978 1 62703 011 3

METHODS IN MOLECULAR BIOLOGY™
Series Editor
John M. Walker
School of Life Sciences
University of Hertfordshire
Hatfield, Hertfordshire, AL10 9AB, UK
For further volumes:

http://www.springer.com/series/7651
Genomic Imprinting
Methods and Protocols
Edited by
Nora Engel
Fels Institute/Biochemistry, School of Medicine, Temple University, Philadelphia, PA, USA
Editor
Nora Engel
Fels Institute/Biochemistry
School of Medicine
Temple University
Philadelphia, PA, USA
ISSN 1064-3745 ISSN 1940-6029 (electronic)

ISBN 978-1-62703-010-6 ISBN 978-1-62703-011-3 (eBook)
DOI 10.1007/978-1-62703-011-3
Springer New York Heidelberg Dordrecht London
Library of Congress Control Number: 2012943828
© Springer Science+Business Media, LLC 2012

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. Exempted from this
legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for
the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.
Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publisher’s location, in its current version, and permission for use must always be obtained from Springer. Permissions
for use may be obtained through RightsLink at theCopyright Clearance Center. Violations are liable to prosecution
under the respective Copyright Law.
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.
While the advice and information in this book are believed to be true and accurate at the date of publication, neither
the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be
made. The publisher makes no warranty, express or implied, with respect to the material contained herein.
Printed on acid-free paper
Humana Press is a brand of Springer

Springer is part of Springer Science+Business Media (www.springer.com)
Preface
While timorous knowledge stands considering, audacious ignorance hath done the
deed.
-Samuel Daniel
Genomic imprinting has been fascinating us for over three decades and has provided many
emerging scientists with the chance to hit their stride in a frontier posing many unexpected
questions and even more surprising answers. Imprinting is the process by which the non-
equivalence of the paternal and maternal genomes is established, leading to parent-of-
origin-specific effects. The most ostensible effects in mammals of parental-specific
marks—and to date, the most accessible to study—are the differential outcomes in gene
expression between the paternal and maternal alleles. During the first two decades, the field
grew hand in hand with technological innovations in embryology and gene targeting,
mainly in the mouse. In fact, advances in imprinting and other unique regulatory mecha-
nisms were instrumental in establishing Epigenetics as the “umbrella organization,” as
Davor Solter so wittily calls it (1). Many of the broader principles of epigenetic regulation
were unearthed by studying imprinted domains (2) and their alterations in cancer and
developmental diseases. As technology has moved forward into the “genome-wide” and
“high-throughput” arenas, many imprinted regions have been even more fully character-
ized—with an abundance of information on the epigenetic modifications occurring at
specific domains and throughout development. The availability of genome sequences and
their variations have moved the field forward enormously. We now know that imprinted
genes tend to occur in clusters, that the mechanisms by which the inactive genes are silenced
vary from one region to another, that establishment and erasure of the imprints occur at
different developmental stages for male and female germ cells, and that DNA methylation
is the most consistent candidate for the imprint, at least in the embryo. Clusters of imprinted
genes are regulated in cis by long-range control elements, designated as imprinting control
regions, and these are the sequences bearing the memory of parental origin. Moreover,
noncoding RNAs with regulatory roles are present in all imprinted domains.
It is interesting to note, however, that we have yet to answer some of the fundamental
questions that the discovery of imprinting posed when it was first described—i.e., how
widespread is imprinting across the animal and plant kingdoms, how does the imprinting
process vary across genotypes and species, how is the imprint targeted to specific DNA
sequences, how is the marking erased, what is the mechanism of tissue-specific and stage-
specific imprinting (3), and what is the functional role and origin of imprinting (4). The
huge amounts of genome-wide epigenetic data are correlative and have not provided an
answer to the question of whether the marks are the cause or consequence of gene expres-
sion state, nor have we gained insight into how chromatin-modifying enzymes are targeted
to specific sequences. Still to be achieved is the feat of conferring imprinting on a normal
gene by transferring a specific sequence into its vicinity. A host of candidate imprinted genes
await validation by site-specific molecular studies. Taking advantage of the combined
v
vi Preface
genomic and epigenomic data, we now need more detailed mechanistic models to be tested.
In addition, new questions have emerged on the variability of imprinting marks in the
population, the effects of culture and in vitro fertilization on imprints, the nature of imprint-
ing in extraembryonic tissues, and the role of noncoding RNAs, among others.
Genomic Imprinting: Methods and Protocols is a survey of the technologies that are being
applied to advance the study of imprinting. It includes new technologies that are accelerat-
ing the pace of discovery of imprinted genes and characterization of their epigenetic profile,
bioinformatic procedures for prediction and comparative analyses of imprinted genes, as
well as methods in embryology and basic molecular biology that have been employed for
many years, some appearing in new versions for small cell numbers. Undoubtedly, focusing
on individual imprinting clusters has uncovered many novel mechanisms in gene regula-
tion, and doing so with traditional but ever more sensitive molecular biology tools will
continue to be essential in elucidating the molecular logic of imprint establishment and
erasure.
Since many of the compelling questions of the field will require querying very small
numbers of cells, we anticipate that the newer technologies will eventually be scaled down
to meet this requirement. Also, bioinformatics will continue to expand its influence in the
field to bring new insights into the evolutionary history of imprinting. Hopefully, we will
also begin to see more of an impact of our imprinting research on other parent-of-origin
effects (5). Although attempts are continuously being made to synthesize and generalize
our knowledge of imprinted genes, the fact remains that each imprinted domain is unique
in some respects, and there is still much to be explored at the molecular level. There is no
doubt the next few years will unveil both much-awaited answers and new questions to keep
us busy for many exciting years to come.
I thank all the authors for their outstanding contributions to this volume.
Philadelphia, PA, USA Nora Engel
References
1. Solter D (1998) Imprinting. Int J Dev Biol 4. Hurst LD (1997) Evolutionary theories of
42:951–4 genomic imprinting. In: Reik W, Surani A (ed)
2. Barlow DP (2011) Genomic imprinting: a Genomic imprinting. Frontiers in molecular
mammalian epigenetic discovery model. Annu biology, 18. IRL Press
Rev Genet 45:379–403 5. Pardo-Manuel de Villena F, de la Casa-Esperon
3. Latham KE (1995) Stage-specific and cell type- E and Sapienza C (2000) Natural selection and
specific aspects of genomic imprinting effects in the function of genome imprinting: beyond the
mammals. Differentiation 59:269–82 silenced minority. Trends Genet 16:573–579
Contents
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
PART I PARENT-OF-ORIGIN EFFECTS
1 Uniparental Embryos in the Study of Genomic Imprinting . . . . . . . . . . . . . . . 3

Yong Cheng, Dasari Amarnath, and Keith E. Latham
2 Derivation of Induced Pluripotent Stem Cells by Retroviral
Gene Transduction in Mammalian Species. . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Masanori Imamura, Hironobu Okuno, Ikuo Tomioka,
Yoshimi Kawamura, Zachary Yu-Ching Lin, Ryusuke Nakajima,
Wado Akamatsu, Hirotaka James Okano, Yumi Matsuzaki,
Erika Sasaki, and Hideyuki Okano
3 Generation of Trophoblast Stem Cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Michael C. Golding
4 Immunomagnetic Purification of Murine Primordial Germ Cells . . . . . . . . . . . 61
Emily Y. Smith and James L. Resnick
PART II IDENTIFYING IMPRINTED GENES
5 Whole Genome Methylation Profiling by Immunoprecipitation

of Methylated DNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
Andrew J. Sharp
6 Identification of Imprinted Loci by Transcriptome Sequencing . . . . . . . . . . . . 79
Tomas Babak
7 Data Mining as a Discovery Tool for Imprinted Genes . . . . . . . . . . . . . . . . . . 89
Chelsea Brideau and Paul Soloway
PART III IDENTIFYING THE REGULATORY FEATURES OF IMPRINTED DOMAINS
8 Engineering of Large Deletions and Duplications In Vivo . . . . . . . . . . . . . . . . 137

Louis Lefebvre
PART IV EPIGENETICS OF IMPRINTED REGIONS
9 Methylated DNA Immunoprecipitation (MeDIP)

from Low Amounts of Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Julie Borgel, Sylvain Guibert, and Michael Weber
vii
viii Contents
10 Chromatin Immunoprecipitation to Characterize

the Epigenetic Profiles of Imprinted Domains . . . . . . . . . . . . . . . . . . . . . . . . . 159
Purnima Singh and Piroska E. Szabó
11 Quantitative Chromosome Conformation Capture . . . . . . . . . . . . . . . . . . . . . 173
Raffaella Nativio, Yoko Ito, and Adele Murrell
12 Genome-Wide Analysis of DNA Methylation in Low Cell Numbers
by Reduced Representation Bisulfite Sequencing . . . . . . . . . . . . . . . . . . . . . . . 187
Sébastien A. Smallwood and Gavin Kelsey
PART V ANALYSIS OF IMPRINTED EXPRESSION
13 Isolation of RNA and DNA from Single Preimplantation

Embryos and a Small Number of Mammalian Oocytes
for Imprinting Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Sarah Rose Huffman, Md Almamun, and Rocío Melissa Rivera
14 Generation of cDNA Libraries from RNP-Derived Regulatory
Noncoding RNAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Mathieu Rederstorff
15 Co-Immunoprecipitation of Long Noncoding RNAs . . . . . . . . . . . . . . . . . . . 219
Victoria A. Moran, Courtney N. Niland, and Ahmad M. Khalil
PART VI IMPRINTING IN PLANTS
16 Specialized Technologies for Epigenetics in Plants . . . . . . . . . . . . . . . . . . . . . . 231

