol 13:790–795
other arthropods. Philos Trans R Soc B Biol 91. Bennett VA, Pruitt NL, Lee RE (1997) Sea-
Sci 326:613–633 sonal changes in fatty acid composition asso-
78. Ansart A, Vernon P (2003) Cold hardiness in ciated with cold-hardening in third instar
molluscs. Acta Oecol 24:95–102 larvae of Eurosta solidaginis. J Comp
79. Burke RL, Hussain AA, Storey JM et al Physiol B, Biochem Syst Environ Physiol
(2002) Freeze tolerance and supercooling 167:249–255
ability in the Italian wall lizard, Podarcis 92. Liu C-J, Tsai M-M, Hung P-S et al (2010)
sicula, introduced to Long Island, miR-31 ablates expression of the HIF regu-
New York. Copeia 2002:836–842 latory factor FIH to activate the HIF pathway
in head and neck carcinoma. Cancer Res responses of metabolic rate depression. J
70:1635–1644 Mol Cell Biol 3:167–175
93. Morin P, McMullen DC, Storey KB (2005) 106. Krivoruchko A, Storey KB (2015) Turtle
HIF-1α involvement in low temperature and anoxia tolerance: biochemistry and gene reg-
anoxia survival by a freeze tolerant insect. Mol ulation. Biochim Biophys Acta
Cell Biochem 280:99–106 1850:1188–1196
94. Alyokhin A, Baker M, Mota-Sanchez D et al 107. Storey KB (2007) Anoxia tolerance in turtles:
(2008) Colorado potato beetle resistance to metabolic regulation and gene expression.
insecticides. Am J Potato Res 85:395–413 Comp Biochem Physiol A Mol Integr Physiol
95. Zhang J, Khan SA, Hasse C et al (2015) Full 147:263–276
crop protection from an insect pest by expres- 108. Lutz PL, Nilsson GE (1997) Contrasting
sion of long double-stranded RNAs in plas- strategies for anoxic brain survival--glycolysis
tids. Science 347:991–994 up or down. J Exp Biol 200:411–419
96. Palli SR (2014) RNA interference in Color- 109. Zhang J, Biggar KK, Storey KB (2013) Reg-
ado potato beetle: steps toward development ulation of p53 by reversible post-
of dsRNA as a commercial insecticide. Curr transcriptional and post-translational
Opin Insect Sci 6:1–8 mechanisms in liver and skeletal muscle of an
97. Zhu F, Xu J, Palli R et al (2011) Ingested anoxia tolerant turtle, Trachemys scripta ele-
RNA interference for managing the popula- gans. Gene 513:147–155
tions of the Colorado potato beetle, Leptino- 110. Hefler J, Wu C-W, Storey KB (2015) Tran-
tarsa decemlineata. Pest Manag Sci scriptional activation of p53 during cold
67:175–182 induced torpor in the 13-lined ground squir-
98. Shan F, Li J, Huang QY (2014) HIF-1 alpha- rel Ictidomys tridecemlineatus. Biochem Res
induced up-regulation of miR-9 contributes Int 2015:731595
to phenotypic modulation in pulmonary 111. Luu BE, Wijenayake S, Zhang J et al (2018)
artery smooth muscle cells during hypoxia. J Strategies of biochemical adaptation for
Cell Physiol 229:1511–1520 hibernation in a south American marsupial,
99. Wang H, Bian S, Yang CS (2011) Green tea Dromiciops gliroides: 3. Activation of
polyphenol EGCG suppresses lung cancer cell pro-survival response pathways. Comp Bio-
growth through upregulating miR-210 chem Physiol B Biochem Mol Biol 224:26–31
expression caused by stabilizing HIF-1α. Car- 112. Greenway SC, Storey KB (2000) Mitogen-
cinogenesis 32:1881–1889 activated protein kinases and anoxia tolerance
100. Jia LF, Huang YP, Zheng YF et al (2014) in turtles. J Exp Zool 287:477–484
MiR-29b suppresses proliferation, migration, 113. Krivoruchko A, Storey KB (2013) Activation
and invasion of tongue squamous cell carci- of the unfolded protein response during
noma through PTEN-AKT signaling pathway anoxia exposure in the turtle Trachemys
by targeting Sp1. Oral Oncol 50:1062–1071 scripta elegans. Mol Cell Biochem
101. Rubinsky B, Wong ST, Hong JS et al (1994) 374:91–103
1H magnetic resonance imaging of freezing 114. G€ade G (1984) Effects of oxygen deprivation
and thawing in freeze-tolerant frogs. Am J during anoxia and muscular work on the
Phys 266:R1771–R1777 energy metabolism of the crayfish, Orco-
102. Padilla PA, Nystul TG, Zager RA et al (2002) nectes limosus. Comp Biochem Physiol A
Dephosphorylation of cell cycle-regulated Mol Integr Physiol 77:495–502
proteins correlates with anoxia-induced sus- 115. Lant B, Storey KB (2011) Glucose-6-phos-
pended animation in Caenorhabditis elegans. phate dehydrogenase regulation in anoxia tol-
Mol Biol Cell 13:1473–1483 erance of the freshwater crayfish Orconectes
103. Yang F, Chen S, Dai Z-M et al (2013) Regu- virilis. Enzyme Res 2011:524906
lation of trehalase expression inhibits apopto- 116. Chan YC, Banerjee J, Choi SY et al (2012)
sis in diapause cysts of Artemia. Biochem J miR-210: the master hypoxamir. Microcircu-
456:185–194 lation 19:215–223
104. Lyons PJ, Lang-Ouellette D, Morin PJ 117. Liscovitch-Brauer N, Alon S, Porath HT et al
(2013) CryomiRs: towards the identification (2017) Trade-off between transcriptome plas-
of a cold-associated family of microRNAs. ticity and genome evolution in cephalopods.
Comp Biochem Physiol Part D Genomics Cell 169:191–202.e11
Proteomics 8:358–364 118. Rosa R, Seibel BA (2010) Metabolic physiol-
105. Biggar KK, Storey KB (2011) The emerging ogy of the Humboldt squid, Dosidicus gigas:
roles of microRNAs in the molecular implications for vertical migration in a
pronounced oxygen minimum zone. Prog hypovolemia, and salt load. J Exp Zool
Oceanogr 86:72–80 218:305–308
119. Xin H, Li Y, Liu Z et al (2013) MiR-133b 132. Storey KB, Storey JM (2010) Metabolic reg-
promotes neural plasticity and functional ulation and gene expression during
recovery after treatment of stroke with multi- aestivation. In: Navas CA, Carvalho JE (eds)
potent mesenchymal stromal cells in rats via Aestivation Mol. Physiol. Asp. Springer, Ber-
transfer of exosome-enriched extracellular lin Heidelberg, pp 25–45
particles. Stem Cells 31:2737–2746 133. Han Q, Keesing JK, Liu D (2016) A review of
120. Zhou Y, He Y, Wang C et al (2015) Charac- sea cucumber aquaculture, ranching, and
terization of miRNAs from hydrothermal stock enhancement in China. Rev Fish Sci
vent shrimp Rimicaris exoculata. Mar Geno- Aquac 24:326–341
mics 24:371–378 134. Xu K, Yu Q, Zhang J et al (2018) Cell loss by
121. Wheeler BM, Heimberg AM, Moy VN et al apoptosis is involved in the intestinal degen-
(2009) The deep evolution of metazoan eration that occurs during aestivation in the
microRNAs. Evol Dev 11:50–68 sea cucumber Apostichopus japonicus. Comp
122. Xu C, Lu Y, Pan Z et al (2011) The muscle- Biochem Physiol B Biochem Mol Biol
specific microRNAs miR-1 and miR-133 pro- 216:25–31
duce opposing effects on apoptosis by target- 135. Yang H, Yuan X, Zhou Y et al (2005) Effects
ing HSP60, HSP70 and caspase-9 in of body size and water temperature on food
cardiomyocytes. J Cell Sci 124:3187 consumption and growth in the sea cucumber
123. Zhang W, Cohen SM (2013) The hippo path- Apostichopus japonicus (Selenka) with special
way acts via p53 and microRNAs to control reference to aestivation. Aquac Res
proliferation and proapoptotic gene expres- 36:1085–1092
sion during tissue growth. Biol Open 136. Chen M, Wang S, Li X et al (2018) The
2:822–828 potential contribution of miRNA-200-3p to
124. Carè A, Catalucci D, Felicetti F et al (2007) the fatty acid metabolism by regulating
MicroRNA-133 controls cardiac hypertro- AjEHHADH during aestivation in sea
phy. Nat Med 13:613–618 cucumber. PeerJ 6:e5703
125. Hillman SS (1978) The roles of oxygen deliv- 137. Mendell JT (2008) miRiad roles for the
ery and electrolyte levels in the dehydrational miR-17-92 cluster in development and dis-
death of Xenopus laevis. J Comp Physiol B ease. Cell 133:217–222
128:169–175 138. Yang S, Zhang J, Wang S et al (2017) Identi-
126. Malik AI, Storey KB (2009) Activation of fication and profiling of miRNAs in overwin-
extracellular signal-regulated kinases during tering Lissorhoptrus oryzophilus via next-
dehydration in the African clawed frog, Xeno- generation sequencing. Cryobiology
pus laevis. J Exp Biol 212:2595–2603 74:68–76
127. Balinsky JB, Choritz EL, Coe CG et al (1967) 139. Wang J, Zhang G, Fang X et al (2012) The
Amino acid metabolism and urea synthesis in oyster genome reveals stress adaptation and
naturally aestivating Xenopus laevis. Comp complexity of shell formation. Nature
Biochem Physiol 22:59–68 490:49–54
128. Balinsky JB, Cragg MM, Baldwin E (1961) 140. Rouble AN, Hefler J, Mamady H et al (2013)
The adaptation of amphibian waste nitrogen Anti-apoptotic signaling as a cytoprotective
excretion to dehydration. Comp Biochem mechanism in mammalian hibernation. PeerJ
Physiol 3:236–244 1:e29
129. Hillman SS (1978) Some effects of dehydra- 141. Fu W, Hu H, Dang K et al (2016) Remark-
tion on internal distributions of water and able preservation of Ca2+homeostasis and
solutes in Xenopus laevis. Comp Biochem inhibition of apoptosis contribute to anti-
Physiol Part A Physiol 61:303–307 muscle atrophy effect in hibernating Daurian
130. Jokumsen A, Weber RE (1980) ground squirrels. Sci Rep 6:1–13
Haemoglobin-oxygen binding properties in 142. Gerber VEM, Wijenayake S, Storey KB
the blood of Xenopus laevis, with special ref- (2016) Anti-apoptotic response during
erence to the influences of aestivation and of anoxia and recovery in a freeze-tolerant
temperature and salinity acclimation. J Exp wood frog (Rana sylvatica). PeerJ 4:e1834
Biol 86:19–37 143. Yi SX, Moore CW, Lee RE (2007) Rapid
131. Hillman SS, Sommerfeldt RW (1981) Micro- cold-hardening protects Drosophila melano-
sphere studies of amphibian systemic blood gaster from cold-induced apoptosis. Apopto-
flow redistribution during dehydration, sis 12:1183–1193
144. Ha M, Kim VN (2014) Regulation of micro- 153. Donker RB, Mouillet J-FF, Nelson DM et al
RNA biogenesis. Nat Rev Mol Cell Biol (2007) The expression of Argonaute2 and
15:509–524 related microRNA biogenesis proteins in nor-
145. Olejniczak M, Kotowska-Zimmer A, Krzyzo- mal and hypoxic trophoblasts. Mol Hum
siak W (2018) Stress-induced changes in Reprod 13:273–279
miRNA biogenesis and functioning. Cell 154. Davis BN, Hilyard AC, Lagna G et al (2008)
Mol Life Sci 75:177–191 SMAD proteins control DROSHA-mediated
146. Emde A, Hornstein E (2014) miRNAs at the microRNA maturation. Nature 454:56–61
interface of cellular stress and disease. EMBO 155. Aguilar OA, Hadj-Moussa H, Storey KB
J 33:1428–1437 (2016) Regulation of SMAD transcription
147. Kucherenko MM, Shcherbata HR (2018) factors during freezing in the freeze tolerant
miRNA targeting and alternative splicing in wood frog, Rana sylvatica. Comp Biochem
the stress response - events hosted by Physiol B Biochem Mol Biol 201:64–71
membrane-less compartments. J Cell Sci 156. Semenza GL (2012) Hypoxia-inducible fac-
131:jcs202002 tors in physiology and medicine. Cell
148. Ho JJD, Metcalf JL, Yan MS et al (2012) 148:399–408
Functional importance of Dicer protein in 157. Wu C, So J, Davis-Dusenbery BN et al (2011)
the adaptive cellular response to hypoxia. J Hypoxia potentiates microRNA-mediated
Biol Chem 287:29003–29020 gene silencing through posttranslational
149. Yang Q, Li W, She H et al (2015) Stress modification of Argonaute2. Mol Cell Biol
induces p38 MAPK-mediated phosphoryla- 31:4760–4774
tion and inhibition of drosha-dependent cell 158. Zeng Y, Sankala H, Zhang X et al (2008)
survival. Mol Cell 57:721–734 Phosphorylation of Argonaute 2 at serine-
150. Stoecklin G, Kedersha N (2013) Relationship 387 facilitates its localization to processing
of GW/P-bodies with stress granules. Adv bodies. Biochem J 413:429–436
Exp Med Biol 768:197–211 159. Murchison EP, Partridge JF, Tam OH et al
151. Caruso P, MacLean MR, Khanin R et al (2005) Characterization of Dicer-deficient
(2010) Dynamic changes in lung microRNA murine embryonic stem cells. Proc Natl
profiles during the development of pulmo- Acad Sci U S A 102:12135–12140
nary hypertension due to chronic hypoxia 160. Tang X, Wen S, Zheng D et al (2013) Acety-
and monocrotaline. Arterioscler Thromb lation of Drosha on the N-terminus inhibits
Vasc Biol 30:716–723 its degradation by ubiquitination. PLoS One
152. Kozakowska M, Ciesla M, Stefanska A et al 8:e72503
(2012) Heme oxygenase-1 inhibits myoblast 161. Winter J, Jung S, Keller S et al (2009) Many
differentiation by targeting myomirs. Anti- roads to maturity: microRNA biogenesis
oxid Redox Signal 16:113–127 pathways and their regulation. Nat Cell Biol
11:228–234
Chapter 17
Computational and Bioinformatics Methods for MicroRNA

Gene Prediction
Ege Riza Karagur, Sakir Akgun, and Hakan Akca
Abstract
MicroRNAs (miRNAs) are 20–24-nucleotide-long noncoding RNAs that bind to the untranslated region
(30 UTR) of their target mRNAs. The importance of miRNAs in medicine has grown rapidly in the 20 years
since the discovery of them. As the regulatory function of miRNAs on biological processes was discovered,
they were advocated to play a role in the underlying mechanisms of human pathogenesis. Functional studies
have confirmed that miRNAs are promising in preclinical development through deregulation of genes
targeted by miRNAs in many cancer cases. In this chapter, we summarize the miRNAs identified for some
specific types of cancer and their functions. Besides, miRNAs function as cancer biomarker and their
benefits to diagnosis and treatment of cancer are also discussed.
Key words Biomarkers, Cancer, microRNAs, Exosome, Circulating miRNAs
1 Introduction
Cancer is a leading cause of death, globally. However, some

advances have been made in the management of cancer treatment.
Early detection of cancer is important reducing recurrence and
mortality of patients. Early detection consists of two stages: early
diagnosis and screening. To monitor and diagnose cancer, there are
lots of techniques including computerized tomography (CT) scan,
magnetic resonance imaging (MRI), positron emission tomogra-
phy (PET) scan, radiography, ultrasound, and X-ray. Screening and
diagnosis are routinely supported by histopathological examina-
tions of resected tumor tissue and biopsy. Despite the development
of new techniques for early detection of disease, the survival rates of
patients with cancer remains poor. Even detection with these diag-
nosis applications, tumor has locally outgrown and spread to lymph
nodes (stage III) or metastasized (stage IV) at diagnosis, and hence,
the disease could not be effectively treated due to the late diagnosis
[1]. Therefore, the identification of reliable and effective strategies
349
350 Ege Riza Karagur et al.
for early diagnosis of cancer is crucial to promote the survival rates

of patients.
In this view, biomarkers present an advantage for early diagno-
sis of cancer. A biomarker can provide assessment of cancer patients
including diagnosis and prediction of progression and response to
therapy. In addition, it has been discovered that working with
multiple biomarkers plays an important role in the treatment of
cancer patients. Traditional tumor biomarkers circulating in blood
and serum are proteins, sugars, and glycans. These blood-based
biomarkers provide great opportunity for screening of cancer with-
out injury for patients. Although there are traditional blood-based
tumor biomarkers such as CA15-3 for breast cancer, alpha-
fetoprotein (AFP) for liver cancer, CA19-9 for pancreatic cancer,
CA-125 for ovarian cancer, carcinoembryonic antigen for colorec-
tal cancer, and prostate specific antigen (PSA) for prostate cancer,
many of them exhibit low sensitivity and have potential of false
negative/positive rates [2]. The discovery of new and specific bio-
markers that can be used to diagnose a disease, screen tumor
progression, and predict therapy response as well as predict patient
survival, would contribute to obviate these challenges and improve
outcomes for cancer patients.
miRNAs, small noncoding RNA molecules with 18–25 nucleo-
tides in length, have opened an important path to understanding
cell biology and pathology since their discovery. miRNAs may
function as both oncogenes and tumor suppressor in carcinogenesis
[3]. Numerous studies have demonstrated that miRNAs are abnor-
mally expressed in many diseases including cancer. Aberrant expres-
sion of miRNAs was found in tissues and serum of patients with
malignant tumors [4]. Aberrant miRNA expression and function
significantly impact the prognosis and pathogenesis of cancer.
miRNAs are stably present in a detectable manner not only in
tumor tissue but also in body fluids including bloodstream, sputum,
breast milk, cerebral spinal fluid, and urine. Thus, they can be used
as prognostic and diagnostic biomarkers for cancer using a noninva-
sive to minimally invasive approach, allowing clinicians to take
repeated samples from patients. This enables circulating miRNAs
during cancer progression and treatment to be monitored [5–9].
Circulating miRNAs are found in extracellular organelles such
as exosomes and microvesicles (MVs) or bound in protein complex
such as nucleophosmin (NPM1), argonaute 2 protein (AGO2), or
high-density lipoprotein (HDL) to prevent RNase digestion in
bloodstream [21]. The first circulating miRNA was discovered as
a biomarker obtained from serum of patients with diffuse large
B-cell lymphoma [10].
In this chapter, we present current studies concerning miRNAs
found as potential clinical biomarker for detection, diagnosis, prog-
nosis, and staging of cancer.
MicroRNAs as Biomarkers in Cancer 351
2 Lung Cancer
Lung cancer is the most aggressive type of cancer with high inci-
dence, metastatic ability, and risk of recurrence. It is the most
common cause of cancer-related deaths. Lung cancer is mainly
classified into two histological groups: non–small-cell lung cancer
(NSCLC) representing 85% of cases and small cell lung cancer
(SCLC) representing 15% of cases. NSCLC is subdivided into
adenocarcinoma, squamous cell carcinoma and large cell carci-
noma. These subgroups accounts for 40%, 30% and 10% of lung
cancer cases, respectively [11]. The overall 5-year survival rates of
lung cancer patients in late stages are less than 20%, while 5-year
survival rates in the patients with lung cancer diagnosed at an early-
stage are approximately 70%. There are a lot of techniques includ-
ing chest radiography, computed tomography, and magnetic reso-
nance imaging to monitor and diagnose lung cancer
[12]. Unfortunately, the current techniques and methods are
often insufficient to detect early diagnosis of lung cancer reliably,
and therefore, most patients can only be diagnosed at advanced
stages of the disease. This leads to considerably decreased survival
rates and life quality. Previous studies have reported that miRNAs
may serve as diagnostic and prognostic biomarkers in lung cancer
[13–15]. Furthermore, miRNAs are stably found in paraffinized
cancer tissue, bloodstream, other body fluids such as sputum and
urine, allowing them to be obtained by minimal to noninvasive
methods in early diagnosis of lung cancer [5–7, 9].
The classification of lung cancer has routinely depended on the
histopathological examination of resected tumor tissue derived
from patients [15]. The examination of the resected tissues by
diagnostic molecular pathology approach is critical to decide the
personalized and targeted treatment strategies of lung cancer
patients [16]. Furthermore, available studies have suggested that
miRNA profiling in resected fresh or paraffinized tumor samples
could be useful to estimate relapse, survival, diagnosis and progno-
sis in lung cancer [17–19]. One study has identified a five-miRNA
signature in patients using quantitative polymerase chain reaction
(qPCR) and this five miRNA signature (let-7a, miR-221, miR-137,
miR-372, and miR-182) is correlated with survival and recurrence
risk in lung cancer patients (marked in red in Fig. 1). In the same
study, it has been also shown that lung cancer patients with a high-
risk score for this five microRNA signature in their resected fresh
tumor samples increased relapse potential and shortened survival
rates [18]. In another study, miRNA-seq, a next-generation
sequencing (NGS) technique to identify miRNA biomarkers, was
employed. miRNAs could be associated with lymph node metasta-
sis and poor survival in sixty-four resected frozen tissue specimens
from lung adenocarcinoma patients. It has been indicated that the
Fig. 1 An overview of miRNAs in the different stage and process of lung cancer. During the development of
lung cancer and therapy process, miRNAs have a role as diagnostic miRNA (blue rounded rectangle),
prognostic miRNAs (pink rounded rectangle), or miRNAs with multiple functions (diagnosis, prognosis,
prediction of therapy outcome, early-stage (green rounded rectangle), survival, recurrence (red rounded
rectangle), chemoresistance and metastasis (brown rounded rectangle))
elevated expression of miR-31 could be an indication for the pres-

ence of lymph node metastasis and correlated with adverse outcome
in lung cancer patients (marked in brown in Fig. 1) [20]. Profiling
miRNAs isolated from FFPE (formalin-fixed, paraffin-embedded)
specimens are a significant and alternative approach to predict the
prognosis of lung cancer patients, after it is used for histopatholo-
gical diagnosis [21, 22]. It has been reported that high levels of
miR-92a-2* from FFPE-SCLC tumor specimens are associated
with chemoresistance and with decreased survival in patients with
SCLC. SCLC is the most aggressive subtype of lung cancer and
often cannot be diagnosed at an early-stage in patients (marked in
brown in Fig. 1) [23, 24]. Furthermore, it has been indicated that
low expression levels of miR-34a derived from FFPE tumor tissue
could be used as a novel prognostic biomarker to estimate the
relapse risk of disease. Additionally, it is a useful tool to guide

treatment strategies regarding adjuvant therapy (marked in pink
in Fig. 1) [25].
Recently, several studies have reported that circulating miRNAs
derived from serum or plasma are becoming a potential clinical tool
in early detection, response to therapy, and screening of lung cancer
[13, 26]. It has been shown in a current study conducted using
eight miRNAs that miR-29c, miR-30b, miR-146a, and miR-205
serve as notable diagnostic tool in early-stage of NSCLC patients
(Area Under the ROC Curve (AUC) ¼ 0.96) with 96.05% sensi-
tivity and 82.98% specificity and in late stages of NSCLC patients
(AUC ¼ 0.95) with 94.23% sensitivity and 82.98% specificity when
used together [27]. It has been reported in another study in which
12 miRNAs were evaluated that miR-21, miR-126, miR-210, and
miR-486-5p are a powerful and reliable fingerprint to diagnose
NSCLC patients at stage I (early-stage) with 73.33% sensitivity
and 96.55% specificity (marked in green in Fig. 1) [28]. In a
previous study, small noncoding RNAs were sequenced using
Solexa which is a high-throughput, largely accurate, and reproduc-
ible methodology for sequencing. The sequences were used for
assessing potential miRNA biomarkers in lung cancer. Further-
more, this method could quantify and detect the expression levels
of miRNAs even if they were expressed at low levels. In this study,
highly altered ratios between miR-25 and miR-223 in the serum of
lung cancer patients compared to ones of healthy individuals were
found. By using Solexa, miR-25 and miR-223, which are found to
be specific signature for lung cancer diagnosis were also confirmed
in 152 cancer and 75 healthy volunteers by qPCR. It has been
shown that serum levels of miR-25 and miR-223 are dramatically
higher in lung cancer patients compared to healthy individuals,
indicating that circulating miR-25 and miR-223 are easily detect-
able and reliable biomarkers for lung cancer (marked in blue in
Fig. 1) [29]. Circulating miRNAs cannot only serve as diagnostic
biomarkers, but also might help to predict whether several perso-
nalized therapies in lung cancer are effective. It has been recently
demonstrated that high plasma levels of miR-30b and miR-30c in
epidermal growth factor receptor (EGFR)-mutated lung cancer
patients could reduce the sensitivity to EGFR tyrosine kinase inhi-
bitors, leading to poor outcomes and survival rates (marked in
brown in Fig. 1) [30].
Exosomal miRNAs released from tumor cells are more stable
than other circulating miRNAs bound to protein complex and
provide rich sources for the early diagnosis of lung cancer by a
nonsurgical approach from bodily fluids such as plasma, serum,
sputum, and urine [31]. Therefore, exosomal miRNAs are used as
a nonsurgical and highly sensitive diagnostic tool to improve the
early diagnosis of lung cancer and outcome of the patients. A recent
study performed miRNA-seq using tumor-derived exosomes
isolated from the plasma of 46 stage I NSCLC patients and

42 healthy volunteers. This study indicates that let-7, miR-21,
miR-24, and miR-486, which have previously been proven to
have diagnostic potential, could be used as a biomarker with an
AUC value of 0.899 for the early diagnosis of NSCLC patients.
Moreover, the same study showed that miR-181-5p, miR-30a-3p,
miR-30e-3p, and miR-361-5p are specific biomarkers for lung
adenocarcinoma, and that miR-10b-5p, miR-15b-5p, and
miR-320b are specific biomarkers for squamous lung carcinoma
with AUC values of 0.936 and 0.911, respectively (marked in
blue in Fig. 1) [31]. Furthermore, a study reported that circulating
exosomal miR-222-3p levels might be prognostic for estimating
the response to gemcitabine in NSCLC patients (marked in pink in
Fig. 1) [32].
Sputum is an easily obtainable body fluid and contains exfo-
liated bronchial epithelial cells. Although there is no evidence that
sputum specimens derived from the patients is an acceptable tool in
diagnosis of lung cancer, a few studies suggest that the profiling of
sputum-derived miRNAs might contribute to the diagnosis of lung
cancer [33]. Elevated expression of miR-21 was observed in 16 of
23 sputum specimens derived from lung cancer and in 0 of 17 spu-
tum specimens derived from healthy persons, referring to 69.66%
sensitivity and 100.00% specificity for diagnosis of lung cancer
(marked in blue in Fig. 1) [34]. Low-dose computer tomography
(LDCT) is one of the most important methods used in the early
diagnosis of lung cancer and contributes to the reduction of mor-
tality. Unfortunately, LDCT generates a lot of indeterminate soli-
tary pulmonary nodules (SPNs) in asymptomatic persons, while
only a small part of SPNs have the potential to be lung tumors,
causing overdiagnosis. It has been indicated that a sputum-derived
three miRNA signature (miR-21, miR-31, and miR-210) could be
used to diagnose early-stage lung cancer among SPNs with 82.93%
sensitivity and 87.84% specificity (marked in green in Fig. 1)
[35]. Furthermore, it has been thought that exosomal miR-7,
miR-21, miR-126, Let-7a, miR-17, and miR-19 derived from
bronchoalveolar lavage fluid (BAL) could have a clinical value in
early diagnosis of lung cancer [36]. A list of miRNAs identified as
potential biomarkers in lung cancer are given in Table 1.
3 Breast Cancer
Breast cancer (BC) is the most frequent type of cancer among

women, with an estimated 2.1 million new cases per year [37]. It
is also the primary cause of cancer death among women. Therefore,
studies on novel biomarkers for the prognostic, diagnostic, and
prediction have a great importance to BC. Up to now CA15.3
and BR27.29 biomarkers have been used for the diagnosis of
Table 1
miRNAs in lung cancer, their source and expression status
Diseases Source miRNAs Expression in Diseases References

Lung cancer Tissue Let-7a Upregulated [18]
miR-221 Downregulated
miR-137 Upregulated
miR-372 Upregulated
miR-182 Upregulated
miR-31 Upregulated [20]
miR-92-2a* Upregulated [23, 24]
miR-34a Downregulated [25]
Serum miR-29c Upregulated [27]
miR-30b Upregulated
miR-146a Upregulated
miR-205 Upregulated
miR-223 Upregulated
Plasma miR-21 Upregulated [28]
miR-126 Upregulated
miR-210 Upregulated
miR-486-5p Upregulated
miR-30b Upregulated [30]
miR-30c Upregulated
Let-7 Upregulated [31]
miR-24 Upregulated
miR-486 Upregulated
miR-181 Upregulated
miR-30a-3p Downregulated
miR-30e-3p Downregulated
miR-10b-5p Downregulated
miR-15-5p Downregulated
miR-320b Upregulated
miR-222-3p Upregulated [32]
Sputum miR-21 Upregulated [34, 35]
miR-210 Upregulated
Bronchoalveolar lavage fluid miR-7 Upregulated [36]
miR-21 Upregulated
miR-126 Upregulated
Let-7a Upregulated
miR-17 Upregulated
miR-19 Upregulated
BC. However, these biomarkers have not enough sensitivity to

detect BC for screening disease progression and for proper disease
management [38].
Circulating miRNAs have a big potential as diagnostic, prog-
nostic, and predictive biomarkers. The miRNAs have several advan-
tages such as resistance to endogenous RNase activity, sensitivity,
Fig. 2 An overview of miRNAs in the different stage and process of breast cancer. During the development of
breast cancer and therapy process, miRNAs have a role as predictive miRNA (aqua rounded rectangle), or
miRNAs with multiple functions (survival, (red rounded rectangle), metastasis (brown rounded rectangle))
and stability. Moreover, easy access from body fluids including

saliva, urine, breast milk, blood, and plasma without the require-
ment of a biopsy sample of the tumor enables them to be aa
valuable biomarker [29, 39, 40]. In the following, we describe
breast cancer examples associated with circulating miRNAs, focus-
ing on their biomarker potentials. Different investigator groups
have extensively studied circulating miRNAs as prognostic and
predictive biomarkers [41–43].
Freres and colleagues evaluated the expression of 188 circulat-
ing miRNAs in the plasma of 25 patients after and before Neoad-
juvant Chemotherapy (NAC) by using qPCR. They have shown
that miR-34a and miR-122 expressions were dramatically higher
than other miRNAs after NAC. The expression of mir-34a and
miR-122 were also evaluated in different status such as in tumor
tissue of patients with pathological partial response (pPR) to NAC,
in the plasma of patients with adjuvant chemotherapy (AC), and
plasma of patients who did not receive any chemotherapy. The
authors inferred that circulating miRNAs could be predictive bio-
markers in breast cancer in the light of these assessment (marked in
aqua in Fig. 2) [41]. Another study assessed the expression profiles
of extracellular vesicles and tissue miRNAs from patients with early-
stage breast cancer by miRNA sequencing. The analysis results
indicate that the expression levels of miR-24-2-5p, miR-375,
miR-376b-5p, miR-548b-5p, and miR-655-3p were similar in tis-
sue and extracellular vesicles when compared to the healthy group.
The high expression levels of miRNA-548b-5p and miRNA-376b-
5p and the low expression levels of miRNA-375 and miRNA-24-2-
5p are positively associated with survival (marked in red in Fig. 2)
[44]. A recent study explored the potential of five miRNAs as
metastatic biomarkers and prognostic factors, including miR-21
(related to metastasis), miR-23b, miR-190 (related to tumor dor-

mancy), miR-200b and miR-200c (related to epithelial mesenchy-
mal transition), in the plasma of patients with early and metastatic
breast cancer (MBC). Metastatic patients have higher expression
levels of miR-21, miR-23b, miR-200b and miR-200c than early
breast cancer. Also it has been shown that expression level of
miR-200b play a role as predictive biomarker for the overall survival
(OS) status in the human epidermal growth factor receptor
2 (HER2) negative subgroup (marked in aqua in Fig. 2) [45]. In
another study, the aim was to determine the expression levels of
miRNAs in the serum of patients with breast cancer and healthy
volunteers. The data of the study suggests that miR-25 and
miR-133 have high expression levels, whereas miR-17 has a low
expression in patients with breast cancer. When the expressions of
miRNAs were evaluated with clinical data, it could be observed that
high expression levels of miR-17 was related to increased tumor
size and poor survival. Low expression levels of miR-133 and
miR-25 were associated with clinical stage, metastasis, and survival
time of patients with breast cancer (marked in red and brown in
Fig. 2) [46].
4 Pancreatic Cancer
Pancreatic cancer (PC) is an aggressive malignancy with a high

mortality rate [47]. 80% of patients already have local invasions or
metastatic tumors at the time of diagnosis. Diagnosis of this cancer
is difficult because of the lack of symptoms and the complex anato-
mical position of the pancreas. The latter makes it difficult to
perform direct and invasive methods. Even so, patients who
undergo successful resection and adjuvant therapy have a 5-year
survival rates of approximately 20% [48]. Therefore, an early diag-
nosis and monitoring of cancer stage are very important for effec-
tive treatment and improved overall survival. New strategies such as
circulating tumor DNA, autoantibodies, epigenetics, and miRNAs
to the most effective treatment and early diagnosis are being con-
sidered. Therefore, miRNAs have a unique potential in terms of
their capacity to be stable, predictive, diagnostic, and prognostic
biomarkers.
In a study performed by Wang et al. in 2019, where they
investigated prognostic miRNA profiles in patients with pancreatic
adenocarcinoma (PAAD), miRNAs expressed differentially
between normal samples and PAAD tumor samples were identified
[49]. A four-miRNA signature was generated to predict overall
survival by using survival analysis for patients with PAAD. In addi-
tion, this study evaluated the potential function of the found miR-
NAs using a bioinformatic approach. Among the 17 abnormally
expressed miRNAs; miR-4772, miR-3613, miR-424 and miR-126
were associated with the overall survival of PAAD patients. These

four miRNAs as an independent prognostic agent in patients with
PAAD have been statically confirmed by performing multivariate
Cox’s proportional hazards regression analysis. Pathway enrich-
ment analyzes present that these four miRNAs may be associated
with genes affecting focal adhesion and some signaling pathways
including Wnt and PI3K-Akt. These findings emphasize that these
four miRNAs can be an important prognostic biomarker in pre-
dicting the survival of PAAD patients in the clinic. In another study,
the genome-wide expression of serum miRNAs in pancreas cancer
and biliary tract cancer (BTC) patients was assessed and new bio-
marker candidates were identified. Statistical analysis of the expres-
sion levels of miRNA in 55 patient shows significant differences
between cancer patients and healthy controls, but the two cancers
could not be distinguished using the miRNA panel. Three of the
miRNAs that show the highest differential expression were able to
distinguish cancer patients from controls with 92.7% accuracy.
Furthermore, the dysregulation of these three cancer specific miR-
NAs was analyzed by qPCR in an independent sample group. The
results showed miR-744-5p, miR-409-3p, and miR-128-3p were
three candidate miRNAs as potential therapeutic biomarkers for
pancreas cancer and also BTC diagnosis [50]. Mou et al., on the
other hand, investigated the expression levels of miR-345-5p using
qPCR in both tissues of patients with PC and in cell lines, and
found that miR-345-5p was deliberately overexpressed in cell lines
to evaluate its effect on proliferation and invasiveness in PC. To find
out the underlying mechanisms of the pathology, they used western
blotting. To investigate changes in PANC-1 cells’ skeleton in
response to modulation of miR-345-5p, immunofluorescence
staining was performed. Besides, to identify the target and regula-
tion mechanism of miR-345-5p, luciferase assays were used. The
findings indicate that expression of miR-345-5p was downregu-
lated in PDAC cells and tissues, while upregulation of miR-345-5p
inhibited proliferation and metastasis of PDAC cells. C-C motif
chemokine ligand 8 CCL8 is a direct target of miR-345-5p, and
expression of CCL8 was negatively correlated with expression of
miR-345 in PDAC. This suggests that CCL8 increases proliferation
and invasiveness of PDAC cells via NF-κB signaling pathway activa-
tion. The results showed that miR-345-5p inhibits PDAC progres-
sion via the NF-κB signal. Thus, miR-345-5p is thought to be a
potential biomarker for the treatment of PC patients [51]. In
another study, to determine the potential clinical prognostic value
of miR-1231, exosomes from volunteer individuals diagnosed with
PC, and from both pancreatic cancer cell lines (MIA PaCa-2,
PANC-1, SW1990, AsPC-1, and BxPc-3) and noncancer epithelial
cells (HPDE and human primary pancreatic epithelial cell), were
isolated, then miR-1231 expression levels were examined by qPCR.
The qPCR analysis revealed that the miR-1231 expression levels in
the plasma exosomes obtained from PC patients was significantly

lower than that in healthy controls. Exosomal miR-1231 levels
were significantly higher in patients without stage I–II, distant
metastasis, and lymph node metastasis than those with stage III–
IV, distant metastasis, and lymph node metastasis, respectively. In
addition, no correlation was found between exosomal miR-1231
expression and each of age, sex, smoking history, CA19-9 levels,
and tumor sites. Moreover, the exosomal miR-1231 expression
levels in pancreatic cancer cell lines was found to be significantly
lower than in normal pancreatic epithelial cells. These results were
clearly indicate that initiation of pancreatic cancer and it develop-
ment can be regulate by miR-1231 in both plasma of pancreatic
cancer patients and in the pancreatic cancer cell lines [52].Another
study performed in 2019 by Wu et al. evaluated the role of bone
marrow mesenchymal stem cell (BMSC) derived exosomal
miR-126-3p in PC. They primarily screened predicted miRNA
and associated genes related to PC. The regulatory roles of PC
cells were then investigated by transfecting PANC-1 cells with
miR-126-3p or silencing them with a disintegrin and a
metalloproteinase-9 (ADAM9). In addition, exosomes derived
from BMSCs and cocultured with PC cells to lighten the effects
of exosomes in PC were isolated. Subsequently, whether overex-
pression of miR-126-3p in BMSCs affected proliferation, migra-
tion, invasion, apoptosis, tumor growth, and metastasis of PC cells
in vivo was analyzed. Modulated miR-126-3p was suppressed by
ADAM9 which is downregulated in PC. In particular, it was
claimed that overexpression of miR-126-3p can inhibit prolifera-
tion, invasion, metastasis, and promotes apoptosis on PC cells
in vitro and in vivo conditions. Thus, the findings of the study
suggest that overexpressed miR-126-3p derived from BMSCs inhi-
bits the development of pancreatic cancer by downregulation of
ADAM9. This suggests miR-126-3p as a new potential therapeutic
biomarker for the diagnosis of PC [53]. In another study con-
ducted to investigate the underlying mechanism of the pathology
of PC and to understand the molecular mechanisms of miR-205,
microarray analysis was performed to examine the miRNA expres-
sion profiles, and miR-205 expression levels were confirmed by
qPCR in PC. To search the potential target genes of miR-205
and their potential functions, target prediction and functional
enrichment analysis was performed. Luciferase reporter assays fur-
ther confirmed the target sites of miR-205 and adenomatous poly-
posis choline (APC). Multiple miRNAs with aberrant expression in
PC were determined. The miR-205 was found to be significantly
upregulated and APC was found to be a target of miR-205 (also
downregulated) in PC. The authors also indicated that the results
of proliferation experiments demonstrated that miR-205 could
induce cell proliferation in PC by targeting APC [54]. A list of
miRNAs identified as potential biomarkers in PC are given in
Table 2.
Table 2
miRNAs in breast cancer, their source and expression status