Wenyan Xiao
PART VII EVOLUTION OF IMPRINTED GENES
17 Computational Studies of Imprinted Genes . . . . . . . . . . . . . . . . . . . . . . . . . . . 251

Martina Paulsen
18 Insights on Imprinting from Beyond Mice and Men . . . . . . . . . . . . . . . . . . . . 263
Andrew Pask
19 Nonmammalian Parent-of-Origin Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Elena de la Casa-Esperón
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
Contributors
WADO AKAMATSU • Department of Physiology, School of Medicine, Keio University,

Tokyo, Japan
MD ALMAMUN • Division of Animal Sciences, University of Missouri, Columbia,
MO, USA
DASARI AMARNATH • Department of Biochemistry, The Fels Institute for Cancer
Research and Molecular Biology, Temple University School of Medicine,
Philadelphia, PA, USA
TOMAS BABAK • Department of Biology, Stanford University, Stanford, CA, USA
JULIE BORGEL • Institute of Molecular Genetics, UMR 5535,
Université Montpellier 2, Université Montpellier 1, CNRS, Montpellier, France
CHELSEA BRIDEAU • Nuffield Department of Surgical Sciences, Gray Institute
for Radiation Oncology and Biology, The University of Oxford, Oxford, UK
ELENA DE LA CASA-ESPERÓN • Albacete Science and Technology Park, Regional Center
for Biomedical Research (C.R.I.B.), University of Castilla-La Mancha,
Albacete, Spain
YONG CHENG • Department of Biochemistry, The Fels Institute for Cancer Research
and Molecular Biology, Temple University School of Medicine, Philadelphia, PA, USA
MICHAEL C. GOLDING • Veterinary Physiology and Pharmacology, Texas A&M
University, College Station, TX, USA
SYLVAIN GUIBERT • Institute of Molecular Genetics, UMR 5535,
Université Montpellier 2, Université Montpellier 1, CNRS, Montpellier, France;
UMR 7242 Biotechnology and Cell Signalling, Université de Strasbourg, CNRS,
ESBS, Illkirch, France
SARAH ROSE HUFFMAN • Division of Animal Sciences, University of Missouri,
Columbia, MO, USA
MASANORI IMAMURA • Department of Physiology, School of Medicine, Keio University,
Tokyo, Japan
YOKO ITO • Department of Oncology, CRUK Cambridge Research Institute,
University of Cambridge, Cambridge, UK
YOSHIMI KAWAMURA • Department of Physiology, School of Medicine, Keio University,
Tokyo, Japan
GAVIN KELSEY • Epigenetics Programme, The Babraham Institute, Cambridge, UK;
Centre for Trophoblast Research, University of Cambridge, Cambridge, UK
AHMAD M. KHALIL • Department of Genetics, Center for RNA Molecular Biology,
Case Western Reserve University School of Medicine, Cleveland, OH, USA
KEITH E. LATHAM • Department of Biochemistry, The Fels Institute for Cancer Research
& Molecular Biology, Temple University School of Medicine, Philadelphia, PA, USA
ix
x Contributors
LOUIS LEFEBVRE • Department of Medical Genetics, Molecular Epigenetics Group,

Life Sciences Institute, University of British Columbia, Vancouver, BC, Canada
ZACHARY YU-CHING LIN • Department of Physiology, School of Medicine,
Keio University, Tokyo, Japan
YUMI MATSUZAKI • Department of Physiology, School of Medicine, Keio University,
Tokyo, Japan
VICTORIA A. MORAN • Department of Genetics, Center for RNA Molecular Biology,
ADELE MURRELL • Department of Oncology, CRUK Cambridge Research Institute,
University of Cambridge, Cambridge, UK
RYUSUKE NAKAJIMA • Department of Physiology, School of Medicine, Keio University,
Tokyo, Japan
RAFFAELLA NATIVIO • Department of Oncology, CRUK Cambridge Research Institute,
University of Cambridge, Cambridge, UK; Laboratory of Receptor Biology
and Gene Expression, National Cancer Institute, Bethesda, MD, USA
COURTNEY N. NILAND • Department of Genetics, Center for RNA Molecular Biology,
HIDEYUKI OKANO • Department of Physiology, School of Medicine, Keio University,
Tokyo, Japan
HIROTAKA JAMES OKANO • Department of Physiology, School of Medicine,
Keio University, Tokyo, Japan
HIRONOBU OKUNO • Department of Physiology, School of Medicine, Keio University,
Tokyo, Japan
ANDREW PASK • Department of Molecular and Cell Biology, University of Connecticut,
Storrs, CT, USA
MARTINA PAULSEN • Life Sciences, Saarland University, Saarbrücken, Germany
MATHIEU REDERSTORFF • Université de Lorraine, Biopôle, CNRS UMR 7214 AREMS,
Vandoeuvre-lès-Nancy
JAMES L. RESNICK • Department of Molecular Genetics and Microbiology,
College of Medicine, University of Florida, Gainesville, FL, USA
ROCÍO MELISSA RIVERA • Division of Animal Sciences, University of Missouri,
Columbia, MO, USA
ERIKA SASAKI • Department of Physiology, School of Medicine, Keio University, Tokyo,
Japan; Laboratory of Applied Developmental Biology, Marmoset Research
Department, Central Institute for Experimental Animals, Kawasaki, Japan;
PRESTO Japan Science and Technology Agency, Tokyo, Japan
ANDREW J. SHARP • Department of Genetics and Genomic Sciences, Mount Sinai
School of Medicine, New York, NY, USA
PURNIMA SINGH • Department of Molecular and Cellular Biology, City of Hope
National Medical Center and Beckman Research Institute, Duarte, CA, USA
SÉBASTIEN A. SMALLWOOD • Epigenetics Programme, The Babraham Institute,
Cambridge, UK; Centre for Trophoblast Research, University of Cambridge,
Cambridge, UK
EMILY Y. SMITH • Department of Molecular Genetics and Microbiology,
College of Medicine, University of Florida, Gainesville, FL, USA
Contributors xi
PAUL SOLOWAY • Division of Nutritional Sciences, Cornell University, Ithaca, NY, USA
PIROSKA E. SZABÓ • Department of Molecular and Cellular Biology, City of Hope
National Medical Center and Beckman Research Institute, Duarte, CA, USA
IKUO TOMIOKA • Department of Physiology, School of Medicine, Keio University, Tokyo,
Japan; Laboratory of Applied Developmental Biology, Marmoset Research
Department, Central Institute for Experimental Animals, Kawasaki, Japan
MICHAEL WEBER • Institute of Molecular Genetics, UMR 5535, Université Montpellier 2,
Université Montpellier 1, CNRS, Montpellier, France; UMR 7242 Biotechnology
and Cell Signalling, Université de Strasbourg, CNRS, ESBS, Illkirch, France
WENYAN XIAO • Department of Biology, Saint Louis University, St. Louis, MO, USA
Part I
Parent-of-Origin Effects
Chapter 1
Uniparental Embryos in the Study of Genomic Imprinting

Yong Cheng, Dasari Amarnath, and Keith E. Latham
Abstract
Nuclear transplantation has been used to study genomic imprinting. Available nuclear transfer methods
include pronuclear transfer (PNT), intracytoplasmic sperm injection, and round spermatid injection.
By generating uniparental embryos that have exclusively paternal or maternal genomes, it is possible to
study the functions of the parental genomes separately. It is possible to compare functions in haploid and
diploid states. In addition, nuclear transfer allows the effects of the ooplasm, including mitochondria, to
be distinguished from effects of the maternally inherited chromosomes. PNTs can also be used to study
epigenetic modifications of the parental genomes by the ooplasm. This chapter reviews the methods
employed to generate uniparental embryonic constructs for these purposes.
Key words: Pronuclear transfer, Androgenone, Gynogenone, Parthenogenone, Uniparental embryo,