Breast cancer Tissue miR-34a Upregulated [41]
miR-122 Upregulated
Serum miR-25 Upregulated [46]
miR-133 Upregulated
Plasma miR-458b-5p Upregulated [44]
miR-24-2–5p Downregulated
miR-23 Upregulated
miR-200c Upregulated
5 Colorectal Cancer
Colorectal cancer (CRC) is one of the most common malignant

tumors. Albeit preventable, it ranks third in cancer-related deaths
worldwide. The CRC incidence in males is higher compared to that
in females and increases in an age-dependent manner. Nowadays,
screening is reported to be implemented at earlier ages (i.e., age
40–50 years) considering the increasing presence of CRC cases at
younger than age 55 years [55]. More than 90% of colorectal
carcinomas are adenocarcinomas. The incidence and mortality
rate of CRC varies across the world. Like many other cancers, it
results in poor prognosis when diagnosis takes place in advanced
stage. Regular screening and the slow progression of tumor in
cancer precursor lesions can enable the detection of CRC at an
early-stage, although 60–70% of CRC is diagnosed at an advanced
stage in symptomatic patients who do not undergo screening tests
[56]. Early detection is of paramount importance since it reduces
the mortality and recurrence rate. With the recent technical devel-
opments, the screening methods of CRC are divided into two
groups, namely, the noninvasive methods including fecal, blood,
and radiological tests and the invasive methods including flexible
sigmoidoscopy (FS) and colorectaloscopy. In spite of their high
sensitivity and specificity, invasive methods, which are accepted as
the gold standard, are painful and expensive [57]. Stool-based tests
including fecal occult blood test (FOBT), fecal immunochemical
test (FIT), and the latest one, fecal DNA test, have a high rate of
false-positive results and low sensitivity in detecting preneoplasia
lesions.
The emergence and development of CRC occurs as a result of

genetic and epigenetic changes. These changes bring about expres-
sion differences in miRNAs between cancer and healthy neighbor-
ing tissues. These miRNAs with different expression profiles have
the potential to become novel biomarkers for the diagnosis, treat-
ment, and recurrence detection of CRC. Abnormal miRNA expres-
sion profiles are observed not only in the tumor tissues of CRC
patients but also in their stool and serum/plasma samples. Besides,
miRNAs, which have the potential of becoming biomarkers in
CRC, were first discovered in stool samples. miR-143 and
miR-145 expression levels were lower in the fecal samples taken
from CRC patients compared to healthy subjects [58].
The expression levels of miR-20a, miR-21, miR-106a,
miR-181b, and miR-203 were reported at higher levels in CRC
tumor tissues compared to normal tissues. Expression levels were
determined using the TaqMan qPCR method on CRC tumors and
normal tissue samples collected from 106 black and 239 white
subjects (marked in pink in Fig. 3) [59]. The results indicated
that high miR-181b expression was only associated with the poor
survival of stage III black patients. In line with these results, the
miRNAs were denoted to vary in CRC depending on the stage of
the disease and ethnicity [59]. The analyses made with RNA-seq
data obtained from the Cancer Genome Atlas (TCGA) demon-
strated that the expression profiles of miR-32, miR-181b,
miR-193b, miR-195, and miR-411 were significantly different
between the tumor tissues of T1 and T2 CRC patients without
lymph node metastasis and corresponding neighboring healthy
tissue samples (marked in brown in Fig. 3) [60]. Thus, these
miRNAs may be useful in detecting the presence of lymph node
metastasis from biopsy specimens taken before treatment, thereby
reducing the number of unnecessary operations.
MiRNAs have a high potential in determining resistance to
CRC treatment. miR-195-5p and miR-497-5p were detected to
be low in the tumor tissues of advanced CRC patients. Increasing
expressions of miR-195-5p and miR-497-5p in CRC cell lines was
found to enhance the sensitivity of drugs such as 5-FU and cisplatin
for antitumor treatment (marked in orange in Fig. 3)
[61, 62]. MiR-214, which is present in the blood and CRC cell
lines of CRC patients at significantly low levels, promotes radio-
sensitivity by inhibiting autophagy in CRC cells (marked in orange
in Fig. 3) [63]. In addition, miRNAs can be used as biomarkers in
detecting tumor node metastasis and determining the recurrence
potential of a tumor. A genome-wide study by Kandimalli et al.
indicates that miR-744, miR-429, miR-362, miR-200b, miR-191,
miR-30c2, miR-30b, and miR-33a are biomarkers to predict tumor
recurrence in both in silico discovery and clinical training for stage
II and III CRC patients with AUC values of 0.79 and 0.77 respec-
tively (marked in red in Fig. 3) [64]. Based on clinical results, they
suggested that these miRNAs, which were also validated in fresh

frozen tissues, could effectively stratify stage II and III CRC
patients into high- and low-risk groups [64].
Following the discovery of miRNAs, numerous studies were
performed to identify circulating miRNAs in CRC. MiR-17-3p and
miR-92a were the first miRNAs to be detected at significantly
higher levels in the plasma of CRC patients. They were ascertained
to have a potential for being noninvasive biomarkers with high
sensitivity and specificity for CRC [65]. These results were con-
firmed by many studies [66]. The meta-analyzes including
938 CRC patients and 638 healthy individuals revealed that
miR-17 is a potential biomarker for colorectal cancer (AUC ¼ 0.76,
sensitivity ¼ 0.75, specificity ¼ 68%) [67]. Recently, 95 miRNAs
were identified to be expressed differently according to the RNA
sequence results of small extracellular vesicles obtained from the
plasma samples of 15 early-stage CRC patients and 10 normal
control subjects [14]. By comparing the tissue results obtained
from TCGA data of 95 miRNAs, let-7b-3p, miR-139-3p,
miR-145-3p, and miR150-3p were selected, and these miRNAs
were confirmed in an independent group including 58 CRC
patients and 76 normal control subjects (marked in blue in Fig. 3)
[68]. A list of miRNAs identified as potential biomarkers in CRC is
given in Table 3.
6 Prostate Cancer
Prostate cancer is a type of malignant cancer in males. It is the

second main cause of cancer-related death in men after skin cancer
[69]. The diagnosis of the disease is mainly made by investigating
the presence of prostate specific antigen (PSA) levels in the blood
and/or via physical examination of the prostate gland (digital rectal
examination—DRE) [70]. However, it is known that the PSA
method is not sensitive enough to diagnose the disease
[71]. Today, prostate cancer treatment is performed with the
open, laparoscopic or the robotic assisted radical prostatectomy
methods [72]. Yet the prevention, diagnosis, and treatment of
prostate cancer remain controversial. Therefore, there is an urgent
need to find new noninvasive factors that can diagnose the disease
with high sensitivity, help with treatment, and predict postoperative
recurrence.
In a study targeted to investigate miRNA expressions in plasma
and tissue of prostate cancer patients, researchers observed that
patients with prostate cancer had high miR-130a-3p and
miR-150-5p expression levels and a low miR-148a-3p expression
level in both tissue and plasma samples. Evaluation of the results by
the Delta Delta CT method confirmed that miR-150 has potential
to be a biomarker. The circulating miRNA displayed fold changes
Fig. 3 An overview of miRNAs in the different stage and process of colorectal cancer. During the development
of colorectal cancer and therapy process, miRNAs have a role as diagnostic miRNA (blue rounded rectangle),
prognostic miRNAs (pink rounded rectangle), or miRNAs with multiple functions (prediction of therapy outcome
chemosensitivity and radiosensitivity (orange rounded rectangle), recurrence (red rounded rectangle), metas-
tasis (brown rounded rectangle))
of 2.697 and 1.693 in tissue and plasma, respectively ( p < 0.001)

[69]. Moya et al. have identified four miRNAs: miR-4289,
miR-326, miR-152-3p, and miR-98-5p that increased significantly
in plasma from patients with prostate cancer compared with control
groups. The panel they used demonstrated high specificity and
sensitivity in the diagnosis of prostate cancer (AUC ¼ 0.88). Fur-
ther analysis on the safety of these four miRNAs as biomarkers for
prostate cancer is being conducted. Many studies found that these
miRNAs are downregulated and are often related to cancer pro-
gression and poorer overall survival. In summary, the prognostic
potential of these four miRNAs should be further investigated
[73]. Zhang et al. observed that the expression levels of
miR-515-5p were significantly reduced in prostate cancer tissues
Table 3
miRNAs in colorectal cancer, their source and expression status

Colorectal cancer Tissue miR-21 Upregulated [59]
miR-106a Upregulated
miR-203 Upregulated
miR-195-5p Downregulated [61, 62]
miR-429 Upregulated
miR-362 Upregulated
miR-191 Upregulated
miR-30c2 Upregulated
miR-30b Upregulated
miR-33a Upregulated
Plasma miR-214 Upregulated [63]
miR-17-3p Upregulated [65]
miR-92a Upregulated
Let-7b-3p Downregulated [68]
Fecal miR-143 Downregulated [58]
and cancer cell lines compared to matched control groups.

MiR-515-5p levels are inversely correlated with the thyroid hor-
mone receptor interactor 13 (TRIP13) mRNA and also its protein
expression levels in prostate cancer tissues. The TRIP13 30 -UTR is
targeted by miR-515-5p directly, explaining reduced expression for
this mRNA. MiR-515-5p could modulate cell proliferation, migra-
tion and invasion by targeting TRIP13. Therefore, the potential of
miR-515-5p to serve as a tumor suppressor was investigated. Sur-
vival analysis showed that cases of prostate cancer with high
miR-515-5p expression display a better prognosis than patients
with low miR-515-5p expression. In conclusion, this study empha-
sizes the clinical role of miR-515-5p as a tumor biomarker in
prostate cancer [74].
7 Melanoma Cancer
Melanoma, the most malignant and highly fatal form of skin cancer,
is known to have a poor prognosis [75]. In the last decades of the
twentieth century, the incidence of melanoma rose with an increase
of 3–7% per year [76]. Despite recent therapeutic progress in

melanoma treatment, early diagnosis remains challenging. At the
same time, early diagnosis appears to be the main aspects of mela-
noma treatment [77]. Therefore, it is urgent to find new therapeu-
tic biomarkers, especially for the early detection of the disease. In
their study, Xin et al. examined the miRNA expression profiles of
80 uveal melanoma (UM) patients according to the cancer genome
atlas (TCGA) database, a total of 204 miRNAs were examined by
the Cox proportional hazard regression model in their training set
[78]. Then, they investigated potential miRNA target genes using
bioinformatics analysis. As a result, there are nine miRNAs that can
divide UM patients into high and low risk groups, provided they are
independent of each other: miR-7702, miR-4709, miR-873,
miR-513c, miR-452, miR-365b, miR-365a, miR-224, and
miR-195. Furthermore, the ROC analysis confirmed that patients
in the low-risk group had significantly longer survival than those in
the high-risk group. When they performed Gene Ontology
(GO) analysis, they observed that the estimated target genes of
these miRNAs are involved in the regulation of protein processing
and expression. In addition to these results, Kyoto Gene and
Genome Encyclopedia (KEGG) analysis showed that these target
genes are involved in multiple signaling pathways associated with
carcinogenesis. In conclusion, in this study, the authors concentrate
on nine miRNAs (miR-7702, miR-4709, miR-873, miR-513c,
miR-452, miR-365b, miR-365a, miR-224, and miR-195) and
these miRNAs were found to be tumor-specific biomarker panel
[78]. In a study performed by Kang et al. showed that miR-326 is
dramatically downregulated both in melanoma cells and melanoma
tissues. By using bioinformatics analysis, Kristen rat sarcoma 2 viral
oncogene homolog (KRAS) was determined as a potential target of
miR-326 and luciferase reporter assay confirmed this prediction.
Besides, a negative correlation between decreased KRAS expression
in melanoma tissues and miR-326 expression was found. The
authors reported that they were able to significantly reverse the
antitumor effects caused by overexpression of miR-326 in mela-
noma cells by alteration of KRAS expression. Additionally, they
found that miR-326 blocks the AKT and ERK signaling pathways
in melanoma and demonstrated that miR-326 acts as a tumor
suppressor in melanoma by targeting KRAS. Therefore, miR-326
may be a promising therapeutic tool that can be used for melanoma
patients [79]. In another study conducted by Shi et al. in 2019,
which aimed to understand the function and molecular mechanism
of miR-22 in melanoma, it was shown by qPCR analysis that
miR-22 expression was significantly reduced in melanoma tissues.
Decreased miR-22 expression in melanoma tissues was correlated
with adverse clinical prognosis and poor prognosis. It was reported
that miR-22 can be a tumor suppressor in the progression of
melanoma and may be a novel therapeutic target and prognostic
biomarker in the treatment of melanoma [80]. In another study

with exosomes isolated from 30 melanoma and 30 healthy indivi-
duals, five miRNAs, miRNA-532-5p, miRNA-210, miRNA-200c,
miRNA-199a-5p, and miRNA-106b, were examined by qPCR.
These miRNAs were investigated in serum samples of 95 healthy
volunteers and 95 melanoma patients both individually and in
combination. Furthermore, in a blind test of 25 healthy individuals
and 25 melanoma patients, 22 healthy individuals with 88.0%
sensitivity and 23 melanoma patients with 92.0% sensitivity were
identified via this miRNA panel. The exosomal miRNA panel used
in the study differentiated between patients with and without
metastasis, patients with stages I and II of the disease from patients
with stages III and IV as well as untreated patients receiving pem-
brolizumab treatment. In conclusion, the authors claim that exo-
somal miRNAs in serum, particularly exosomal miRNA-532-5p
and exosomal miRNA-106b, may be used for the diagnosis of
melanoma. In addition to tissues, there are also studies with miR-
NAs in the serum of melanoma patients [81]. For example, in one
study, after upregulation of miR-203 in the serum of melanoma
patients and also upregulation of melanospheres in the model of
stem cells, overexpression of melanoma cells on stemness potential
and migration ability was examined. The investigators found similar
expression of miR-203 in serum and melanoma stem cells, but the
opposite in tumor melanospheres. In conclusion, the researchers
suggest that miR-203 is downregulated in melanoma tissues and
that overexpressed miR-203 can be a therapeutic target for early
detection of metastasis [82].
8 Endometrial Cancer
Human endometrial cancer (EC) is a gynecologic malignancy that

is the most common in developed countries [83]. There are two
subgroups which consist of type I (estrogen-dependent) and type II
(non–estrogen-dependent) EC [84]. Worldwide 380,000 new
cases are diagnosed each year [85]. Although it is benign in
women of reproductive age, it is associated with primary or second-
ary infertility in 30% of these women [86]. Physicians have used
different therapeutic ways such as ultrasonography, surgical resec-
tion, radiotherapy, brachytherapy, and chemotherapy in the process
of diagnosis and treatment of EC, but these methods are inade-
quate for diagnosis. Therefore, urgent new therapeutic targets are
needed for diagnosis and treatment. Previous studies have shown
that miRNAs have diagnostic and prognostic biomarker potential
for EC [84, 87].
In the literature, miR-200c, a member of the miR-200 family, is

known to be pivotal in tumor progression. In a study conducted by
Wilczynski et al. in 2018, firstly miR-200c expression levels were
confirmed using real time PCR [88]. 90 archived formalin-fixed
paraffin-embedded FFPE tissue samples from endometrioid endo-
metrial cancer (EEC) and 10 normal endometrium samples were
used. miR-200c expression levels were found to be higher in EEC
samples compared to controls. miR-200c expression was evaluated
according to the stages of EEC and was significantly lower in later
stages of EC compared to early-stages. Considering these results,
different expression levels of miR-200c may be correlated with
clinicopathological characteristics and survival rates [88]. Wu
et al. evaluated the significance of miRNA biomarkers in EC by
investigating miRNA data from GSE35794 and the TCGA data-
base. Using unpaired t-test in limma package in R [84], they
identified differentially expressed miRNAs (DE-miRNAs) and
determined their potential targets using miRWalk. The target func-
tional enrichment assays and protein–protein interaction network
were then visualized using Cytoscape. According to these data,
EEC specimens detected 24 overlapping DE-miRNAs when com-
pared nonendometrioid endometrial specimens. They established a
miRNA-target regulatory network for 11 upregulated miRNAs
including miR-200a, miR-200b, and miR-200c, as well as five
downregulated miRNAs including miR-449a, miR-145-5p and
miR-145-3p [84]. As a result, the researchers predicted that
lymphocyte-enhancing factor-1 (LEF1) is a target of miR-449a
and that SRY-Box Transcription Factor 11 (SOX11) is a target of
miR-145-5p. In addition, survival analysis showed that miR-449a,
miR-145-5p, and LEF1 were associated with the overall survival
prognosis of EC patients. These data suggest that the downregula-
tion of miR-449a and miR-145-5p may have a role in the clinical
pathogenesis of EC. Therefore, the aforementioned miRNAs may
serve as prognostic biomarkers for EC patients [84].
In another study, which aimed to understand the effect and
related mechanism of miR-582-5p on proliferation and apoptosis
of endometrial carcinoma, miR-582-5p and v-akt murine thymoma
viral oncogene homolog 3 (AKT3) expression was detected in
human tissue samples and cells using qPCR [89]. In this study,
cell proliferation was evaluated by CyQuant and 2D colony tests,
and protein expression was determined by Western blot. Targets of
miR-582-5p and mir-582-5p were identified by luciferase reporter
assay, and Annexin V was used to determine cell apoptosis. As a
result of these experiments, the authors found that miR-582-5p
expression was significantly decreased in tissues with human endo-
metrial carcinoma and that upregulation of mi-582-5p in human
endometrial carcinoma cells inhibited cell proliferation and induced
apoptosis. Furthermore, miR-582-5p was confirmed to target
AKT3, and AKT3 expression was found to be negatively correlated
with miR-582-5p expression. Upregulation of AKT3 eliminates the

effect of miR-582-5p; therefore, miR-582-5p regulates apoptosis
and cell proliferation via AKT3 in human endometrial carcinoma.
This suggests miR-582-5p as a potential biomarker that requires
further investigation in human endometrial carcinoma [89].
In another study that investigated the High Mobility Group
AT-hook 2 (HMGA2) gene and related miRNAs, primarily, the
binding sites of miR-302a-5p and miR-367-3p in HMGA2
mRNA were identified using bioinformatics [90]. The miRNA
expression levels were further determined by qPCR and confirmed
by using in situ hybridization experiments. In addition, HMGA2
and epithelial mesenchymal transition (EMT)-related protein levels
were evaluated by immunohistochemistry analysis. Consequently,
the roles of miR-302a-5p and miR-367-3p in HMGA regulation in
EC were investigated in vitro. Afterwards, 40 EC, 37 normal endo-
metrial tissues and 99 FFPE tissues including 80 EC and 19 normal
endometrial tissues were used. Patients with liver and kidney disease
and patients who did not receive preoperative radiotherapy and
chemotherapy were excluded. The levels of miR-302a-5p and
miR-367-3p were also investigated in these tissues, and their
expression was consistently low. As a result, the researchers identi-
fied two new abnormally expressed miRNAs, miR-302a-5p and
miR-367-3p, which change the malignant state of endometrial
carcinoma cells by suppression of HMGA2 expression. It is antici-
pated that potential therapies targeting the miRNA/HMGA2 axis
should be explored as new strategies for treating EC. Thus, it is
recommended that potential therapies targeting the miRNA/
HMGA2 axis should be studied in depth [90].
In a study carried out by Roman-Canal et al. in 2019, a new
approach was developed to clarify sensitive and specific biomarkers
in endometrial carcinoma [91]. In this study, extracellular vesicles
(EVs) of peritoneal lattices and acidic fluids were isolated from
25 EC patients and 25 control patients for comparison. The Taq-
man OpenArray technology was used to detect expression of
371 EV-associated miRNAs. Finally, the study demonstrated that
114 miRNAs were significantly dysregulated in EC patients. Eight
of them are miR-10b-5p, miR-34b-3p, miR-34c-3p, miR-34c-5p,
miR-200b-3p, miR-383-5p, miR-449b-5p, and miR-2110
[91]. These findings may pave the way for the use of peritoneal
lavages of EV-associated miRNAs as a source of biomarkers which
have not been used for EC yet. In another study, which aimed to
investigate the function and the underlying molecular mechanism
of miR-326 in EC, miR-326 expression in EC tissues and cell lines
was assessed by qPCR, and TWIST and EMT-related protein levels
were examined [92]. The results obtained using a dual-lucifer
reporter system showed that TWIST was a direct target of
miR-326. Furthermore, in vitro experiments revealed that knock-
down of the Twist-related protein 1 (TWIST1) gene inhibited EC
cell migration, invasion, and EMT. Finally, miR-326 levels demon-

strated the tumor suppressor role by targeting TWIST1. As a result,
the researchers demonstrated that miR-326 acts as a tumor sup-
pressor by targeting TWIST1 and that it can be used as a new
biomarker for patients with endometrial cancer [92].
9 Discussion
Numerous studies indicated that miRNAs are expressed at different

levels in malignant tissues compared to normal tissues. This dissim-
ilarity is quite convenient for clinical diagnosis. The differential
expression of miRNAs found in body fluids and tissues renders
their use as biomarkers in the diagnosis of cancer appealing.
Screening using miRNAs is economical compared to other
techniques, less invasive, and does not require exposure to risk
factors such as radiation. These are the advantages of using miRNAs
derived from body fluids as biomarkers for the determining diag-
nosis and prognosis of cancer. Also, miRNAs in body fluids escape
from degradation, since they are found in vesicles and exosomes.
Although the stability of miRNAs obtained from tissue and body
fluids promotes their potential to be used as biomarkers, it con-
duces several disadvantages. Some miRNAs were validated in vari-
ous malignancies for prognostic and diagnostic purposes in various
studies. However, the majority of these studies did not take the
factors such as pretreatment, age, and gender into consideration.
Furthermore, studies on circulating miRNAs have not yet pro-
duced a validated disease marker with high specificity. The expres-
sion levels of patient-specific miRNAs may show variation. The
factors behind these expression dissimilarities are not fully under-
stood. The miRNomes belonging to miRNAs in tissue and circula-
tion should be generated in larger and independent groups for each
cancer. In addition, miRNA detection and analysis methods should
be optimized. The lack of a stable internal reference gene for each
type of cancer hinders optimization. The aforementioned disadvan-
tages limit the potential of miRNAs to become good biomarkers.
We suggest that these problems can be overcome with developing
technology [93] and increasing scientific studies and thereby cer-
tainly increase the role of miRNAs in diagnosing and screening
diseases.
References
1. Neal RD et al (2015) Is increased time to diag- 16. Oberndorfer F, Müllauer L (2018) Molecular
nosis and treatment in symptomatic cancer pathology of lung cancer: current status and
associated with poorer outcomes? Systematic perspectives. Curr Opin Oncol 30(2):69–76
review. Br J Cancer 112(Suppl 1):S92–S107 17. Raponi M et al (2009) MicroRNA classifiers for
2. Diamandis EP (2012) The failure of protein predicting prognosis of squamous cell lung
cancer biomarkers to reach the clinic: why, cancer. Cancer Res 69(14):5776–5783
and what can be done to address the problem? 18. Yu S-L et al (2008) MicroRNA signature pre-
BMC Med 10(1):87 dicts survival and relapse in lung cancer. Cancer
3. Zhang B et al (2007) microRNAs as oncogenes Cell 13(1):48–57
and tumor suppressors. Dev Biol 302(1):1–12 19. Yanaihara N et al (2006) Unique microRNA
4. Wang J, Sen S (2011) MicroRNA functional molecular profiles in lung cancer diagnosis and
network in pancreatic cancer: from biology to prognosis. Cancer Cell 9(3):189–198
biomarkers of disease. J Biosci 36(3):481–491 20. Meng W et al (2013) MicroRNA-31 predicts
5. Florczuk M, Szpechcinski A, Chorostowska- the presence of lymph node metastases and
Wynimko J (2017) miRNAs as biomarkers survival in patients with lung adenocarcinoma.
and therapeutic targets in non-small cell lung Clin Cancer Res 19(19):5423–5433
cancer: current perspectives. Target Oncol 12 21. Skrzypski M et al (2014) Prognostic value of
(2):179–200 microRNA expression in operable non-small
6. Leidinger P, Keller A, Meese E (2012) Micro- cell lung cancer patients. Br J Cancer 110:991
RNAs—important molecules in lung cancer 22. Schuster C et al (2011) MicroRNA expression
research. Front Genet 2:104 profiling of specific cells in complex archival
7. Qin X et al (2015) The tumor cytosol miRNAs, tissue stained by immunohistochemistry. Lab
fluid miRNAs, and exosome miRNAs in lung Investig 91(1):157–165
cancer. Front Oncol 4:357 23. Ranade AR et al (2010) MicroRNA 92a-2*: a
8. Weber JA et al (2010) The microRNA spec- biomarker predictive for Chemoresistance and
trum in 12 body fluids. Clin Chem 56 prognostic for survival in patients with small
(11):1733–1741 cell lung cancer. J Thorac Oncol 5
9. Yu H et al (2018) Circulating microRNA bio- (8):1273–1278
markers for lung cancer detection in Western 24. Huang J, Peng J, Guo L (2015) Non-coding
populations. Cancer Med 7(10):4849–4862 RNA: a new tool for the diagnosis, prognosis,
10. Lawrie CH et al (2007) MicroRNA expression and therapy of small cell lung cancer. J Thorac
distinguishes between germinal center B cell- Oncol 10(1):28–37
like and activated B cell-like subtypes of diffuse 25. Gallardo E et al (2009) miR-34a as a prognos-
large B cell lymphoma. Int J Cancer 121 tic marker of relapse in surgically resected non-
(5):1156–1161 small-cell lung cancer. Carcinogenesis 30
11. Travis WD et al (2015) The 2015 World (11):1903–1909
Health Organization classification of lung 26. Bianchi F et al (2011) A serum circulating
tumors: impact of genetic, clinical and radiolo- miRNA diagnostic test to identify asymptom-
gic advances since the 2004 classification. J atic high-risk individuals with early stage lung
Thorac Oncol 10(9):1243–1260 cancer. EMBO Mol Med 3(8):495–503
12. Khalil A et al (2016) Contribution of magnetic 27. Yang X et al (2019) Serum microRNA signa-
resonance imaging in lung cancer imaging. ture is capable of early diagnosis for non-small
Diagn Interv Imaging 97(10):991–1002 cell lung cancer. Int J Biol Sci 15
13. Inamura K (2017) Diagnostic and therapeutic (8):1712–1722
potential of MicroRNAs in lung cancer. Can- 28. Shen J et al (2011) Plasma microRNAs as
cers 9(5):49 potential biomarkers for non-small-cell lung
14. Zhang Y et al (2018) A 5-microRNA signature cancer. Lab Investig 91(4):579–587
identified from serum microRNA profiling pre- 29. Chen X et al (2008) Characterization of micro-
dicts survival in patients with advanced stage RNAs in serum: a novel class of biomarkers for
non-small cell lung cancer. Carcinogenesis 40 diagnosis of cancer and other diseases. Cell Res
(5):643–650 18(10):997–1006
15. Del Vescovo V et al (2014) MicroRNAs as lung 30. Hojbjerg JA et al (2019) Circulating miR-30b
cancer biomarkers. World J Clin Oncol 5 and miR-30c predict erlotinib response in
(4):604–620
<em>EGFR</em>mutated non-small cell 44. Yan C et al (2019) Plasma extracellular vesicle-

lung cancer patients. Lung Cancer 135:92–96 packaged microRNAs as candidate diagnostic
31. Jin X et al (2017) Evaluation of tumor-derived biomarkers for earlystage breast cancer. Mol
Exosomal miRNA as potential diagnostic bio- Med Rep 20(5):3991–4002
markers for early-stage non–small cell lung can- 45. Papadaki C et al (2019) Circulating miRNAs as
cer using next-generation sequencing. Clin a marker of metastatic disease and prognostic
Cancer Res 23(17):5311–5319 factor in metastatic breast cancer. Oncotarget
32. Wei F et al (2017) Exosomes derived from 10(9):966–981
gemcitabine-resistant cells transfer malignant 46. Hesari A et al (2018) Expression of circulating
phenotypic traits via delivery of miRNA-222- miR-17, miR-25, and miR-133 in breast cancer
3p. Mol Cancer 16(1):132 patients. J Cell Biochem
33. Hubers AJ et al (2013) Molecular sputum anal- 47. Kaur S et al (2012) Early diagnosis of pancre-
ysis for the diagnosis of lung cancer. Br J Can- atic cancer: challenges and new developments.
cer 109(3):530–537 Biomark Med 6(5):597–612
34. Xie Y et al (2010) Altered miRNA expression in 48. Sirri E et al (2016) Recent trends in survival of
sputum for diagnosis of non-small cell lung patients with pancreatic cancer in Germany and
cancer. Lung Cancer 67(2):170–176 the United States. Pancreas 45(6):908–914
35. Xing L et al (2015) Sputum microRNA bio- 49. Wang ZX, Deng TX, Ma Z (2019) Identifica-
markers for identifying lung cancer in indeter- tion of a 4-miRNA signature as a potential
minate solitary pulmonary nodules. Clin prognostic biomarker for pancreatic adenocar-
Cancer Res 21(2):484–489 cinoma. J Cell Biochem 120
36. Kim JE et al (2018) Diagnostic value of micro- (10):16416–16426
RNAs derived from exosomes in bronchoal- 50. Kim K et al (2019) Identification of potential
veolar lavage fluid of early-stage lung biomarkers for diagnosis of pancreatic and bili-
adenocarcinoma: a pilot study. Thorac Cancer ary tract cancers by sequencing of serum micro-
9(8):911–915 RNAs. BMC Med Genet 12(1):62
37. Bray F et al (2018) Global cancer statistics 51. Mou T et al (2019) MiR-345-5p functions as a
2018: GLOBOCAN estimates of incidence tumor suppressor in pancreatic cancer by
and mortality worldwide for 36 cancers in directly targeting CCL8. Biomed Pharmac-
185 countries. CA Cancer J Clin 68 other 111:891–900
(6):394–424 52. Chen SL et al (2019) Clinical significance of
38. Chan M et al (2013) Identification of circulat- exosomal miR-1231 in pancreatic cancer.
ing microRNA signatures for breast cancer Zhonghua Zhong Liu Za Zhi 41(1):46–49
detection. Clin Cancer Res 19(16):4477–4487 53. Wu DM et al (2019) Bone marrow mesenchy-
39. Gilad S et al (2008) Serum microRNAs are mal stem cell-derived Exosomal MicroRNA-
promising novel biomarkers. PLoS One 3(9): 126-3p inhibits pancreatic cancer development
e3148 by targeting ADAM9. Mol Ther Nucleic Acids
40. Mitchell PS et al (2008) Circulating micro- 16:229–245
RNAs as stable blood-based markers for cancer 54. Qin RF et al (2019) MiR-205 mediated APC
detection. Proc Natl Acad Sci U S A 105 regulation contributes to pancreatic cancer cell
(30):10513–10518 proliferation. World J Gastroenterol 25
41. Freres P et al (2015) Neoadjuvant chemother- (28):3775–3786
apy in breast cancer patients induces miR-34a 55. Siegel RL et al (2017) Colorectal cancer inci-
and miR-122 expression. J Cell Physiol 230 dence patterns in the United States, 1974-
(2):473–481 2013. J Natl Cancer Inst 109(8):djw322
42. Saleh AA et al (2019) Potential value of circu- 56. Maida M et al (2017) Screening of colorectal
latory microRNA122 gene expression as a cancer: present and future. Expert Rev Anti-
prognostic and metastatic prediction marker cancer Ther 17(12):1131–1146
for breast cancer. Mol Biol Rep 46 57. Issa IA, Noureddine M (2017) Colorectal can-
(3):2809–2818 cer screening: an updated review of the avail-
43. Uen Y et al (2018) Mining of potential micro- able options. World J Gastroenterol 23
RNAs with clinical correlation - regulation of (28):5086–5096
syndecan-1 expression by miR-122-5p altered 58. Li JM et al (2012) Down-regulation of fecal
mobility of breast cancer cells and possible cor- miR-143 and miR-145 as potential markers for
relation with liver injury. Oncotarget 9 colorectal cancer. Saudi Med J 33(1):24–29
(46):28165–28175
59. Bovell LC et al (2013) The prognostic value of with prostate cancer. Medicine (Baltimore) 98
MicroRNAs varies with patient race/ethnicity (36):e16705
and stage of colorectal cancer. Clin Cancer Res 73. Moya L et al (2019) Assessment of miR-98-5p,
19(14):3955–3965 miR-152-3p, miR-326 and miR-4289 expres-
60. Ozawa T et al (2018) A MicroRNA signature sion as biomarker for prostate cancer diagnosis.
associated with metastasis of T1 colorectal can- Int J Mol Sci 20(5):1154
cers to lymph nodes. Gastroenterology 154 74. Zhang X et al (2019) MiR-515-5p acts as a
(4):844–848.e7 tumor suppressor via targeting TRIP13 in
61. Poel D et al (2019) Proteomic analysis of prostate cancer. Int J Biol Macromol
miR-195 and miR-497 replacement reveals 129:227–232
potential candidates that increase sensitivity to 75. Sabarimurugan S et al (2018) Systematic
Oxaliplatin in MSI/P53wt colorectal cancer review and meta-analysis of the prognostic sig-
cells. Cells 8(9):1111 nificance of miRNAs in melanoma patients.
62. Pouliot LM et al (2012) Cisplatin sensitivity Mol Diagn Ther 22(6):653–669
mediated by WEE1 and CHK1 is mediated by 76. Xu Y et al (2020) Novel MicroRNA biomar-
miR-155 and the miR-15 family. Cancer Res kers, miR-142-5p, miR-550a, miR-1826, and
72(22):5945–5955 miR-1201, were identified for primary mela-
63. Hu JL et al (2018) Inhibition of ATG12- noma. J Comput Biol 27(5):815–824
mediated autophagy by miR-214 enhances 77. Mohammadpour A et al (2019) Melanoma:
radiosensitivity in colorectal cancer. Oncogen- where we are and where we go. J Cell Physiol
esis 7(2):16 234(4):3307–3320
64. Kandimalla R et al (2018) Genome-wide dis- 78. Xin X et al (2019) Identification of a nine-
covery and identification of a novel miRNA miRNA signature for the prognosis of uveal
signature for recurrence prediction in stage II melanoma. Exp Eye Res 180:242–249
and III colorectal cancer. Clin Cancer Res 24 79. Kang K et al (2018) MicroRNA-326 inhibits
(16):3867–3877 melanoma progression by targeting KRAS and
65. Ng EK et al (2009) Differential expression of suppressing the AKT and ERK signalling path-
microRNAs in plasma of patients with colorec- ways. Oncol Rep 39(1):401–410
tal cancer: a potential marker for colorectal 80. Shi L et al (2019) MicroRNA-22 targets
cancer screening. Gut 58(10):1375–1381 FMNL2 to inhibit melanoma progression via
66. Herreros-Villanueva M et al (2019) Plasma the regulation of the Wnt/beta-catenin signal-
MicroRNA signature validation for early detec- ing pathway and epithelial-mesenchymal tran-
tion of colorectal cancer. Clin Transl Gastroen- sition. Eur Rev Med Pharmacol Sci 23
terol 10(1):e00003 (12):5332–5342
67. Liu T et al (2018) A meta-analysis of 81. Tengda L et al (2018) Serum exosomal micro-
microRNA-17 as a potential biomarker in diag- RNAs as potent circulating biomarkers for mel-
nosis of colorectal cancer. Cell Mol Biol anoma. Melanoma Res 28(4):295–303
(Noisy-le-Grand) 64(6):86–93 82. Sahranavardfard P et al (2019) MicroRNA-203
68. Min L et al (2019) Evaluation of circulating reinforces stemness properties in melanoma
small extracellular vesicles derived miRNAs as and augments tumorigenesis in vivo. J Cell
biomarkers of early colon cancer: a comparison Physiol 234(11):20193–20205
with plasma total miRNAs. J Extracell Vesicles 83. Wang Q, Zhu W (2019) MicroRNA-873 inhi-
8(1):1643670 bits the proliferation and invasion of endome-
69. Paunescu IA et al (2019) Biomarker potential trial cancer cells by directly targeting
of plasma MicroRNA-150-5p in prostate can- hepatoma-derived growth factor. Exp Ther
cer. Medicina (Kaunas) 55(9):564 Med 18(2):1291–1298
70. Barcelo M et al (2019) Semen miRNAs 84. Wu X et al (2020) Downregulations of
contained in exosomes as non-invasive biomar- miR-449a and miR-145-5p act as prognostic
kers for prostate cancer diagnosis. Sci Rep 9 biomarkers for endometrial cancer. J Comput
(1):13772 Biol 27(5):834–844
71. Mottet N et al (2017) EAU-ESTRO-SIOG 85. Delangle R et al (2019) The use of microRNAs
guidelines on prostate cancer. Part 1: screen- in the Management of Endometrial Cancer: a
ing, diagnosis, and local treatment with cura- meta-analysis. Cancers (Basel) 11(6):832
tive intent. Eur Urol 71(4):618–629 86. Acien P, Velasco I (2013) Endometriosis: a
72. Bi CW et al (2019) Increased expression of disease that remains enigmatic. ISRN Obstet
miR-153 predicts poor prognosis for patients Gynecol 2013:242149
87. Giglio S et al (2019) miRNAs as candidate 91. Roman-Canal B et al (2019) EV-associated

biomarker for the accurate detection of atypical miRNAs from peritoneal lavage are a source
endometrial hyperplasia/endometrial intrae- of biomarkers in endometrial cancer. Cancers
pithelial neoplasia. Front Oncol 9:526 11(6):839
88. Wilczynski M et al (2018) Association of 92. Liu W et al (2017) miR-326 regulates EMT
microRNA-200c expression levels with clinico- and metastasis of endometrial cancer through
pathological factors and prognosis in endome- targeting TWIST1. Eur Rev Med Pharmacol
trioid endometrial cancer. Acta Obstet Gynecol Sci 21(17):3787–3793
Scand 97(5):560–569 93. Akgül B, Stadler PF, Hawkins LJ, Hadj-
89. Li L, Ma L (2018) Upregulation of miR-582- Moussa H, Storey KB, Ergin K, Çetinkaya R,
5p regulates cell proliferation and apoptosis by Paschoal AR, Nachtigall PG, Tutar Y, Yousef
targeting AKT3 in human endometrial carci- M, Allmer J (2021) 44 Current challenges in
noma. Saudi J Biol Sci 25(5):965–970 miRNomics. In: Allmer J, Yousef M (eds) miR-
90. Ma J et al (2018) miR-302a-5p/367-3p- Nomics: microRNA biology and computa-
HMGA2 axis regulates malignant processes tional analysis. Methods in molecular biology,
during endometrial cancer development. J vol 2257. Springer, New York
Exp Clin Cancer Res 37(1):19
Chapter 18
The Role of MiRNA in Cancer: Pathogenesis, Diagnosis,

and Treatment
Erez Uzuner, Gizem Tugçe Ulu, Sevim Beyza Gürler, and Yusuf Baran
Abstract
Cancer is also determined by the alterations of oncogenes and tumor suppressor genes. These gene
expressions can be regulated by microRNAs (miRNA). At this point, researchers focus on addressing two
main questions: “How are oncogenes and/or tumor suppressor genes regulated by miRNAs?” and “Which
other mechanisms in cancer cells are regulated by miRNAs?” In this work we focus on gathering the
publications answering these questions. The expression of miRNAs is affected by amplification, deletion or
mutation. These processes are controlled by oncogenes and tumor suppressor genes, which regulate
different mechanisms of cancer initiation and progression including cell proliferation, cell growth, apopto-
sis, DNA repair, invasion, angiogenesis, metastasis, drug resistance, metabolic regulation, and immune
response regulation in cancer cells. In addition, profiling of miRNA is an important step in developing a new
therapeutic approach for cancer.
Key words miRNA, Oncogenic miRNA, Tumor suppressor miRNA, miRNAs in proliferation,
miRNAs in growth, miRNAs in apoptosis, miRNA-based cancer treatment
1 Introduction
Cancer is uncontrolled cell division that is caused by internal and

external factors including chemicals, environmental toxins, radia-
tion, viruses, and genetic abnormalities [1]. Although it is known
that various factors can cause cancer, there remain many innumera-
ble open questions regarding the causes of many cancers. As
reported by various studies, miRNAs are one of the players in
cancer initiation and progression. The miRNAs are small noncod-
ing RNAs that can regulate gene expression [2]. Expression levels
of certain miRNAs could be different between cancerous and
healthy cells, which serves as an important tool for classification,
diagnosis, and treatment of cancer [3].
The miRNA profiling is used to characterize cancer-specific
miRNAs to offer a new perspective on their role in cancer
[4]. This information not only opens doors for new treatments
375
376 Erez Uzuner et al.
but is also used in diagnosis and prognosis. In addition, high-

throughput technology in miRNA expression level profiling leads
to the discovery of novel cancer-associated miRNA biomarkers.
The newly discovered biomarkers will help control and monitor
cancer stages [5]. On the other hand, miRNAs themselves are used
in cancer treatment. For this purpose, small molecule inhibitors to
block biogenesis and function of miRNAs and nanocarriers to use
targeted drugs delivery into cancer are used for miRNA replace-
ment or inhibition [6, 7]. This novel knowledge promises to
improve treatment approaches and perspective in preclinical and
clinical studies of cancer.
2 MicroRNAs in Cancers
MicroRNAs (miRNAs) play important roles in the regulation of

gene and protein expression levels at the posttranscriptional and
translational level [8, 9]. The expression level of miRNAs becomes
balanced to provide cellular homeostasis in biological processes
[10]. Abnormalities in the transcriptional control of miRNAs,
amplification or deletion of miRNAs causes epigenetic changes
and may induce upregulation/downregulation of miRNAs
[11, 12]. The miRNA biogenesis consists of a cascade of processes
that manage the regulation of miRNAs and their maturation
[13]. The miRNA biogenesis and activity are mediated by transcrip-
tion and maturation of miRNAs via multi-protein complexes
[14]. Mutation or dysregulation of miRNA biogenesis causes for-
mation of oncogenesis [15]. Hence, profiling the expression levels
of miRNAs is used to determine cancer progression [16]. On the
other hand, some miRNAs, called tumor suppressor miRNAs,
inhibit tumor formation and cancer progression. Both oncogenic
and tumor suppressor miRNAs regulate many biological processes
such as cell cycle, cell proliferation, cell resistance and sensitivity,
cell death, cell differentiation, cell invasion, cell migration, as well as
angiogenesis. In these processes, miRNAs function as oncogenes
and tumor suppressors [17–20].
2.1 Oncogenic Oncogenic miRNAs induce oncogenesis by inhibiting tumor sup-

microRNAs pressor genes. Oncogenic miRNAs increase cell proliferation and
cell survival that promote cancer progression [21]. The biogenesis
of oncogenic and tumor suppressor miRNAs is regulated by two
different processes, canonical and noncanonical miRNA biogenesis
pathways. The biogenesis of oncogenic and tumor suppressor miR-
NAs is regulated by two different processes, canonical and nonca-
nonical miRNA biogenesis pathways. In the canonical pathway,
miRNAs bind to the 30 untranslated regions (30 -UTR) of mRNAs
(messenger RNAs), preventing the translation of mRNAs. Next,
the untranslated mRNA induces mRNA degradation or
Profiling miRNA in Cancer Pathogenesis and Treatment 377
suppression of translation that regulate protein expression level. In

the canonical pathway, pre-miRNAs are formed by the mRNA
splicing machinery without Drosha digestion [2]. The transcription
of long primary RNAs (pri-miRNAs) by RNA polymerase II is the
first step of miRNA biogenesis. Next, the pri-miRNAs are cleaved
by Drosha (microprocessor complex) to form precursor miRNAs
(pre-miRNAs). The pre-miRNAs are transported from the nucleus
to the cytoplasm by the exportin 5 complex. Following transporta-
tion, pre-miRNAs are cleaved by Dicer to form a miRNA duplex.
After these processes, two different ways are found either transla-
tional repression or mRNA cleavage. These processes are regulated
by the AGO protein family and the RNA-induced silencing com-
plex (RISC). The differences between canonical and noncanonical
pathways lie in their different combination of proteins, namely
Drosha, Dicer, exportin 5, and AGO proteins [22, 23].
Abnormal miRNA expression results from changes in the
amplification, translocation of genes, and that modulate many cel-
lular events in different cancer types such as lymphocyte, prostate,
gastric, ovarian cancer, and leukemia. For example, the miR-17-92
cluster affects oncogenic events by targeting signaling pathways in
tumor-suppressive proteins of cancer cells that are seen as over-
expressed in different cancer types [24]. The miR-17-92 cluster
modulates B-cell development and resistance-sensitivity by regula-
tion of the PI3K signaling pathway [25]. Besides, miR-17-92
targets the tumor necrosis factor receptor-associated factor
3 (TRAF3) in gastric cancer cells that induce cell migration and
invasion; it is itself regulated by the NF-κB signaling pathway
[26]. MiR-17-92 has a role in the modulation of chemoresistance
to cisplatin in prostate cancer [27]. The copy number of DNA
induces cancer formation that have a potential role in miRNA
expression level. Such as amplification of 3q26 chromosome causes
an increase in miR551b-3p expression level that regulates STAT3
transcription. The increase of STAT3 expression levels stimulates
cell proliferation in ovarian cancer [28]. The deletion of a 13q14.3
locus on chromosome causes downregulation of miR-15a and
miR-16-1 in chronic lymphocytic leukemia (CLL) [29].
The inhibition of oncogenic miRNAs is an important point to
focus on in targeting abnormal miRNA biogenesis to find a new
therapeutic approach for cancer. Different small molecules are used
as therapeutic inhibitors. Cyclic peptidomimetics are used as inhi-
bitors of miR-155 biogenesis by blocking maturation of pre-miR-
155 through inhibition of Dicer-mediated pre-miR155 processing.
This process induces cell apoptosis in breast cancer cells [30]. In
another study, inhibitors also are used for blocking the abnormal
function of RNA binding proteins. The novel small inhibitor of
KCB3602 that is modified and identified by Lim D. et al. to block
the binding of pre-let-7 and Lin28 that avoid maturation of let-7
miRNA formation. The inhibition of pre-let-7 and Lin28
interaction will supply new pharmacological approaches to target

blocking of cancer stem cell formation [31]. For the first time,
Lawrie et al. demonstrated the presence of miRNAs in the serum
of patients with diffuse large B-cell lymphoma. They also suggested
using miRNAs as diagnostic biomarkers [32]. In addition to that,
miR-210, miR-21, and miR-126 were used as diagnostic biomar-
kers to screen for and diagnose colorectal cancer in that study. The
expression levels of miR-210 and miR-21 are increased while
miR-126’s is downregulated in adenocarcinomas and adenoma,
compared to healthy cells. These miRNAs have a role in cancer
initiation by dysregulation of the VEGF signaling pathway. Thus,
these miRNAs can be novel biomarkers for early detection of cancer
and determination of cancer progression. In addition, Sabry et al.
suggested that miR-210, miR-21, and miR-126 can be used as
novel diagnostic biomarkers in the early stages of colorectal cancer.
Thus, their diagnostic value is high in terms of with sensitivity and
specificity [33]. As a result, oncogenic miRNAs have a critical role
in cancer initiation and progression. Thus, the understanding of the
mechanism of miRNA biogenesis provides insight into “How does
miRNA biogenesis regulate tumorigenesis?” and “How do miR-
NAs improve or find new therapeutic approaches for cancer?”
2.2 Tumor Several studies show that miRNAs regulate cell proliferation and
Suppressor apoptosis, central processes in cancer. Different types of miRNAs
microRNAs which are categorized according to their specificity and properties,
such as abnormal expressions of oncogenic and tumor suppressor
miRNAs, have a role in cancer progression [34]. The inhibition of
oncogenic miRNAs or activation of tumor suppressor miRNAs
reduces cancer initiation and progression. The general properties
of tumor suppressor miRNAs are the induction of apoptosis and
suppression of tumor volume by reduction of the mRNA and
protein levels of FOXM1, MBD3, CCND1, KLK10, and CASP2,
which are activators of cell survival. One tumor suppressor miRNA,
miR-8073, inhibits cancer progression in colorectal cancer. The use
of synthetic miR-8073 reduced tumor volume to 43% in an in vivo
study [35]. Other studies focused on the effect and mechanism of
miR-498 in triple negative breast cancer, which has a poor progno-
sis and limited therapy options. Their results show that miR-498
binds the 30 untranslated region of PTEN and suppresses protein
levels of PTEN. PTEN has roles in cell survival and cell migration
by regulation of the PI3K/Akt signaling pathway [36]. miR-101 is
another tumor suppressor that regulates leukemogenesis. Acute
myeloid leukemia (AML) patients have a low survival rate and
limited treatment. MiR-101 inhibits leukemogenesis by regulating
the EZH2/Wnt/β-catenin signaling pathways directly
[37]. Understanding this mechanism is crucial to developing
novel approaches to target cancer. Accordingly, ongoing research
focuses on this subject to develop new targeted therapy or
investigate new drugs for cancer. miRNAs have different role in

cancer progression according to different types of cancer. One such
example is miR-125b, which plays an oncogenic role in hematolog-
ical cancer but a tumor suppressor role in solid tumors [38, 39]. A
similar dual-property is seen in the mechanism of mir-155.
MiR-155 acts as an inflammation-associated, oncogenic miRNA
in hematological malignancies and solid tumors [40]. On the
other hand, miR-155 has tumor suppressor properties by dysregu-
lation of the small nucleolar RNA host gene 5 (SNHG5) that
induces apoptosis in melanoma [41]. Accordingly, it is critical to
determine a new approach for cancer therapy. To this end, various
studies focus on targeted cancer therapy using miRNAs directly or
indirectly. Hepatocellular carcinoma (HCC) is a type of liver cancer
that has a poor prognosis and limited therapy. Tran et al. high-
lighted new treatment strategies involving targeting the myc gene
by downregulating Linc00176, long intergenic noncoding RNA.
The depletion of Linc00176 has a role in expression of more tha
200 genes which regulate cell proliferation and death signal path-
ways by sponge function of tumor suppressor miR-9 and miR-185
[42]. In addition, tumor suppressor miRNAs mediate cancer pro-
gression by inhibiting tumorigenesis. The mechanism or regulation
of tumor suppressor miRNAs has a critical role in cancer treatment.
2.3 MicroRNAs in Because of the role of microRNAs as tumor suppressors or onco-

Proliferation and genes, some microRNAs promote the proliferation and growth of
Growth cancer cells while others suppress or inhibit proliferation and/or
growth of cancer cells. For instance, a study showed that miR-802
inhibits cell proliferation and promotes apoptosis in human cervical
cancer cell lines and tissues by directly suppressing the serine/
arginine-rich splicing factor 9 expression [43] and that
miR-449b-5p suppresses cell growth and invasion of breast cancer
cells and tissues by inhibiting CREPT-mediated Wnt/β-catenin
signaling [44]. MiR-376c-3p promotes proliferation of hepatocel-
lular carcinoma cells in tissue and mice models by suppressing the
AT-Rich interaction domain 2, [45] while miR-371a-3p enhances
cell proliferation, migration, and invasion in gastric cancer cells by
directly targeting TOB1 [46].
2.4 MicroRNAs in Apoptosis is the programmed cell death which can be triggered by
Apoptosis intrinsic (apoptotic cascades in mitochondria) and extrinsic (death
receptor-mediated) signaling pathways. The purposes of apoptosis
include embryonic development, cell turnover in an organism,
proper regulation of immune cells, and the inhibition of tumori-
genesis [47]. During these cellular events, the activation of cas-
pases, cysteine protease family members, and the recruitments of
effectors and initiators are involved in the progression of pro-
grammed cell death to destruct cellular morphology and DNA
structure [48].
The intrinsic pathway of apoptosis is triggered by intracellular

signals, which affect electron transport system modulation on the
mitochondrial membrane and thereby result in cytochrome c (cyt c)
release into the cytoplasm [49]. Pro-apoptotic Bcl-2 family proteins
located on mitochondria membranes, such as Bax (Bcl-2-associated
X protein), Bak (Bcl-2 homologous antagonist/killer), BH3-only
proteins (Bid and Bim), and all pro-apoptotic proteins regulate
apoptosis induction by forming pores in the mitochondrial outer
membrane (MOM), resulting in the release of cyt
c [50]. Conversely, cancerous cells inhibit the release of cyt c by
overexpressing anti-apoptotic proteins such as Bcl-2 (B-cell
lymphoma-2), Bcl-xL (B-cell lymphoma-extra-large) and MCL-1
(myeloid cell leukemia 1) [51].
Bcl-2 related studies revealed the functions of novel micro-
RNAs in the progression of cancer cells. In human colon cancer
SW480 and HCT116 cell lines, the expression levels of microRNA-
184 were lower in cancerous cells than in adjacent cells, and ectopic
expression of miRNA-184 could stimulate apoptosis of these cells
and inhibit their proliferation ability by negatively regulating Bcl-2
and c-Myc expression levels [52]. The unfavorable response for
cisplatin-based chemotherapy was observed in non-small cell lung
cancer (NSCLC) patients with lower miR-630 and higher Bcl-2
expression levels. The increase in miR-630 expression might inhibit
the chemo-resistance in NSCLC by downregulating Bcl-2
[53]. Similarly, miR-181b might overcome this chemoresistance
of small cell lung cancer cells by directly binding the 30 UTR of
Bcl-2 and thereby inhibiting Bcl-2 expression [54]. LINC00461, a
long noncoding RNA, is highly expressed in multiple myeloma
(MM) patients. It has binding sites for miR-15a/miR-16 to inhibit
their tumor suppressive effects. Thereby, lncRNA overexpression
results in BCL-2 overexpression in U266 and OPM-2 MM cell
lines by abolishing the negative effects of miR-15a/miR-16
[55]. The overexpression of miR-203b-3p and miR-203a-3p
modulated by c-Myc decreases the antiapoptotic Bcl-xL expression
to inhibit the cellular proliferation of breast cancer cells [56]. The
inhibition of these antiapoptotic proteins in the mitochondria can-
not prevent the release of cyt c. Therefore, cancerous cells under-
went programmed cell death. MiR-145 ectopic expression in
NSCLC inhibited proliferation, invasion and migration of A549
lung cancer cells and promoted apoptosis of these cells by raising
Bax expression, thus elevating Bax/Bcl-2 ratio and caspase-3 activ-
ity in these cells [57]. MiR-125b raises the resistance of ovarian and
breast cancer cells to cisplatin or paclitaxel by suppressing proapop-
totic Bak1 expression [58, 59]. Overall, the upregulation of miR-
NAs might inhibit or promote the activities of pro-apoptotic
proteins on the MOM.
MiRNAs are also involved in the modulation of the extrinsic

apoptotic pathway in tumor cells. The extrinsic, or death receptor-
mediated, pathway differs from the intrinsic pathway in that the
tumor necrosis factor (TNF) family receptors on the MOM such as
Fas and TNF receptors initiate apoptotic signaling after binding of
the death ligands such as FAS-L (Fas ligand), TNF-related apopto-
sis-inducing ligand (TRAIL) and TNF. The incoming signals from
receptor–ligand interaction are conveyed using adaptors such as
TNF receptor-associated death domain (TRADD) and
Fas-associated death domain (FADD) [48]. These adaptor proteins
bind to initiator procaspases to construct the DISC, the death-
inducing signaling complex for activation of the executioner cas-
pases [48, 60]. The activation of these caspases starts protein cleav-
age and cytoskeleton alterations, thereby resulting in cell death
[60]. Liver cancer stem cells (LCSCs) highly express miRNA-21-
3p compared to non-LCSCs, which protects the cells from TRAIL-
induced apoptosis. The miRNA-21-3p have role in tumorigenesis
and its suppression gives rise to the enhancement of PTEN levels in
Huh7-LCSCs toward TRAIL-induced cell death by PI3K/Akt/
Bad cascade [61]. MicroRNA-145 and TNF-α treatment stimulate
TNF-α induced cell death in triple-negative breast cancer (TNBC)
cells by forming the RIP1-FADD-caspase-8 complex. Moreover,
the degradation of cIAP1 (important for negative regulation of
RIP) induced by miR-145 is crucial for TNF-α induced cell death
in MDA-MB-231 cells [62]. A study on Fas/FasL pathway related
miRNAs documented that the ectopic expression of miR-143 in
SKM-1 AML cells induces apoptosis toward the Fas/FasL pathway
and inhibits cellular proliferation thereby arresting cells in the
G0/G1 phase. Additionally, miR-143 suppresses MLLT3/AF9
expression by directly binding to its 30 UTR to block the AML
tumorigenesis [63]. A FADD protein related study revealed that
the oncogenic miR-633 is highly expressed in gastric cancer cells
and tissue and has a compulsory correlation with the FADD pro-
tein. MiR-633 inhibition sharply increased FADD levels, doxoru-
bicin, and cisplatin-induced apoptosis in vitro. Indeed, the
forkhead box O 3 (Foxo3a) directly represses miR-633 transcrip-
tion by binding to its promoter region. The combination of
miR-633 antagonist that might mimic Foxo3a and doxorubicin
in vivo gives rise to tumor suppression by increasing FADD expres-
sion in gastric cancer [64]. miR-30c-2-3p specifically targets both
the TNF receptor-associated death domain (TRADD) and the cell
cycle protein Cyclin E1 (CCNE1), thereby regulating NF-kB sig-
naling. Its ectopic expression decreases the cellular proliferation by
downregulating IL6, IL8, and CXCL1. A recent breast cancer
patient study revealed that the overexpression of this miRNA was
associated with a better survival rate in estrogen receptor-negative
breast cancer cells [65]. Mir-200a overexpression inhibits TNF-α
induced protein (A20) expression and stimulates the cleavage of
caspase 8 in DISC induced by TRAIL triggered apoptosis in gastric

cancer cells [66]. MiR-142-5p overexpression suppressed the
X-linked inhibitor of apoptosis (XIAP) expression by specifically
binding to its 30 UTR region in ovarian cancer OVCAR3 and in
SKOV3 cells. Additionally, miR-142-5p increases the sensitivity of
cancer cells to cisplatin-induced cell death by specifically targeting
Bcl-2 and Mcl1 as well [67].
2.5 MicroRNAs in DNA repair is an important mechanism to avoid DNA damage by

DNA Repair protecting genome stability and cellular processes that are regulated
by signaling pathways. Many proteins mediate multiple signaling
pathways in the DNA repair system [68]. The expression level of
miRNAs is also a response to regulate DNA repair through inhibi-
tion/activation of different types of DNA repair proteins at the
post-transcriptional level to provide genome stability [69]. DNA
repair consists of four main repair systems: homologous recombi-
nation (HR), nonhomologous end joining (NHEJ), nucleotide
excision repair (NER), and mismatch repair (MMR). These systems
are regulated by different proteins and factors [70–72]. Under-
standing the mechanism of interaction between miRNAs and the
DNA repair system helps promote the development of cancer treat-
ment. Patel et al. highlighted that miR-15a/miR-16 have a signifi-
cant role in increasing the drug sensitivity of breast cancer cells by
BMI1 dependent- histone H2A ubiquitination. The BMI1 (B cell-
specific Molony murine leukemia virus integration site-1) is a com-
plex protein that regulates homologous recombination [73]. Fur-
thermore, miRNAs have a role in conferring resistance against
inhibitors that are used in cancer treatment. The overexpression
of miRNA-622 increases PARP inhibitors and cisplatin resistance in
ovarian cancer with BRCA1/2 by inhibition of NHEJ. miRNA-
622 controls the balance of HR and NHEJ in the cell cycle through
regulation of the Ku complex protein [74]. Also, ERCC1 has a role
in cisplatin resistance. According to recent studies, ERCC1 regu-
lates the nucleotide excision repair mechanism by removing of
DNA adducts. Li et al. found that interaction of miRNA-196a
and cisplatin leads to resistance by expression of ERCC1. The
expression level of ERCC1 decreases in the anti-miRNA-196a
(inhibited) group of non–small-cell lung cancer cells
[75]. miR-29c also mediates factors and proteins, including the
transcription factor specificity protein 1 (SP1) and MSH6 in DNA
mismatch repair. The expression levels of SP1 and MSH6 regulate
chemoresistance in glioma cells. miR-29c could be a means to
overcome the chemoresistance of glioma cells [76].
In recent years, DNA repair mechanisms have become a useful

tool to manage DNA stability and for targeted cancer therapy. The
miRNA-targeted proteins in the DNA repair system provide a novel
therapeutic opportunity to treat cancer cells that increase the effi-
ciency of drug application and sensitivity of traditional chemother-
apy in cancer.
2.6 MicroRNAs in Cancer cells have migration and invasion abilities to occupy nearby
Invasion, tissues or other organs. Some microRNAs show tumor suppressor
Angiogenesis, and effects and inhibit or suppress migration, invasion and metastasis,
Metastasis or inhibit angiogenesis, while some promote cancer progression.
For instance, miR-140-5p inhibits invasion and angiogenesis
through inhibiting VEGF-A expression in breast cancer cells and
tissues [77] while miR-23a promotes invasion and glial-
mesenchymal transition through directly reducing HOXD10 in
glioblastoma [78]. Another tumor-suppressor microRNA
miR-33a suppresses the epithelial-mesenchymal transition (EMT)
and the ATP drug transporter gene ABCA1 expression. However,
miR-33a is downregulated by Ras association domain family mem-
ber 1 (RASSF1) oncogene in lung cancer cells [79]. Besides,
P53-induced miR-1249 inhibits tumor growth, migration, inva-
sion, metastasis, and angiogenesis by targeting VEGFA and
HMGA2 both in vitro and in vivo [80].
2.7 MicroRNAs in Drug resistance is a major and very important issue in cancer
Drug Resistance treatment as it can cause relapse and/or death. Because cancer
treatment becomes difficult with drug resistance, current therapies
and studies aim to overcome drug resistance problems in cancer
patients. Regulation of microRNAs has opened a new window in
drug resistance. Some microRNAs suppress drug resistance and
induce apoptosis in cancer cells, while upregulation of some micro-
RNAs causes or enhances drug resistance. For instance, miR-383-
5p targets TRIM27 and suppresses cell proliferation and paclitaxel
resistance in ovarian cancer [81], whereas microRNA-98, which is
upregulated in bladder cancer tissues, causes doxorubicin and cis-
platin drug resistance and reduces apoptosis by downregulating the
LASS2 gene and by regulating Drp1 signaling [82]. In addition,
miR-148a showed anticancer and antitumor effects through regu-
lating β-catenin signaling in cisplatin-resistant colorectal cancer
cells, tissues, and mice models, [83]. Another example is
miR-503-5p which regulated metastasis and paclitaxel drug resis-
tance through CD97-mediated JAK2/STAT3 signaling pathways
in ovarian cancer cells [84]. MicroRNA-193a-3p upregulated in
CD44(+) gastric cancer cells caused cisplatin resistance [85] while
downregulation of miR-9 enhances daunorubicin chemoresistance
through the EIF5A2/MCL-1 axis in AML cells [86]. microRNA-
455-3p regulates gemcitabine resistance through regulating TAZ
in pancreatic cancer cells [87]. Another study showed that
increased levels of miR-498 enhance the cisplatin-sensitivity

through downregulation of Bmi in gastric cancer, both in vitro
and in vivo [88], and miR-135b-5p downregulates the anterior
gradient 2 (AG2), which enhances doxorubicin resistance of breast
cancer [89]. Moreover, cisplatin sensitivity was enhanced by
miR-29a in non-small cell lung cancer, [90] while miR-122 targets
an oncogene, SerpinB3, and thus regulates the sorafenib resistance
in hepatocellular carcinoma cells [91]. MicroRNA-22 was found to
be a tumor suppressor microRNAs due to the fact that it increased
paclitaxel sensitivity of breast cancer cells through inhibition of a
pro-oncogene NRAS [92].
In addition to drug resistance, microRNAs also regulate the
radioresistance in cancer cells. For instance, miR-372 enhances
radiation sensitivity as well as it inhibits invasion and metastasis by
activation of the PBK-dependent p53 signaling pathway in naso-
pharyngeal carcinoma, [93] while miR-205 enhances the response
to radiotherapy by repressing PKCε-EGFR-DNA-PK in both pros-
tate cancer cells and xenograft mice models [94].
2.8 MicroRNAs in MicroRNAs are crucial members of oncogenic regulatory net-

Metabolic Regulation works, modulating the metabolic regulation in cancer cells. Altera-
tions in the cellular metabolism of cancer cells provides essential
adaptations during the proliferation process. The metabolic regula-
tion in cancer cells includes miRNA related regulations in glucose
metabolism, lipid metabolism, and signaling pathways
[95, 96]. Cancer cells have reprogrammed their metabolic activ-
ities, especially glucose metabolism, in which they produce lactic
acid through aerobic glycolysis instead of producing carbon dioxide
and water, even if oxygen is present in the cellular environment
[97]. Cancer cells with these changes, called the Warburg effect,
might tolerate inefficient ATP production by overexpressing the
glucose transporters in the cellular membrane to uptake more
glucose molecules [98]. One of these glucose transporters,
GLUT1 is highly expressed in cancer cells, and microRNAs are
recruited in the modulation of GLUT1 expression [99]. GLUT1
is directly or indirectly related to several microRNAs (Table 1).
In addition to increased glucose uptake via the overexpressed
GLUT1 protein in the membrane, the enzymes recruited in glycol-
ysis: hexokinases (HKs), phosphofructokinase (PFK), and pyruvate
kinases (PKs) are overexpressed and their regulator miRNAs are
downregulated or upregulated in cancer cells to compensate for
reduced ATP production due to the Warburg effect [99]. Hexoki-
nases are enzymes that phosphorylate glucose to produce glucose-
6-phosphate (G6P), thereby resulting in the trapping of glucose in
the cytosol of cancer cells. Cancer cells overexpress HK2 by down-
regulating or overexpressing numerous microRNAs for regulation
of cancer cell metabolism (Table 2).
Table 1
Glucose metabolism related microRNAs via GLUT1 expression directly or indirectly
Target microRNA Expression Pathway Cancer type References

GLUT1 miR-218 Downregulated – Bladder cancer [220]
miR- Upregulated – Ovarian squamous cell [221]
1204 carcinoma
micR- Downregulated – Renal cell carcinoma [222]
1291
IGF-1R miR-342- Downregulated PI3K/AKT/GLUT1 HCC [223]
3p
INSR, miR-128 Downregulated PI3K/AKT/GLUT1 TNBC [224]
IRS1
CAB 39 miR-451 Downregulated LKB1/AMPK/PI3K/ Glioma [225]
AKT/GLUT1
Table 2
Hexokinase 2 HK2 related microRNAs in various cancer types directly or indirectly
Target microRNA Expression Pathway Cancer type References