Imprinting, Ooplasm
1. Introduction
Green algae, carrots, salamanders, frogs, sea urchins, and mammals—

all of these organisms have been used in a remarkable series of
experiments dating back for over a century to study nuclear potency
via nuclear transplantation, embryo splitting, and cellular repro-
gramming (1), leading ultimately to the demonstration of nuclear
totipotency of somatic cells by the end of the last century (2, 3).
These studies demonstrated that in most organisms the hereditary
material is contained within the nucleus and remains intact during
development, establishing the basic foundation for our current
concept of cellular differentiation via epigenetic regulation of
the genome.
Epigenetic regulation of the embryonic genome begins at con-
ception and proceeds throughout the life of the organism. The earli-
est steps in this long legacy are mediated by the ooplasm acting upon
the maternal and paternal genomes within the newly formed zygote.
Nora Engel (ed.), Genomic Imprinting: Methods and Protocols, Methods in Molecular Biology, vol. 925,
DOI 10.1007/978-1-62703-011-3_1, © Springer Science+Business Media, LLC 2012
3
4 Y. Cheng et al.
These early epigenetic processes are responsible for creating and

activating the embryonic genome, maintaining parent-of-origin-
specific information (imprints), modifying imprints, establishing
or modifying epigenetic information, providing adaptive
responses to environmental stressors, and initiating the develop-
mental program.
While we accept that these processes occur in the early embryo,
the nature and extent of the resulting modifications remains an
active and interesting area of study. The techniques of somatic cell
hybridization, heterokaryons, and induction of pluripotency using
combinations of transcription factors (4, 5) have provided insight
into mechanisms that can be employed in vitro to modify epigenetic
information. The degree to which such in vitro methods recapitulate
normal embryonic processes is uncertain. Nuclear transplantation
remains a key approach to understanding epigenetic mechanisms
and processes during early development.
With nuclear transplantation, it is possible to dissect the various
roles played by the ooplasm, mitochondria, maternal genome, and
paternal genome in early development: (1) nuclear transplantation
allows the epigenetic aspects of the parental genomes to be studied
within the context of a normally fertilized cell; more specifically,
it permits the ontogeny of imprinting information to be studied;
(2) it allows genetic and evolutionary questions to be addressed by
creating interstrain or interspecies nuclear-cytoplasmic hybrids;
(3) matrilineal effects can be characterized as either ooplasmic or
genomic; (4) interactions between the nucleus and other cellu-
lar organelles and structures can be examined; (5) the reversibility
and timing of epigenetic changes can be determined; and (6) it
potentially allows the effects of reduced ooplasm quality to be
overcome.
Thus, nuclear transplantation remains an important tool for
understanding developmental changes in the epigenetic control
of genome function. This chapter provides examples of different
applications of nuclear transplantation methods to study epigenetic
genome regulation.
1.1. Uniparental The existence of genomic imprinting in mammals was initially

Embryos discovered through a combination of pronuclear transfer (PNT)
studies and genetic studies. Genetic studies involving mice bearing
Robertsonian translocations revealed that certain chromosome
segments affected phenotype differently when of maternal or paternal
origin (6). PNT studies revealed lethal phenotypes in embryos
possessing exclusively maternal or paternal chromosomes (7–9).
Studies of parthenogenetic embryos likewise indicated defects
related to the absence of a paternal set of chromosomes.
Through application of single PNT methods developed ini-
tially by McGrath and Solter (10, 11), it was found that diploid
androgenones (exclusively paternal chromosomes) and gynogenones
1 Nuclear Transfer for Uniparental Embryos 5
(exclusively maternal chromosomes) are unable to develop to term

(7–9). Abnormal phenotypes are somewhat complementary, with
the extraembryonic tissues being most highly affected in gynog-
enones, as compared to greater deficiencies in blastocyst develop-
ment and development of the embryo proper with androgenesis.
Pronuclear transplantation and uniparental diploid embryos
also provided new insight into the biology of imprinting when
individual genes were examined at early stages of development. For
example, the Igf2r gene, which is expressed from the maternal
allele in most somatic tissues of the mouse, is nevertheless expressed
in androgenetic preimplantation stage embryos (12). The Ascl2
gene (a.k.a. Mash2) was likewise biallelically expressed in preim-
plantation stage uniparental embryos (13). Subsequent studies
confirmed stage-dependent and/or tissue-dependent gene silencing
of many imprinted genes. More recent studies revealed acquisition
of imprinted gene methylation and histone acetylation patterns
during preimplantation development (14, 15) as well as changes in
non-imprinted genes (14).
More recently, nuclear transfer has been used to study the
ontogeny of imprints during gametogenesis. Nuclear transfer in
early stage oocytes yielded oocytes that could be activated parthe-
nogenetically and display enhanced developmental potential.
Combined with targeted gene modifications, term development of
parthenogenetic mice was achieved (16, 17). These studies, com-
bined with developmental studies of acquisition of DNA methy-
lation patterns (18), provided novel insight into the timing of
establishment of imprints during oogenesis, indicating that indi-
vidual gene imprints are acquired at different times. In contrast,
nuclear transfer using round spermatids and both primary and
secondary spermatocyte nuclei into oocytes supported term devel-
opment, indicating that paternal imprinting is established at least
by the primary spermatocyte stage (19, 20), although other devel-
opmental factors affect faithful chromosome segregation and can
limit early development (19).
The ontogeny of imprinting information is also relevant to
understanding post-fertilization modification of epigenetic infor-
mation. It is well established that epigenetic inheritance is modified
after fertilization, including, for example, strain-specific oocyte
modifiers of gene function (21), active global DNA demethylation
state and changes in other chromatin properties of the paternal
genome (22–28), and differential modification of microinjected
transgenes regulated by the initial methylation state (29). The ques-
tion invariably arises whether such post-fertilization modifications
are related to imprinting, as many choose to define imprinting as a
strictly gametogenic process. However, differential modification of
parental genomes (which remain physically separated) after fertil-
ization will yield the equivalent result of gametogenic imprints,
namely, parental chromosome-specific modifications.
6 Y. Cheng et al.
PNT yielded early evidence for differential modification of the

paternal genome by the ooplasm. Androgenones display differ-
ences in developmental potential determined by the ooplasm
strain-of-origin (C57Bl/6 and DBA/2) (21). This effect is not
observed with gynogenetic embryos, and is due to a stable
modification imposed by the time pronuclei are formed, indicating
that the ooplasm specifically modifies the paternal genome during
the period immediately following fertilization. This difference was
mapped to two separate genetic loci (30, 31). A separate set of
genetic loci controlling effects of Balb/c oocyte modifiers on trans-
genes was mapped (32). These observations collectively support a
model wherein genomic imprinting information from the father
may be subject to an editing function of the ooplasm, possibly to
compensate for genetic variation in maternal imprints. More
recently ooplasm transfer combined with intracytoplasmic sperm
injection (ICSI) to generate diploid androgenones revealed that
the developmental potential of the paternal genome could be
affected by transferring ooplasm from the low developing strain,
but the reciprocal enhancement or rescue with ooplasm from the
high developing stain could not be achieved (33).
Uniparental embryos also have been valuable in studying X
chromosome regulation. Examining mouse 2-cell stage embryos
revealed early expression of the paternal allele of the Xist gene
(34). Subsequent studies in androgenetic and gynogenetic embryos
revealed early repression of genes lying near the paternal X chro-
mosome inactivation center, with spreading to more distal regions
as development proceeded (34), manifested as differences in gene
expression between the androgenetic, gynogenetic, and fertilized
control embryos. Another area in which nuclear transfer has been
useful has been to investigate the interaction between parent of
origin and haploidy (35).
One very interesting, emerging area of study relates to under-
standing transgenerational inheritance. Mouse 2-cell stage embryos
display a genetic variation in predisposition to blastomere fragmen-
tation. PNT to vary the combination of maternal and paternal origin
of ooplasm, maternal genome, and paternal genome revealed that
the maternal pronucleus was the main determinant of fragmenta-
tion (36). Interestingly, different effects were seen for reciprocal
F1 hybrid maternal pronuclei, indicating an effect of the maternal
grandpaternal allele. Such insights can be realized when the effects
of maternal genome and ooplasm are separated microsurgically.
1.2. Production of The above discussion illustrates the value of uniparental embryos
Uniparental Embryos in studying epigenetic processes during early development, partic-
ularly genomic imprinting. There are many different ways to pro-
duce uniparental embryos. Parthenogenesis has been employed to
study maternal imprinting and to search for novel imprinted genes
(37). Though production of diploid parthenogenones is generally
simple to perform in comparison to microsurgery, consideration

needs to be given to the possible effects of the chosen method of
oocyte activation on gene regulation (38). Androgenones and
gynogenones prepared by PNT are an attractive, classical source of
uniparental embryos for study. Androgenones can also be produced
by removing the oocyte spindle–chromosome complex (SCC) and
then injecting either two sperm or two spermatids. The following
sections describe routine methodologies for PNT, sperm injection,
and spermatocyte/spermatid nuclear transfer.
2. Materials
The methods described here are all well established and widely
applied, and have been the subject of many recent laboratory
protocol publications. However, considerable variability exists in
the specific equipment, solutions, media, and procedural details
that can be incorporated. Our goal here is to describe procedures
that will be effective, offer choices in some of these details, and
provide information about the potential impact of some of these
procedural variations.
2.1. Equipment 1. Stereo microscope (see Note 1).

2. Inverted microscope with micromanipulators (see Note 2).
3. Microforge (see Note 3).
4. Pipet Puller (see Note 4).
5. Pipet beveler (see Note 5).
6. Microinjectors (see Note 6).
7. Piezo pipet driver (see Notes 7 and 8).
8. Electrofusion device (e.g., ECM 2001, BTX Inc., San Diego,
CA, USA).
9. Temperature-controlled, humidified CO2 cell culture incubator.
10. Billups-Rothenberg (Del Mar, CA) modular incubators or
equivalent.
2.2. Culture and 1. HEPES-buffered CZB (HCZB) or modified M2 medium as

Culture Media (See described (39).
Notes 9 and 10) 2. KSOM medium or sequential media (e.g., CZB followed by
Whitten’s medium or M16; see refs. 40–44).
3. Activation medium: Ca2+-free CZB or KSOM supplemented
with 10 mM SrCl2.
4. PVP supplemented media: HCZB with10% PVP or 7% PVP.
5. Electrofusion medium: 275 mM mannitol, 0.05 mM CaCl2,
0.1 mM MgSO4, and 0.3% BSA.
8 Y. Cheng et al.
6. Dulbecco’s PBS containing 5.6 mM glucose and 5.4 mM

sodium lactate (GL-PBS).
7. Erythrocyte lysis buffer (ELB): 155 mM NH4Cl, 10 mM
NaHCO3, 2 mM EDTA, pH 7.2.
2.3. Solutions 1. Cytochalasin B (Sigma, 5 mg/ml 1000 ´ stock in ethanol).

and Chemicals 2. Demecolcine (Sigma, 0.2 mg/ml 1000 ´ stock).
(See Note 11)
3. Equine (pregnant mare) chorionic gonadotropin (eCG, a.k.a.
PMSG) (Calbiochem, EMD Chemicals, Gibbstown, NJ).
4. Human Chorionic gonadotropin (hCG) (Sigma, St. Louis,
MO).
2.4. Pipets 1. Embryo transfer pipet connected to aspiration device.