HK2 miR-98 Downregulated – Colon [226]
cancer
AKT1/ miR-124 Downregulated AKT1/AKT2/GLUT1 / NSCLC [227]
AKT2 HK2
PTEN miR-214 Overexpressed PTEN/AKT/mTOR/ HK2 NSCLC [228]
HK2 miR-125a Downregulated – HCC [229]
Phosphofructokinase (PFK) is an enzyme that converts fruc-

tose-6-phosphate through fructose-1,6 bisphosphate. PFK liver
type (PFKL) is predominantly targeted by miR-128 in lung cancer
cells for glucose metabolism. The overexpression of miR-128
decreases glucose uptake and the increased PFKL expression gives
rise to poor overall survival in patients [100]. Pyruvate kinase M2 is
the last crucial enzyme in glycolysis, which converts phosphoenol-
pyruvate to pyruvate. miR-139-5p downregulation causes the inhi-
bition of PKM2 expression by directly binding, thereby leading to a
decrease in the glucose consumption in gallbladder carcinoma
patients [101]. Similarly, miR-148a and miR-326 directly target
PKM2 in thyroid cancer. The downregulation of these tumor-
suppressive microRNAs activates the PKM2 for thyroid carcino-
genesis and migration as well [102]. Lactate dehydrogenase A
(LDHA) is an essential enzyme that regulates aerobic glycolysis in
various human cancers. MiR-30a-5p blocks LDHA expression by
directly binding its 30 UTR to decrease lactate production and
glucose uptake in normal cells. However, miR-30a-5p downregula-

tion in breast cancer cells causes the upregulation of LDHA for
breast cancer growth and metastatic ability in patients [103].
In addition to the PTEN/AKT/mTOR pathway discussed
above with other interacting proteins, decreased expression of
miR-155 causes the inhibition of glucose uptake into the cytosol
and glycolysis in breast cancer cells by negatively regulating
PIK3R1-FOXO3a-cMyc signaling pathway. This study, involving
50 TNBC patients, revealed that miR-155 represses FOXO3a and
PIK3R1 by directly binding to these mRNAs for elevated glucose
uptake [104]. MiR-203 is upregulated in ovarian cancer tissues
compared to noncancerous ovarian cells, and its overexpression is
crucial for tumor cell growth and migration of the cells of in vivo
nude mice models. Noncancerous cells might inhibit miR-203
regulation by p53, but miR-203 inhibits pyruvate dehydrogenase
B (PDHB) by directly binding its 30 UTR for higher glucose uptake
by stimulating glycolysis in ovarian cancerous cells [105].
Another hallmark of cancer cell is the reprogramming of the
lipid metabolism by stimulating the de novo synthesis of fatty acids
(FA), raising lipolysis and lipid uptake from the cellular environ-
ment. The increased lipid uptake or presence of lipid in the cytosol
may provide cancer cells adaptive responses to unfavorable condi-
tions [106]. One experiment showed that PM2.5 liposoluble
extracts, a collection of FA promoted the immediate upregulation
of FA translocase known as CD36 in liver HepG2 cells by inversely
regulating miR-26a expression. This miRNA plays a role in the
regulation of lipid accumulation in non-cancerous cells by directly
targeting the 30 UTR of the CD36 mRNA [107]. Similarly, nega-
tive regulation of the SIRT1 protein, which inhibits lipid accumu-
lation in hepatocytes by miR-23b directly binding to its mRNA,
increases lipid accumulation in HepG2 cancer cells [108]. Repro-
gramming of the de novo lipogenesis by upregulation of miR-182
causes tumorigenesis of human lung cancer and an increase in
cellular proliferation. MiR-182 upregulation dramatically inhibits
the activity of pyruvate dehydrogenase (PDH) kinase 4 (PDK4) by
binding its 30 UTR, which can block PDH by phosphorylation to
mediate the entry of the TCA cycle as a metabolic switch from
pyruvate to acetyl-CoA. The upregulation of miR-182 and down-
regulation of PDK4 provides the increased levels of triglycerides for
lipogenesis in lung cancer cells. The silencing or chemical inhibition
of the fatty acid synthase (FASN) demonstrated miR-182 deregu-
lation and upregulation of PDK4 in cancer cells, thereby resulting
in diminished cellular growth. Indeed, miR-182 upregulation
causes stimulation of JNK phosphorylation for the proliferation of
lung cancer cells via the JNK signaling pathway [109]. MiR-148a is
expressed in the normal adult liver to modulate cholesterol, triglyc-
eride and lipoprotein levels. However, its deficiency is in parallel
with poor survival in hepatocellular carcinoma patients because of a
regulator of lipid metabolism. The knockout mice model of

miR-148a documented higher expression levels of genes recruited
in lipogenesis and fatty acid uptake and the increase in the choles-
terol level and its synthesis [110].
Collectively, cancer cells are able to reprogram the metabolic
regulation by downregulating or upregulating microRNAs during
cancer cell progression. The activity of miRNAs in various cancer
types can improve the poor survival rate of cancer patients. The
expression levels of these miRNAs might be helpful as a therapeutic
target for the specific cancer types.
2.9 MicroRNAs in MicroRNAs are functional in reprogramming the tumor immune

Immune Response response to escape from eradication. There is increasing evidence
Regulation that microRNAs can contribute to the regulation of antigen pre-
senting cells such as dendritic cells, macrophages and B cells;
tumor-related immune cells such as T cells, Natural killer
(NK) cells, T regulatory (Treg) cells, tumor-associated macro-
phages (TAMs), myeloid-derived suppressor cells (MDSCs), and
cancer-associated fibroblasts (CAFs). Indeed, microRNAs can reg-
ulate immune checkpoint interaction such as PD-1-PD-L1 via
signaling pathways [111].
Dendritic cells (DC) are antigen-presenting cells that can regu-
late tumor immunity by changing the microRNA network in mul-
tiple myeloma (MM). MiR-29b is only dramatically downregulated
in tumor related DCs and significantly upregulated in normal den-
dritic cells. The upregulation of miR-29b gives rise to an increase of
proliferative molecules such as NF-kB and cytokine/chemokine
signaling by altering Th17 polarization through the bone marrow
cells in a co-cultured environment. This study revealed that MM
can reprogram DCs by downregulating miR-29b to obtain these
proliferative molecules for their cellular growth and survival [112].
Macrophages function as immune regulators or effectors and
types of scavengers in the cellular environment. They can be classi-
fied into two groups: M1 macrophages with anti-tumorigenic roles
and M2 macrophages with pro-tumorigenic roles. The transforma-
tion of M1 to M2 in the tumor microenvironment can be regulated
by microRNAs [113]. MiR-21a has role in inhibition of PTEN
expression and upregulation of the CD206 expression in primary
macrophages to modulate the M1 to M2 transformation. More-
over, miR-21a can positively regulate the DNA methyltransferase,
thereby inhibiting miR-200c expression by hypermethylation the
promoter region of miR-200c to prevent PTEN expression.
MiR-200c directly targets the 30 UTR of PTEN for higher expres-
sion of PTEN in primary macrophages. Therefore, miR-21a over-
expression regulates the M2 macrophage transformation via
miR-200c hypermethylation and PTEN-related molecules involved
in the pAKT/HIF1α/VEGF pathway (Fig. 1) [114].
Fig. 1 The role of microRNA in PTEN signaling pathway and macrophage polarization
B cells are another antigen presenting cell that contain immu-

noglobulin A (IgA+) in many tumor cells for higher expression of
suppressive molecules in immune regulation. The expression levels
of miR-15A and miR-16-1 are lower in colon cancer cells than their
non-tumor adjacent cells. IgA+ B cells in the tumor tissue with a
knockout of these miRNAs are much larger and their number is
higher than other mice with only the tumor. These B-cells have
overexpressed immune regulatory molecules and have released
ınterleukin-10, TGF-β and programmed death ligand-1 to repress
the activation of proliferation of cytotoxic (CD8+) T cells. There-
fore, colon cancer cells can survive better with reduced levels of
miR-15A and miR-16-1. The ectopic expression level of these
miRNAs in non-tumor intestinal epithelial cells gives rise to a
decrease in NF-kB and STAT1 expression and chemotaxis activity
of IgA+ B cells. The lower level of IgA+ B cells and higher expres-
sion of the miRNAs in colon cancer tissues have increased the
survival of patients. These data reveal that miR-15A and miR-16-
1 reduce the number of IgA+ B cells in colon cancer tissues by
activating the signaling pathways for B-cell-based immune
suppression [115].
T cells in an immune regulation system have various types to
protect the organism exogenous factors. The tumor cells can mod-
ulate the immune system by the activity of microRNAs. MiR-141 in
breast cancer cells is downregulated to promote cellular prolifera-
tion by overexpressing cyclooxygenase-2 (COX-2), TNF-α and
prostaglandin E2 (PGE2) and inhibiting IL-10. Furthermore, the
downregulation gives rise to upregulation of mitogen-activated
protein kinase kinase kinase kinase 4 (MAP4K4). This upregulation
decreases the activity of miR-141 on the effectivity of cytotoxic CD

+ T cells on MCF-7 breast cancer cells. The downregulation of
miR-141 provides the cancer cells a mode of evasion from the
cytotoxic CD4+ T cells [116]. Another type of T cell called Treg
or T regulatory cells are immune regulatory cells that mainly sup-
press the activity of effector T cells to provide an immune escape of
cancer cells while Th17 cells block tumor immunity. miR-155 is
overexpressed in the cervical cancer tissues compared to the normal
cervical tissues, and thereby increased the Th17/Treg proportion
by inducing Th17-related cytokines; IL-17A, IL-6, and RORγt
protein in cervical cancer tissues [117]. SOCS1, suppressor of
cytokine signaling is a protein which is recruited in the modulation
of the JAK/STAT pathway for the regulation of Th17 and Treg
[118]. The increase in the expression of these factors and down-
regulation of SOCS1 protein activity reveal that miR-155 is crucial
for the tumorigenesis of cervical cancer throughout the inhibition
of SOCS1 expression directly binding its 30 UTR [118] and induc-
ing the differentiation of Th17 and Treg cells in CD4+ T
cells [117].
Natural killer (NK) cells are other tumor-fighting cells that
block the growth of cancer cells. Oxaliplatin treatment causes resis-
tance to oxaliplatin in colorectal cancer cell lines expressing the
WBSCR22 gene. The coculture study of colorectal cancer cells
and NK cells have demonstrated the significant decrease in the
growth of oxaliplatin resistant cancerous cells in in vitro and
in vivo studies. This NK cell treatment causes the upregulation of
miR-146b-5p by decreasing the WBSCR22 gene expression in
colorectal cancer cells. Deeper analysis has revealed that
miR-146b-5p directly targets the 30 UTR of WBSCR22 mRNA
for inhibition of cellular growth in colorectal cancer cells [119].
Myeloid-derived suppressor cells (MDSCs) are immunosup-
pressive cells that can regulate tumor environments in cancer
types such as glioma patients. The examination of hypoxia-
stimulated glioma-derived exosomes (GDEs) revealed the higher
ability of GDEs to stimulate MDSCs. The higher levels of miR-10a
and miR-21 in these exosomes induce the activation of MDSCs
throughout RAR-related orphan receptor alpha (RORA) and
PTEN pathways, respectively. Therefore, glioma cells raise MDSC
activation via hypoxia-stimulated exosomes bearing these miRNAs
to regulate their immune suppression for their survival and
propagation [120].
Cancer-associated fibroblasts (CAFs) found in the tumor
microenvironment contribute to tumor progression and resistance.
High-grade serous ovarian cancer (HGSOC) contains the accumu-
lation of the CAF-S1 subpopulation for immune suppression. The
chemokine (C-X-C motif) ligand 12 isoform β (CXCL12β) is
expressed by CAF-S1 subpopulation to attract the regulatory
T-cells and promote their activities in ovarian tumors based on
the co-culture study of CAF-S1 and CD4+CD25+ T cells. A

CXCL12β related miRNA study has demonstrated that CXCL12β
is directly targeted by miR-141 and miR-200a in CAF-S4 cells for
CXCL12β accumulation in CAF-S1 cells. The higher amount of
CAF-S1 in this ovarian cancer type can increase the resistance level
to immunotherapies [121].
In addition to tumor-associated cells, immune checkpoint
interactions such as PD-1-PD-L1 can be regulated by microRNAs
via signaling pathways. The deficiency of miR-21 in macrophages
stimulates the polarization of macrophages through M1 macro-
phages for antitumor immunity in mice models. Mir-21 expression
is stimulated by polarization stimuli such as coculturing of macro-
phages and tumor cells, and its expression inhibits STAT1 signaling
induced by IFN-γ that is related to M1 polarization. However,
enhanced STAT1 expression in the deficiency of miR-21 promotes
PD-L1 upregulation in macrophages and this gives rise to inhibi-
tion of the anti-tumor activity. Therefore, this article states that
PD-1 antibodies combined with miR-21 knockout macrophages
increase anti-tumor activity through the polarization of tumor-
associated macrophages toward M1 type macrophages to enhance
the host’s immune system [111].
Collectively, tumor-associated immune cells are functional to
regulate the immune responses of tumor cells. MicroRNAs play
major roles in the immune response by regulating the tumor
microenvironment for immune cells and the interactions between
tumor cells and immune cells.
3 Therapeutic Potentials of microRNAs in Cancerous Cells
3.1 MicroRNAs in According to the National Institute of Health, 1,735,350 new

Diagnosis and cancer cases were expected to be diagnosed and 609,640 people
Prognosis were projected to die from cancer, only in the United States in 2018
(www.cancer.gov/about-cancer/understanding/statistics). There-
fore, early diagnosis of cancer is critical for the intervention of
oncologists and the treatment of cancer patients. In this regard,
microRNAs can be useful.
One study revealed the diagnostic value of three microRNAs in
colorectal carcinoma. It was shown that the upregulated expression
of miR-210 and miR-21, and downregulated expression of
miRNA-126 could be used in the clinic as a non-invasive novel
marker for early diagnosis, screening of colorectal carcinoma
[33]. Another study showed that a combination of miR-141,
miR-21, and miR-375 could be used as a diagnostic marker panel
for prostate cancer patients [122].
MicroRNAs, especially serum and plasma microRNAs have

been shown to be valuable diagnostic biomarkers for human hepa-
tocellular carcinoma [123], colorectal cancer [124, 125], esopha-
geal squamous cell carcinoma [126], pancreatic cancer [127],
gastric cancer [128], breast cancer [129], and ovarian cancer
[130, 131]. For instance, serum miR-16, which was shown to be
a tumor suppressor in several studies, was revealed as a potential
cancer biomarker for various cancers [132]. miR-20a is a novel
promising biomarker for human cancers [133]. Another study
showed that blood microRNAs miR-33a-5p and miR-128-3p
were demonstrated to be potential biomarkers for early diagnosis
of lung cancer [134].
During cancer management, treatment strategies can change
based on the prognosis of disease. For this, prognostic markers are
valuable to predict problems and make provisions for disease treat-
ment during its progression. The prognostic value of microRNAs
on survival outcomes of patients with hepatocellular carcinoma
[135], osteosarcoma [136], lung adenocarcinoma [137, 138],
small cell lung cancer [139], central nervous system lymphoma
[140], nasopharyngeal carcinoma [141], and cervical squamous
cell carcinoma [142] has been evidenced. For instance, increased
miR-98 expression was shown to be a good prognostic factor for
acute myeloid leukemia patients who were only treated with che-
motherapy, [143] while downregulated miR-1297 reflected poor
prognosis in gastric cancer cells [144]. A study showed that the
expression of microRNAs is correlated with overall survival and
relapse of oral squamous cell carcinoma (OSCCs) [145], while
miR-23a-3p has been correlated with better prognosis in OSCCs
[146]. Another study showed that the expression level of miR-141
was related to poor outcome, and it is a prognostic factor in
prostate cancer patients [147]. For instance, serum miR-22 was
shown to be a prognostic marker related to poor clinical outcome in
diffuse large B-cell lymphoma (DLBCL), [148] while another
study connected poor prognosis with downregulation of circulat-
ing miR-638 in colon cancer patients [149] and downregulation of
microRNA-141 to poor outcome in breast cancer through resis-
tance to trastuzumab [150].
3.2 Biomarker Biomarkers play an important role in diagnosis, monitoring,

Profiling of MicroRNAs and/or prognosis of a disease in clinical settings, as well as for
in Cancer Progression drug development. Biomarkers can be derived from cells, tissues
and body fluids. Nowadays, microRNAs opened a new window as
biomarkers in cancer diagnosis, progression, and treatment. The
expression patterns of microRNAs vary with the type and stage of
the cancer and the patient. Therefore, cancer diagnosis and treat-
ment of patients can be shaped through microRNA profiling and
via the identification of potential microRNA biomarkers.
For instance, differences between the microRNA expressions

between leukemia stem cells (CD34+ CD38) and leukemic pro-
genitor cells (CD34+ CD38+) were identified in Philadelphia
chromosome-positive acute lymphoblastic leukemia (Ph+ ALL)
by using high-throughput whole genome sequencing. Multiple
differently expressed microRNAs were identified in leukemia stem
cells and leukemic progenitor cells. Thus, it was suggested that for
the identification of LSC-related miRNAs can be helpful in diagno-
sis and treatment of Ph+ ALL and can help to develop new strate-
gies for therapy and help elucidate possible drug resistance in Ph+
ALL [151]. Another study showed that miR-181a expression levels
increased while miR-27a, miR-107, and miR-195 were downregu-
lated in ER-positive breast cancer when compared to ER-negative
breast cancer patients [152]. Based on this study, miR-181a,
miR-27a, miR-107, and miR-195 expression levels can be used as
clinical markers to diagnose ER-positive or –negative breast cancer
so that they can aid patient stratification. Moreover, two miRNAs,
miR-125b-2-3p and miR-933, were revealed to be a signature for
advanced colorectal cancer patients through predicting colorectal
cancer cell sensitivity against first-line chemotherapy
[153]. Another study showed that circulating miR-130b-5p,
miR-151a-5p, miR-206, and miR-222-3p were downregulated
after surgery in early breast cancer patients. Thus, circulating
miR-130b-5p, miR-151a-5p, miR-206, and miR-222-3p can be a
diagnostic and prognostic markers for breast cancer patients since
they are indicators of early breast cancer progression
[154]. MiRNA-3653 was revealed to be potential tissue biomarker
for pancreatic neuroendocrine tumors since upregulation of
miR-3653 can reflect increased metastatic risk [155].
In addition, profiling of circulating microRNAs demonstrated
that microRNA-196a can be used as a biomarker in the diagnosis of
early-stage gastric cancer [156] while profiling of serum miR-320a,
miR-665, miR-1275, miR-3185, miR-3184-5p, miR-3195,
miR-4459, miR-4640-5p, miR-6076, and miR-6717-5p revealed
them to be diagnostic biomarkers for ovarian cancer patients
[157]. Another microRNA profiling showed that miR-186-5p
expression was upregulated in the serum of non-metastatic and
metastatic prostate cancer patients and that miR-186-5p regulates
cell proliferation, growth, and invasion in cell lines [158].
Ongoing and future studies aim to identify new potential
microRNAs and genes related to microRNA expression for use in
diagnosis and progression monitoring of different cancer types
using bioinformatics analysis. For instance, microRNAs were pro-
filed to predict progression and recurrence of primary esophageal
cancer based on data in the TCGA database [159]. Serum micro-
RNAs for prostate cancer were discovered through transcriptomic
analysis [160], while novel miRNA biomarkers were identified and
mRNA-miRNA interactions were considered for five molecular
subtypes of breast cancer by using system biology [161].
3.3 Circulating Since microRNAs contribute to cancer development as tumor sup-

microRNAs as Novel pressors or oncogenes, studies have aimed to identify new and
Biomarkers potential biomarkers that are circulating in serum, plasma, or
other body fluids to use in cancer diagnosis or therapy. In this
part, we review some recent studies pertaining to this line of
research.
In studies on circulating placental nucleic acids in maternal
blood, circulating microRNAs provide to determine pregnancy
complications to use predict pregnancy disorders [162–164]. Cir-
culating microRNAs were the subject of several subsequent studies
that demonstrated that circulating microRNAs are not only derived
from cell lysis of cancer cells but also from extracellular secretion
from cancer cells [165–167]. Mitchell et al. demonstrated that
circulating microRNAs, which are blood-derived markers, are sta-
ble markers for the detection of cancer because they are protected
from endogenous RNase activity [166]. It was demonstrated that
the stability and transport of circulating microRNAs in blood are
provided by argonaute-2 (Ago-2) [168], high-density lipoproteins
(HDL) [169], or microvesicles [170].
Expression levels and effects of circulating microRNAs depend
on the cancer type and the biological or pathological stage of the
cancer. Recent studies revealed that circulating microRNAs can be
diagnostic markers for the specific and early detection of several
cancer types including prostate cancer [171, 172], lung cancer
[173, 174], glioblastoma [175], bladder cancer [176], breast can-
cer [177], esophageal squamous cell carcinoma [178], gastric can-
cer, pancreatic cancer [179], chronic lymphocytic leukemia [180],
and childhood acute lymphoblastic leukemia [181]. Based on these
studies, the circulating miR-99 family can be used as a liquid biopsy
marker for pancreatic cancer, [179, 182] while circulating
miR-106b-3p, miR-101-3p, and miR-1246 can be used as biomar-
kers for hepatocellular carcinoma detection [183]. Another study
showed that the level of circulating hsa-miR-1273g-3p can be used
as a biomarker for epithelial ovarian cancer detection [184]. Addi-
tionally, it has been revealed that exosomal microRNA-486-5p,
microRNA-181a-5p, and microRNA-30d-5p were circulating bio-
markers for high-risk rectal cancer patients [185]. The circulating
microRNA expression also could be specific to a population and can
be used in diagnosis. For instance, circulating miR-215-5p and
miR-642a-5p can be potential biomarkers for osteosarcoma in the
Mexican population [186].
Weber et al. tried to understand whether circulating micro-
RNAs are suitable and specific to use in early detection of malignant
mesothelioma [187]. They found that the miRNAs miR-132-3p,
miR-126-3p, and miR-103a-3p were suitable for detection from
prediagnostic plasma samples with 98% specificity. Although the
study has some limitations, based on this specificity they concluded
that miR-132-3p, miR-126-3p, and miR-103a-3p can be used for
early detection of malignant mesothelioma [187]. Another study

demonstrated that seven circulating miRNAs (miR-182-5p,
miR-328-3p, miR-339-5p, miR-340-5p, miR-485-3p, miR-486-
5p, and miR-543) can be used in non-invasive glioblastoma
diagnosis [175].
Circulating microRNAs, especially serum microRNAs, can be
isolated from patients painlessly and facilitate easy and cost-effective
cancer detection. Recent studies demonstrated that serum micro-
RNAs are like a signature of patients. For instance, increased serum
levels of miR-320 vary significantly among prostate cancer patients
and are associated with clinical parameters based on age and cancer
stage of patients [188]. In addition, serum miR-1915-3p and
miR-455-3p were identified as biomarkers for breast cancer
patients [189]; and miR-17, miR-25, and miR-133 expressions
can be used as biomarkers since they are related with clinical
stage, poor survival, and metastasis of breast cancer patients
[190]. Moreover, serum miR-206 could predict chemoradiother-
apy resistance in advanced-stage cervical squamous cell carcinoma
[191], while serum miR-30c levels were correlated with cardiotoxi-
city in bevacizumab therapy in non-small cell lung cancer patients
[192]. In addition to serum microRNAs, plasma microRNAs can
be used as biomarkers for early detection and monitoring of breast
cancer [193, 194].
Although circulating microRNAs can be helpful in early diag-
nosis of cancer, the following questions should be raised: Are
circulating microRNAs suitable for diagnosis, do they really reflect
cancer in patients, and can these circulating microRNAs be used for
all patient types? Do hormones change the expression pattern of
microRNAs? What is the role of circulating microRNAs in cancer
treatment with novel bioactive compounds? Ongoing and future
studies should consider these questions to find more specific and
stable microRNAs to use in diagnosis. These questions can be
answered using different microRNA panels. Thus, circulating
microRNAs can open a new window for personalized therapy in
cancer treatment.
3.4 Small Molecules The regulation of miRNA expression levels is critical for controlling
Inhibitors of miRNAs cancer initiation and progression. Many therapeutic approaches in
(SMIRs) cancer are based on the regulation of the miRNA biogenesis. The
microRNAs, which are related to cancer processes are modulated by
small molecules. These small molecules have a potential role in
targeting cancer related-miRNAs [195].
Different approaches are being investigated to control the
expression levels of miRNAs, includinggenome editing by
CRISPR/Cas9-base, antagomirs, miRNA sponges, and small mol-
ecule inhibitors of miRNA (SMIRs). The novel technology of the
CRISPR/Cas9 system depends on genome editing that provides
the change of miRNA expression level in cancer cells [196]. The
antagomiRs that are also known as antisense oligonucleotides

inhibit oncogenic miRNAs to repress cancer formation
[197]. Also, overexpressed oncogenic miRNAs can be inhibited
by miRNA sponges, which have multiple target sites to bind miR-
NAs [198]. Lastly, the small molecule inhibitors of miRNA
(SMIRs) are chemical compounds that influence miRNA
biogenesis [6].
In targeting cancer-related miRNAs, new approaches for small
molecules to inhibit miRNA biogenesis or manage dysregulated
miRNAs are examined for more effective and stable cancer treat-
ment. Through the development of high-throughput screening
that is used to determine or categorize chemical compounds in
terms of pharmacological properties and biological activities, new
specific drugs targeting miRNAs are being developed for cancer
[199]. For this purpose, Arachchilage et al. designed locked nucleic
acid (LNA) to inhibit pre-miRNA maturation. Treatment with
LNA on non––small-cell lung cancer induces PTEN expression
levels that inhibit miR-92b expression. This group mentioned
that the use of LNAs will be a powerful approach to targeted cancer
therapy [200]. Another small inhibitor is peptidic-kanamycin ami-
nosugar, which specifically binds pre-miRNA-21, which has a role
in cancer progression and metastasis. The use of peptidic-
kanamycin amino sugars in breast cancer cells inhibits miR-21
maturation. The inhibition of miR-21 regulates PTEN, Ki-67
expression levels, and the transition of epithelial to mesenchymal
to inhibit cancer [201]. LIN28 is an oncogenic protein that reg-
ulates let-7 maturation through the interaction of let-7 and LIN28.
The inhibition of the LIN28 domain causes dysregulation of let-7
biogenesis in leukemia cells and embryonic stem cells [202].
3.5 Delivery of MiRNAs are endogenous non-coding RNAs that modulate many
microRNAs with cellular events such as cell growth, proliferation, and differentia-
Carriers tion. Their expression levels are used as a diagnostic, prognostic
biomarker and as a therapeutic modulator to regulate cancer [203–
205]. In addition, recent studies demonstrated that miRNAs are
used for tumor gene therapy in the multiple targeted drug
approach with exosomes and nanoparticles [206, 207]. Nanomedi-
cine is important for therapeutic approaches in cancer and drug
delivery systems with low toxicity and easy modification to target
cancer cells while resolving the problem of side effects involved in
chemotherapy [208]. Three categories of nanocarriers, polymeric,
inorganic and lipid nanocarriers, are found to specifically target
cancer [7].
Many researchers target cancer cells based on gene interference
therapy. The expression levels of miRNAs affect regulation of gene
expression, so transfection with miRNAs using nanocarriers pro-
vides modulation of gene expression. The resistance of breast stem
cells to chemotherapeutic drugs is a barrier to treatment of breast
cancer. Also, the expression level of miR-200c induces sensitivity of

breast cancer cells by dysregulation of microtubule formation. The
solid lipid nanoparticle (SLN) are used as delivery of miR-200c. In
addition, SLN- miR-200c, and paclitaxel are used as combination
therapy to treat breast cancer. Transfection with SLN- miR-200c
disrupts microtubule formation by downregulation of the
β-tubulin expression and the use of paclitaxel loaded-nanocarriers
induces cytotoxicity in breast cancer [209]. In addition, miR-31
induces the downregulation of mtEF4 (mitochondrial elongation
factor 4) that induces apoptosis. The conjugation of doxorubicin-
loaded silica nanoparticles, miR-31 bind targeted mtEF4 to trigger
apoptosis, and the combination of doxorubicin provides a synergis-
tic effect on human cervical cancer cells [210]. Following that,
polyarginine peptide (R11)-labeled BPEI-SS nanoparticles that
are non-cytotoxic are used to deliver miR-145. The use of conju-
gates of R11-labeled BPEI-SS and miR-145 are taken up by pros-
tate cancer cells. According to in vivo studies, R11-labeled BPEI-SS
nanocarriers of mir-145 inhibit tumor growth and increase survival
rate. As a result, this type of nanocarrier will be used as a novel
targeted therapy for prostate cancers [211].
Different approaches to treat cancer are available such as sur-
gery and radiotherapy for local tumor and chemotherapy and hor-
mone therapy for metastatic cancers [212, 213]. However, these
treatments are still being further developed. For this purpose,
different designed nanocarriers are used to deliver drugs. These
nanocarriers provide drug stability, selective targeting of cancer
cells, and regulation of extravasation of nanocarriers on blood
vessels. The determination of molecular pathways of miRNAs
exposes novel therapeutic approaches on cancer.
3.6 Preclinical and Although last decades have passed since the discovery of miRNAs,
Clinical Studies of they are target topics of not only studies but also therapeutic
microRNA Applications applications. Preclinical studies on microRNAs are in fashion to
examine their toxicity, efficacy, and safety in in vitro and in vivo
experiments. The above-indicated information about tumor sup-
pressive miRNAs, oncomiRs, metabolism, and immune regulatory
miRNAs and DNA repair system related miRNAs provide preclini-
cal studies related to therapeutic approaches employing miRNAs
invarious cancers [214]. The results of these studies allow the
translation from preclinical studies to clinical applications. In addi-
tion to the above studies, microRNAs function in cardiotoxicity,
which is the leading cause of drug attrition and a major problem in
the regulation of preclinical safety testing of novel drugs.
Doxorubicin (DOX) is an effective chemotherapeutic drug that
gives rise to numerous side effects such as cardiotoxicity. The
leading cause for this side effect can be related to the regulation
of microRNAs in cardiovascular tissues. The in vitro model of DOX
in H9C2 cells demonstrated that DOX application decreases
cellular growth and proliferation of these cells. The in vivo model of

DOX in rat hearts documented significantly increased levels of
miR-140-5p upon DOX treatment. Deeper analysis found that
miR-140-5p directly targets Nrf2 and Sirt2 mRNAs, thereby pro-
moting the DOX-induced myocardial oxidative damage that causes
cardiotoxicity. The potential use of a miR-140-5p inhibitor
through its targets may be a novel therapeutic way of
DOX-enhanced cardiotoxicity after DOX application in clinical
studies of various cancers [215]. The analysis of circulating miR-
NAs in female mice plasma and miRNA expression in female mice
cardiac tissue after DOX treatment for 2 weeks demonstrated the
deregulation of miR-34a-5p and miR-451a levels, thereby resulting
in the cardiotoxicity for the first time in both plasma and heart
tissues. This article sheds light on specific replies of drug applica-
tions in patients, implying the recruitment of various players and
pathways [216].
The improvements in microRNA-based analyses provides vari-
ous potential therapeutic targets to inhibit tumor progression. One
of these miRNA-based therapeutics, which has passed into clinical
studies, is miR-34a-based liposomal miR-34a mimic drug MRX34.
This miRNA is included in tumorigenesis since its downregulation
causes the enhancement of more than 30 oncogenes in various
tumor types. The MRX34 application in 14 hepatocellular carci-
noma patients was given intravenously twice a week for 3 weeks of a
4-week duration for the first-in-human phase I study. The most
common adverse effects were fever, fatigue, and back pain, which
can be eliminated by taking dexamethasone as a premedication.
The dual application of these agents indicated the increased evi-
dence for their antitumor activity in the patient population with
advanced solid carcinomas [217]. Another miRNA-based drug
candidate, TargomiRs, are minicells which contain miR-16 mimic
miRNAs, whic are targeted to epidermal growth factor receptors
(EGFRs), and help diminish the effects of loss of the miR-15 and
miR-16 family in preclinical studies of a type of lung carcinoma,
malignant pleural mesothelioma. The first-in-man phase I Targo-
miR study in three different cancer centers in Sydney demonstrated
that dose-related toxicity in 26 patients who took at least one dose
of TargomiR caused various side effects such as transient lympho-
penia in 25 patients and cardiac events in five patients. The maxi-
mum tolerated dose is 5 109 TargomiRs in a week with
dexamethasone for the malignant pleural mesothelioma patients.
Twenty-one patients died during the TargomiR trial: 20 of these
deaths were related to tumor progression and one was due to
inflammatory bowel perforation. Collectively, TargomiRs should
be further developed through various studies combining with
immune checkpoint inhibitors and chemotherapy [218].
Improved drug candidates must proceed with their results for

better therapeutic methods with fewer adverse side effects. MiR-
NAs are small regulatory molecules, but they can be functional in
each step of tumorigenesis and further steps of cancerous cells. The
presence of these phase I studies sheds lights on the improvement
of miRNA based therapeutic studies, and these studies might pro-
ceed through phase IV by combining immune regulation, thereby
solving adverse side effect problems.
4 Conclusion and Future Perspectives
The expression levels of microRNAs maintain the balance of many

cellular processes including cell proliferation, cell growth, cell dif-
ferentiation, and cell death. These microRNAs are called as onco-
genic and tumor suppressor microRNAs. Recent studies about
oncogenic and tumor suppressor microRNAs are compiled in
Tables 3, 4, 5 and 6. The dysregulation of microRNAs (Fig. 2)
causes abnormalities in gene and protein expression levels because
microRNAs are involved in many regulatory processes that induce
or inhibit oncogenes and tumor suppressor genes, respectively.
This situation disrupts homeostasis in cells that cause cancer
formation by dysregulating signaling pathways. Recently, many
researchers have focused on the role of microRNAs in cancer initia-
tion, progression, and therapy as an important point of study for
cancer targeted treatment. The manipulation of microRNAs by
using inhibitors, drugs, and nanocarriers modulates cellular pro-
cesses. In addition, preclinical and clinical studies show that micro-
RNAs have a potential role in cancer treatment. According to
preclinical studies, the role of microRNAs in cancer cells are begin-
ning to be understood in respect to cancer progression and for
future clinical application. In addition, clinical studies continue
microRNA profiling of cancer patients to determine cancer pro-
gression and treatment response. Accordingly, studying the poten-
tial role of microRNAs in cancer involves understanding as well as
monitoring, controlling, and treating cancer. On the other hand,
many cancer researchers focus on using microRNA as a biomarker.
The expression level of different microRNAs is used as a biomarker
for diagnosis and prognosis of cancer. The cancer-associated micro-
RNAs in the biofluid of patients are detected by developing high-
throughput screening technology. At the same time, this helps
develop high-throughput screening technologies [219]. Both
these situations are critical to increasing diagnosis efficiency in the
early stages of cancer.
Table 3
Tumor suppressor microRNAs targets and effects in cellular proliferation and growth in cancer cells
microRNAs Cancer type Pathways References

MicroRNA- Osteosarcoma Breaks Akt3 activity [230]
1258
MicroRNA- Oral squamous cell Inhibition of IKKβ/NF-κB signaling pathway [231]
199a-5p
MicroRNA- Colorectal cancer cells Regulation of Smad3 [232]
140
miR-29a-3p Papillary thyroid carcinoma Suppressing NF-kB signaling [233]
MicroRNA- Gastric cancer cells, tissues Repression of WW domain-containing E3 [234]
129 and mice model ubiquitin protein ligase 1 (WWP1)
(miR-129-
5p
And
miR-129-
3p)
MicroRNA- Glioblastoma cells Regulation of c-MET/AKT pathway [235]
562
MicroRNA- Small cell lung cancer Suppression of the myeloid cell leukemia [236]
26b 1 expression
MicroRNA- Gastric cancer cell lines and Downregulation of FAK, MMP12, and JUN [237]
593-5p mice models protein expression.
MicroRNA- Hepatocellular Inhibition of cyclin B1 (CCNB1) expression [238]
144 Carcinoma cells
MicroRNA- Non–small-cell lung cancer Regulation of MYO6 expression [239]
5195-3p cell lines and tissues
MicroRNA- Breast cancer cells Targeting truncated neurokinin-1 receptor [240]
22 and ERα
miR-1254 Gastric cancer cell lines, Downregulation of Smurf1 and PI3K/Akt [241]
tissues and mice model signaling pathway
MicroRNA- Non–small-cell lung cancer Targeting the PI3K/AKT pathway and [242]
4458 cells HMGA1
MiR-542- Epithelial ovarian cancer Targeting cyclin-dependent kinase [67]
3p cells, tissues and mice 14 (CDK14)
model
cells and tissues
miR-377-5p Lung cancer cells Targeting AKT1 signaling [243]
microRNA- Non–small-cell lung cancer Downregulation of SOS2 expression and [244]
148a-3p modulation of Ras/MAPK/Erk signaling
MicroRNA- Hepatocellular carcinoma Downregulating regulators of G-protein [245]
199 cells and mice models signaling (RGS) RGS17
(continued)
Table 3
(continued)

miR-28-5p Colorectal cancer cells and Negatively regulation of structure-specific [246]
mice models recognition protein 1 (SSRP1)
MiR-219- Ovarian cancer cells and mice Targeting of high-mobility group A2 [247]
5p models (HMGA2) protein
MicroRNA- Colorectal cancer cell lines Decreasing expression of specificity protein [248]
382 and tissues 1 (SP1)
MicroRNA- Non-small cell lung cancer Targeting transforming growth factor beta [249]
98-5p cells and mice model receptor 1 (TGFBR1)
microRNAs- Breast cancer cells, tissues Targeting high mobility group protein B1 [250]
107 and mice model (HMGB1)
MicroRNA- Endometrial glandular Inhibition of Frizzled-7 (FZD7), cyclinD1, [251]
488 epithelial cells and mice β-catenin, and c-Myc and Wnt
model
MiR-183 Osteosarcoma cells and mice Downregulation of LRP6 Wnt/β-catenin [252]
models signaling pathway
MicroRNA- Colon cancer cells Downregulation of vimentin and CXCR4 [253]
193a-5p
miR-29b Gastric cancer cells and mice Negatively regulates MMP2 [254]
models
MiR-1287- Triple-negative breast cells Regulating phosphoinositide 3-kinase CB [255]
5p and mice models
MicroRNA- Breast cancer cells, tissues Downregulating CDC25A [256]
99a-5p and mice models
MicroRNA- Colorectal carcinoma cells Promotes cell apoptosis via claudin-1 [257]
98 and tissues
microRNA- Multiple myeloma cells Downregulation of MAPK/ERK signaling [258]
497 pathway and Raf-1
MiR-101- Glioblastoma Downregulates MMP-2, MMP-9 and [259]
3p mesenchymal markers and targeting
TRIM44
MicroRNA- Ovarian cancer cell Targeting a molecular chaperone, [260]
598 unconventional prefoldin RPB5 interactor
(URI)
miRNA-641 Cervical cancer cells, tissues Directly targeting zinc finger E-box binding [261]
and mice models homeobox 1 (ZEB1)
MiR-223- Lung squamous cell Directly targets P53 [262]
3p Carcinoma cells, tissues and
mice models
MicroRNA- Cervical cancer cells and Inhibition of PI3K/AKT/mTOR signaling [115]
99b tissues pathway
miR-216a- Small cell lung cancer cells Regulation of the Bcl-2 family proteins [263]
5p and mice model
(continued)
Table 3
(continued)

MiR-194 Nasopharyngeal carcinoma Suppression of MAP3K3 expression [264]
cell lines and tissues
MiR-130a- Breast cancer stem cell-like Suppression of RAB5B [265]
3p cells
MiR-1-5p Gallbladder carcinoma Targeting neurogenic locus notch homolog [266]
protein 2 (Notch2)
MicroRNA- Lung cancer cells and tissues Downregulation of sex determining region [267]
181b Y-related high mobility group-box
6 (Sox6)
Micro Gastric cancer cells, tissues Negatively regulation of MDM2 [268]
RNA-518 and mice models
MiR-125a- Bladder cancer cells Suppression of FUT4 [269]
5p
MiR-128 Thyroid carcinoma cells and Negatively regulation of sphingosine kinase-1 [270]
tissues (SPHK1)
miR-23c Hepatocellular carcinoma Suppression of erbb2 interacting protein [271]
cells, tissues and mice (ERBB2IP)
models
miR-133b Esophageal squamous cell Decreases EGFR and suppresses PI3K/AKT [272]
carcinoma and MAP/ERK signaling pathways
MicroRNA- Osteosarcoma cells Suppressing vesicle-associated membrane [273]
185 protein 2
MiR-1 Prostate cancer cells Negaticely regulated c-met/ Akt/ mTOR [274]
signaling
MicroRNA- Gastric cancer cells Downregulation of zeste homolog 2 (ZH2) [275]
31 expression
microRNA- Non–small cell lung cancer Wnt/β-catenin signaling pathway [276]
383 cells and mice models
MicroRNA- Laryngeal cancer cells and Directly targeting Wnt-induced secreted [277]
384 tissues protein-1 (WISP1) gene
miR-9-5p Pancreatic cancer cells and Targeting GOT1 expression [278]
tissues
miR-769 Colorectal cancer cells, Directly targeting cyclin-dependent kinase 1 [279]
tissues and mice models
miR-503 Cervical cancer cells Inhibiting AKT2 expression [280]
miR-23c Hepatocellular carcinoma Negatively regulating erbb2 interacting [271]
cells, tissues and mice protein (ERBB2IP)
model
MiR-34a Hepatocellular carcinoma Repressing hexokinase-1 [281]
cells, tissues and mice
models
Table 4
Oncogenic microRNAs targets and effects in cellular proliferation and growth in cancer cells

miR-135b Triple-negative breast cancer Downregulation of APC [282]
cells and tissues
MicroRNA Clear cell renal cell carcinoma Suppression of Forkhead box O3 [283]
122 cell lines and mice models
miR-183 Ovarian cancer cell lines and Regulation of TGF-β/Smad4 signaling [284]
tissues pathway
MicroRNA- Gastric cancer cells Targeting of Cullin-5 (CUL5) [285]
19a
MiR-940 Gastric cancer cells Upregulation of programmed death ligand-1 [286]
(PD-L1) expression
MicroRNA- Gastric cancer cells Increasing of PTEN/Akt signaling Axis [287]
21
MicroRNA- Endometrial Activating the Wnt/β-catenin pathway [288]
373 Cancer
miR-196a- Colorectal cancer cells and Inhibition of Nf-kB signaling [289]
5p mice models
MicroRNA- Cervical cancer cells Targeting thrombospondin-2 [290]
20a
MicroRNA- Colorectal cancer Suppressing pyruvate dehydrogenase [291]
23a lipoamide kinase isozyme 4 (PDK4)
MicroRNA- Hepatocellular carcinoma Regulation of SRY-related HMG-box [292]
645 cells, tissues and mice 30 (SOX30)
models
MicroRNA- Hepatocellular carcinoma cell Regulation of VGLL4 [293]
301a-3p and tissues
MicroRNA- Osteosarcoma cells Suppression of SASH1 expression and [294]
17 activation of PI3K/Akt signaling
miR-423- Prostate cancer cells, tissues Suppressing gene associated with retinoid- [295]
5p and mice models interferon-induced mortality 19 (GRIM-
19)
miR-183 Glioblastoma cells Downregulating leucine-rich repeats and [296]
immunoglobulin-like domains protein
1 (LRIG1)
MiR-214 Esophageal Regulation of PI3K/AKT/mTOR signaling [297]
Squamous cell carcinoma cell pathway
lines, tissues and mice
models
miR-500 Non-small cell lung Suppressing inhibitor of growth 1 (ING1) [298]
and Cancer
miR-628
(continued)
Table 4
(continued)

MicroRNA- Papillary thyroid cancer cell Upregulating suppressor of cancer cell [299]
1270 lines, tissues and mice invasion (SCAI)
models
miR-150- Cervical carcinoma cells Suppression SRC kinase signaling inhibitor [300]
5p 1 (SRCIN1)
miR-210 Lung adenocarcinoma Targeting lysyl oxidase-like 4 [301]
MicroRNA- Osteosarcoma Targeting CDKN1B/p27 [302]
221
miR-501 Cervical cancer cells Targeting cylindromatosis (CYLD) [303]
microRNA- Pancreatic ductal Downregulation of Glypican-5 [304]
4295 adenocarcinoma cells
Table 5
Tumor suppressor microRNAs targets and effects in invasion and migration in cancer cells
microRNAs Cancer type Targets of microRNA References

MicroRNA-26b Small cell lung cancer Suppression of the myeloid cell leukemia [236]
1 expression
MicroRNA Glioblastoma cells Downregulation of TOP2A expression [305]
144-3p
MicroRNA Gastric cancer cells, tissues Repression of WW domain-containing E3 [234]
129 and mice model ubiquitin protein ligase 1 (WWP1)
(miR-129-5p
and -3p)
4458 cells HMGA1
MicroRNA- Non–small-cell lung cancer Regulation of MYO6 expression [239]
5195-3p cell lines and tissues
MicroRNA-22 Breast cancer Targeting truncated neurokinin-1 [240]
receptor and ERα
MicroRNA- Gastric cancer cell lines and Downregulation of FAK, MMP12, and [237]
593-5p mice models JUN protein expression.
MicroRNA-144 Hepatocellular carcinoma Inhibition of cyclin B1 (CCNB1) [238]
cell lines and tissues expression
MicroRNA-140 Colorectal cancer cells and Regulation of Smad3 [232]
mice models
miR-1254 Gastric cancer cell lines, Downregulation of Smurf1 and PI3K/Akt [241]
tissues and mice model signaling pathway
(continued)
Table 5
(continued)

4458 cells HMGA1
MiR-542-3p Epithelial ovarian cancer Targeting cyclin-dependent kinase [67]
cells, tissues and mice 14 (CDK14)
model
miR-377-5p Lung cancer cells and tissues Targeting AKT1 signaling [243]
MicroRNA- Non–small-cell lung cancer Downregulation of SOS2 expression and [244]
148a-3p modulation of Ras/MAPK/ERK
signaling
MicroRNA-199 Hepatocellular carcinoma Downregulating regulators of G-protein [245]
cells and mice models signaling (RGS) RGS17
miR-145 Non–small cell lung cancer Negatively regulates ZEB2 gene [306]
MicroRNA-30e Prostate cancer cells and Downregulation of CHRM3 and [307]
tissues inhibition of MAPK signaling pathway
miR-28-5p Colorectal cancer cells and Negatively regulation of structure-specific [246]
mice models recognition protein 1 (SSRP1)
MiR-219-5p Ovarian cancer cells Targeting of high-mobility group A2 [247]
(HMGA2) protein
MicroRNA-588 Gastric cancer cells, tissues Suppression of EIF5A2 mRNA expression [308]
and mice models
MicroRNA-382 Colorectal cancer cell lines Decreasing expression of specificity [248]
and tissues protein 1 (SP1)
MicroRNA-98- Non-small cell lung cancer Targeting transforming growth factor beta [249]
5p cells and mice model receptor 1 (TGFBR1)
microRNAs- Breast cancer cells, tissues Targeting high mobility group protein B1 [250]
107 and mice model (HMGB1)
MicroRNA-488 Endometrial glandular Inhibition of Frizzled-7 (FZD7), [251]
epithelial cells and mice cyclinD1, β-catenin, and c-Myc and
model Wnt
MiR-183 Osteosarcoma cells and mice Downregulating LRP6-Wnt/β-catenin [252]
models signaling pathway
miR-29b Gastric cancer cells and mice Negatively regulates MMP2 [254]
models
MicroRNA-98 Colorectal carcinoma cells Promotes cell apoptosis via claudin-1 [257]
and tissues
MicroRNA-497 Multiple myeloma cells Downregulation of MAPK/ERK [258]
signaling pathway and Raf-1
(continued)
Table 5
(continued)

MiR-101-3p Glioblastoma Downregulates MMP-2, MMP-9, and [259]
mesenchymal markers and targeting
TRIM44
MicroRNA486- Colorectal cancer cells Directly targets BH3-only family of [309]
3p proapoptotic proteins BIK
MicroRNA-31 Gastric cancer cells Downregulation of zeste homolog [275]
2 (ZH2) expression
MicroRNA-598 Ovarian cancer cell Targeting a molecular chaperone, [260]
unconventional prefoldin RPB5
interactor (URI)
miR-125a Gastric cancer cells Suppression of STAT3 [310]
MicroRNA- Oral squamous cell Targeting SP1 and inhibition of PI3K/ [311]
1258 carcinoma cells, tissues AKT and ERK signaling pathway
and mice models
miRNA-641 Cervical cancer cells, tissues Directly targeting zinc finger E-box [261]
and mice models binding homeobox 1 (ZEB1)
MicroRNA- Endometrial cancer cells and Suppressing steroid receptor [312]
449a tissues Coactivator (SRC) and inactivation of
AKT/ERK1/2
MiR-194 Nasopharyngeal carcinoma Suppression of MAP3K3 expression [264]
miR-216a-5p Small cell lung cancer cells Regulation of the Bcl-2 family proteins [263]
and mice model
miR-9-5p Pancreatic cancer cells and Targeting GOT1 expression [278]
tissues
MicroRNA-99b Cervical cancer cells and Inhibition of PI3K/AKT/mTOR [115]
tissues signaling pathway
MiR-223-3p Lung squamous cell Directly targets P53 [262]
Carcinoma cells, tissues and
mice models
MiR-130a-3p Breast cancer stem cell-like Suppression of RAB5B [265]
cells
MiR-873-5p Colorectal cancer cells Targeting zinc finger E-box-binding [313]
homeobox 1 (ZEB1)
MiR-1-5p Gallbladder carcinoma cells Targeting neurogenic locus notch [266]
homolog protein 2 (Notch2)
MicroRNA- Lung cancer cells and tissues Downregulation of sex determining [267]
181b region Y-related high mobility group-
box 6 (Sox6)
(continued)
Table 5
(continued)

MicroRNA- Colorectal cancer cells Negatively regulates transcription factor [314]
302c AP-4
MiR-125a-5p Bladder cancer cells Suppression of FUT4 [269]
MiR-128 Thyroid carcinoma cells and Negatively regulation of sphingosine [270]
tissues kinase-1 (SPHK1)
miR-139-5p Prostate cancer cells and Repressing sex-determining region Y-box [315]
tissues protein 5 (SOX5) expression
MicroRNA-31 Gastric cancer cells Downregulation of zeste homolog [275]
2 (ZH2) expression
miR-769 Colorectal cancer cells, Directly targeting cyclin-dependent kinase [279]
tissues and mice models 1
MiR-515-5p Prostate cancer cells Directly regulating thyroid hormone [316]
receptor interactor 13 (TRIP13)
MicroRNA-185 Osteosarcoma cells Suppressing vesicle-associated [273]
membraneprotein 2
microRNA- Bladder cancer cells Regulation of HMGB3/Wnt/β-catenin [317]
532-5p signaling
microRNA- Nasopharyngeal carcinoma Negatively regulating FOXQ1 [318]
342-3p cells, tissues and mice
models
miR-506-3p Osteosarcoma cells Inhibiting sphingosine kinase 1 (SPHK1) [319]
miR-133b Esophageal squamous cell Decreases EGFR and suppresses PI3K/ [272]
carcinoma AKT and MAP/ERK signaling
pathways
Table 6
Oncogenic microRNAs targets and effects in invasion and migration in cancer cells

miR-135b Triple-negative breast cancer cells and Downregulation of APC [282]
tissues
MicroRNA- Gastric cancer cells Targeting of Cullin-5 (CUL5) [285]
19a
MicroRNA- Gastric cancer cells and mice models Upregulation of programmed death [286]
940 ligand-1 (PD-L1) expression
MicroRNA- Gastric cancer cell Increasing of PTEN/Akt signaling [287]
21 Axis
(continued)
Table 6
(continued)

MicroRNA- Endometrial cancer Activating the Wnt/β-catenin [288]
373 pathway
MicroRNA- Bladder cancer Targeting PEDF [320]
93
MicroRNA Clear cell renal cell carcinoma cell Suppression of Forkhead box O3 [282]
122 lines and mice models
MicroRNA- Nasopharyngeal carcinoma cells and Negatively regulation of Sox7 [320]
494-3p tissues
miR-196a- Colorectal cancer cells and mice Inhibition of Nf-kB signaling [288]
5p models
MicroRNA- Osteosarcoma cells Suppression of SASH1 expression and [293]
17 activation of PI3K/Akt signaling
MiR-214 Esophageal squamous cell carcinoma Regulation of PI3K/AKT/mTOR [296]
cell lines, tissues and mice models signaling pathway
miR-500 Non-small cell lung Suppressing inhibitor of growth [297]
and Cancer 1 (ING1)
miR-628
MicroRNA- Papillary thyroid cancer cell lines, Upregulating suppressor of cancer cell [298]
1270 tissues and mice models invasion (SCAI)
miR-3191 Colorectal cancer Downregulating transforming growth [321]
factor BETA (TGF-BETA)
miR-150- Cervical carcinoma cells Suppression SRC kinase signaling [299]
5p inhibitor 1 (SRCIN1)
miR-210 Lung adenocarcinoma Targeting lysyl oxidase-like 4 [300]
MicroRNA- Osteosarcoma Targeting CDKN1B/p27 [301]
221
miR-501 Cervical cancer cells Targeting cylindromatosis (CYLD) [302]
microRNA- Pancreatic ductal adenocarcinoma Downregulation of Glypican-5 [303]
4295 cells
miR-532- Hepatocellular carcinoma cells and Targeting receptor protein tyrosine [322]
3p tissues phosphatase T (PTPRT)
miR-376c- Hepatocellular carcinoma cells, Suppressing AT-rich [45]
3p tissues and mice models Interaction domain 2
Fig. 2 The dysregulation of microRNA biogenesis in cancer pathogenesis
References
1. Anand P, Kunnumakkara AB, Sundaram C diagnostic marker- a review. Adv Biomarker

et al (2008) Cancer is a preventable disease Sci Technol 1:1–8
that requires major lifestyle changes. Pharm 4. Farazi TA, Hoell JI, Morozov P et al (2013)
Res 25:2097–2116 MicroRNAs in human cancer. Adv Exp Med
2. Oliveto S, Mancino M, Manfrini N et al Biol 774:1–20
(2017) Role of microRNAs in translation reg- 5. Wang J, Yu J-T, Tan L et al (2015) Genome-
ulation and cancer. World J Biol Chem wide circulating microRNA expression
8:45–56 profiling indicates biomarkers for epilepsy.
3. Bhat SA, Majid S, Hassan T (2019) Micro- Sci Rep 5:9522
RNAs and its emerging role as breast cancer
6. Monroig PDC, Chen L, Zhang S et al (2015) matrix Metalloproteinase-14. Cancer Res

Small molecule compounds targeting miR- 75:2674–2685
NAs for cancer therapy. Adv Drug Deliv Rev 21. O’Bryan S, Dong S, Mathis JM et al (2017)
81:104–116 The roles of oncogenic miRNAs and their
7. Fernandez-Piñeiro I, Badiola I, Sanchez A therapeutic importance in breast cancer. Eur
(2017) Nanocarriers for microRNA delivery J Cancer 72:1–11
in cancer medicine. Biotechnol Adv 22. Staedel C, Tran TPA, Giraud J et al (2018)
35:350–360 Modulation of oncogenic miRNA biogenesis
8. Dykes IM, Emanueli C (2017) Transcrip- using functionalized polyamines. Sci Rep
tional and post-transcriptional gene regula- 8:1667
tion by long non-coding RNA. Genomics 23. O’Brien J, Hayder H, Zayed Y et al (2018)
Proteomics Bioinformatics 15:177–186 Overview of MicroRNA biogenesis, mechan-
9. Lee KT, Nam JW (2017) Post-transcriptional isms of actions, and circulation. Front Endo-
and translational regulation of mRNA-like crinol 9:402
long non-coding RNAs by microRNAs in 24. Fuziwara CS, Kimura ET (2015) Insights into
early developmental stages of zebrafish regulation of the miR-17-92 cluster of miR-
embryos. BMB Rep 50:226–231 NAs in cancer. Front Med 2:64
10. Wang Y, Lee CGL (2009) MicroRNA and 25. Labi V, Schoeler K, Melamed D (2019)
cancer--focus on apoptosis. J Cell Mol Med miR-1792 in lymphocyte development and
13:12–23 lymphomagenesis. Cancer Lett 446:73–80
11. Peng Y, Croce CM (2016) The role of Micro- 26. Liu F, Cheng L, Xu J et al (2018) miR-17-92
RNAs in human cancer. Signal Transduct Tar- functions as an oncogene and modulates
get Ther 1:15004 NF-kappaB signaling by targeting TRAF3 in
12. Ramassone A, Pagotto S, Veronese A et al MGC-803 human gastric cancer cells. Int J
(2018) Epigenetics and MicroRNAs in can- Oncol 53:2241–2257
cer. Int J Mol Sci 19 27. Zhou P, Ma L, Zhou J et al (2016) miR-17-
13. Olejniczak M, Kotowska-Zimmer A, Krzyzo- 92 plays an oncogenic role and conveys
siak W (2018) Stress-induced changes in chemo-resistance to cisplatin in human pros-
miRNA biogenesis and functioning. Cell tate cancer cells. Int J Oncol 48:1737–1748
Mol Life Sci 75:177–191 28. Chaluvally-Raghavan P, Jeong KJ, Pradeep S
14. Bracken CP, Scott HS, Goodall GJ (2016) A et al (2016) Direct upregulation of STAT3 by
network-biology perspective of microRNA MicroRNA-551b-3p deregulates growth and
function and dysfunction in cancer. Nat Rev metastasis of ovarian cancer. Cell Rep
Genet 17:719–732 15:1493–1504
15. Lin S, Gregory RI (2015) MicroRNA biogen- 29. Acunzo M, Croce CM (2016) Downregula-
esis pathways in cancer. Nat Rev Cancer tion of miR-15a and miR-16-1 at 13q14 in
15:321–333 chronic lymphocytic leukemia. Clin Chem
16. Davis BN, Hata A (2009) Regulation of 62:655–656
MicroRNA biogenesis: a miRiad of mechan- 30. Yan H, Zhou M, Bhattarai U et al (2019)
isms. Cell Commun Signal 7:18–18 Cyclic Peptidomimetics as inhibitor for
17. Venkatadri R, Muni T, Iyer AKV et al (2016) miR-155 biogenesis. Mol Pharm 16:914–920
Role of apoptosis-related miRNAs in 31. Lim D, Byun WG, Park SB (2018) Restoring
resveratrol-induced breast cancer cell death. Let-7 microRNA biogenesis using a small-
Cell Death Dis 7:e2104 molecule inhibitor of the protein–RNA inter-
18. Xu Y, Huang J, Ma L et al (2016) MicroRNA- action. ACS Med Chem Lett 9:1181–1185
122 confers sorafenib resistance to hepatocel- 32. Lawrie CH, Gal S, Dunlop HM et al (2008)
lular carcinoma cells by targeting IGF-1R to Detection of elevated levels of tumour-
regulate RAS/RAF/ERK signaling pathways. associated microRNAs in serum of patients
Cancer Lett 371:171–181 with diffuse large B-cell lymphoma. Br J Hae-
19. Morgado AL, Rodrigues CM, Solá S (2016) matol 141:672–675
MicroRNA-145 regulates neural stem cell dif- 33. Sabry D, El-Deek SEM, Maher M et al (2018)
ferentiation through the Sox2-Lin28/let-7 Role of miRNA-210, miRNA-21 and
signaling pathway. Stem Cells 34:1386–1395 miRNA-126 as diagnostic biomarkers in colo-
20. Li Y, Kuscu C, Banach A et al (2015) rectal carcinoma: impact of HIF-1alpha-
miR-181a-5p inhibits cancer cell migration VEGF signaling pathway. Mol Cell Biochem
and angiogenesis via downregulation of 454(1–2):177–189. https://doi.org/10.
1007/s11010-018-3462-1
34. Anastasiadou E, Jacob LS, Slack FJ (2017) repressing AT-rich interaction domain 2. J
Non-coding RNA networks in cancer. Nat Cancer 9:4187–4196
Rev Cancer 18:5 46. Guo H, Ji F, Zhao X et al (2019) MicroRNA-
35. Mizoguchi A, Takayama A, Arai T et al (2018) 371a-3p promotes progression of gastric can-
MicroRNA-8073: tumor suppressor and cer by targeting TOB1. Cancer Lett
potential therapeutic treatment. PLoS One 443:179–188
13:e0209750 47. Elmore S (2007) Apoptosis: a review of pro-
36. Wang B, Chai C, Wu H et al (2018) grammed cell death. Toxicol Pathol
MicroRNA-498 promotes proliferation and 35:495–516
migration by targeting the tumor suppressor 48. Goldar S, Khaniani MS, Derakhshan SM et al
PTEN in breast cancer cells. Carcinogenesis (2015) Molecular mechanisms of apoptosis
39:1185–1196 and roles in cancer development and treat-
37. Gonzales-Aloy E, Grebert-Wade DG, Wang ment. Asian Pac J Cancer Prev 16:2129–2144
JY (2018) Abstract 484: tumor suppressor 49. Flusberg DA, Sorger PK (2015) Surviving
miRNA-101 modulates leukemogenesis by apoptosis: life-death signaling in single cells.
targeting the EZH2/Wnt/β-catenin signal- Trends Cell Biol 25:446–458
ing pathways. Cancer Res 78:484–484 50. Birkinshaw RW, Czabotar PE (2017) The
38. Bowles KM, Rushworth SA (2018) Targeting BCL-2 family of proteins and mitochondrial
the KEAP1/NRF2 pathway to manipulate outer membrane permeabilisation. Semin Cell
the expression of oncogenic and oncosuppres- Dev Biol 72:152–162
sive miRNAs in human leukemia. Mol Cell 51. Edlich F (2018) BCL-2 proteins and apopto-
Oncol 5:e988484 sis: recent insights and unknowns. Biochem
39. Song S, Yang Y, Liu M et al (2018) MiR-125b Biophys Res Commun 500:26–34
attenuates human hepatocellular carcinoma 52. Wang YB, Zhao XH, Li G et al (2018)
malignancy through targeting SIRT6. Am J MicroRNA-184 inhibits proliferation and
Cancer Res 8:993–1007 promotes apoptosis of human colon cancer
40. Witten LW, Cheng CJ, Slack FJ (2018) SW480 and HCT116 cells by downregulating
miR-155 drives oncogenesis by promoting C-MYC and BCL-2. J Cell Biochem
and cooperating with mutations in the c-Kit 119:1702–1715
oncogene. Oncogene 38(12):2151–2161. 53. Chen MJ, Wu DW, Wang GC et al (2018)
https://doi.org/10.1038/s41388-018- MicroRNA-630 may confer favorable
0571-y cisplatin-based chemotherapy and clinical
41. Yan L, Wang S, Li Y et al (2018) SNHG5 outcomes in non-small cell lung cancer by
promotes proliferation and induces apoptosis targeting Bcl-2. Oncotarget 9:13758–13767
in melanoma by sponging miR-155. RSC Adv 54. Liu HN, Qie P, Yang G et al (2018)
8:6160–6168 miR-181b inhibits chemoresistance in
42. Tran DDH, Kessler C, Niehus SE et al (2017) cisplatin-resistant H446 small cell lung cancer
Myc target gene, long intergenic noncoding cells by targeting Bcl-2. Arch Med Sci
RNA, Linc00176 in hepatocellular carcinoma 14:745–751
regulates cell cycle and cell survival by titrat- 55. Deng M, Yuan H, Liu S et al (2019)
ing tumor suppressor microRNAs. Oncogene Exosome-transmitted LINC00461 promotes
37:75 multiple myeloma cell proliferation and sup-
43. Zhang Q, Lv R, Guo W et al (2018) presses apoptosis by modulating microRNA/
microRNA-802 inhibits cell proliferation and BCL-2 expression. Cytotherapy 21:96–106
induces apoptosis in human cervical cancer by 56. Aakko S, Straume AH, Birkeland EE et al
targeting serine/arginine-rich splicing factor (2019) MYC-induced miR-203b-3p and
9. J Cell Biochem 120(6):10370–10379. miR-203a-3p control Bcl-xL expression and
https://doi.org/10.1002/jcb.28321 paclitaxel sensitivity in tumor cells. Transl
44. Jiang J, Yang X, He X et al (2019) Micro- Oncol 12:170–179
RNA-449b-5p suppresses the growth and 57. Pan Y, Ye C, Tian Q et al (2018) miR-145
invasion of breast cancer cells via inhibiting suppresses the proliferation, invasion and
CREPT-mediated Wnt/beta-catenin signal- migration of NSCLC cells by regulating the
ing. Chem Biol Interact 302:74–82 BAX/BCL-2 ratio and the caspase-3 cascade.
45. Wang Y, Chang W, Chang W et al (2018) Oncol Lett 15:4337–4343
MicroRNA-376c-3p facilitates human hepa- 58. Zhou M, Liu Z, Zhao Y et al (2010)
tocellular carcinoma progression via MicroRNA-125b confers the resistance of
breast cancer cells to paclitaxel through
suppression of pro-apoptotic Bcl-2 antagonist and related processes. Chromosoma

killer 1 (Bak1) expression. J Biol Chem 127:187–214
285:21496–21507 71. Sugasawa K (2019) Molecular mechanism of
59. Kong F, Sun C, Wang Z et al (2011) DNA damage recognition for global genomic
miR-125b confers resistance of ovarian cancer nucleotide excision repair: a defense system
cells to cisplatin by targeting pro-apoptotic against UV-induced skin cancer. In:
Bcl-2 antagonist killer 1. J Huazhong Univ Nishigori C, Sugasawa K (eds) DNA repair
Sci Technolog Med Sci 31:543 disorders. Springer Singapore, Singapore, pp
60. Zaman S, Wang R, Gandhi V (2014) Target- 1–23
ing the apoptosis pathway in hematologic 72. Rudolph C, Melau C, Nielsen JE et al (2017)
malignancies. Leuk Lymphoma Involvement of the DNA mismatch repair sys-
55:1980–1992 tem in cisplatin sensitivity of testicular germ
61. Zhu Y, Tang H, Zhang L et al (2019) Sup- cell tumours. Cell Oncol 40:341–355
pression of miR-21-3p enhances TRAIL- 73. Patel N, Garikapati KR, Pandita RK et al
mediated apoptosis in liver cancer stem cells (2017) miR-15a/miR-16 down-regulates
by suppressing the PI3K/Akt/bad cascade via BMI1, impacting Ub-H2A mediated DNA
regulating PTEN. Cancer Manag Res repair and breast cancer cell sensitivity to
11:955–968 doxorubicin. Sci Rep 7:4263
62. Zheng M, Wu Z, Wu A et al (2016) MiR-145 74. Choi YE, Meghani K, Brault M-E et al (2016)
promotes TNF-α-induced apoptosis by facil- Platinum and PARP inhibitor resistance due
itating the formation of RIP1-FADDcaspase- to overexpression of MicroRNA-622 in
8 complex in triple-negative breast cancer. BRCA1-mutant ovarian cancer. Cell Rep
Tumor Biol 37:8599–8607 14:429–439
63. Cui J, Wei C, Deng L et al (2018) Micro- 75. Li J-H, Luo N, Zhong M-Z et al (2016)
RNA143 increases cell apoptosis in myelodys- Inhibition of microRNA-196a might reverse
plastic syndrome through the Fas/FasL cisplatin resistance of A549/DDP non–small-
pathway both in vitro and in vivo. Int J cell lung cancer cell line. Tumor Biol
Oncol 53:2191–2199 37:2387–2394
64. Pang X, Zhou Z, Yu Z et al (2019) Foxo3a- 76. Du P, Zhao H, Peng R et al (2017) LncRNA-
dependent miR-633 regulates chemothera- XIST interacts with <em>miR-29c</em>
peutic sensitivity in gastric cancer by targeting to modulate the chemoresistance of glioma
Fas-associated death domain. RNA Biol cell to TMZ through DNA mismatch repair
16:233–248 pathway. Biosci Rep 37:BSR20170696
65. Shukla K, Sharma AK, Ward A et al (2015) 77. Lu Y, Qin T, Li J et al (2017) MicroRNA-
MicroRNA-30c-2-3p negatively regulates 140-5p inhibits invasion and angiogenesis
NF-kappaB signaling and cell cycle progres- through targeting VEGF-A in breast cancer.
sion through downregulation of TRADD and Cancer Gene Ther 24:386–392
CCNE1 in breast cancer. Mol Oncol 78. Yachi K, Tsuda M, Kohsaka S et al (2018)
9:1106–1119 miR-23a promotes invasion of glioblastoma
66. Guo T, Zhang Y, Qu X et al (2018) miR-200a via HOXD10-regulated glial-mesenchymal
enhances TRAIL-induced apoptosis in gastric transition. Signal Transduct Target Ther 3:33
cancer cells by targeting A20. Cell Biol Int 79. Amaar YG, Reeves ME (2019) RASSF1C reg-
42:506–514 ulates miR-33a and EMT marker gene expres-
67. Li X, Chen W, Jin Y et al (2019) miR-142-5p sion in lung cancer cells. Oncotarget
enhances cisplatin-induced apoptosis in ovar- 10:123–132
ian cancer cells by targeting multiple anti- 80. Chen X, Zeng K, Xu M et al (2019)
apoptotic genes. Biochem Pharmacol P53-induced miR-1249 inhibits tumor
161:98–112 growth, metastasis, and angiogenesis by tar-
68. Jeggo PA, Pearl LH, Carr AM (2015) DNA geting VEGFA and HMGA2. Cell Death Dis
repair, genome stability and cancer: a histori- 10:131
cal perspective. Nat Rev Cancer 16:35 81. Jiang J, Xie C, Liu Y et al (2019)
69. Hu H, Gatti RA (2010) MicroRNAs: new Up-regulation of miR-383-5p suppresses pro-
players in the DNA damage response. J Mol liferation and enhances chemosensitivity in
Cell Biol 3:151–158 ovarian cancer cells by targeting TRIM27.
70. Ranjha L, Howard SM, Cejka P (2018) Main Biomed Pharmacother 109:595–601
steps in DNA double-strand break repair: an 82. Luan T, Fu S, Huang L et al (2018)
introduction to homologous recombination MicroRNA-98 promotes drug resistance and
regulates mitochondrial dynamics by target- 94. El Bezawy R, Tinelli S, Tortoreto M et al