(See Note 12) 2. Holding pipets (see Note 13).
3. Spindle removal pipets (see Note 14).
4. PNT pipets (see Note 15).
5. Sperm ICSI pipets (see Note 16).
6. Spermatocyte and round spermatid nuclear transfer pipets.
3. Methods
3.1. Oocyte Isolation 1. Isolate MII stage oocytes from females after either spontaneous
and Culture ovulation or, more commonly, induced superovulation (5 IU
eCG followed 46–48 h later with 5 IU hCG). Oocytes are best
isolated near the time of ovulation at approximately 14 h post-
hCG injection, and then manipulated promptly, followed by
embryo culture or activation procedure if needed.
2. Release oocytes from the ampullae into either HCZB or M2
medium containing 4.16 mM bicarbonate.
3. For microsurgical manipulations, remove cumulus cells by brief,
gentle treatment with hyaluronidase (Sigma, H3506, stock
concentration 600 U/ml diluted to 100 U/ml when applied)
at room temperature as rapidly as possible. Oocytes are then
cultured in the medium of choice (e.g., CZB medium). Once
manipulated and activated, the constructs are washed and cul-
tured in the appropriate medium, depending on embryo type.
3.2. Embryo Isolation 1. Using similar procedures to those described for oocytes, fertil-
and Culture ized zygotes are isolated from mated females, typically at
19–20 h post hCG injection.
2. Culture embryos in medium of choice (e.g., KSOM). Select high-
quality fertilized embryos (most easily recognized by the presence
of pronuclei) of appropriate morphology and granularity for

manipulation. Microsurgical manipulations can be performed
in HCZB, M2, or HEPES-buffered KSOM.
3.3. Pronuclear 1. The basic PNT technique (10) involves removing a plasma
Transfer membrane-bound “karyoplast” containing one pronucleus and
placing it under the zona pellucida of the recipient zygote, fol-
lowed by fusion to complete the PNT. The following setup
and procedure are presented as appropriate for constructing
androgenones and gynogenones using an inverted microscope
system and electrofusion to accomplish karyoplast fusion (see
Note 17). Variations in setup can be made as needed for other
purposes. Major steps in the procedure are shown in Fig. 1.
Fig. 1. Pronuclear transfer procedure. Panels show embryos before manipulation (a) using
blunt pipet or (b) beveled pipet, followed by pronucleus aspiration using (c) a blunt pipet
or (d) a beveled pipet, (e) karyoplast inserted into the perivitelline space, and (f) embryos
after karyoplast fusion.
10 Y. Cheng et al.
2. Place two rows of drops of HEPES-buffered manipulation

medium (e.g., HCZB containing 5 μg/ml CB and 0.2 μg/ml
demecolcine) on the plastic dish under mineral oil (see Note
18), most conveniently in a staggered arrangement to allow
easy access to each drop from both sides.
3. Before manipulation, treat zygotes for at least 30 min with
5 μg/ml CB and 0.2 μg/ml demecolcine at 37 °C in the incu-
bator. One zygote is added to each drop. It is preferable that
all embryos loaded be manipulated in no longer than 30 min
time on the microscope. This often equates to 10–12 drops of
embryos per round.
4. Hold the first zygote on the holding pipet with a slight negative
pressure. Rotate the zygote until it is oriented so that both pro-
nuclei are visible and in the same plane of focus, and the polar
body between 10 and 2 o’clock positions. The maternal
pronucleus is typically smaller and located closer to the polar
body than the paternal pronucleus, although the size difference
may vary with strain. The pronucleus to be removed should be
oriented for easy access to the tip of the PNT pipet (e.g.,
between 3 and 6 o’clock with PNT pipet on right hand side).
5. While maintaining slight to moderate negative pressure on the
holding pipet, insert the PNT pipet (bevel oriented toward the
6 o’clock position) through the zona pellucida without pene-
trating the plasma membrane. This can be accomplished with
a flat-tip pipet inserted through a slit cut in the zona pellu-
cida with a sharp glass needle, with a beveled pipet sharpened
using the “broken spike” method, or with a piezo pipet driver
and unsharpened, beveled pipet. In the latter case, care must
be taken to keep the intensity of pulses as low as possible and
to avoid transmitting the pulse to the oolemma, which would
lyse the cell.
6. Once through the zona, press the tip of the PNT pipet inward
and position it adjacent to the target pronucleus. The pronu-
cleus can be nudged to ensure that it is in position.
7. Apply negative pressure to draw the intervening plasma mem-
brane, minimal cytoplasm, and pronucleus gradually into the
pipet. Withdraw the pipet from the perivitelline space. The
karyoplast and plasma membrane will seal themselves. The first
pronucleus removed in the experiment is discarded, thus form-
ing the first recipient. Simultaneously with removal of each
pronucleus, the operator can remove the polar body if desired.
8. Using the same approach, obtain a donor karyoplast from the
next zygote. Once the karyoplast is drawn into the pipet, a
small volume of medium is drawn in to prevent the karyoplast
from contacting the mineral oil between drops. The operator
returns to the previously manipulated zygote (now the recipient).
The recipient zygote is reacquired onto the holding pipet so
that the original opening in the zona pellucida is oriented at

the 3 o’clock position. The opening is often easily visible by a
small strand of cytoplasm, or as a slit seen in profile. The tip of
the PNT pipet is reinserted into the perivitelline space, and the
karyoplast is gently expelled. If desired, the PNT pipet tip can
be withdrawn from the perivitelline space as soon as the pronu-
cleus portion of the karyoplast is expelled in order to reduce
transfer of any excess cytoplasm, which can then be expelled
separately.
9. Alternately transfer maternal and paternal pronuclei to successive
recipients to yield both gynogenones and androgenones. If no
loss occurs, this leaves androgenones in one row and gynog-
enones in the other, with one haploid zygote to serve as recipi-
ent for the next round of embryos. It is recommended that the
operator maintain an ongoing record to confirm the constructs
in each drop during each round of manipulations.
10. Wash PNT constructs and return to embryo culture without
cytoskeletal inhibitors for at least 30 min at 37 °C. Electrofusion
is then performed using a suitable apparatus and a dish con-
taining electrodes about 1 mm apart. Because of density differ-
ences between the electrofusion and embryo culture media,
embryos should be washed through electrofusion medium or
equilibrated with this medium inside the pipet before loading
between electrodes. With the BTX system, a brief AC pulse
(50 V/cm) can be given to orient the constructs between the
electrodes (membranes at point of contact between karyoplast
and recipient cell parallel to the electrodes). Immediately after
orientation, a single DC pulse of 900 V/cm is delivered (see
Note 19). Embryos are then washed through several drops of
culture medium and allowed to recover in the incubator.
Fusion should be completed within about 1 h.
3.4. Intracytoplasmic ICSI involves injecting a spermatozoon into the ooplasm of