ing LASS2 in bladder cancer cells. Exp Cell (2019) miR-205 enhances radiation sensitiv-
Res 373:188–197 ity of prostate cancer cells by impairing DNA
83. Shi L, Xi J, Xu X et al (2019) MiR-148a damage repair through PKCepsilon and
suppressed cell invasion and migration via tar- ZEB1 inhibition. J Exp Clin Cancer Res
geting WNT10b and modulating beta- 38:51
catenin signaling in cisplatin-resistant colorec- 95. Pinweha P, Rattanapornsompong K, Char-
tal cancer cells. Biomed Pharmacother oensawan V et al (2016) MicroRNAs and
109:902–909 oncogenic transcriptional regulatory net-
84. Park GB, Kim D (2019) MicroRNA-503-5p works controlling metabolic reprogramming
inhibits the CD97-mediated JAK2/STAT3 in cancers. Comput Struct Biotechnol J
pathway in metastatic or paclitaxel-resistant 14:223–233
ovarian cancer cells. Neoplasia 21:206–215 96. Tomasetti M, Amati M, Santarelli L et al
85. Lee SD, Yu D, Lee DY et al (2019) Upregu- (2016) MicroRNA in metabolic
lated microRNA-193a-3p is responsible for re-programming and their role in tumorigen-
cisplatin resistance in CD44(+) gastric cancer esis. Int J Mol Sci 17:754
cells. Cancer Sci 110:662–673 97. Warburg O, Wind F, Negelein E (1927) The
86. Liu Y, Lei P, Qiao H et al (2019) miR-9 metabolism of tumors in the body. J Gen
enhances the Chemosensitivity of AML cells Physiol 8:519
to Daunorubicin by targeting the EIF5A2/ 98. Vander Heiden MG, Cantley LC, Thompson
MCL-1 Axis. Int J Biol Sci 15:579–586 CB (2009) Understanding the Warburg
87. Zhan T, Huang X, Tian X et al (2018) Down- effect: the metabolic requirements of cell pro-
regulation of MicroRNA-455-3p links to proliferation. Science 324:1029
liferation and drug resistance of pancreatic 99. Shankaraiah RC, Veronese A, Sabbioni S et al
cancer cells via targeting TAZ. Mol Ther (2018) Non-coding RNAs in the reprogram-
Nucleic Acids 10:215–226 ming of glucose metabolism in cancer. Cancer
88. Zhao T, Chen Y, Sheng S et al (2018) Upre- Lett 419:167–174
gulating microRNA-498 inhibits gastric can- 100. Yang J, Li J, Le Y et al (2016) PFKL/miR-
cer proliferation invasion and chemoresistance 128 axis regulates glycolysis by inhibiting
through inverse interaction of Bmi1. Cancer AKT phosphorylation and predicts poor sur-
Gene Ther 26(11–12):366–373. https://doi. vival in lung cancer. Am J Cancer Res
org/10.1038/s41417-018-0065-7 6:473–485
89. Zhang Y, Xia F, Zhang F et al (2019) 101. Chen J, Yu Y, Chen X et al (2018) MiR-139-
miR-135b-5p enhances doxorubicin- 5p is associated with poor prognosis and reg-
sensitivity of breast cancer cells through tar- ulates glycolysis by repressing PKM2 in gall-
geting anterior gradient 2. J Exp Clin Cancer bladder carcinoma. Cell Prolif 51:e12510
Res 38:26 102. Yu G, Sun W, Shen Y et al (2018) PKM2
90. Chen X, Zhu H, Ye W et al (2019) Micro- functions as a potential oncogene and is a
RNA29a enhances cisplatin sensitivity in non- crucial target of miR-148a and miR-326 in
small cell lung cancer through the regulation thyroid tumorigenesis. Am J Transl Res
of REV3L. Mol Med Rep 19:831–840 10:1793–1801
91. Turato C, Fornari F, Pollutri D et al (2019) 103. Li L, Kang L, Zhao W et al (2017) miR-30a-
MiR-122 targets SerpinB3 and is involved in 5p suppresses breast tumor growth and
Sorafenib resistance in hepatocellular carci- metastasis through inhibition of LDHA-
noma. J Clin Med 8:171 mediated Warburg effect. Cancer Lett
92. Song YK, Wang Y, Wen YY et al (2018) 400:89–98
MicroRNA-22 suppresses breast cancer cell 104. Kim S, Lee E, Jung J et al (2018) microRNA-
growth and increases paclitaxel sensitivity by 155 positively regulates glucose metabolism
targeting NRAS. Technol Cancer Res Treat via PIK3R1-FOXO3a-cMYC axis in breast
17:1533033818809997 cancer. Oncogene 37:2982–2991
93. Wang Z, Mao JW, Liu GY et al (2019) 105. Xiaohong Z, Lichun F, Na X et al (2016)
MicroRNA-372 enhances radiosensitivity MiR-203 promotes the growth and migration
while inhibiting cell invasion and metastasis of ovarian cancer cells by enhancing glycolytic
in nasopharyngeal carcinoma through activat- pathway. Tumour Biol 37:14989–14997
ing the PBK-dependent p53 signaling path- 106. Liu Q, Luo Q, Halim A et al (2017) Target-
way. Cancer Med 8:712–728 ing lipid metabolism of cancer cells: a
promising therapeutic strategy for cancer. 119. Zhao H, Su W, Kang Q et al (2018) Natural
Cancer Lett 401:39–45 killer cells inhibit oxaliplatin-resistant colorec-
107. Ding D, Ye G, Lin Y et al (2019) MicroRNA- tal cancer by repressing WBSCR22 via upre-
26a-CD36 signaling pathway: pivotal role in gulating microRNA-146b-5p. Am J Cancer
lipid accumulation in hepatocytes induced by Res 8:824–834
PM2.5 liposoluble extracts. Environ Pollut 120. Guo X, Qiu W, Liu Q et al (2018) Immuno-
248:269–278 suppressive effects of hypoxia-induced glioma
108. Borji M, Nourbakhsh M, Shafiee SM et al exosomes through myeloid-derived suppres-
(2019) Down-regulation of SIRT1 expres- sor cells via the miR-10a/Rora and miR-21/
sion by mir-23b contributes to lipid accumu- Pten pathways. Oncogene 37:4239–4259
lation in HepG2 cells. Biochem Genet 57 121. Givel AM, Kieffer Y, Scholer-Dahirel A et al
(4):507–521. https://doi.org/10.1007/ (2018) miR200-regulated CXCL12beta pro-
s10528-019-09905-5 motes fibroblast heterogeneity and immuno-
109. Li G, Li M, Hu J et al (2017) The microRNA- suppression in ovarian cancers. Nat Commun
182-PDK4 axis regulates lung tumorigenesis 9:1056
by modulating pyruvate dehydrogenase and 122. Porzycki P, Ciszkowicz E, Semik M et al
lipogenesis. Oncogene 36:989–998 (2018) Combination of three miRNA
110. Cheng L, Zhu Y, Han H et al (2017) (miR-141, miR-21, and miR-375) as poten-
MicroRNA-148a deficiency promotes hepatic tial diagnostic tool for prostate cancer recog-
lipid metabolism and hepatocarcinogenesis in nition. Int Urol Nephrol 50:1619–1626
mice. Cell Death Dis 8:e2916 123. Bai X, Liu Z, Shao X et al (2019) The hetero-
111. Xi J, Huang Q, Wang L et al (2018) miR-21 geneity of plasma miRNA profiles in hepato-
depletion in macrophages promotes tumori- cellular carcinoma patients and the
cidal polarization and enhances PD-1 immu- exploration of diagnostic circulating miRNAs
notherapy. Oncogene 37:3151–3165 for hepatocellular carcinoma. PLoS One 14:
112. Botta C, Cucè M, Pitari MR et al (2017) e0211581
MiR-29b antagonizes the pro-inflammatory 124. Liu X, Xu X, Pan B et al (2019) Circulating
tumor-promoting activity of multiple miR-1290 and miR-320d as novel diagnostic
myeloma-educated dendritic cells. Leukemia biomarkers of human colorectal cancer. J
32:1003 Cancer 10:43–50
113. Graff JW, Dickson AM, Clay G et al (2012) 125. Wang X, Liang JQ, Zhang LH et al (2019)
Identifying functional microRNAs in macro- C8orf76 promotes gastric tumorigenicity and
phages with polarized phenotypes. J Biol metastasis by directly inducing lncRNA
Chem 287:21816–21825 DUSP5P1 and associates with patient out-
114. Li N, Qin JF, Han X et al (2018) miR-21a comes. Clin Cancer Res 25(10):3128–3140.
negatively modulates tumor suppressor genes https://doi.org/10.1158/1078-0432.Ccr-
PTEN and miR-200c and further promotes 18-2804
the transformation of M2 macrophages. 126. Yao C, Liu HN, Wu H et al (2018) Diagnos-
Immunol Cell Biol 96:68–80 tic and prognostic value of circulating Micro-
115. Liu R, Lu Z, Gu J et al (2018) MicroRNAs RNAs for esophageal squamous cell
15A and 16-1 activate signaling pathways that carcinoma: a systematic review and meta-
mediate chemotaxis of immune regulatory B analysis. J Cancer 9:2876–2884
cells to colorectal tumors. Gastroenterology 127. Goto T, Fujiya M, Konishi H et al (2018) An
154:637–651 e637 elevated expression of serum exosomal
116. Zhang Q, Xin H, Fen T (2018) Function of microRNA-191, 21, 451a of pancreatic
microRNA141 in human breast cancer neoplasm is considered to be efficient diag-
through cytotoxic CD4+ T cells regulated by nostic marker. BMC Cancer 18:116
MAP4K4 expression. Mol Med Rep 128. Liu HN, Wu H, Tseng YJ et al (2018) Serum
17:7893–7901 microRNA signatures and metabolomics have
117. Zhang Y, Wang ZC, Zhang ZS et al (2018) high diagnostic value in gastric cancer. BMC
MicroRNA-155 regulates cervical cancer via Cancer 18:415
inducing Th17/Treg imbalance. Eur Rev 129. Swellam M, El Magdoub HM, Hassan NM
Med Pharmacol Sci 22:3719–3726 et al (2018) Potential diagnostic role of circu-
118. Wang D, Tang M, Zong P et al (2018) lating MiRNAs in breast cancer: implications
MiRNA-155 regulates the Th17/Treg ratio on clinicopathological characters. Clin Bio-
by targeting SOCS1 in severe acute pancreati- chem 56:47–54
tis. Front Physiol 9:686–686
130. Su YY, Sun L, Guo ZR et al (2019) Upregu- outcome in nasopharyngeal carcinoma. PLoS
lated expression of serum exosomal miR-375 One 14:e0209760
and miR-1307 enhance the diagnostic power 142. Xiong J, Guo S, Bing Z et al (2018) A com-
of CA125 for ovarian cancer. J Ovarian Res prehensive RNA expression signature for cer-
12:6 vical squamous cell carcinoma prognosis.
131. Prahm KP, Hogdall C, Karlsen MA et al Front Genet 9:696
(2018) Identification and validation of poten- 143. Hu N, Cheng Z, Pang Y et al (2019) High
tial prognostic and predictive miRNAs of epi- expression of MiR-98 is a good prognostic
thelial ovarian cancer. PLoS One 13: factor in acute myeloid leukemia patients trea-
e0207319 ted with chemotherapy alone. J Cancer
132. Huang Z, Chen W, Du Y et al (2019) Serum 10:178–185
miR-16 as a potential biomarker for human 144. Gao W, Cao Y, Guo P et al (2018) Down-
cancer diagnosis: results from a large-scale regulation of MiR-1297 predicts poor prog-
population. J Cancer Res Clin Oncol nosis and enhances gastric cancer cell growth
145:787–796 by targeting CREB1. Biomed Pharmacother
133. Huang D, Peng Y, Ma K et al (2018) 105:413–419
MiR-20a, a novel promising biomarker to 145. Moratin J, Hartmann S, Brands RC et al
predict prognosis in human cancer: a meta- (2019) MicroRNA expression correlates
analysis. BMC Cancer 18:1189 with disease recurrence and overall survival
134. Pan J, Zhou C, Zhao X et al (2018) A in oral squamous cell carcinoma. J Cranio-
two-miRNA signature (miR-33a-5p and maxillofac Surg 47:523–529
miR-128-3p) in whole blood as potential bio- 146. Chen F, Qi S, Zhang X et al (2018) miR-23a-
marker for early diagnosis of lung cancer. Sci 3p suppresses cell proliferation in oral squa-
Rep 8:16699 mous cell carcinomas by targeting FGF2 and
135. Li P, Fan H, He Q (2018) Investigation of the correlates with a better prognosis: miR-23a-
clinical significance and prognostic value of 3p inhibits OSCC growth by targeting FGF2.
microRNA-145 in human hepatocellular car- Pathol Res Pract 215(4):660–667. https://
cinoma. Medicine (Baltimore) 97:e13715 doi.org/10.1016/j.prp.2018.12.021
136. Wang W, Hu S, Chang J et al (2018) Down- 147. Richardsen E, Andersen S, Melbo-Jorgensen
regulated microRNA-34a expression as a C et al (2019) MicroRNA 141 is associated to
prognostic marker for poor osteosarcoma in outcome and aggressive tumor characteristics
mice: a systematic review and meta-analysis. J in prostate cancer. Sci Rep 9:386
Cancer 9:4179–4186 148. Marchesi F, Regazzo G, Palombi F et al
137. Zhang N, Li Y, Zheng Y et al (2018) miR-608 (2018) Serum miR-22 as potential
and miR-4513 significantly contribute to the non-invasive predictor of poor clinical out-
prognosis of lung adenocarcinoma treated come in newly diagnosed, uniformly treated
with EGFR-TKIs. Lab Invest 99 patients with diffuse large B-cell lymphoma:
(4):568–576. https://doi.org/10.1038/ an explorative pilot study. J Exp Clin Cancer
s41374-018-0164-y Res 37:95
138. Ren ZP, Hou XB, Tian XD et al (2019) Iden- 149. Yan S, Dang G, Zhang X et al (2017) Down-
tification of nine microRNAs as potential bio- regulation of circulating exosomal miR-638
markers for lung adenocarcinoma. FEBS predicts poor prognosis in colon cancer
Open Bio 9:315–327 patients. Oncotarget 8:72220–72226
139. Yan H, Xin S, Ma J et al (2018) A three 150. Han G, Qiu N, Luo K et al (2019) Down-
microRNA-based prognostic signature for regulation of miroRNA-141 mediates
small cell lung cancer overall survival. J Cell acquired resistance to trastuzumab and is
Biochem. https://doi.org/10.1002/jcb. associated with poor outcome in breast cancer
28159 by upregulating the expression of ERBB4. J
140. Takashima Y, Kawaguchi A, Iwadate Y et al Cell Biochem. https://doi.org/10.1002/jcb.
(2019) MicroRNA signature constituted of 28416
miR-30d, miR-93, and miR-181b is a 151. Valiollahi E, Ribera JM, Genesca E et al
promising prognostic marker in primary cen- (2019) Genome-wide identification of micro-
tral nervous system lymphoma. PLoS One 14: RNA signatures associated with stem/pro-
e0210400 genitor cells in Philadelphia chromosome-
141. Sabarimurugan S, Kumarasamy C, Baxi S et al positive acute lymphoblastic leukemia. Mol
(2019) Systematic review and meta-analysis of Biol Rep 46(1):1295–1306. https://doi.
prognostic microRNA biomarkers for survival org/10.1007/s11033-019-04600-5
152. Tsunoda Y, Sasaki A, Sakamoto N et al (2018) noninvasive prenatal diagnosis and monitor-
Protein expression profile and microRNA ing. Clin Chem 51:1786–1795
expression signature in estrogen receptor- 163. Chim SS, Shing TK, Hung EC et al (2008)
positive and -negative breast cancers: report Detection and characterization of placental
of two cases. Breast Cancer 10:195–199 microRNAs in maternal plasma. Clin Chem
153. Lu JH, Zuo ZX, Wang W et al (2018) A two- 54:482–490
microRNA-based signature predicts first-line 164. Ng EK, Tsui NB, Lau TK et al (2003) mRNA
chemotherapy outcomes in advanced colorec- of placental origin is readily detectable in
tal cancer patients. Cell Death Discov 4:116 maternal plasma. Proc Natl Acad Sci U S A
154. Wang Y, Yin W, Lin Y et al (2018) Down- 100:4748–4753
regulated circulating microRNAs after sur- 165. Pigati L, Yaddanapudi SC, Iyengar R et al
gery: potential noninvasive biomarkers for (2010) Selective release of microRNA species
diagnosis and prognosis of early breast cancer. from normal and malignant mammary epithe-
Cell Death Discov 4:21 lial cells. PLoS One 5:e13515
155. Gill P, Kim E, Chua TC et al (2019) MiRNA- 166. Mitchell PS, Parkin RK, Kroh EM et al
3653 is a potential tissue biomarker for (2008) Circulating microRNAs as stable
increased metastatic risk in pancreatic neuro- blood-based markers for cancer detection.
endocrine Tumours. Endocr Pathol 30 Proc Natl Acad Sci U S A 105:10513–10518
(2):128–133. https://doi.org/10.1007/ 167. Valadi H, Ekstrom K, Bossios A et al (2007)
s12022-019-9570-y Exosome-mediated transfer of mRNAs and
156. Chen TH, Lee C, Chiu CT et al (2018) Cir- microRNAs is a novel mechanism of genetic
culating microRNA-196a is an early gastric exchange between cells. Nat Cell Biol
cancer biomarker. Oncotarget 9:654–659
9:10317–10323 168. Arroyo JD, Chevillet JR, Kroh EM et al
157. Yokoi A, Matsuzaki J, Yamamoto Y et al (2011) Argonaute2 complexes carry a popu-
(2018) Integrated extracellular microRNA lation of circulating microRNAs independent
profiling for ovarian cancer screening. Nat of vesicles in human plasma. Proc Natl Acad
Commun 9:4319 Sci U S A 108:5003–5008
158. Jones DZ, Schmidt ML, Suman S et al (2018) 169. Vickers KC, Palmisano BT, Shoucri BM et al
Micro-RNA-186-5p inhibition attenuates (2011) MicroRNAs are transported in plasma
proliferation, anchorage independent growth and delivered to recipient cells by high-
and invasion in metastatic prostate cancer density lipoproteins. Nat Cell Biol
cells. BMC Cancer 18:421 13:423–433
159. Chen F, Zhou H, Wu C et al (2018) Identifi- 170. Kosaka N, Iguchi H, Yoshioka Y et al (2010)
cation of miRNA profiling in prediction of Secretory mechanisms and intercellular trans-
tumor recurrence and progress and bioinfor- fer of microRNAs in living cells. J Biol Chem
matics analysis for patients with primary 285:17442–17452
esophageal cancer: study based on TCGA 171. Mello-Grand M, Gregnanin I, Sacchetto L
database. Pathol Res Pract 214:2081–2086 et al (2018) Circulating microRNAs com-
160. Dhondt B, De Bleser E, Claeys T et al (2018) bined with PSA for accurate and
Discovery and validation of a serum micro- non-invasive prostate cancer detection. Carci-
RNA signature to characterize oligo- and nogenesis 40(2):246–253. https://doi.org/
polymetastatic prostate cancer: not ready for 10.1093/carcin/bgy167
prime time. World J Urol 37 172. Farran B, Dyson G, Craig D et al (2018) A
(12):2557–2564. https://doi.org/10.1007/ study of circulating microRNAs identifies a
s00345-018-2609-8 new potential biomarker panel to distinguish
161. Adhami M, MotieGhader H, Haghdoost AA aggressive prostate cancer. Carcinogenesis
et al (2019) Gene co-expression network 39:556–561
approach for predicting prognostic micro- 173. Sun Y, Mei H, Xu C et al (2018) Circulating
RNA biomarkers in different subtypes of microRNA-339-5p and 21 in plasma as an
breast cancer. Genomics 112(1):135–143. early detection predictors of lung adenocarci-
https://doi.org/10.1016/j.ygeno.2019.01. noma. Pathol Res Pract 214:119–125
010
174. Yu H, Guan Z, Cuk K et al (2018) Circulating
162. Wong BC, Chiu RW, Tsui NB et al (2005) microRNA biomarkers for lung cancer detec-
Circulating placental RNA in maternal plasma tion in Western populations. Cancer Med
is associated with a preponderance of 50 7:4849–4862
mRNA fragments: implications for
175. Ebrahimkhani S, Vafaee F, Hallal S et al biomarker for diagnosis of osteosarcoma in

(2018) Deep sequencing of circulating exo- Mexican population. Hum Cell 31:292–299
somal microRNA allows non-invasive glio- 187. Weber DG, Brik A, Casjens S et al (2019) Are
blastoma diagnosis. NPJ Precis Oncol 2:28 circulating microRNAs suitable for the early
176. Usuba W, Urabe F, Yamamoto Y et al (2019) detection of malignant mesothelioma?
Circulating miRNA panels for specific and Results from a nested case-control study.
early detection in bladder cancer. Cancer Sci BMC Res Notes 12:77
110:408–419 188. Lieb V, Weigelt K, Scheinost L et al (2018)
177. Yu X, Liang J, Xu J et al (2018) Identification Serum levels of miR-320 family members are
and validation of circulating MicroRNA sig- associated with clinical parameters and diag-
natures for breast cancer early detection based nosis in prostate cancer patients. Oncotarget
on large scale tissue-derived data. J Breast 9:10402–10416
Cancer 21:363–370 189. Guo J, Liu C, Wang W et al (2018) Identifi-
178. Zheng D, Ding Y, Ma Q et al (2018) Identi- cation of serum miR-1915-3p and miR-455-
fication of serum MicroRNAs as novel bio- 3p as biomarkers for breast cancer. PLoS One
markers in esophageal squamous cell 13:e0200716
carcinoma using feature selection algorithms. 190. Hesari A, Azizian M, Darabi H et al (2018)
Front Oncol 8:674 Expression of circulating miR-17, miR-25,
179. Stroese AJ, Ullerich H, Koehler G et al and miR-133 in breast cancer patients. J Cell
(2018) Circulating microRNA-99 family as Biochem. https://doi.org/10.1002/jcb.
liquid biopsy marker in pancreatic adenocar- 27984
cinoma. J Cancer Res Clin Oncol 191. Han Y, Liu M, Wang Z et al (2017) Serum
144:2377–2390 MicroRNAs related with Chemoradiotherapy
180. Alharthi A, Beck D, Howard DR et al (2018) resistance in advanced-stage cervical squa-
An increased fraction of circulating miR-363 mous cell carcinoma. Transl Oncol
and miR-16 is particle bound in patients with 10:378–384
chronic lymphocytic leukaemia as compared 192. Zhou F, Lu X, Zhang X (2018) Serum
to normal subjects. BMC Res Notes 11:280 miR-30c level predicted cardiotoxicity in
181. Swellam M, Hashim M, Mahmoud MS et al non-small cell lung cancer patients treated
(2018) Aberrant expression of some circulat- with bevacizumab. Cardiovasc Toxicol
ing miRNAs in childhood acute lymphoblas- 18:284–289
tic leukemia. Biochem Genet 56:283–294 193. Khalighfard S, Alizadeh AM, Irani S et al
182. Jayaraj R, Kumarasamy C, Sabarimurugan S (2018) Plasma miR-21, miR-155, miR-10b,
et al (2018) Commentary: blood-derived and let-7a as the potential biomarkers for the
microRNAs for pancreatic cancer diagnosis: monitoring of breast cancer patients. Sci Rep
a narrative review and meta-analysis. Front 8:17981
Physiol 9:1896 194. Fang R, Zhu Y, Hu L et al (2018) Plasma
183. Moshiri F, Salvi A, Gramantieri L et al (2018) MicroRNA pair panels as novel biomarkers
Circulating miR-106b-3p, miR-101-3p and for detection of early stage breast cancer.
miR-1246 as diagnostic biomarkers of hepa- Front Physiol 9:1879
tocellular carcinoma. Oncotarget 195. Shah MY, Ferrajoli A, Sood AK et al (2016)
9:15350–15364 microRNA therapeutics in cancer — an
184. Gunel T, Gumusoglu E, Dogan B et al (2018) emerging concept. EBioMedicine 12:34–42
Potential biomarker of circulating hsa-miR- 196. Yoshino H, Yonemori M, Miyamoto K et al
1273g-3p level for detection of recurrent epi- (2017) microRNA-210-3p depletion by
thelial ovarian cancer. Arch Gynecol Obstet CRISPR/Cas9 promoted tumorigenesis
298:1173–1180 through revival of TWIST1 in renal cell carci-
185. Bjornetro T, Redalen KR, Meltzer S et al noma. Oncotarget 8:20881–20894
(2019) An experimental strategy unveiling 197. Acunzo M, Romano G, Wernicke D et al
exosomal microRNAs 486-5p, 181a-5p and (2015) MicroRNA and cancer—a brief over-
30d-5p from hypoxic tumour cells as circulat- view. Adv Biol Regul 57:1–9
ing indicators of high-risk rectal cancer. J 198. Chakraborty C, Sharma AR, Sharma G et al
Extracell Vesicles 8:1567219 (2018) The novel strategies for next-
186. Monterde-Cruz L, Ramirez-Salazar EG, generation cancer treatment: miRNA com-
Rico-Martinez G et al (2018) Circulating bined with chemotherapeutic agents for the
miR-215-5p and miR-642a-5p as potential treatment of cancer. Oncotarget
9:10164–10174
199. Di Giorgio A, Tran TPA, Duca M (2016) MtEF4 cancer. ACS Appl Mater Interfaces
Small-molecule approaches toward the target- 10:22767–22775
ing of oncogenic miRNAs: roadmap for the 211. Zhang T, Xue X, He D et al (2015) A prostate
discovery of RNA modulators. Future Med cancer-targeted polyarginine-disulfide linked
Chem 8:803–816 PEI nanocarrier for delivery of microRNA.
200. Mirihana Arachchilage G, Kharel P, Reid J Cancer Lett 365:156–165
et al (2018) Targeting of G-Quadruplex har- 212. Sabolch A, Else T, Griffith KA et al (2015)
boring pre-miRNA 92b by LNA rescues Adjuvant radiation therapy improves local
PTEN expression in NSCL cancer cells. ACS control after surgical resection in patients
Chem Biol 13:909–914 with localized adrenocortical carcinoma. Int
201. Ghosh A, Degyatoreva N, Kukielski C et al J Radiat Oncol Biol Phys 92:252–259
(2018) Targeting miRNA by tunable small 213. Jacquet E, Lardy-Cléaud A, Pistilli B et al
molecule binders: peptidic aminosugar (2018) Endocrine therapy or chemotherapy
mediated interference in miR-21 biogenesis as first-line therapy in hormone receptor–po-
reverts epithelial to mesenchymal transition. sitive HER2-negative metastatic breast cancer
MedChemComm 9:1147–1154 patients. Eur J Cancer 95:93–101
202. Wang L, Rowe RG, Jaimes A et al (2018) 214. Rupaimoole R, Slack FJ (2017) MicroRNA
Small-molecule inhibitors disrupt let-7 Oli- therapeutics: towards a new era for the man-
gouridylation and release the selective block- agement of cancer and other diseases. Nat Rev
ade of let-7 processing by LIN28. Cell Rep Drug Discov 16:203–222
23:3091–3101 215. Zhao L, Qi Y, Xu L et al (2018) MicroRNA-
203. Yuan H-L, Wang T, Zhang K-H (2018) 140-5p aggravates doxorubicin-induced car-
MicroRNAs as potential biomarkers for diag- diotoxicity by promoting myocardial oxida-
nosis, therapy and prognosis of gastric cancer. tive stress via targeting Nrf2 and Sirt2.
Onco Targets Ther 11:3891–3900 Redox Biol 15:284–296
204. Kahraman M, Röske A, Laufer T et al (2018) 216. Gioffre S, Ricci V, Vavassori C et al (2019)
MicroRNA in diagnosis and therapy monitor- Plasmatic and chamber-specific modulation of
ing of early-stage triple-negative breast can- cardiac microRNAs in an acute model of
cer. Sci Rep 8:11584 DOX-induced cardiotoxicity. Biomed Phar-
205. Mirzaei H, Fathullahzadeh S, Khanmoham- macother 110:1–8
madi R et al (2018) State of the art in micro- 217. Beg MS, Brenner AJ, Sachdev J et al (2017)
RNA as diagnostic and therapeutic Phase I study of MRX34, a liposomal
biomarkers in chronic lymphocytic leukemia. miR-34a mimic, administered twice weekly
J Cell Physiol 233:888–900 in patients with advanced solid tumors. Inves-
206. Di C, Zhang Q, Wang Y et al (2018) Exo- tig New Drugs 35:180–188
somes as drug carriers for clinical application. 218. van Zandwijk N, Pavlakis N, Kao SC et al
Artif Cells Nanomed Biotechnol 46(sup3): (2017) Safety and activity of microRNA-
S564–S570. https://doi.org/10.1080/ loaded minicells in patients with recurrent
21691401.2018.1501381 malignant pleural mesothelioma: a first-in-
207. Amreddy N, Babu A, Muralidharan R et al man, phase 1, open-label, dose-escalation
(2018) Chapter five - recent advances in study. Lancet Oncol 18:1386–1396
nanoparticle-based cancer drug and gene 219. Akgül B, Stadler PF, Hawkins LJ, Hadj-
delivery. In: Tew KD, Fisher PB (eds) Moussa H, Storey KB, Ergin K, Çetinkaya R,
Advances in cancer research. Academic press, Paschoal AR, Nachtigall PG, Tutar Y, Yousef
Cambridge, pp 115–170 M, Allmer J (2021) 44 Current challenges in
208. Wang K, Kievit FM, Zhang M (2016) Nano- miRNomics. In: Allmer J, Yousef M (eds)
particles for cancer gene therapy: recent miRNomics: microRNA biology and compu-
advances, challenges, and strategies. Pharma- tational analysis. Methods in molecular biol-
col Res 114:56–66 ogy, vol 2257. Springer, New York
209. Liu J, Meng T, Yuan M et al (2016) 220. Li P, Yang X, Cheng Y et al (2017)
MicroRNA-200c delivered by solid lipid MicroRNA-218 increases the sensitivity of
nanoparticles enhances the effect of paclitaxel bladder cancer to cisplatin by targeting
on breast cancer stem cell. Int J Nanomedi- Glut1. Cell Physiol Biochem 41:921–932
cine 11:6713–6725 221. Xu J, Gu X, Yang X et al (2019) MiR-1204
210. Wang F, Zhang L, Bai X et al (2018) Stimuli- promotes ovarian squamous cell carcinoma
responsive Nanocarrier for co-delivery of growth by increasing glucose uptake. Biosci
MiR-31 and doxorubicin to suppress high Biotechnol Biochem 83:123–128
222. Yamasaki T, Seki N, Yoshino H et al (2013) 234. Ma L, Chen X, Li C et al (2018) miR-129-5p

Tumor-suppressive microRNA-1291 directly and -3p co-target WWP1 to suppress gastric
regulates glucose transporter 1 in renal cell cancer proliferation and migration. J Cell Bio-
carcinoma. Cancer Sci 104:1411–1419 chem. https://doi.org/10.1002/jcb.28027
223. Liu W, Kang L, Han J et al (2018) miR-342- 235. Nie X, Su Z, Yan R et al (2019) MicroRNA-
3p suppresses hepatocellular carcinoma pro- 562 negatively regulated c-MET/AKT path-
liferation through inhibition of IGF-1R- way in the growth of glioblastoma cells. Onco
mediated Warburg effect. Onco Targets Targets Ther 12:41–49
Ther 11:1643–1653 236. Yu JG, Ji CH, Shi MH (2018) MicroRNA-
224. Xiao M, Lou C, Xiao H et al (2018) MiR-128 26b suppresses tumorigenicity and promotes
regulation of glucose metabolism and cell apoptosis in small cell lung cancer cells by
proliferation in triple-negative breast cancer. targeting myeloid cell leukemia 1 protein.
Br J Surg 105:75–85 Kaohsiung J Med Sci 34:593–605
225. Guo H, Nan Y, Zhen Y et al (2016) miRNA- 237. Ma C-h, Zhang Y-x, Tang L-h et al (2018)
451 inhibits glioma cell proliferation and MicroRNA-1469, a p53-responsive micro-
invasion by downregulating glucose trans- RNA promotes Genistein induced apoptosis
porter 1. Tumor Biol 37:13751–13761 by targeting Mcl1 in human laryngeal cancer
226. Zhu W, Huang Y, Pan Q et al (2017) cells. Biomed Pharmacother 106:665–671
MicroRNA-98 suppress Warburg effect by 238. Gu J, Liu X, Li J et al (2019) MicroRNA-144
targeting HK2 in colon cancer cells. Dig Dis inhibits cell proliferation, migration and inva-
Sci 62:660–668 sion in human hepatocellular carcinoma by
227. Zhao X, Lu C, Chu W et al (2017) targeting CCNB1. Cancer Cell Int 19:15
MicroRNA-124 suppresses proliferation and 239. Yang Q (2019) MicroRNA-5195-3p plays a
glycolysis in non-small cell lung cancer cells by suppressive role in cell proliferation, migra-
targeting AKT-GLUT1/HKII. Tumour Biol tion and invasion by targeting MYO6 in
39:1010428317706215 human non-small cell lung cancer. Biosci Bio-
228. Zhang K, Zhang M, Jiang H et al (2018) technol Biochem 83:212–220
Down-regulation of miR-214 inhibits prolif- 240. Liu X, Zhang L, Tong Y et al (2019)
eration and glycolysis in non–small-cell lung MicroRNA-22 inhibits proliferation, invasion
cancer cells via down-regulating the expres- and metastasis of breast cancer cells through
sion of hexokinase 2 and pyruvate kinase iso- targeting truncated neurokinin-1 receptor
zyme M2. Biomed Pharmacother and ERalpha. Life Sci 217:57–69
105:545–552 241. Jiang M, Shi L, Yang C et al (2019)
229. Jin F, Wang Y, Zhu Y et al (2017) The miR-1254 inhibits cell proliferation, migra-
miR-125a/HK2 axis regulates cancer cell tion, and invasion by down-regulating
energy metabolism reprogramming in hepa- Smurf1 in gastric cancer. Cell Death Dis
tocellular carcinoma. Sci Rep 7:3089 10:32
230. Liu W, Zhou Z, Zhang Q et al (2019) Over- 242. Ma Y, Li X, Chen S et al (2019) MicroRNA-
expression of miR-1258 inhibits cell prolifer- 4458 suppresses migration and epithelial-
ation by targeting AKT3 in osteosarcoma. mesenchymal transition via targeting
Biochem Biophys Res Commun HMGA1 in non–small-cell lung cancer cells.
510:479–486 Cancer Manag Res 11:637–649
231. Wei D, Shen B, Wang W et al (2019) Micro- 243. Wu H, Liu HY, Liu WJ et al (2018) miR-377-
RNA199a5p functions as a tumor suppressor 5p inhibits lung cancer cell proliferation, inva-
in oral squamous cell carcinoma via targeting sion, and cell cycle progression by targeting
the IKKbeta/NFkappaB signaling pathway. AKT1 signaling. J Cell Biochem. https://doi.
Int J Mol Med 43(4):1585–1596. https:// org/10.1002/jcb.28091
doi.org/10.3892/ijmm.2019.4083 244. Xie Q, Yu Z, Lu Y et al (2018) microRNA-
232. Li J, Zou K, Yu L et al (2018) MicroRNA- 148a-3p inhibited the proliferation and
140 inhibits the epithelial-mesenchymal tran- epithelial-mesenchymal transition progres-
sition and metastasis in colorectal cancer. Mol sion of non–small-cell lung cancer via modu-
Ther Nucleic Acids 10:426–437 lating Ras/MAPK/Erk signaling. J Cell
233. Ma Y, Sun Y (2019) miR-29a-3p inhibits Physiol 234(8):12786–12799. https://doi.
growth, proliferation, and invasion of papil- org/10.1002/jcp.27899
lary thyroid carcinoma by suppressing 245. Zhang W, Qian S, Yang G et al (2018)
NF-kappaB signaling via direct targeting of MicroRNA-199 suppresses cell proliferation,
OTUB2. Cancer Manag Res 11:13–23 migration and invasion by downregulating
RGS17 in hepatocellular carcinoma. Gene proliferation and promotes cell apoptosis via
659:22–28 claudin-1 in human colorectal carcinoma. J
246. Wu W, He K, Guo Q et al (2019) SSRP1 Cell Biochem 120:6090–6105
promotes colorectal cancer progression and 258. Ye CY, Zheng CP, Ying WW et al (2018)
is negatively regulated by miR-28-5p. J Cell Up-regulation of microRNA-497 inhibits
Mol Med 23(5):3118–3129. https://doi. the proliferation, migration and invasion but
org/10.1111/jcmm.14134 increases the apoptosis of multiple myeloma
247. Xing F, Song Z, He Y (2018) MiR-219-5p cells through the MAPK/ERK signaling
inhibits growth and metastasis of ovarian can- pathway by targeting Raf-1. Cell Cycle 17
cer cells by targeting HMGA2. Biol Res 51:50 (24):2666–2683. https://doi.org/10.1080/
248. Ren Y, Zhang H, Jiang P (2018) MicroRNA- 15384101.2018.1542895
382 inhibits cell growth and migration in 259. Li L, Shao MY, Zou SC et al (2019)
colorectal cancer by targeting SP1. Biol Res MiR-101-3p inhibits EMT to attenuate pro-
51:51 liferation and metastasis in glioblastoma by
249. Jiang F, Yu Q, Chu Y et al (2019) MicroRNA- targeting TRIM44. J Neuro-Oncol
98-5p inhibits proliferation and metastasis in 141:19–30
non-small cell lung cancer by targeting 260. Xing F, Wang S, Zhou J (2019) The expres-
TGFBR1. Int J Oncol 54:128–138 sion of MicroRNA-598 inhibits ovarian can-
250. Ai H, Zhou W, Wang Z et al (2018) cer cell proliferation and metastasis by
microRNAs-107 inhibited autophagy, prolif- targeting URI. Mol Ther Oncolytics 12:9–15
eration, and migration of breast cancer cells 261. Yao R, Zheng H, Wu L et al (2018) miRNA-
by targeting HMGB1. J Cell Biochem. 641 inhibits the proliferation, migration, and
https://doi.org/10.1002/jcb.28157 invasion and induces apoptosis of cervical can-
251. Zhu H, Cao XX, Liu J et al (2019) cer cells by directly targeting ZEB1. Onco
MicroRNA-488 inhibits endometrial glandu- Targets Ther 11:8965–8976
lar epithelial cell proliferation, migration, and 262. Luo P, Wang Q, Ye Y et al (2019) MiR-223-
invasion in endometriosis mice via Wnt by 3p functions as a tumor suppressor in lung
inhibiting FZD7. J Cell Mol Med 23 squamous cell carcinoma by miR-223-3p-
(4):2419–2430. https://doi.org/10.1111/ mutant p53 regulatory feedback loop. J Exp
jcmm.14078 Clin Cancer Res 38:74
252. Yang X, Wang L, Wang Q et al (2018) 263. Sun Y, Hu B, Wang Y et al (2018) miR-216a-
MiR-183 inhibits osteosarcoma cell growth 5p inhibits malignant progression in small cell
and invasion by regulating LRP6-Wnt/beta- lung cancer: involvement of the Bcl-2 family
catenin signaling pathway. Biochem Biophys proteins. Cancer Manag Res 10:4735–4745
Res Commun 496:1197–1203 264. Yin W, Shi L, Mao Y (2019) MiR-194 regu-
253. Shirafkan N, Shomali N, Kazemi T et al lates nasopharyngeal carcinoma progression
(2018) microRNA-193a-5p inhibits migra- by modulating MAP3K3 expression. FEBS
tion of human HT-29 colon cancer cells via Open Bio 9:43–52
suppression of metastasis pathway. J Cell Bio- 265. Kong X, Zhang J, Li J et al (2018) MiR-130a-
chem. https://doi.org/10.1002/jcb.28164 3p inhibits migration and invasion by regulat-
254. Wang T, Hou J, Jian S et al (2018) miR-29b ing RAB5B in human breast cancer stem cell-
negatively regulates MMP2 to impact gastric like cells. Biochem Biophys Res Commun
cancer development by suppress gastric cancer 501:486–493
cell migration and tumor growth. J Cancer 266. Hua CB, Song SB, Ma HL et al (2019)
9:3776–3786 MiR-1-5p is down-regulated in gallbladder
255. Schwarzenbacher D, Klec C, Pasculli B et al carcinoma and suppresses cell proliferation,
(2019) MiR-1287-5p inhibits triple negative migration and invasion by targeting Notch2.
breast cancer growth by interaction with Pathol Res Pract 215:200–208
phosphoinositide 3-kinase CB, thereby sensi- 267. Zhou Y, Zheng X, Chen LJ et al (2019)
tizing cells for PI3Kinase inhibitors. Breast microRNA-181b suppresses the metastasis of
Cancer Res 21:20 lung cancer cells by targeting sex determining
256. Qin H, Liu W (2019) MicroRNA-99a-5p region Y-related high mobility group-box
suppresses breast cancer progression and 6 (Sox6). Pathol Res Pract 215:335–342
cell-cycle pathway through downregulating 268. Feng C, Xian Q, Liu S (2018) Micro
CDC25A. J Cell Physiol 234:3526–3537 RNA-518 inhibits gastric cancer cell growth
257. Zheng YF, Luo J, Gan GL et al (2019) Over- by inducing apoptosis via targeting MDM2.
expression of microRNA-98 inhibits cell Biomed Pharmacother 97:1595–1602
269. Zhang Y, Zhang D, Lv J et al (2018) 280. Fu Y, Meng Y, Gu X et al (2019) miR-503

MiR-125a-5p suppresses bladder cancer pro- expression is downregulated in cervical cancer
gression through targeting FUT4. Biomed and suppresses tumor growth by targeting
Pharmacother 108:1039–1047 AKT2. J Cell Biochem. https://doi.org/10.
270. Cao XZ, Bin H, Zang ZN (2019) MiR-128 1002/jcb.28099
suppresses the growth of thyroid carcinoma 281. Zhou Y, Liu K, Liu Y et al (2018) MicroRNA-
by negatively regulating SPHK1. Biomed 34a inhibit hepatocellular carcinoma progres-
Pharmacother 109:1960–1966 sion by repressing hexokinase-1. J Cell Bio-
271. Zhang L, Wang Y, Wang L et al (2018) chem. https://doi.org/10.1002/jcb.27988
miR-23c suppresses tumor growth of human 282. Lv ZD, Xin HN, Yang ZC et al (2019)
hepatocellular carcinoma by attenuating miR-135b promotes proliferation and metas-
ERBB2IP. Biomed Pharmacother tasis by targeting APC in triple-negative
107:424–432 breast cancer. J Cell Physiol 234
272. Zeng W, Zhu JF, Liu JY et al (2018) (7):10819–10826. https://doi.org/10.
miR-133b inhibits cell proliferation, migra- 1002/jcp.27906
tion and invasion of esophageal squamous 283. Nie W, Ni D, Ma X et al (2019) miR122
cell carcinoma by targeting EGFR. Biomed promotes proliferation and invasion of clear
Pharmacother 111:476–484 cell renal cell carcinoma by suppressing Fork-
273. Li L, Wang X, Liu D (2019) MicroRNA-185 head box O3. Int J Oncol 54:559–571
inhibits proliferation, migration and invasion 284. Zhou J, Zhang C, Zhou B et al (2019)
in human osteosarcoma MG63 cells by target- miR-183 modulated cell proliferation and
ing vesicle-associated membrane protein apoptosis in ovarian cancer through the
2. Gene 696:80–87 TGF-beta/Smad4 signaling pathway. Int J
274. Gao S, Zhao Z, Wu R et al (2019) MiR-1 Mol Med 43(4):1734–1746. https://doi.
inhibits prostate cancer PC3 cells prolifera- org/10.3892/ijmm.2019.4082
tion through the Akt/mTOR signaling path- 285. Zhu Y, Li L, Hou D et al (2019) MicroRNA-
way by binding to c-met. Biomed 19a regulates the proliferation, migration and
Pharmacother 109:1406–1410 invasion of human gastric cancer cells by tar-
275. Sun KK, Shen XJ, Yang D et al (2019) geting CUL5. Arch Biochem Biophys
MicroRNA-31 triggers G2/M cell cycle 662:93–100
arrest, enhances the chemosensitivity and 286. Fan Y, Che X, Hou K et al (2018) MiR-940
inhibits migration and invasion of human gas- promotes the proliferation and migration of
tric cancer cells by downregulating the expres- gastric cancer cells through up-regulation of
sion of zeste homolog 2 (ZH2). Arch programmed death ligand-1 expression. Exp
Biochem Biophys 663:269–275 Cell Res 373:180–187
276. Gu B, Wang J, Song Y et al (2018) 287. Zhou H, Liu H, Jiang M et al (2019) Target-
microRNA-383 regulates cell viability and ing MicroRNA-21 suppresses gastric cancer
apoptosis by mediating Wnt/beta-catenin cell proliferation and migration via PTEN/
signaling pathway in non-small cell lung can- Akt signaling axis. Cell Transplant 28
cer. J Cell Biochem. https://doi.org/10. (3):306–317. https://doi.org/10.1177/
1002/jcb.28069 0963689719825573
277. Wang L, Sun J, Cao H (2019) MicroRNA- 288. Li Y, Sun D, Gao J et al (2018) MicroRNA-
384 regulates cell proliferation and apoptosis 373 promotes the development of endome-
through directly targeting WISP1 in laryngeal trial cancer by targeting LATS2 and activating
cancer. J Cell Biochem 120:3018–3026 the Wnt/beta-catenin pathway. J Cell Bio-
278. Wang J, Wang B, Ren H et al (2019) miR-9- chem. https://doi.org/10.1002/jcb.28149
5p inhibits pancreatic cancer cell proliferation, 289. Xin H, Wang C, Liu Z (2019) miR-196a-5p
invasion and glutamine metabolism by target- promotes metastasis of colorectal cancer via
ing GOT1. Biochem Biophys Res Commun targeting IkappaBalpha. BMC Cancer 19:30
509:241–248 290. Zhou Q, Dong J, Luo R et al (2019)
279. Yang H, Sui Y, Guo X et al (2018) Endostar MicroRNA-20a regulates cell proliferation,
continuous intravenous infusion combined apoptosis and autophagy by targeting throm-
with S-1 and oxaliplatin chemotherapy could bospondin 2 in cervical cancer. Eur J Pharma-
be effective in treating liver metastasis from col 844:102–109
gastric cancer. J Cancer Res Ther 14: 291. Deng YH, Deng ZH, Hao H et al (2018)
S1148–S1151 MicroRNA-23a promotes colorectal cancer
cell survival by targeting PDK4. Exp Cell Res 303. Sanches JGP, Xu Y, Yabasin IB et al (2018)
373:171–179 miR-501 is upregulated in cervical cancer and
292. Tao J, Liu Z, Wang Y et al (2019) MicroRNA- promotes cell proliferation, migration and
645 represses hepatocellular carcinoma pro- invasion by targeting CYLD. Chem Biol
gression by inhibiting SOX30-mediated p53 Interact 285:85–95
transcriptional activation. Int J Biol Macro- 304. Yuan Q, Zhang Y, Li J et al (2018) High
mol 121:214–222 expression of microRNA-4295 contributes
293. Hu J, Ruan J, Liu X et al (2018) MicroRNA- to cell proliferation and invasion of pancreatic
301a-3p suppressed the progression of hepa- ductal adenocarcinoma by the down-
tocellular carcinoma via targeting VGLL4. regulation of Glypican-5. Biochem Biophys
Pathol Res Pract 214:2039–2045 Res Commun 497:73–79
294. Wu D, Zhang H, Ji F et al (2019) MicroRNA- 305. Song J, Ma Q, Hu M et al (2018) The inhibi-
17 promotes osteosarcoma cells proliferation tion of miR-144-3p on cell proliferation and
and migration and inhibits apoptosis by reg- metastasis by targeting TOP2A in HCMV-
ulating SASH1 expression. Pathol Res Pract positive glioblastoma cells. Molecules
215:115–120 23:3259
295. Lin H, Lin T, Lin J et al (2019) Inhibition of 306. Liu Q, Chen J, Wang B et al (2018) miR-145
miR-423-5p suppressed prostate cancer modulates epithelial-mesenchymal transition
through targeting GRIM-19. Gene and invasion by targeting ZEB2 in non-small
688:93–97 cell lung cancer cell lines. J Cell Biochem.
296. Fan H, Yuan R, Cheng S et al (2018) Over- https://doi.org/10.1002/jcb.28126
expressed miR-183 promoted glioblastoma 307. Zheng XM, Zhang P, Liu MH et al (2019)
radioresistance via down-regulating LRIG1. MicroRNA-30e inhibits adhesion, migration,
Biomed Pharmacother 97:1554–1563 invasion and cell cycle progression of prostate
297. Guanen Q, Junjie S, Baolin W et al (2018) cancer cells via inhibition of the activation of
MiR-214 promotes cell meastasis and inhi- the MAPK signaling pathway by downregu-
bites apoptosis of esophageal squamous cell lating CHRM3. Int J Oncol 54:443–454
carcinoma via PI3K/AKT/mTOR signaling 308. Zhou X, Xu M, Guo Y et al (2018)
pathway. Biomed Pharmacother MicroRNA-588 regulates invasion, migration
105:350–361 and epithelial-mesenchymal transition via tar-
298. Jiang M, Zhou LY, Xu N et al (2018) Down- geting EIF5A2 pathway in gastric cancer.
regulation of miR-500 and miR-628 suppress Cancer Manag Res 10:5187–5197
non-small cell lung cancer proliferation, 309. Feng L, Jing L, Han J et al (2018) MicroRNA
migration and invasion by targeting ING1. 486-3p directly targets BIK and regulates
Biomed Pharmacother 108:1628–1639 apoptosis and invasion in colorectal cancer
299. Yi T, Zhou X, Sang K et al (2019) cells. Onco Targets Ther 11:8791–8801
MicroRNA-1270 modulates papillary thyroid 310. Yang L, Zhang S, Guo K et al (2019)
cancer cell development by regulating SCAI. miR-125a restrains cell migration and inva-
Biomed Pharmacother 109:2357–2364 sion by targeting STAT3 in gastric cancer
300. Zhu J, Han S (2019) miR-150-5p promotes cells. Onco Targets Ther 12:205–215
the proliferation and epithelial-mesenchymal 311. Zhang H, Jiang S, Guo L et al (2019)
transition of cervical carcinoma cells via tar- MicroRNA-1258, regulated by c-Myb, inhi-
geting SRCIN1. Pathol Res Pract 215 bits growth and epithelial-to-mesenchymal
(4):738–747. https://doi.org/10.1016/j. transition phenotype via targeting SP1 in
prp.2019.01.004 oral squamous cell carcinoma. J Cell Mol
301. Xie S, Liu G, Huang J et al (2019) miR-210 Med 23(4):2813–2821. https://doi.org/10.
promotes lung adenocarcinoma proliferation, 1111/jcmm.14189
migration, and invasion by targeting lysyl 312. Hu Y, Wu AY, Xu C et al (2019) MicroRNA-
oxidase-like 4. J Cell Physiol 234 449a inhibits tumor metastasis through
(8):14050–14057. https://doi.org/10. AKT/ERK1/2 inactivation by targeting ste-
1002/jcp.28093 roid receptor coactivator (SRC) in endome-
302. Hu XH, Zhao ZX, Dai J et al (2019) trial cancer. J Cancer 10:547–555
MicroRNA-221 regulates osteosarcoma cell 313. Li G, Xu Y, Wang S et al (2019) MiR-873-5p
proliferation, apoptosis, migration, and inva- inhibits cell migration, invasion and
sion by targeting CDKN1B/p27. J Cell Bio- epithelial-mesenchymal transition in colorec-
chem 120:4665–4674 tal cancer via targeting ZEB1. Pathol Res
Pract 215:34–39
314. Ma W, Liu B, Li J et al (2018) MicroRNA- epithelial transition and suppresses autophagy

302c represses epithelial-mesenchymal transi- in osteosarcoma cells by directly targeting
tion and metastasis by targeting transcription SPHK1. Biosci Biotechnol Biochem 83
factor AP-4 in colorectal cancer. Biomed (5):836–844. https://doi.org/10.1080/
Pharmacother 105:670–676 09168451.2019.1569496
315. Yang B, Zhang W, Sun D et al (2019) Down- 320. Jiang H, Bu Q, Zeng M et al (2019)
regulation of miR-139-5p promotes prostate MicroRNA-93 promotes bladder cancer pro-
cancer progression through regulation of liferation and invasion by targeting PEDF.
SOX5. Biomed Pharmacother Urol Oncol 37:150–157
109:2128–2135 321. He H, Liao X, Yang Q et al (2018) Micro-
316. Zhang X, Zhou J, Xue D et al (2019) RNA-494-3p promotes cell growth, migra-
MiR-515-5p acts as a tumor suppressor via tion, and invasion of nasopharyngeal
targeting TRIP13 in prostate cancer. Int J carcinoma by targeting Sox7. Technol Cancer
Biol Macromol 129:227–232 Res Treat 17:1533033818809993
317. Xie X, Pan J, Han X et al (2019) Downregula- 322. He H, Zhao X, Zhu Z et al (2019)
tion of microRNA-532-5p promotes the pro- MicroRNA-3191 promotes migration and
liferation and invasion of bladder cancer cells invasion by downregulating TGFBR2 in colo-
through promotion of HMGB3/Wnt/beta- rectal cancer. J Biochem Mol Toxicol 33(6):
catenin signaling. Chem Biol Interact e22308. https://doi.org/10.1002/jbt.
300:73–81 22308
318. Cui Z, Zhao Y (2019) microRNA-342-3p 323. Wang Y, Yang Z, Wang L et al (2019)
targets FOXQ1 to suppress the aggressive miR-532-3p promotes hepatocellular carci-
phenotype of nasopharyngeal carcinoma noma progression by targeting PTPRT.
cells. BMC Cancer 19:104 Biomed Pharmacother 109:991–999
319. Wang D, Bao F, Teng Y et al (2019) Micro-
RNA-506-3p initiates mesenchymal-to-
Chapter 19
44 Current Challenges in miRNomics

Bünyamin Akgül, Peter F. Stadler, Liam J. Hawkins,
Hanane Hadj-Moussa, Kenneth B. Storey, Kemal Ergin,
Rahmi Çetinkaya, Alexandre R. Paschoal, Pedro G. Nachtigall,
Yusuf Tutar, Malik Yousef, and Jens Allmer
Abstract
Mature microRNAs (miRNAs) are short RNA sequences about 18–24 nucleotide long, which provide the
recognition key within RISC for the posttranscriptional regulation of target RNAs. Considering the
canonical pathway, mature miRNAs are produced via a multistep process. Their transcription
(pri-miRNAs) and first processing step via the microprocessor complex (pre-miRNAs) occur in the nucleus.
Then they are exported into the cytosol, processed again by Dicer (dsRNA) and finally a single strand
(mature miRNA) is incorporated into RISC (miRISC). The sequence of the incorporated miRNA provides
the function of RNA target recognition via hybridization. Following binding of the target, the mRNA is
either degraded or translation is inhibited, which ultimately leads to less protein production. Conversely, it
has been shown that binding within the 50 UTR of the mRNA can lead to an increase in protein product.
Regulation of homeostasis is very important for a cell; therefore, all steps in the miRNA-based regulation
pathway, from transcription to the incorporation of the mature miRNA into RISC, are under tight control.
While much research effort has been exerted in this area, the knowledgebase is not sufficient for accurately
modelling miRNA regulation computationally. The computational prediction of miRNAs is, however,
necessary because it is not feasible to investigate all possible pairs of a miRNA and its target, let alone
miRNAs and their targets. We here point out open challenges important for computational modelling or
for our general understanding of miRNA-based regulation and show how their investigation is beneficial. It
is our hope that this collection of challenges will lead to their resolution in the near future.
Key words miRNomics, Challenges, miRNA prediction, miRNA targeting, Mature miRNA, RISC
1 Introduction
Cell homeostasis is vital; therefore, gene expression (RNA/protein)

is under tight control [1]. The dysregulation of gene expression is
associated with various disorders such as Parkinson’s [2] and can-
cers [3, 4]. Regulatory interventions are manifold. For example,
transcript abundance is influenced by cis- and trans-acting tran-
scriptional regulators and epigenetics mechanisms while protein
lifetime is controlled via, for example, ubiquitination [5]. One
423
424 Bünyamin Akgül et al.
regulatory pathway which was discovered relatively recently [6]

controls protein abundance following transcription and export of
the mRNA templates. This process, termed posttranscriptional
regulation, is currently under heavy investigation due to the prom-
ise it holds of generating easily accessible biomarkers and persona-
lized therapeutic interventions [4, 7].
Checks and balances need to be in place to ensure that any
regulatory mechanism cannot wreak havoc on cell homeostasis.
Therefore, miRNA biogenesis and targeting are under strict regu-
lation themselves [8]. MicroRNA biogenesis occurs in a multistep
process [9, 10] starting with the transcription of pri-miRNAs. This
initial transcription step is controlled by genetic and epigenetic
mechanisms. Transcriptional control is more relevant to miRNAs
produced from their own loci while other miRNAs might be
coproducts of other transcripts such as mRNAs or long noncoding
RNAs (lncRNAs). Mirtrons, microRNAs encoded within introns,
are sideproducts of genes and do not influence the abundance of
the host gene they are encoded in directly. This is different for
miRNAs which are encoded within exons. A relevant example in
this context is a miRNA (not listed in miRBase yet) encoded within
the sixth exon (or seventh intron; depending on gene version) of
the Di George Critical Region Eight (DGCR8) [11]. This presents
a direct feedback loop for the abundance of the microprocessor and
produces a miRNA which has several putative targets [12]. Both
events, however, are not direct and the decreased amount of mRNA
for DGCR8 will take effect during its translation and the resulting
pre-miRNA needs to proceed through the miRNA biogenesis and
targeting pathways to take effect. Here emerges a very important
question, what is the abundance of the transcripts involved and how
much ncRNA/protein will they produce? Obviously, this question
is too complex to be answered in a straightforward manner. There-
fore, we will break down questions into smaller, manageable chunks
in the following.
2 Open Challenges
2.1 Transcription For intergenic miRNAs the primary transcript frequently remains
and Transcriptional unannotated, thus making it difficult to systematically investigate
Regulation miRNAs under similar transcriptional control. The situation is also
far from ideal for intronic miRNAs. Here the question remains
whether miRNAs are processed equivalently from alternative over-
lapping transcripts. Furthermore, there are indications for a cou-
pling of alternative splicing and miRNA processing in plants
[13, 14]. Disentangling such connections will require a systematic
analysis of matching sRNA-seq and RNA-seq data to see the rela-
tionships between precursors and processing products. A wealth of
such data is presumably available as part of the large-scale cancer
Current Challenges in miRNomics 425
sequencing projects, but does not seem to have been mined with
such generic questions in mind. From a computational perspective,
prediction of RNA-Polymerase binding sites and their resulting
transcription efficiency would be beneficial to pinpointing miRNAs
from their own loci. Attempts have been made to analyze large scale
ChIP-seq data leading to the detection of many transcription factor
binding sites [15]. However, not all possible stress responses, envi-
ronmental conditions, tissues, and so on, can be queried via ChIP-
seq; therefore, a computational tool for predicting such binding
sites is needed. This leads to the following list of queries.
1. Improve miRNA annotation [16], for example, for the primary
transcript of intergenic loci.
2. Define whether there is an effective expression difference
between exons and miRNAs encoded in the introns of the
same genes, considering alternative transcripts.
3. What is the effect of miRNA processing on alternative splicing?
4. Mine the large amount of available sRNA- and RNA-seq data
from different perspectives, for example, to answer (1–3).
5. Detect or predict RNA-polymerase binding sites to aid predic-
tion of pri-miRNAs.
6. Predict RNA-polymerase efficiency (based on binding sites) to
predict the abundance of pri-miRNA transcripts that can be
expected.
In summary, points (1–6) pertain to the questions of whether
and how much initial transcript can be expected from a locus. With
more than a hundred million hairpin structures that can be pre-
dicted from the human genome defining whether they are likely to
be expressed in large enough amounts would aid in filtering many
of the candidate miRNAs.
2.2 Steps In order to better understand miRNA-based regulation, a number

in the miRNA of important investigations should be performed.
Biogenesis
7. Define the specificity of Drosha.
and Targeting Pathway
8. Define the specificity of Drosha cleavage.
9. Define the specificity of Exportin-5.
10. Define the specificity of Dicer.
11. Define the possible types of mature miRNAs derivable from a
pre-miRNA.
12. Define incorporation into RISC.
13. Define targeting specificity and effect.
14. Define number and intracytoplasmic localization of RISC
per cell.
15. Define mRNA translation efficiency.

16. Define how often loaded RISC can modulate translation.
17. Define dissociation of miRNA from RISC.
7. Some of the challenges (7–16) are still complex and can be

further broken down. For example, defining Drosha, includes
whether there are structural and/or sequence requirements for
hairpin recognition. Structural questions can be answered rela-
tively easily by carefully designing a set of RNAs including one
hairpin each, with varying structural features. Incubating them
with Drosha for a set period of time and determining the ratio
of educts to products can define the enzyme’s specificity. Given
knowledge from the previous experiment, the hairpins can be
varied in terms of sequence while the structure stays intact.
Performing the same experiment as before will show whether
Drosha has a preference for sequence features. Determining
these rules, will not only benefit the prediction of pre-miRNAs
from a genome but will further allow for quantitative
modelling.
8. It is not only important to understand which hairpins (if any
bias) Drosha processes but also what the outcome may
be. There are accounts in literature ranging from blunt cutting
of the RNA duplex, to several nucleotide overhang on either
side. Knowledge about this could be a side effect of experi-
ments performed for (7) since sequencing the products of the
known educts will lead to clearly defined cleavage rules whether
based on structure, sequence, or both.
9. Any hairpin excised by Drosha or created as a miRTron, needs
to be exported from the nucleus into the cytosol to progress in
the miRNA biogenesis pathway. Incubating purified Exportin-
5 with hairpins and performing a pull-down experiment is one
approach that could be taken. Whether inserting Exportin-5
into vesicles along with other essential proteins, filled with
pre-miRNAs and then measuring efflux is an easy experiment,
is unlikely. Perhaps similar approaches might be performable.
Unless some hairpins are rejected from transport due to struc-
tural or sequence features, which seems unlikely, working
knowledge about Exportin-5 will allow quantitative modelling
of the transport.
10. Similar to Drosha, it needs to be tested whether Dicer has a
specificity in terms of structure, sequence, or both and what
the exact outcome of the processing is. A similar approach can
be taken as for Drosha (Challenges 7 and 8). In addition, the
position of the cleavage event needs to be determined precisely
and whether it is dependent on structure and/or sequence of
the stem-loop. Determining the cleavage rules will strongly

enhance the prediction of the mature sequence from the
stem-loop precursor.
11. In addition to mature microRNAs derived from the 50 -arm or
the 30 -arm of the hairpin, there is evidence that the loop
region can also give rise to a third mature microRNA called
“loop-miR” [17, 18]. These loop-miRs have been found in
different species and have been shown to be functionally
active. The only known determinant of loop-miR production
is the size of the loop, that is, the distance between the Dicer
processing sites at the 30 -end of the 50 -arm and the 50 -end of
the 30 -arm. For the future, it would be important to investi-
gate, how many and which pre-miRNAs can give rise to this
third mature microRNA, to establish their functional role and
target spectrum and to elucidate whether they could be used
in biotechnology to create multifunctional microRNA pre-
cursors giving rise to multiple independent microRNA
sequences.
12. There are different accounts of incorporation of mature
sequences into RISC. One mature sequence, both with
some bias, or without can be assembled into miRISC. For
computational prediction and for the understanding of the
process, it is important to know which mature sequence or in
which ratio they are incorporated into RISC. Here current
methodologies such as HITS-CLIP and CLASH [19, 20] can
help to define the amount of mature miRNAs bound in
miRISC.
13. Currently, knowledge about targeting is based on a small
number of early studies. There a seed sequence and an
out-seed have been defined. This needs to be tested on a
larger scale. For example, miRISC with different mature
sequences could be used to pull down random RNAs which
can then be sequenced. Thereby, the binding affinity can be
defined. In a follow-up experiment, overexpression experi-
ments can be performed and the ratio of translational repres-
sion to mRNA degradation can be measured. When done for
a large variety of miRNAs and targets, this can help detect any
sequence and structure biases for the miRISC-mRNA system.
Furthermore, such experiments will allow for a better under-
standing of the targeting process and perhaps more precise
computational models.
14. There exist estimates for the number of RISC complexes in a
cell. Here it would be interesting to analyze the RISC content
among different tissues and species. Another important
finding would be the ratio of miRISC versus free RISC. Add-
ing to this complexity is the intracytoplasmic re-localization of
miRISC complexes, which might add another layer to the
regulation of miRNA activity [21, 22].
15. It is important to understand how effective an mRNA is trans-

lated without miRNA influence. This is something that is being
investigated computationally in the DREAM challenges in
proteogenomics. Then the effects of miRNA on the translation
can be better understood and perhaps delineated based on
binding site motifs, location of binding sites, and binding site
multiplicity. Initially, more combined investigations of mRNA
and protein abundance are needed.
16. After RISC binds an mRNA, can the same RISC complex
dissociate from the mRNA (perhaps after processing the
mRNA) and bind another mRNA for cleavage or translational
repression? This is an important question which goes hand in
hand with the following one.
17. Whether an assembled miRISC complex can revert to free
RISC and bind another miRNA and what are the association
and dissociation constants?
Many functions can derive from the knowledge queried in
(7–16), for example, when designing miRNA mimics to control
diseases such as cancer. This list above pertains only to the canonical
processing of miRNAs. However, there are many alternatives that
only partially overlap with the canonical pathway [23]. It would be
important to understand (a) how to identify the sequence proces-
sing steps in each case, (b) to what extent they can vary and admit
alternatives (e.g., loop miRNAs processed from precursors that also
produce canonical Dicer-cleaved products, or the production of
microRNA-offset RNAs, moRNAs, from overly long precursor
hairpin [24–26], and (c) to what extent the usage of pathways
changes throughout evolution. One should also keep in mind
that details of the relevant pathways will not have been invariant
over large evolutionary time scales. We suspect, therefore, that the
answers to the questions above (7–17) will show subtle differences
between clades. It will be important in particular for computational
approaches to understand this variability. The fact that it is possible
to recognize a clade by the k-mer distribution of microRNAs
[27, 28] is a clear indication for such an effect. MicroRNAs seem
to have emerged independently several times in the history of
Eukaryotes, thereby using the RNAi pathway(s) as an additional
layer of gene regulation. What are the commonalities and differ-
ences between these innovation events? To what extent are homol-
ogous proteins used in the pertinent pathways, and to what extent
have different pathways been co-opted into this new function. Is
the observed plurality of processing pathways in animals a reflection
of the ease with which the RNAi pathway(s) can be accessed by
feeding them with custom-produced RNAs? Therefore, the ques-
tions (7–17) need to be investigated in different species from
different clades to gain a better understanding of functionality
and miRNA evolution.
2.3 MicroRNA Experimental miRNA and miRNA target detection has improved
and Target Prediction with many novel technologies enabling the capturing of actual
miRNA target interactions [29]. Nonetheless, it is not feasible to
investigate miRNA, mRNA, and protein expression under all devel-
opmental stages, all tissue types, and all different stressors. There-
fore, computational prediction of miRNAs and their targets is
indispensable [20]. While pre-miRNA prediction is at a quite
mature state [30], the prediction of mature miRNAs [3] from
these pre-miRNAs and the prediction of miRNA targets needs
further improvement [31]. A deeper understanding of how the
various enzymes in the miRNA biogenesis and targeting pathway
work would be very beneficial for computational modeling.
18. Does the Microprocessor complex have a structural or
sequential bias?
19. What are the exact cleavage rules of the Microprocessor com-
plex, and are they dependent on sequence or structure?
20. Is there a structural or sequential bias for pre-miRNA export
into the cytosol?
21. Are all transported pre-miRNAs associated with Dicer or is
there a structural, sequential bias or a competition among
pre-miRNAs?
22. What are the exact cleavage rules of Dicer, and are they
dependent on sequence or structure?
23. On which factors does the mature miRNA incorporation into
RISC depend?
18. Similarly to questions (7) and (8), question (18) investigates

pri-miRNA processing into pre-miRNA. It is likely that the
Microprocessor complex has a structural bias and for further
investigation in this direction more cocrystals have to be pro-
duced and linked to processing efficiency. Binding often
changes the structures of the partners and, therefore, the
RNA structure before binding is another clue needed to better
understand miRNA genesis. Whether apart from structural
biases sequence biases play a role can be investigated by analyz-
ing processing efficiency of a large number of pri-miRNAs
(perhaps containing many pre-miRNAs) where the predicted
structures are largely similar but the sequence varies.
19. seems similar to (18) but here the question is where the
pre-miRNA is cleaved to produce the pre-miRNA. Is there a
structural dependence such as cleavage n bases following the
opening of the stem? Maybe the distance is measured from the
loop since only a limited size structure fits into the Micropro-
cessor complex. Perhaps some kind of sequential or energetical
motif is required for the cleavage to occur. What determines the

actual cleavage leading to blunt products or products with an
overhang? Answering these questions would greatly reduce the
amount of false positive pre-miRNA predictions.
20. Following the excision of pre-miRNAs, are they all exported or
do some remain in the nucleus and are perhaps degraded there?
Here markers could help determine the cellular localization of
miRNAs. Perhaps the flux of pre-miRNAs from overexpressed
pri-miRNAs can be monitored live using markers and confocal
laser scanning microscopy.
21. Another important question is whether hairpins are free-
floating in the cytosol or whether they are handed from one
protein complex to the next, for example, Microproces-
sor!Exportin!Dicer. Whether free floating or not, is there a
competition among pre-miRNAs for Dicer processing and then
incorporation into RISC? If there is competition, is it structural
or sequence-based? Perhaps it is based on the binding energy
distribution in the stem or maybe just miRNA abundance.
22. How does Dicer cleave the stem-loop exactly? Are there over-
hangs or blunt ends? Is the cleavage dependent on sequential or
structural features? These questions can be queried in vitro.
23. Similarly, how is the incorporation of the mature miRNA into
RISC achieved? There seems to be a bias for one of the strands
of the ds-RNA following Dicer processing. What are the rules
driving the selection of the strands? Are they based on binding
energy, sequence-based, or structural? Given a set of miRNAs
with perfect complementarity, thereby excluding structural
components, sequence and energetical components can be
investigated.
2.4 MiRNA Tools MicroRNAs have seen a lot of experimental investigation in the past
and Databases decades which led to humongous amounts of data available for
further analysis. However, it is a challenge to incorporate all this
information into a comprehensive resource. Having this in mind,
we focused on exploring the available literature and providing
useful and practical guidance on the miRNA database and tools
[32]. From a bioinformatics perspective, the prediction of
pre-miRNAs, mature miRNAs, miRNA-targeting are important
areas that need improvement. Also, integrating the various data-
bases which include miRNAs, targets, expression, and function and
other information separately, would be beneficial. Despite all
advances in the miRNA and miRNA-target prediction field, the
next-generation of predictors should incorporate the following
tasks.
24. Improve precision, by increasing the number of true positives

and decreasing the number of false positives in the predictions
for all pre-miRNA, mature, and target prediction.
25. Integrate the prediction into the regulatory networks and
pathways by considering the expressed genes in the tissue/
condition being analyzed.
26. Detect when the action of a miRNA on a specific target is
direct or indirect (i.e., a miRNA targets a gene and its genes
regulate another gene that also has a binding site, but this
binding site is not functional).
27. Take into account the transcript isoforms present of the spe-
cific target. In this sense, the predictors should consider the
existence of alternative poly-A signals in the 30 UTR
sequence, and, when possible, the tools should use transcrip-
tome data to confirm the 30 UTR isoform present in the
tissue/condition being analyzed as well as consider alternative
splicing.
28. The miRNA target predictors should consider the genomic
context of the UTR and the region of the binding site therein
(i.e., the position of the binding site, secondary structure of
the UTR, etc.).
29. Predictors for targeting should consider target multiplicity,
distance of targets to the stop signal, and other information
that could be integrated.
30. Current miRNA databases are largely dependent on miRBase
for naming miRNAs and they are interlinked. Both are good,
but an ontology integrating all information and allowing
logical reasoning other the data would do much to improve
on the current knowledge.
31. Enable the prediction of noncanonical mature miRNAs such
as loop-miRs.
32. Establish whether sequence information is sufficient to pre-
dict the mature miRNA.
31. As pointed out in challenge (11), it is important to discover

noncanonical mature miRNAs such as loop-miRs that are
located in the loop part. Yet those mature miRNAs are ignored
by current tools that predict the location of the mature miRNA
[3]. Here the challenge is to develop an algorithm enabling the
identification of noncanonical mature miRNAs. At the same
time it is important to annotate miRNAs more precisely in
existing databases to support algorithm development.
32. Recent studies show that it is possible to predict the microRNA
precursor considering only sequence information, more
specifically, utilizing a vector of k-mers [27, 33, 34]. Studies

considering the mature miRNAs’ targets similarly led to some
success when just employing sequence information
[35, 36]. Here the challenge is to explore the idea of the
prediction of the location of the mature part considering just
the sequence information. This can also answer whether
sequence information is a determining factor for the selection
of mature miRNAs during RISC incorporation.
Points (24 to 30) bring us to miRNA function which (see query
25) is not a singular process void of interactions with other
processes [12].
2.5 MicroRNA MicroRNAs are under spatiotemporal control and so are messenger
Function RNAs. Posttranscriptional control involving miRNAs can only
occur when both the miRNA and at least one of its targets is
coexpressed in the same space at the same time. Whether both are
expressed in the same space can be confirmed relatively easily using
RNA-seq. MicroRNAs can also be transported to their targets via
exosomes [37]. Posttranscriptional regulation can only be
measured on the transcriptional level for those regulations where
the mRNA is degraded, however, it may be hard to differentiate
among RNAs specifically degraded in response to miRNA regula-
tion and other degradation events. For those miRNA regulatory
events that modulate protein abundance without degrading the
mRNA, protein abundance needs to be correlated with mRNA
abundance and miRNA abundance and its overall occupancy status
in RISC. Additionally, miRNAs with many coexpressed targets may
not cause a measurable effect while the same miRNA coexpressed
with only a few of its targets may cause strong regulation. Addi-
tionally, miRNA sponges (see below) can modulate miRNA regu-
lation. Even for two slightly similar miRNAs (X and Y) a complex
interaction network can unfold (Fig. 1). The hypothetical regu-
latory network in Fig. 1 also shows that some target sites may be
shared among miRNAs while others are not. The location of target
sites may also be important (exon mRNA D) or in 30 UTRs (A–D).
The lncRNA in the example will likely hide any gene regulatory
effect of miR-X while not significantly affecting miR-Y (Fig. 1).
Therefore, all possible target sites of a miRNA need to be
monitored. MicroRNAs can be derived from genes and, therefore,
they are bound to the same regulation as the gene. To elucidate this
complex situation, a number of statistics need to be defined.
33. How much RISC is available in a cell?
34. What is the association rate and dissociation rate of the mature
sequence and RISC, that is, how long is a miRISC complex
active before the mature miRNA is replaced or the miRISC is
degraded?
Fig. 1 Two miRNAs (green) and three of their mRNA targets (blue) as well as one
long noncoding RNA acting as a sponge (red). Target sites are red boxes and
connections show binding of the two miRNAs. 30 UTRs are in gray
35. How many proteins are translated from one mRNA before it
is degraded?
36. When miRISC binds an mRNA, but does not degrade it, will
it dissociate from the mRNA and allow translation in another
round or will it stay bound until either is degraded?
37. Do all miRNAs have the same chance to be incorporated in
RISC or is there a structural bias?
33. while this seems trite at first glance, it cannot be expected that
the RISC amount is similar for all cell types and, therefore, this
needs to be investigated for various tissues. Perhaps the RISC
amount also varies in response to stress and other stimuli.
Initially (see 4), overall expression levels could be established
mining publicly available mRNA and protein expression data.
34. Similarly to (17), it is important to understand whether RISC
can be reused and whether the mature miRNA goes back into
the pool of possible miRISC partners or is degraded. How fast
do mature miRNAs associate and dissociate to RISC (if they
do). Once loaded, is the miRISC only used to regulate one
mRNA or is it reused and if it is how long does the mature
miRNA stay associated with miRISC?
35. When miRISC inhibits translation; it is important to know
whether the same mRNA would be translated multiple times.
If so, does the inhibition affect all possible translation events of
the mRNA?
36. Very similar to (33), assuming that there are multiple rounds of
translation from on mRNA and that there is only translational
repression. Does miRISC dissociate from the mRNA after

failed translation and bind another mRNA or is it bound to
the mRNA and degraded together with it?
37. Similarly to processing with Drosha and Dicer, is there a
sequence-based or energetical bias for incorporation of the
mature miRNA into RISC? This can be investigated by mining
publicly available expression data.
2.6 MicroRNA MicroRNAs, which are important micromanagers of gene expres-