Sperm Injection matured eggs to achieve fertilization. ICSI bypasses the need for
sperm motility, zona penetration, binding, and fusion to the oocyte.
ICSI in the mouse requires piezo-actuated micromanipulation to
avoid lysing the oocyte (45). ICSI can be used to produce unipa-
rental androgenone embryos by removing the maternal chromo-
somes and injecting one (haploid) or two (diploid) spermatozoa.
It is also useful where a high efficiency of fertilization is needed, or
where IVF may be problematic.
1. Prepare injection pipet. The diameter of injection pipet for
sperm varies with different strains. For example, DBA/2 strain
sperm heads are a little larger than C57BL/6 sperm heads.
Correspondingly, the diameter of the injection pipet for inject-
ing DBA/2 sperm is bigger than that for injecting C57BL/6
sperm (see Note 16). A small bead of mercury is introduced
into the injection pipet in order to increase the mass loaded
12 Y. Cheng et al.
and decrease the lateral oscillations that may damage the oocyte
when using the piezo driver.
2. Enucleate MII stage oocytes (for preparing androgenones):
Place eggs in the drop of M2 medium with 5 μg/ml CB for
3 min. The SCC in matured eggs is visible as a nongranular
“clear” region within the ooplasm under Hoffman modulation
contrast optics. Gently aspirate the egg onto the holding pipet,
rotate it (can use fluid flow in and out of holding pipet and
contact with the enucleation pipet to turn the oocyte) to the
position with the spindle at 3 o’clock, and then stabilize the
position using negative pressure in the holding pipet. Move
the enucleation pipet to the outer surface of zona pellucida at
3 o’clock. A couple of piezo pulses are applied to allow the
enucleation pipet to penetrate through the zona pellucida into
the perivitelline space. Piezo pulses should be terminated as
the inner surface of the zona pellucida is approached to avoid
lysing the oocyte. Position the enucleation pipet adjacent to
the spindle, and observe the spindle move as this is achieved.
Gently increase the negative pressure on the spindle removal
pipet to aspirate the spindle into the pipet. Withdraw the pipet
from the perivitelline space to remove the SCC as a membrane-
bound “karyoplast.” Push the spindle out of the pipet and
release the enucleated oocyte (cytoplast) to complete one enu-
cleation procedure. Continuously remove the spindles of other
oocytes as rapidly as possible, within about 10 min for experi-
enced operator to remove 20–30 spindles in one round.
Completely rinse the cytoplasts in fresh CZB medium and
allow them to recover in the incubator at least 15 min.
3. Prepare capacitated sperm for injection. We suggest using
capacitated sperm from adult males for ICSI in order to obtain
highest fertilization rates. A 200 μl CZB medium drop in
65 mm diameter culture dish is covered with the mineral oil
and equilibrated in the incubator for at least 30 min. The cauda
epididymes are dissected from one adult male and immediately
placed in the CZB drop to allow the sperm to swim out freely.
It is helpful to squeeze the cauda epididymes with a pair of
sterile fine forceps to increase sperm quantity. After the sperm
become active, sperm at the edge of the medium drop are col-
lected and transferred into 7% PVP drop in the manipulation
plate for injection.
4. Prepare the ICSI micromanipulation dish. Manipulation solu-
tions in the dish for ICSI consist of three kinds of drops. One
drop of 10% PVP is used to lubricate the inner wall of the injec-
tion pipet by repeated aspiration (see Note 20). A couple of 7%
PVP drops are added, in which to place capacitated sperm.
Several HCZB medium drops are applied to the dish, in which
to place the enucleated cytoplasts. In 7% PVP solution, sperm
swim gently and slowly and can be captured easily.
5. Perform sperm head injection. Aspirate the sperm tail first using
the injection pipet and apply several pulses immediately at the
junction between the sperm head and principal piece of tail to
separate sperm head. Blow out the tail and aspirate sperm heads
individually into the injection pipet. It is important not to
accumulate sperm heads touching each other. Holding the
cytoplast with the holding pipet, insert the injection pipet
through the zona pellucida by applying a couple of pulses at
intensity of 3–6 and frequency at 2. Once the injection pipet
passes into the perivitelline space, put the injection tip gently
touching the ooplasm membrane and at the same time, pushing
sperm heads forward to the pipet tip. Press the injection pipet
against the ooplasm membrane and then toward the opposite
side near the holding pipette. Promptly give the pulse (inten-
sity and frequency settings of “1”) to penetrate the ooplasm
membrane. Operator should observe backward rebound of the
oocyte membrane to confirm successful penetration. Gently
push sperm head(s) into the ooplasm and immediately with-
draw the pipet to complete the ICSI procedure. Once all
oocytes in the group are injected, leave them in the injection
dish for about 5 min to recover.
6. Collect the manipulated eggs in and rinse with HEPES-free
CZB medium completely. Transfer the embryos to embryo
culture medium, such as CZB or KSOM, to observe pronu-
cleus formation and embryo development.
3.5. Round Spermatid Viable offspring have been produced from round spermatid
Nuclear Transfer injection (ROSI) in mouse, rat, rabbit, and humans (46, 47). Round
spermatids are immature haploid cells characterized by the pres-
ence of a decondensed nucleus. The difference in nuclear status
between spermatid and spermatozoa, which are decondensed and
condensed, respectively, affects ICSI and ROSI protocols. In a
standard ICSI protocol, a spermatozoon is simply injected into an
MII oocyte. In ROSI, however, injected oocytes must be artificially
activated before or after the injection of the round spermatid.
Similar to ICSI, ROSI can be used to produce uniparental andro-
genic embryos by injecting two round spermatid nuclei.
1. Oocyte preparation for ROSI. While ROSI can be used to
make androgenones using SCC-depleted MII oocytes fol-
lowed by chemical activation, the proportion of oocytes sur-
viving injection of two round spermatids is better (48) if
spermatids are injected into preactivated intact MII oocytes
(progressing to telophase) followed by removal of maternal
pronucleus within 4 h after activation. The injections should
be completed within 70–80 min of activation (48, 49).
Activation of mouse oocytes can be achieved efficiently by
exposing oocytes to 5 mM of SrCl2 in Ca2+-free CZB/KSOM
medium for 20 min.
14 Y. Cheng et al.
The procedure for ROSI is almost the same as described above

for ICSI except for the following changes.
2. Prepare ROSI pipets with an inner diameter of 3.5 μm (adjust
for sperm donor strain as needed) and place a small bead of
mercury inside as described under ICSI.
3. Collect the testes from mature males into GL-PBS and remove
the tunica albuginea using a pair of fine forceps. Allow the
seminiferous tubules to spread into the buffer and cut into
minute pieces using a pair of sharp scissors. Avoid contaminat-
ing testis and seminiferous tubules with blood. If necessary
testes and seminiferous tubules may be washed in ELB briefly
(155 mM NH4Cl, 10 mm KHCO3, 2 mM EDTA, pH 7.2)
briefly before placing them in GL-PBS. Gently pipet the mix-
ture repeatedly to disperse spermatozoa and spermatogenic
cells into the collection medium. Filter the cell suspension
through a 50 μm nylon mesh and wash three times by centrifu-
gation at 200 g for 5 min. Resuspend the cells in GL-PBS and
keep them at 4 °C. The isolated cell suspension stored at 4 °C
should be viable for several hours.
4. Prepare the ROSI micromanipulation dish as in ICSI described
above. Mix a small aliquot of 1–3 μl of cell suspension (from
step 3) with 10 μl of PVP as described above in ICSI.
5. Round spermatids can be recognized easily by their small size
(~10 μm) and a distinct centrally located chromatin mass. For
injection, draw a single spermatid into the injection pipet.
Move the spermatid in and out the pipet until the plasma
membrane is ruptured and the spermatid nucleus is separated
from the cytoplasm. While only one or two spermatids are
injected into a single oocyte, loading several (ten or more)
spermatids into the pipet improves speed of production. Inject
the spermatid nucleus into a telophase or metaphase II stage
oocyte using a piezo-driven pipet as described under ICSI.
Release the injected oocyte and repeat with another oocyte
until all the spermatids drawn into the injection pipet are
finished. Usually 10–15 oocytes can be injected within
15–20 min with practice. Allow the injected oocytes to recover
for 5–10 min before returning them to the culture dish con-
taining CZB droplets. Briefly wash the injected oocytes to
remove HEPES before transferring them to the HEPES-free
CZB drops in the incubator.
6. Following ROSI the maternal of Telophase II oocytes the
maternal pronucleus is removed along with second polar body
in the presence of cytochalasin B and demecolcine as described
for PNT. Allow 5–10 min for the recovery of oocytes after pro-
nucleus removal, and then wash in cytochalasin B-free medium
and return to the incubator. Transfer the embryos to embryo

culture medium, such as CZB or KSOM, to observe pronu-
cleus formation and embryo development.
4. Notes
1. Stereomicroscope providing from 10.5× to 105× magnification,

with mirror (e.g., Gimbal mount) suitable for providing oblique
illumination so that intracellular detail (e.g., pronuclei) can be
seen.
2. The different procedures described here may impose different
optical requirements on the microscope to be used for micro-
surgery or microinjection. Fixed stage, upright microscopes
can be used in conjunction with a hanging drop system for
specimen mounting. More commonly, inverted microscopes
are used in conjunction with plastic dishes with specimens in
medium droplets under oil. An excellent system that will be
easy to use for all of the procedures described here is the
Olympus Inverted model IX71 with modulation contrast
optics and available magnifications of 4×, 20×, and 40× objec-
tives and 10× eyepieces.
3. There are two main kinds of microforge that can be used. One
is a DeFonbrune style microforge that provides fine control of
filament temperature using a combination of rheostat, fan
speed, and flow restriction. Alternatively, Narishige provides a
widely used model (MF900) with excellent control for crafting
fine gauge microneedles.
4. We find the Sutter Instrument line of pipet pullers easily pro-
grammable and capable of yielding pulled pipets with consis-
tent geometry, which can be varied easily through choice of
filament size and type, with guidance from the merchant-
supporting literature.
5. We use the Sutter Instruments pipet beveler, which can come
with a choice of grinding surface and is easily maintained.
6. One microinjector is required for aspiration of pronuclei and
spindles, microinjection transfer of karyoplasts or cells beneath
the zona, or injection of nuclei or sperm heads into the ooplasm.
A second microinjector can be used to control pressure on the
holding pipet, or an air-filled 50 cm3 syringe also works well for
this. The Narishige model IM9B/5B injector works well for
microinjection and the IM9A/5A model works well for the
holding pipet.
7. A common choice for micromanipulator has been the Narishige
hydraulic system (MN4). Eppendorf also provides a widely
16 Y. Cheng et al.
used system. Sutter Instruments offers an attractive new system