Sponges sion at the posttranscriptional level, are themselves subject to regu-
lation both at the transcriptional and posttranscriptional levels. One
of the most interesting posttranscriptional regulatory mechanisms
that modulates intracellular miRNA abundance involves sponging
by other transcripts that includes but is not limited to lncRNAs,
circRNAs, and pseudogenes [38]. Although sponging, by defini-
tion, requires the presence of multiple miRNA recognition ele-
ments on the sponging transcript, there are numerous examples
of sponging that involve a single binding site. Thus, the major
challenge in this field is to distinguish among various potential
miRNA-RNA interactions especially when there exists a single
interaction between two transcripts. The best-characterized inter-
action includes miRNA-mediated silencing and/or translational
inhibition of the target transcript. However, the miRNA binding
to its target RNA can induce miRNA degradation as well. More
importantly, miRNA-RNA interaction may simply sequester the
miRNA away from its real target RNA without having any effect
on the sponging RNA. The outcome can be predicted when multi-
ple miRNA binding sites exist, especially on circRNAs. However,
we do not yet have a sufficiently good grasp of sequence informa-
tion and/or complementary rules for reliable predictions with
bioinformatics methods only. Pressing questions in this field are:
38. What is the minimum number of miRNA binding sites for
effective sponging?
39. What is the minimum amount of sequence complementarity
required to induce the sponging mechanism without
compromising the integrity of the miRNA and the
bound RNA?
40. Which protein complexes are involved in the recognition of
sponged miRNAs versus others?
41. What triggers the release of the sequestered or sponged
miRNAs?
42. Are all miRNAs subject to sponging or is there a selective set
of miRNAs targeted for sponging?
43. How do sponging, gene regulation, and miRNA regulation
collaborate to form regulatory circuits?
Answering these (43) questions will significantly forward our

understanding of miRNA-based regulation and will lead to large
improvement for the computational prediction of functional
miRNA interactions. All this will aid experimental studies and
support novel questions such as stress adaptation.
2.7 Studying The identification and characterization of both conserved and

MicroRNAs and novel species-specific miRNAs in non–genome-sequenced organ-
Extreme Stress isms poses numerous challenges that can be circumvented via de
Adaptation novo transcriptome sequencing and assembly and/or through the
development of complex bioinformatics pipelines [39]. To gain
global insights into the unity of miRNA regulation in different
environmental stresses will require expanding the current studies
to more unique species [40]. Increasing the biodiversity of a com-
parative work to more uniquely adapted species is difficult when
working on nonmodel organisms. This is due to the harsh environ-
ments these animals inhabit as well as due to the difficulty of
performing downstream validation (for example siRNA and
knock-out) experiments, in a scientific system that is centered on
model organisms [41].
The field of miRNA research in nonmodel organisms is rapidly
expanding and this can be helped with the adoption of more readily
available next generation sequencing technologies. This expansion
is paving the way for the generation of increasingly reliable and
accurate biomarker profiles for different stresses that could be
applied in biomedical contexts. Examples of this from extreme
animal survival strategies include using the miRNAs implicated in
protecting the brains of anoxia tolerant and for the development of
stroke therapeutics and interventions, as well as the use of freeze
tolerant responsive miRNAs to innovate organ and tissue
preservation.
3 Conclusion
Much has been achieved in the past decades since the first account
of miRNA regulation. MicroRNAs are now recognized as impor-
tant regulators of gene expression and it has become clear that
miRNA-based regulation and transcription factor-based gene reg-
ulation are intertwined and form complex regulatory networks.
Unfortunately, the discourse about and involving miRNAs is
suffering because some terms do not have the same meaning to
all researchers. For example, what is a miRNA? Does this refer to
the mature miRNA or a generic concept involving regulation? Does
the term miRTron refer to a miRNA precisely excised during
splicing or, more generally, to a miRNA encoded in an intron?
Therefore, we want to add one more challenge here:
44. An ontology and a controlled vocabulary defining terms and

actions in miRNomics is needed to ensure a common under-
standing of miRNA-based regulation.
It is clear that trying to investigate all miRNA-target interac-
tions encoded in a genome is a futile endeavor. This is due to both
miRNAs and their targets being under spatiotemporal control.
Some miRNAs may be triggered only under special circumstances
such as in response to stresses. This entails the need for computa-
tional prediction of miRNAs and their targets directly from gen-
omes. Currently, eukaryotic genomes with millions of candidate
hairpins are a formidable challenge. In order to alleviate the situa-
tion, we define (44) challenges in miRNomics which, when solved,
will enhance the accuracy of computational predictions. We hope
that this is an inspiration to researchers in the field of miRNomics to
perhaps slightly modify their experiments to solve one of the chal-
lenges while pursuing their primary aims.
References
1. Ergin K, Çetinkaya R (2021) Regulation of to lin-14. Cell 75:843–854. https://doi.org/
microRNAs. In: Allmer J, Yousef M (eds) miR- 10.1016/0092-8674(93)90529-Y
Nomics: microRNA biology and computa- 7. Uzuner E, Ulu GT, Gürler SB, Baran Y (2021)
tional analysis. Methods in molecular biology, The role of MiRNA in cancer: pathogenesis,
vol 2257. Springer, New York diagnosis, and treatment. In: Allmer J, Yousef
2. Olcum M, Tufekci KU, Genc S (2021) Micro- M (eds) miRNomics: microRNA biology and
RNAs in genetic etiology of human diseases. computational analysis. Methods in molecular
In: Allmer J, Yousef M (eds) miRNomics: biology, vol 2257. Springer, New York
microRNA biology and computational analy- 8. Hamid SM, Akgül B (2014) Master regulators
sis. Methods in molecular biology, vol 2257. of posttranscriptional gene expression are sub-
Springer, New York ject to regulation. Methods Mol Biol
3. Yildiz MT, Tutar L, Giritlioğlu NI, Bayram B, 1107:303–310
Tutar Y (2021) MicroRNAs and heat shock 9. Yaylak B, Akgül B (2021) Experimental micro-
proteins in breast cancer biology. In: Allmer J, RNA detection methods. In: Allmer J, Yousef
Yousef M (eds) miRNomics: microRNA biol- M (eds) miRNomics: microRNA biology and
ogy and computational analysis. Methods in computational analysis. Methods in molecular
molecular biology, vol 2257. Springer, New biology, vol 2257. Springer, New York
York 10. Erson-Bensan AE (2014) Introduction to
4. Karagur ER, Akgun S, Akca H (2021) Compu- MicroRNAs in biological systems. In:
tational and bioinformatics methods for micro- Yousef M, Allmer J (eds) miRNomics: Micro-
RNA gene prediction. In: Allmer J, Yousef M RNA biology and computational analysis, 1st
(eds) miRNomics: microRNA biology and edn. Humana Press, New York, pp 1–14
computational analysis. Methods in molecular 11. Mechtler P, Johnson S, Slabodkin H, Cohanim
biology, vol 2257. Springer, New York AB, Brodsky L, Kandel ES (2017) The evi-
5. Nguyen LK, Dobrzyński M, Fey D, Kholo- dence for a microRNA product of human
denko BN (2014) Polyubiquitin chain assem- DROSHA gene. RNA Biol 14:1508–1513.
bly and organization determine the dynamics https://doi.org/10.1080/15476286.2017.
of protein activation and degradation. Front 1342934
Physiol 5:4. https://doi.org/10.3389/fphys. 12. Saçar Demirci MD, Yousef M, Allmer J (2019)
2014.00004 Computational prediction of functional micro-
6. Lee RC, Feinbaum RL, Ambros V (1993) The RNA–mRNA interactions. Methods Mol Biol
C. elegans heterochronic gene lin-4 encodes 1912:175–196
small RNAs with antisense complementarity
13. Yan K, Liu P, Wu C-A, Yang G-D, Xu R, Guo microRNAs—protocols, pitfalls, and problems.
Q-H, Huang J-G, Zheng C-C (2012) Stress- In: Allmer J, Yousef M (eds) miRNomics:
induced alternative splicing provides a mecha- microRNA biology and computational analy-
nism for the regulation of MicroRNA proces- sis. Methods in molecular biology, vol 2257.
sing in Arabidopsis thaliana. Mol Cell Springer, New York
48:521–531. https://doi.org/10.1016/j. 24. Langenberger D, Bermudez-Santana C,
molcel.2012.08.032 Hertel J, Hoffmann S, Khaitovich P, Stadler
14. Yang X, Zhang H, Li L (2012) Alternative PF (2009) Evidence for human microRNA-
mRNA processing increases the complexity of offset RNAs in small RNA sequencing data.
microRNA-based gene regulation in Arabidop- Bioinformatics 25:2298–2301. https://doi.
sis: alternative splicing of microRNA binding org/10.1093/bioinformatics/btp419
sites. Plant J 70:421–431. https://doi.org/10. 25. Shi W, Hendrix D, Levine M, Haley B (2009)
1111/j.1365-313X.2011.04882.x A distinct class of small RNAs arises from pre-
15. Niu M, Tabari E, Ni P, Su Z (2018) Towards a miRNA-proximal regions in a simple chordate.
map of cis-regulatory sequences in the human Nat Struct Mol Biol 16:183–189. https://doi.
genome. Nucleic Acids Res 46:5395–5409. org/10.1038/nsmb.1536
https://doi.org/10.1093/nar/gky338 26. Zhao J, Schnitzler GR, Iyer LK, Aronovitz MJ,
16. Dweep H, Showe LC, Kossenkov AV (2021) Baur WE, Karas RH (2016) MicroRNA-offset
Functional annotation of microRNAs using RNA alters gene expression and cell prolifera-
existing resources. In: Allmer J, Yousef M tion. PLoS One 11:e0156772. https://doi.
(eds) miRNomics: microRNA biology and org/10.1371/journal.pone.0156772
computational analysis. Methods in molecular 27. Yousef M, Khalifa W, Acar E, Allmer J (2017)
biology, vol 2257. Springer, New York MicroRNA categorization using sequence
17. Winter J, Link S, Witzigmann D, motifs and k-mers. BMC Bioinformatics
Hildenbrand C, Previti C, Diederichs S 18:170. https://doi.org/10.1186/s12859-
(2013) Loop-miRs: active microRNAs gener- 017-1584-1
ated from single-stranded loop regions. 28. Odenthal L, Allmer J, Yousef M (2021)
Nucleic Acids Res 41:5503–5512. https:// Ensemble classifiers for multiclass microRNA
doi.org/10.1093/nar/gkt251 classification. In: Allmer J, Yousef M (eds) miR-
18. Okamura K, Ladewig E, Zhou L, Lai EC Nomics: microRNA biology and computa-
(2013) Functional small RNAs are generated tional analysis. Methods in molecular biology,
from select miRNA hairpin loops in flies and vol 2257. Springer, New York
mammals. Genes Dev 27:778–792. https:// 29. Dedeoğlu BG, Noyan S (2021) Experimental
doi.org/10.1101/gad.211698.112 microRNA targeting validation. In: Allmer J,
19. Ghanbarian H, Yıldız MT, Tutar Y (2021) Yousef M (eds) miRNomics: microRNA biol-
MicroRNA targeting. In: Allmer J, Yousef M ogy and computational analysis. Methods in
(eds) miRNomics: microRNA biology and molecular biology, vol 2257. Springer, New
computational analysis. Methods in molecular York
biology, vol 2257. Springer, New York 30. Demirci MDS (2021) Computational detec-
20. Nachtigall PG, Bovolenta LA (2021) Compu- tion of pre-microRNAs. In: Allmer J, Yousef
tational detection of microRNA targets. In: M (eds) miRNomics: microRNA biology and
Allmer J, Yousef M (eds) miRNomics: micro- computational analysis. Methods in molecular
RNA biology and computational analysis. biology, vol 2257. Springer, New York
Methods in molecular biology, vol 2257. 31. Yousef M, Parveen A, Kumar A (2021)
Springer, New York Computational methods for predicting mature
21. Akgül B, Erdoğan İ (2018) Intracytoplasmic microRNAs. In: Allmer J, Yousef M (eds) miR-
re-localization of miRISC complexes. Front Nomics: microRNA biology and computa-
Genet 9:403. https://doi.org/10.3389/ tional analysis. Methods in molecular biology,
fgene.2018.00403 vol 2257. Springer, New York
22. Cosacak M, Yiğit H, Kizil C, Akgül B (2018) 32. de Amorim TS, Pedro DLF, Paschoal AR
Re-arrangements in the cytoplasmic distribu- (2021) MicroRNA databases and tools. In: All-
tion of small RNAs following the maternal-to- mer J, Yousef M (eds) miRNomics: microRNA
zygotic transition in drosophila embryos. biology and computational analysis. Methods
Genes 9:82. https://doi.org/10.3390/ in molecular biology, vol 2257. Springer, New
genes9020082 York
23. Velandia-Huerto CA, Yazbeck AM, Schor J, 33. Yousef M, Allmer J, Khalifa W (2016) Accurate
Stadle PF (2021) Evolution and phylogeny of plant MicroRNA prediction can be achieved
using sequence motif features. J Intell Learn in cell-to-cell communication. In: Allmer J,
Syst Appl 08:9–22. https://doi.org/10. Yousef M (eds) miRNomics: microRNA biol-
4236/jilsa.2016.81002 ogy and computational analysis. Methods in
34. Yousef M (2019) Hamming distance and molecular biology, vol 2257. Springer, New
K-mer features for classification of pre-cursor York
microRNAs from different species. In: 38. Alkan AH, Akgül B (2021) Endogenous
Benavente-Peces C, Slama SB, Zafar B (eds) miRNA sponges. In: Allmer J, Yousef M (eds)
Proceedings of the 1st international conference miRNomics: microRNA biology and computa-
on smart innovation, ergonomics and applied tional analysis. Methods in molecular biology,
human factors (SEAHF). Springer Interna- vol 2257. Springer, New York
tional Publishing, Cham, pp 180–189 39. Zhang J, Hadj-Moussa H, Storey KB (2016)
35. Yousef M, Khalifa W, Acar İE, Allmer J (2017) Current Progress of high-throughput Micro-
Distinguishing between MicroRNA targets RNA differential expression analysis and ran-
from diverse species using sequence motifs dom Forest gene selection for model and
and K-mers. In: Proceedings of the 10th inter- non-model systems: an R implementation. J
national joint conference on biomedical engi- Integr Bioinform 13:306–306
neering systems and technologies. 40. Hadj-Moussa H, Hawkins LJ, Storey KB
SCITEPRESS - Science and Technology Pub- (2021) Role of microRNAs in extreme animal
lications, Setúbal, pp 133–139 survival strategies. In: Allmer J, Yousef M (eds)
36. Yousef M, Levy D, Allmer J (2018) Species miRNomics: microRNA biology and computa-
categorization via MicroRNAs - based on tional analysis. Methods in molecular biology,
3’UTR target sites using sequence features: vol 2257. Springer, New York
in: proceedings of the 11th international joint 41. Hawkins LJ, Storey KB (2020) Advances and
conference on biomedical engineering systems applications of environmental stress adaptation
and technologies. SCITEPRESS - Science and research. Comp Biochem Physiol A Mol Integr
Technology Publications, Funchal, Madeira, Physiol 240:110623. https://doi.org/10.
Portugal, pp 112–118 1016/j.cbpa.2019.110623
37. Tastan B, Tarakcioglu E, Birinci Y, Park Y,
Genc S (2021) Role of exosomal microRNAs
INDEX
A Body fluids.................................................. 236, 323, 350,

351, 354, 356, 369, 391, 393
Adaptation .................................................. 213, 312–315, Brain....................................................282–284, 316, 320,
318, 323, 329, 331, 333, 338, 340, 384, 435 321, 323, 327, 328, 332–334, 338–340, 435
Adipogenesis.................................................................. 317
Breathing .............................................................. 320, 323
Adipose tissue .............................295, 301, 317, 321, 322 Breathing rate................................................................ 312
Aerobic respiration........................................................ 312 Brown adipose tissue (BAT)....................... 317, 318, 322
African clawed frog (Xenopus laevis)............................. 97,
123, 316, 333, 334 C
Anaerobic fermentation ....................................... 312, 331
Animals ............................................ 9, 58, 79, 80, 82, 92, Calcium.......................................................................... 319
106, 118, 131, 138–141, 144, 146, 147, 150, Carbohydrate................................................................. 271
180, 183, 184, 187–189, 193, 194, 198, 199, Carbohydrate catabolism .............................................. 312
211, 212, 217, 218, 221, 311–341, 428, 435 Cardiac muscle .............................................................. 278
Anoxia .................................................................. 325, 327, Cardiomyopathy............................................................ 326
329–332, 337, 338, 340, 435 Cell cycle..................................................... 3, 19, 22, 301,
Anoxia tolerance......................... 313, 314, 316, 329–333 304, 318, 326, 328–330, 376, 382
Antifreeze molecules ..................................................... 323 Cell death ..................................................... 85, 154, 303,
Antioxidant defenses ............................................ 319, 324 326, 337, 376, 379–382, 398
Antioxidants ................................................ 313, 320, 331 Chaperone proteins........................................24, 313, 321
Anurans................................................................. 334, 335 Cold exposure ...................................................... 315, 321
Apoptosis .................................................. 3, 6, 11, 73, 96, Cold hardiness...................................................... 324, 325
98, 255, 256, 261, 269, 279, 280, 295, 298, Colorado potato beetle (Leptinotarsa
300–302, 317, 319, 321, 324, 328, 330, 333, decemlineata) ............................................ 316, 325
335, 359, 367, 368, 377–383, 396, 400, 404 Common periwinkle (Littorina littorea) ....................316,
Arctic ground squirrel (Spermophilus paryii)..............316, 325, 331, 338
318 Conservation ....................................................... 116, 119,
Argonaute endonucleases (AGO proteins)..................... 2, 171, 192, 196, 197, 219, 221, 224, 227, 253, 315
11–15, 22, 92, 109, 315, 336, 377 Cryoprotectants .......................................... 323, 324, 326
Arthropods .................................................................... 323 Cyclin D1 .................................................... 114, 329, 330
Asia................................................................................. 335 Cytoskeletal reorganization .......................................... 321
ATP ....................................................................... 329, 383
D
ATP production .........................312, 323, 329, 331, 384
Atrophy................................................................. 319, 320 Deadenylation .................................................... 2, 93, 315
Autophagy .................................... 14, 279, 280, 330, 361 Dehydration ......................................................... 333, 334
Axon guidance............................ 315, 320–322, 334, 338 Dehydration tolerance .................................313, 333–336
Dicer .......................................................7, 33, 58, 79, 93,
B 110, 182, 225, 257, 273, 336, 377, 425
Bats ....................................................................... 320, 321 Differential expression ........................................ 159, 200,
Biochemistry......................................................... 312, 318 316, 317, 336, 358
Biological function .............................................. 111, 120, DiGeorge Syndrome Critical Region 8
204, 272, 273, 279, 281, 283, 286, 311, 314 (DGCR8).............................................3–7, 23, 26,
Blood ......................................................21, 50, 259, 269, 33, 92–94, 175, 337–339, 424
270, 283–285, 303, 319, 327, 334, 350, 351, DNA ......................................................... 6, 7, 11, 40–43,
360–362, 391, 393, 396 47–49, 81, 84, 86, 97, 117, 271, 298, 303, 319,
321, 357, 360, 377, 379, 382–383, 387, 396
vol. 2257, https://doi.org/10.1007/978-1-0716-1170-8, © Springer Science+Business Media, LLC, part of Springer Nature 2022
439
MIRNOMICS: MICRORNA BIOLOGY AND COMPUTATIONAL ANALYSIS
440 Index
DP103 ........................................................................... 337 Genetics ..................................................15, 79, 157, 167,
Drosha ....................................................2, 33, 57, 79, 92, 199, 212, 255–263, 313, 361, 375, 424
175, 225, 257, 273, 303, 336, 377, 425 Genome ..................................................1, 43, 63, 79, 94,
dsRNA ...................................................4, 10, 23, 92, 336 110, 142, 170, 176, 199, 212, 294, 322, 358,
382, 425
E Glucose ....................................................... 283, 301, 317,
Embryology................................................................... 333 319, 320, 323, 324, 326, 384–386
Enoyl-CoA hydratase and the 3-hydroxyacyl Glycerol........................................................ 313, 323, 324
Glycogen......................................... 6, 313, 317, 329–331
CoA dehydrogenase (EHHADH) .......... 335, 336
Environmentally-induced hypometabolism................. 316 Glycogenolysis............................................................... 326
Environmental stresses........................................ 312–314, Goldenrod gall fly (Eurosta solidaginis)......................316,
324, 325
333, 335–340, 435
Enzyme ......................................................2–4, 15–19, 25, Gray mouse lemur (Microcebus murinus) ...................316,
26, 44, 45, 48, 86, 92, 113, 175, 182, 235, 255, 322
Guide strand ............................................... 11, 12, 15, 20,
260, 274, 312, 317, 324, 337, 338, 384, 385,
426, 429 58, 92, 93, 314, 336
Equilibrium freezing point ........................................... 325
H
Equilibrium freezing point depression ........................ 323
ERK1/2 signalling cascades ......................................... 334 Heart........................................................... 259, 278, 301,
Estivation .................................... 313, 314, 316, 333–336 315–320, 323, 326–328, 333, 338, 397
Evolutionarily conserved ........................ 79, 91, 157, 313 Heartbeat.............................................................. 312, 323
Exportin-5 ........................................................... 4, 34, 58, Heart rate ............................................................. 320, 334
79, 92, 175, 337, 339, 377, 425, 426 Heat shock proteins ............................270, 293–305, 313
Heme oxygenase-1........................................................ 337
F Hemoglobin affinity...................................................... 334
Fatty acid synthase (FAS) .................................... 317, 386 Hepatopancreas ..........................316, 325, 331, 332, 338
Hibernating marsupials................................................. 322
Feeding ........................................................ 312, 335, 428
Fish............................................................... 214, 323, 326 Hibernation ................................................ 313–322, 327,
5‘ seed region ................................................................ 314 329, 330, 337, 338
Focal adhesion...................................................... 322, 358 HIF1α ..................................................325, 331, 332, 387
Homeostasis ...................................................... vii, 34, 93,
Food scarcity ........................................................ 315, 321
FOXO ............................................................................ 320 187, 269, 271, 295, 301, 312, 317, 336, 376,
Freeze avoidance .................................313, 316, 323–328 398, 423
Humboldt squid (Dosidicus gigas) ..............................316,
Freeze-thaw cycle ........................................ 326, 327, 339
Freeze tolerance .................................................. 313, 314, 331–333, 340
316, 323–328, 330, 331 Hydrothermal vent shrimp (Rimicaris
exoculata) ........................................................... 333
Freezing ............................................................... 320, 323,
325–328, 331, 333, 337, 338, 340 Hydroxylation ........................................................ 13, 338
Frogs .................................................................... 195, 323, Hypertrophy............................................... 260, 263, 279,
280, 318, 319, 326, 333
326–328, 333, 335, 338–340
Fuel reserves .................................................................. 313 Hypometabolism................................................. 313, 320,
327, 330, 334–336, 338
G Hypoxia ............................................................... 302, 313,
331–334, 337, 338, 389
Gall moth (Epiblema scudderiana) ..................... 316, 324 Hypoxia tolerance ...............................314, 316, 329–333
GATA4.................................................................. 318, 319
Gene expression ............................................ vii, 1, 37, 50, I
79, 82, 85, 91–93, 98, 99, 105–108, 136, 145,
167, 187, 188, 250, 255, 258, 259, 261–263, Ice crystallization .......................................................... 323
Ice nucleation ....................................................... 323, 326
273, 294, 295, 311, 313–315, 320, 323, 324,
375, 389, 395, 423, 435 Illumina sequencing...................................................... 318
Gene ontology (GO) ........................................59, 60, 64, Insecticides .................................................................... 325
67, 68, 74, 143, 321, 365 In silico predictions..................................... 198, 201, 322
Insulin ................................................................... 301, 319
Index 441
Insulin resistance .................................................. 318, 319 Microarrays ................................................. 37, 43–45, 48,
Invertebrates............................................... 221, 323–326, 50, 51, 65, 66, 69, 74, 82, 83, 86, 107, 109, 113,
328–331, 333, 338 155, 168, 199, 200, 204, 316, 318, 359
Ischemia....................................................... 320, 327, 340 MicroRNAs (miRNAs) ..............................................1–27,
33–51, 57–76, 79–87, 91–100, 105–123,
K 131–162, 167–172, 175–205, 211–228,
Kidney....................... 315–317, 329, 330, 334, 338, 368 235–253, 255–263, 269–288, 293–305,
Kyoto encyclopedia of genes and genomes 311–341, 349–359, 375–408, 423–436
biogenesis ..............................................2, 5–9, 12, 15,
(KEGG) .................................................34, 61, 64,
68, 69, 74, 143, 305, 321, 334, 365 16, 19, 21, 22, 26, 33, 34, 79, 92–94, 175, 257,
258, 260–262, 273, 313, 336–340, 376–378,
L 394, 395, 408, 424–429
MicroRNA-mRNA duplexing ...................................... 330
Lactate ......................................................... 329, 331, 385 MicroRNA-mRNA interaction........................... 108–111,
Lemurs .................................................................. 315, 322 113, 114, 188, 203, 328
Lipid catabolism .......................................... 312, 321, 335 MicroRNA targeting................................................79–87,
Lipids .................................................... 82, 269–274, 277, 105–123, 187–205
278, 287, 317, 319, 324, 325, 384, 386, 387, Mitochondria........................................................ 379, 380
395, 396 Mitochondrial uncoupling proteins ............................. 325
Little brown bat (Myotis lucifugus) .............................316, Molecular biology .......................................... vii, 167, 305
320, 338 Molluscs....................................................... 323, 325, 331
Liver ................................................. 17, 20, 60, 261, 299, Monito del monte (Dromiciops gliroides) .................... 321
301, 315–319, 321, 322, 326–330, 334, 350, mTOR......................................................... 321, 322, 334,
368, 379, 381, 385, 386 385, 386, 400–402, 405, 407
Locomotion.......................................................... 312, 323 Muscle atrophy.............................................................. 319
Low molecular weight osmolytes ................................ 323 MyomiRs ....................................................................... 320
Myostatin..................................................... 260, 318, 320
M
Madagascar .................................................................... 322 N
Mammalian hibernation ...................................... 314–322 NADPH......................................................................... 331
Mammals ...................................................... 2, 9, 16, 161, Neural activity ............................................................... 312
214, 223, 236, 284, 326 Neuronal differentiation ............................................... 321
MAPK ......................................................... 6, 10, 12, 302, Neuroprotective .................................................. 321, 327,
317, 321, 322, 328, 338, 399, 400, 404 332, 334, 340
Marsupials.....................................................315, 321–322 Next-generation sequencing (NGS) ............................. 44,
MEF2............................................................................. 320 45, 51, 65, 66, 75, 83, 107, 136, 167, 171, 176,
Membrane transport ..................................................... 273 198, 316, 351, 430
messenger RNAs (mRNAs) .......................................1, 33, NFκB ....................................................86, 279, 280, 285,
57, 79, 91, 105, 131, 167, 175, 255, 270, 312, 286, 358, 377, 381, 387, 388, 399, 404
364, 376, 424 Nitrogen waste .............................................................. 334
decay ........................................................................ 315 Non-coding RNAs (ncRNAs) ................................ 57, 73,
degradation........................................... 11, 15, 16, 18, 91, 98, 99, 117, 131–133, 136, 145, 146, 158,
26, 33, 80, 81, 93, 94, 96, 108, 312, 315, 376, 199, 213, 216, 225, 236, 395, 424
427, 434 Northern crayfish (Orconectes virilis) ........................... 332
destabilization ................................................ 108, 315 Nrf2 ............................................279, 280, 319, 320, 397
storage............................................312, 313, 327, 328
translation ............................. 108, 111, 315, 327, 426 O
Metabolic pathways..........................................1, 312, 330
Metabolic rate depression................................... 312, 313, Oxidative damage........................................ 319, 333, 397
Oxygen deprivation.............................................. 329, 331
315, 317, 318, 321–324, 326, 328, 330, 331,
337–340
P
Metabolic reorganization .................................... 312, 331
Metabolism................................................. 295, 296, 298, p53 ..................................................... 3, 7, 257, 328, 330,
319, 321, 324, 325, 330, 331, 384–387, 396 337, 383, 384, 386, 400, 405
442 Index
PACT .........................................................................9, 339 Skin .............................................316, 326, 334, 362, 364
Painted turtle (Chrysemys picta marginata) ....... 328–330 Snails ....................................................325, 331, 335, 338
P-bodies ..........................................................12, 315, 328 Solute carriers ................................................................ 334
Pentose phosphate pathway ......................................... 331 Sorbitol ................................................................. 313, 324
Phosphoglucomutase (PGM)....................................... 330 Species specific microRNA ................................. 168, 176,
Phosphorylation ............................................6, 10, 12–14, 213, 314, 322, 336, 435
262, 263, 319, 321, 329, 337, 338, 386 Spleen.................................................................... 316, 330
PI3K-Akt .............................................................. 321, 358 Squirrels ...................................... 223, 315, 317–320, 338
Plasma viscosity ............................................................. 334 Stresses .....................................................vii, 6, 12–14, 26,
Plasma volume............................................................... 334 151, 155, 261, 269, 278–280, 301, 311–316,
Poikilotherms ................................................................ 323 331, 333–341, 425, 429, 433, 435, 436
PolyADP-ribosylation ............................................ 14, 338 Stress granules .......................................14, 315, 337, 338
Polymerase chain reaction (PCR) ...........................37–41, SUMOylation ................................. 6, 263, 273, 274, 320
44–48, 50, 51, 84, 86, 367 Supercooling.................................................................. 325
Polyols................................................................... 323, 324 Survival strategies .........................................311–341, 435
Post-translational modification (PTM)........... 12–14, 273
Protein .......................................................... 1, 33, 57, 82, T
91, 107, 143, 175, 213, 235, 255, 270, 294, 312,
TAR RNA-binding protein (TRBP) ...........................5, 9,
350, 376, 423 10, 26, 273, 339
Protein synthesis .................................312, 319, 325, 331 Temperature-dependent microRNA targeting............ 340
TGFβ..............................7, 303, 317, 319, 388, 402, 407
R
TGFβ-Smad-miRNA axis.............................................. 319
Reactive oxygen species (ROS) ........................... 319, 337 Thermodynamic modeling ........................................... 328
Red-eared slider (Trachemys scripta elegans) ..............329, Thermogenesis ..................................................... 317, 322
330 Thirteen-lined ground squirrel (Ictidomys
Reptiles ................................................................. 323, 326 tridecemlineatus) ............................................... 315
Resources ................................................ 57–76, 113, 117, 3‘ untranslated region (3‘-UTR) .................................. 83,
118, 123, 132, 133, 146, 158, 197, 198, 201, 93, 94, 112, 376, 378
204, 312, 313, 323, 340, 430 Torpor..................................................315–322, 338, 340
Reverse transcriptase quantitative polymerase Torpor-arousal cycle............................................. 315, 321
chain reaction (qRT-PCR).......... 81, 83, 317, 318 Transcriptional networks .............................................. 314
Reverse transcription (RT-PCR) ............................ 39, 40, Transcription factor (TF)..................................3, 61, 113,
45, 51, 113, 278, 316 145, 154, 157, 262, 286, 295, 318–320,
Rickett’s big-footed bat (Myotis ricketti) ..................... 321 337–338, 367, 382, 406, 425, 435
RNA-induced silencing complex (RISC) ....................... 2, Trehalose ....................................................................... 313
10–15, 19, 23–25, 80, 92, 93, 95, 106, 109–111, Type I collagen prolyl-4-hydroxylase........................... 338
176, 235, 273, 314, 336, 337, 377, 425–430,
432–434 U
RNA interference (RNAi) .........................................9, 81, Ubiquitin (Ub)................................. 6, 98, 320, 399, 403
105, 211, 325, 428
Ubiquitination.............................. 99, 273, 320, 382, 424
Urea ...................................................................... 313, 334
S
Salamanders ................................................................... 323 W
Sea cucumber (Apostichopus japonicus)........................ 335 White adipose tissue (WAT) ...............316–318, 321, 338
Shell ............................................................................... 329 Winter ................................ 315, 320, 323–326, 328–330
Skeletal muscles ................................................... 315–318,
Wood frog (Rana sylvatica) ................................ 326, 338
326–328, 330, 338

Mi RNomics

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

Mi RNomics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mi RNomics

Uploaded by

Copyright:

Available Formats

Methods in

Molecular Biology 2257

For further volumes:

MicroRNA Biology and Computational Analysis

ISSN 1064-3745 ISSN 1940-6029 (electronic)

Mülheim adR, Germany Jens Allmer

16 Role of MicroRNAs in Extreme Animal Survival Strategies. . . . . . . . . . . . . . . . . . . 311

HAKAN AKCA • Department of Medical Biology, School of Medicine, Pamukkale University,

ABHISHEK KUMAR • Institute of Bioinformatics, Bangalore, India; Manipal Academy of

ALI M. YAZBECK • Bioinformatics Group, Department of Computer Science, and

Key words MicroRNA, Biogenesis, RNA-induced silencer complex, Argonaute protein

MicroRNAs (miRNAs), first identified in Caenorhabditis elegans

proteins, on the other hand, employ several molecules that provide

2 Regulation of miRNA Transcription

Primary machines for miRNA regulation at the transcription pro-

genes have similar histone modifications, enhancer regions, and the

3 Nuclear Processing of miRNAs

The long primary transcripts produced by the RNA polymerase II

they form the microprocessor complex [27]. A disease called

3.1.2 DGCR8 DGCR8 is a 86 kDa protein that contains an N-terminal domain,

Posttranslational Posttranslational modifications (especially phosphorylation, acety-

Another modification that changes the affinity of DGCR8

RNA-Binding Proteins RNA-binding proteins (RBPs) can regulate microprocessor activity

complex, regulates pri-miRNA stability by interacting with Drosha

The pre-miRNA must be transported to the cytoplasm to complete

4.1 Posttranslational It is reported that posttranslational modifications of XPO5 can

After processing in the nucleus, miRNAs transported to the cyto-

5.2 Regulation Dicer proteins generally interact with double-stranded RNA-bind-

three different ds-RBP domains. Two of them interact with the

6 RNA-Induced Silencing Complex (RISC)

The RNA duplex, processed by Dicer, is loaded onto the AGO

created for silencing genes. The pre-miRNA, processed by Dicer in

in similar studies, and it was concluded that AGO proteins have a

6.1.2 Post-translational Post-translational modifications of AGO on different residues sig-

modification of AGO, inhibits miRNA loading [12, 78]

temperature [11]. The N-terminal domain of the AGO protein

7 RNA Sequence Alterations

Modifications, mutations, and changes in the sequence of miRNAs

7.1 SNPs (single-nucleotide polymorphisms) occurring in the miRNA-

In addition, the expression of miRNAs can be regulated by

stabilization and miRNA degradation in some cases. Gld2 (Germ-

miR-29a/b, which plays a role in the regulation of the cell

Processing of miRNAs by Drosha or Dicer, RNA editing, or

8 Viral Regulators of miRNA

Viral miRNAs, unlike proteins, are nonimmunogenic and do not

modulates miRNA function by expressing uracyl-rich RNAs

Although miRNAs are mainly localized in the cytoplasm, they were

9.2 Extracellular miRNAs are also be localized in extracellular environments like

proteins. These exosomes can induce the proliferation of cells

10 Regulation of miRNA Target Sites

Modifications in the target sites of miRNAs can also regulate the

The miRNA biogenesis also has three noncanonical pathways:

11.1 The essential components of the noncanonical pathway are the

Fig. 8 The microprocessor or Drosha/DGCR8 complex-independent noncanonical pathway comprises “Mir-

[12]. Therefore, the lariat structure is debranched and folds to form

The microprocessor complex independent but Dicer-

11.2 Most miRNAs must be processed by Dicer to mature, but some