of manipulators that provides excellent motion control and
that avoids problems with leaking of hydraulic fluid.
8. We prefer the PMM piezo-drill micromanipulation controller
PMM-150 (Prime Tech Ltd, Ibaraki, Japan).
9. A widely preferred medium for mouse embryo culture is
KSOM (40). This medium is an excellent choice for embryos
because of documented gains in retaining embryo phenotype
very similar to in vivo developing embryos.
10. Two-step systems can also provide excellent results, such as using
CZB (42) through the 8-cell stage followed by culture in
Whitten’s medium. Other media (e.g., HCZB and M2 medium)
are preferred for culturing MII stage oocytes during in vitro
manipulation. CZB and KSOM medium are used to culture MII
stage oocytes inside incubator before and after manipulation.
11. Cytochalasin B is diluted in 100% ethanol to prepare 1,000×
stock (5 mg/ml). Demecolcine is diluted in Millipore water to
prepare 1,000× stock (0.2 mg/ml). Both of them are stored as
aliquots at −70 °C.
12. Pipets to be used on an inverted microscope are typically bent
on the microforge by about 20° at a distance of about 1–2 mm
from the tip.
13. Holding pipets are constructed by pulling pipets with an elon-
gated geometry, approximately 90–100 μm outer diameter
(O.D.). The pipet needs to be cut on the microforge with an
end that is as close to flat and 90° angle to the sides as possible,
and then heat polished on the microforge to yield an opening
of about 15–25 μm.
14. Spindle removal pipets are prepared with a flat tip and an inner
diameter of 10 μm. Adhesion to the cell membrane can be
reduced by washing with 10% PVP.
15. PNT pipets can be prepared with either flat or beveled tips,
20–30 μm inner diameter (I.D.) depending on operator pref-
erence and skill. Smaller pipets are easier to penetrate the zona
pellucida but larger pipets will offer less adhesion to the
membrane of the karyoplast. The tip is beveled at an angle of
45°. The flat-tip pipet needs to be washed three times in 10%
PVP to lubricate the inner wall prior to use. Beveled pipets are
washed with 20% hydrofluoric acid quickly for three times
(excessive time dissolves too much glass), washed with MilliQ
water for five times, and then washed with 95% ethanol for
three times before use. If no piezo driver is to be used in the
PNT, a very fine spike is drawn at the tip using the micoforge,
and the spike is then broken at its base at the time of use. Prior
to use, the PNT pipet should be treated by aspirating Igepal
CA-630 several times aspirating Igepal CA-630 (Sigma, Cat.

I-3021), then washed at least ten times with vigorous agitation
in a large beaker of MilliQ water, and then air dried. This treat-
ment minimizes adhesion of the karyoplast membrane.
16. Sperm injection pipets are blunt, and cut at a tip with outer
diameter of 6 to 9 μm (varies with strain of sperm) and an
inner diameter slightly larger than diameter of sperm head to
avoid adhesion. The PVP in the sperm suspension medium
reduces adhesion.
17. For PNT, karyoplast fusion was originally accomplished by
McGrath and Solter using inactivated Sendai virus introduced
simultaneously with insertion of the karyoplast under the zona
pellucida. This method works very well with a suitable prepara-
tion of virus. However, in the absence of such a virus prepara-
tion, the electrofusion method also works very well with no
apparent detriment to the construct, provided that repeated
fusogenic pulses are avoided.
18. As with all culture system components, mineral oil should be
quality tested to ensure that it supports maximum embryo
viability. Once suitable lots of oil and other culture system
components are identified, it is recommended that a supply of
these lots sufficient to last for a prolonged period is acquired.
19. For electrofusion only a single pulse should be given. Additional
constructs may fuse with additional pulses. However, the addi-
tional pulses may compromise embryo quality.
20. PVP is added to HCZB media at 10% as a lubricant to mini-
mize adherence of cell membranes inside pipets and to facili-
tate nuclear aspiration prior to injection. We use Sigma PVP
-360 (molecular weight 360,000). PVP containing medium
for suspending nuclei or sperm should contain a maximum of
10% w/v (diluted to about 7% after sperm loading). Due to
possible PVP toxicity, the PVP concentration should be kept as
low as possible.
References
1. Di Berardino MA (1997) Genomimc potential 5. Takahashi K, Yamanaka S (2006) Induction of
of differentiated cells. Columbia University pluripotent stem cells from mouse embryonic
Press, New York and adult fibroblast cultures by defined factors.
2. Campbell KH, McWhir J, Ritchie WA et al Cell 126:663–676
(1996) Sheep cloned by nuclear transfer from 6. Cattanach BM, Kirk M (1985) Differential activ-
a cultured cell line. Nature 380:64–66 ity of maternally and paternally derived chromo-
3. Wakayama T, Perry AC, Zuccotti M et al some regions in mice. Nature 315:496–498
(1998) Full-term development of mice from 7. McGrath J, Solter D (1984) Completion of
enucleated oocytes injected with cumulus cell mouse embryogenesis requires both the mater-
nuclei. Nature 394:369–374 nal and paternal genomes. Cell 37:179–183
4. Davidson RI (1974) Gene expression in 8. Barton SC, Surani MA, Norris ML (1984)
somatic cell hybrids. Annu Rev Genet 8: Role of paternal and maternal genomes in
195–218 mouse development. Nature 311:374–376
18 Y. Cheng et al.
9. Surani MA, Barton SC, Norris ML (1984) 23. Santos F, Peters AH, Otte AP et al (2005)
Development of reconstituted mouse eggs Dynamic chromatin modifications characterise
suggests imprinting of the genome during the first cell cycle in mouse embryos. Dev Biol
gametogenesis. Nature 308:548–550 280:225–236
10. McGrath J, Solter D (1983) Nuclear trans- 24. Santos F, Hendrich B, Reik W et al (2002)
plantation in mouse embryos. J Exp Zool 228: Dynamic reprogramming of DNA methylation
355–362 in the early mouse embryo. Dev Biol 241:
11. McGrath J, Solter D (1983) Nuclear trans- 172–182
plantation in the mouse embryo by microsur- 25. Park JS, Jeong YS, Shin ST et al (2007) Dynamic
gery and cell fusion. Science 220:1300–1302 DNA methylation reprogramming: active
12. Latham KE, Doherty AS, Scott CD et al demethylation and immediate remethylation
(1994) Igf2r and Igf2 gene expression in in the male pronucleus of bovine zygotes. Dev
androgenetic, gynogenetic, and parthenoge- Dyn 236:2523–2533
netic preimplantation mouse embryos: absence 26. van der Heijden GW, Dieker JW, Derijck AA
of regulation by genomic imprinting. Genes et al (2005) Asymmetry in histone H3 variants
Dev 8:290–299 and lysine methylation between paternal and
13. Rossant J, Guillemot F, Tanaka M et al (1998) maternal chromatin of the early mouse zygote.
Mash2 is expressed in oogenesis and preim- Mech Dev 122:1008–1022
plantation development but is not required for 27. McLay DW, Clarke HJ (2003) Remodelling
blastocyst formation. Mech Dev 73:183–191 the paternal chromatin at fertilization in mam-
14. Borgel J, Guibert S, Li Y et al (2010) Targets mals. Reproduction 125:625–633
and dynamics of promoter DNA methylation 28. Lepikhov K, Walter J (2004) Differential
during early mouse development. Nat Genet dynamics of histone H3 methylation at posi-
42:1093–1100 tions K4 and K9 in the mouse zygote. BMC
15. Kim JM, Ogura A (2009) Changes in allele- Dev Biol 4:12
specific association of histone modifications at 29. Howell CY, Steptoe AL, Miller MW et al
the imprinting control regions during mouse (1998) cis-Acting signal for inheritance of
preimplantation development. Genesis 47: imprinted DNA methylation patterns in the
611–616 preimplantation mouse embryo. Mol Cell Biol
16. Kono T (2006) Genomic imprinting is a bar- 18:4149–4156
rier to parthenogenesis in mammals. Cytogenet 30. Latham KE (1994) Strain-specific differences
Genome Res 113:31–35 in mouse oocytes and their contributions to
17. Wu Q, Kumagai T, Kawahara M et al (2006) epigenetic inheritance. Development 120:
Regulated expression of two sets of paternally 3419–3426
imprinted genes is necessary for mouse parthe- 31. Latham KE, Sapienza C (1998) Localization
nogenetic development to term. Reproduction of genes encoding egg modifiers of paternal
131:481–488 genome function to mouse chromosomes one
18. Hiura H, Obata Y, Komiyama J et al (2006) and two. Development 125:929–935
Oocyte growth-dependent progression of 32. Pickard B, Dean W, Engemann S et al (2001)
maternal imprinting in mice. Genes Cells Epigenetic targeting in the mouse zygote
11:353–361 marks DNA for later methylation: a mecha-
19. Kimura Y, Tateno H, Handel MA et al (1998) nism for maternal effects in development.
Factors affecting meiotic and developmental Mech Dev 103:35–47
competence of primary spermatocyte nuclei 33. Liang CG, Han Z, Cheng Y et al (2009) Effects
injected into mouse oocytes. Biol Reprod of ooplasm transfer on paternal genome func-
59:871–877 tion in mice. Hum Reprod 24:2718–2728
20. Ogura A, Yanagimachi R (1995) Spermatids as 34. Latham KE, Rambhatla L (1995) Expression
male gametes. Reprod Fertil Dev 7:155–158, of X-linked genes in androgenetic, gynoge-
discussion 158–159 netic, and normal mouse preimplantation
21. Latham KE, Solter D (1991) Effect of egg embryos. Dev Genet 17:212–222
composition on the developmental capacity of 35. Latham KE, Akutsu H, Patel B et al (2002)
androgenetic mouse embryos. Development Comparison of gene expression during preim-
113:561–568 plantation development between diploid and
22. Yeo S, Lee KK, Han YM et al (2005) haploid mouse embryos. Biol Reprod 67:
Methylation changes of lysine 9 of histone H3 386–392
during preimplantation mouse development. 36. Han Z, Chung YG, Gao S et al (2005) Maternal
Mol Cells 20:423–428 factors controlling blastomere fragmentation
in early mouse embryos. Biol Reprod 72: 43. Chung YG, Gao S, Latham KE (2006)
612–618 Optimization of procedures for cloning by
37. Kaneko-Ishino T, Kuroiwa Y, Miyoshi N et al somatic cell nuclear transfer in mice. Methods
(1995) Peg1/Mest imprinted gene on chro- Mol Biol 348:111–124
mosome 6 identified by cDNA subtraction 44. Latham KE, Westhusin ME (2000) Nuclear
hybridization. Nat Genet 11:52–59 transplantation and cloning in mammals.
38. Ozil JP, Banrezes B, Toth S et al (2006) Ca2+ Methods Mol Biol 136:405–425
oscillatory pattern in fertilized mouse eggs 45. Kimura Y, Yanagimachi R (1995) Intra-
affects gene expression and development to cytoplasmic sperm injection in the mouse. Biol
term. Dev Biol 300:534–544 Reprod 52:709–720
39. Kuretake S, Kimura Y, Hoshi K et al (1996) 46. Ogura A, Ogonuki N, Miki H et al (2005)
Fertilization and development of mouse Microinsemination and nuclear transfer using
oocytes injected with isolated sperm heads. male germ cells. Int Rev Cytol 246:189–229
Biol Reprod 55:789–795 47. Yanagimachi R (2005) Intracytoplasmic injec-
40. Lawitts JA, Biggers JD (1991) Optimization tion of spermatozoa and spermatogenic cells:
of mouse embryo culture media using simplex its biology and applications in humans and ani-
methods. J Reprod Fertil 91:543–556 mals. Reprod Biomed Online 10:247–288
41. Summers MC, Biggers JD (2003) Chemically 48. Miki H, Hirose M, Ogonuki N et al (2009)
defined media and the culture of mammalian Efficient production of androgenetic embryos
preimplantation embryos: historical perspec- by round spermatid injection. Genesis
tive and current issues. Hum Reprod Update 47:155–160
9:557–582 49. Kishigami S, Wakayama S, Nguyen VT et al
42. Chatot CL, Ziomek CA, Bavister BD et al (2004) Similar time restriction for intracytoplas-
(1989) An improved culture medium supports mic sperm injection and round spermatid injec-
development of random-bred 1-cell mouse tion into activated oocytes for efficient offspring
embryos in vitro. J Reprod Fertil 86:679–688 production. Biol Reprod 70:1863–1869
Chapter 2
Derivation of Induced Pluripotent Stem Cells by Retroviral

Gene Transduction in Mammalian Species
Masanori Imamura, Hironobu Okuno, Ikuo Tomioka, Yoshimi Kawamura,
Zachary Yu-Ching Lin, Ryusuke Nakajima, Wado Akamatsu, Hirotaka
James Okano, Yumi Matsuzaki, Erika Sasaki, and Hideyuki Okano
Abstract
Pluripotent stem cells can provide us with an enormous cell source for in vitro model systems for development.
In 2006, new methodology was designed to generate pluripotent stem cells directly from somatic cells, and
these cells were named induced pluripotent stem cells (iPSCs). This method consists of technically simple
procedures: donor cell preparation, gene transduction, and isolation of embryonic stem cell-like colonies.
The iPSC technology enables cell biologists not only to obtain pluripotent stem cells easily but also to
study the reprogramming events themselves. Here, we describe the protocols to generate iPSCs from
somatic origins by using conventional viral vectors. Specifically, we state the usage of three mammalian
species: mouse, common marmoset, and human. As mouse iPSC donors, fibroblasts are easily prepared,
while mesenchymal stem cells are expected to give rise to highly reprogrammed iPSCs efficiently. Common
marmoset (Callithrix jacchus), a nonhuman primate, represents an alternative model to the usual labora-
tory animals. Finally, patient-specific human iPSCs give us an opportunity to examine the pathology and
mechanisms of dysregulated genomic imprinting. The iPSC technology will serve as a valuable method for
studying genomic imprinting, and conversely, the insights from these studies will offer valuable criteria to
assess the potential of iPSCs.
Key words: Genomic imprinting, Induced pluripotent stem cells, Embryonic stem cells, Repro-
gramming, Pluripotency, Epigenetics, Germ cells, Cell culture, Common marmoset, Disease model
1. Introduction
Induced pluripotent stem cells (iPSCs) can be generated by trans-

duction of various sets of defined factors into somatic cells (1, 2).
Molecular and cellular properties of iPSCs are quite similar to those
of embryonic stem cells (ESCs), and they have pluripotency in vivo
and in vitro. Among pluripotent stem cells, there is a great
advantage in utilizing iPSCs: facile derivation from individuals.
21
22 M. Imamura et al.
Because of this, iPSC technology holds an enormous utility as a

cell source for studying many genetic mutants, including those
involved in human disease. In humans, dysregulation of imprinted
genes is correlated with tumorigenesis or various disorders such as
Beckwith–Wiedemann syndrome, Prader–Willi syndrome, and
Angelman syndrome. Therefore, generation of patient-specific iPSCs
could bring us better understanding of the pathology of genomic
imprinting-related disorders (3). Moreover, iPSCs have become
valuable cell sources for parthenogenesis, which is a unique model
for studying genomic imprinting. It has been shown that parthe-
nogenetic blastocysts and ESCs exhibit partial loss of imprinting
(4). Recently, bimaternal parthenogenetic iPSCs were established
from mouse neural stem cells (5). The iPSCs have lost the parthe-
nogenetic imprinting pattern despite their origin, suggesting an
attenuation of parthenogenetic imprinting through reprogram-
ming process. Given that parthenogenetic cells generally exhibit
growth defects, the iPSC technology would also provide ideal
materials and methods to dissect this phenomenon as well.
Epigenetic regulation is intimately tied to the input and output
of artificial reprogramming by somatic cell nuclear transfer (SCNT),
cell fusion, and iPSC technology (6). For example, SCNT some-
times results in abnormal embryogenesis correlated with dysregu-
lated genomic imprinting (7). Since iPSCs with somatic origin do
not receive any germline-derived factors during reprogramming, it
is important to examine the genomic imprinting pattern for quality
assessment of iPSCs. Indeed, recent studies have revealed an impor-
tant role of genomic imprinting in developmental potential of
iPSCs. Gene expression profiling found aberrant silencing of the
Dlk1-Dio3 imprinted locus in mouse iPSCs, although overall gene
expression was indistinguishable with that of ESCs (8, 9). The acti-
vation of the Dlk1-Dio3 imprinted locus is positively correlated
with a fully reprogrammed status, and notably, the developmental
potency in partially reprogrammed iPSCs can be rescued by reacti-
vation of this locus (8). Similarly, human iPSCs also exhibit aberra-
tions in imprinted genes such as H19 and PEG3 in their allele-specific
expression pattern, expression intensity, and DNA methylation
status (10). Thus, the proper genomic imprint could serve as a vital
marker to identify fully reprogrammed and clinically applicable
iPSCs lines.
Obviously, the advantage of pluripotent stem cells in life sciences
is their pluripotency and as an in vitro system to elucidate the
mechanisms of development and differentiation. Since the pioneer-
ing work on germ cell production from ESCs in culture (11–13),
it has been revealed that this potential is commonly observed
among pluripotent stem cell lines (14–16). Furthermore, this
propensity is valid for iPSCs as well; presumptive germ cells can be
induced by in vitro differentiation of mouse and human iPSCs
2 Derivation and Culture of Induced Pluripotent Stem Cells 23
(17–19). In these studies, genomic imprinting is a standard subject

of analyses because germ cells undergo a dynamic alteration of
genomic imprinting status in a developmental phase-specific manner.
Hence, the imprinting status is one of the landmarks used to define
iPSC-derived cells as “germ cells.”
Here, we describe the protocols to obtain iPSCs from somatic
cells derived from three mammalian species: mouse, common mar-
moset (Callithrix jacchus), and human. Because, to date, iPSCs
have been successfully established in various animals using basically
similar protocols (1, 20–25), the current methods could be appli-
cable for other mammals of interest. We believe that this informa-
tion will help accelerate elucidation of genomic imprinting with
molecular and cellular biological approaches.
2. Materials
2.1. Mouse iPSCs 1. Tissue culture plates and dishes: 100-mm, 6-, 24-, and 96-well
from Fibroblasts (BD Falcon).
2. Conical tubes: 15- and 50-ml (BD Falcon).
2.1.1. General Equipment
3. Plastic disposable pipettes: 1-, 5-, 10-, and 25-ml (BD Falcon).
4. 0.22-μm Bottle-top filter (Techno Plastic Products,
Trasadingen, Switzerland).
5. 0.22-μm Pore size filter (Millipore, Billerica, MA, USA).
6. 10-ml Disposable syringe (Terumo, Tokyo, Japan).
7. Cell-freezing container (Nalgene, Rochester, NY, USA).
8. Cryovial (Nunc, Waltham, MA, USA).
2.1.2. Cell Culture 1. PBS.

2. 0.25% (w/v) Trypsin/EDTA solution (Invitrogen).
3. Recovery Cell Culture Freezing Medium (Invitrogen).
4. mDMEM/10% FBS: DMEM containing 4.5 g/l glucose
(Nacalai Tesque, Kyoto, Japan) supplemented with 10% (v/v)
Fetal bovine serum (FBS; Invitrogen, Carlsbad, CA, USA),
50 U/ml penicillin and 50 mg/ml streptomycin (Invitrogen).
Filter with a 0.22-μm bottle-top filter and store at 4°C up to
a week.
5. mESC medium: DMEM supplemented with 15% (v/v) FBS,
2 mM L-Glutamine (Invitrogen), 0.1 mM non-essential amino
acids (Invitrogen), 0.1 mM 2-mercaptoethanol (Invitrogen)
(see Note 1), 50 U/ml penicillin and 50 mg/ml streptomycin,
and 1,000 U/ml ESGRO (Millipore). Filter with a bottle-top
filter and store at 4°C up to a week.
6. Gelatin-coated culture dishes: To prepare 10× stock solution,

dissolve 1 g of gelatin powder (Sigma, St. Louis, MO, USA) in
100 ml of distilled water, autoclave, and store at 4°C for
2 months. To prepare 1× gelatin solution, warm the 10× gela-
tin stock to 37°C, add 50 ml of the stock to 450 ml of distilled
water. Filter the solution with a bottle-top filter and store at
4°C up to 2 weeks. Add 0.1% (w/v) gelatin solution to cover
the entire area of culture dishes. Incubate for at least 30 min at
37°C. Aspirate the solution immediately before plating cells.
2.1.3. Preparation 1. Sterilized forceps and scissors.

of Fibroblasts from
Mouse Embryos and
Adult Mouse Tail
2.1.4. Retrovirus 1. pMXs vectors containing the cDNAs of Oct4 (Plasmid 13366),
Production Sox2 (Plasmid 13367), Klf4 (Plasmid 13370), c-Myc (Plasmid
13375), and DsRed (Plasmid 22724) (Addgene, Cambridge,
MA, USA).
2. Plat-E packaging cells (Available from Dr. Toshio Kitamura at
the University of Tokyo; kitamura@ims.u-tokyo.ac.jp).
3. Puromycin: Dissolve puromycin powder (Sigma) in distilled
water at 10 mg/ml concentration, and filter it through a 0.22-
μm filter (Millipore). Aliquot and store at −20°C.
4. Blastocidin S: Dissolve blastocidin S hydrochloride (Funakoshi,
Tokyo, Japan) in distilled water at 10 mg/ml concentration, and
filter it through a 0.22-μm filter. Aliquot and store at −20°C.
5. 0.05% Trypsin/EDTA: Mix 10 ml of 0.25% (w/v) Trypsin/
EDTA solution (Invitrogen) and 40 ml of PBS. Store at
−20°C.
6. Opti-MEM I Reduced-Serum Medium (Invitrogen).
7. FuGENE 6 transfection reagent (Promega, Madison, WI).
8. 0.45-μm cellulose acetate filter (Schleicher & Schuell, Keene,
NH, USA).
9. Polybrene solution: To prepare the stock solution at 8 mg/ml
concentration, dissolve 80 mg of polybrene (Nacalai Tesque)
in 10 ml of distilled water and filter it through a 0.22-μm filter.
Store at 4°C.
2.1.5. iPSCs’ Derivation 1. SNL medium: DMEM supplemented with 7% (v/v) FBS,
from Mouse Fibroblasts 2 mM L-Glutamine, 50 U/ml penicillin and 50 mg/ml
streptomycin. Filter with a bottle-top filter and store at 4°C up
to a week.
2. SNL feeder cells: SNL cells (SNL 76/7; DS Pharma Biomedical,
Osaka, Japan) are a derivative of STO cells, which express
neomycin-resistant gene and leukemia inhibitory factor (LIF).

Cultivate the cells with SNL medium on gelatin-coated culture
dishes. At 80–90% confluency in 100-mm culture dishes, add
0.3 ml of 0.4 mg/ml mitomycin C (Kyowa Hakko Kirin,
Tokyo, Japan) solution to the culture medium and incubate at
37°C for 2.5 h (see Note 2). After washing with 10 ml of PBS
twice, trypsinize, and count the cell number. Resuspend with
SNL medium and seed the cells on gelatin-coated culture
dishes at 1 × 106 cells per 100-mm culture dish. Use within
2 weeks.
2.2. Mouse iPSCs from 1. Tissue culture plates and dishes: 100- and 60-mm dish (BD
Mesenchymal Stem Falcon).
Cells 2. Conical tubes: 15- and 50-ml (BD Falcon).
2.2.1. General Equipment 3. Plastic disposable pipettes: 1-, 5-, 10-, and 25-ml.
4. 0.22-μm bottle-top filter (Techno Plastic Products).
2.2.2. Animals 1. NanogGFP-IRES-Puro mice (available from RIKEN BioResource

Center, Tsukuba, Japan).

2. 2× PBS containing 4% FBS: 10× PBS is diluted five times with
sterile water and supplemented with 4% (v/v) FBS.
3. HBSS+: HBSS (Nacalai Tesque) supplemented with 2% (v/v)
FBS, 10 mM HEPES, and 50 U/ml penicillin and 50 mg/ml
streptomycin (Invitrogen).
4. Mesenchymal stem cell (MSC) medium: MEM
Alpha + GlutaMAX-I (GIBCO) supplemented with 10% (v/v)
FBS, 10 mM HEPES, and 50 U/ml penicillin and 50 mg/ml
streptomycin.
5. mESC medium: see Subheading 2.1.2, item 5.
2.2.4. Preparation of Bone 1. Enzymatic dissociation solution: 0.2% (w/v) collagenase

Marrow Cell Suspension (Wako, Osaka, Japan) in DMEM containing 1.0 g/l glucose
supplemented with 10 mM HEPES and 50 U/ml penicillin
and 50 mg/ml streptomycin.
2. Cell strainer: 70-μm pore size (BD Falcon).
3. Sterile water.
2.2.5. Purification of MSCs 1. Fluorescently conjugated antibodies (eBioscience, San Diego,

CA, USA): PE-conjugated CD45 (Clone: 30-F11, 12-0451),
TER119 (Clone: TER-119, 12-5921), APC-conjugated
PDGFRα (Clone: APA5, 17-1401), and FITC-conjugated
Sca-1 (Ly6A/E, Clone: D7, 11-5981).
2. Fluorescently conjugated isotype controls (eBioscience): Rat

IgG2b K Isotype Control PE (12-4031), Rat IgG2a K Isotype
Control APC (17-4321), and Rat IgG2a K Isotype Control
FITC (11-4321).
3. Propidium iodide solution (Sigma).
4. Triplelaser cell sorter such as MoFlo (Dako) and JSAN (Bay
Bioscience).
2.2.6. Retrovirus 1. pMXs vectors: see Subheading 2.1.4, item 1.

Production 2. Plat-E packaging cells (available from Dr. Toshio Kitamura at
3. mDMEM/10% FBS: see Subheading 2.1.2, item 4.
4. 0.05% Trypsin/EDTA: see Subheading 2.1.4, item 5.
5. FuGENE 6 transfection reagent (Promega).
6. 0.45-μm Cellulose acetate filter (Schleicher & Schuell).
7. Polybrene solution: see Subheading 2.1.4, item 9.
2.2.7. Induction of iPSCs 1. SNL medium: see Subheading 2.1.5, item 1.

from PαS Cells 2. SNL feeder cells: see Subheading 2.1.5, item 2.
3. Puromycin: see Subheading 2.1.4, item 3.
2.3. Marmsoet iPSCs 1. Tissue culture plates and dishes: 100-mm (Greiner bio-one,
from Fetal Liver Cells Frickenhausen, Germany) and 96-well (Iwaki, Tokyo, Japan).
2.3.1. General Equipment 2. Gelatin-coated culture dishes: 100-mm and 12-well (Iwaki).
Required Through 3. Conical tubes: 15- and 50-ml (BD Falcon).
Experiments 4. Plastic disposable pipettes: 1-, 5-, 10- (BD Falcon, 357551),
and 25-ml (BD Falcon).
2.3.2. Cell Culture 1. cjDMEM/10% FBS: Dulbecco’s modified Eagle’s medium

(Wako) supplemented with 10% (v/v) FBS (Biowest) and 1%
(v/v) antibiotic–antimycotic solution (Invitrogen).
2. cjESC medium: Knockout DMEM (Gibco) supplemented
with 10% (v/v) Knockout Serum Replacement (Invitrogen),
1 mM L-glutamine (Invitrogen), 0.1 mM MEM nonessential
amino acids (Invitrogen), 0.1 mM 2-mercaptoethanol (Sigma),
and 1% (v/v) antibiotic–antimycotic solution (Gibco).
3. Trypsin solution for ESCs: 0.25% (v/v) Difco trypsin 250
(BD, Baltimore, MD, USA), 1 mM CaCl2, and 20% (v/v)
KSR.
4. Hank’s buffered salt solution without calcium or magnesium
(Gibco).
2.3.3. Virus Production 1. pMXs retroviral vectors carrying human OCT4 (Addgene,
Plasmid 17217), SOX2 (Addgene, Plasmid 17218), KLF4
(Addgene, Plasmid 17219), C-MYC (Addgene, Plasmid
17220), NANOG (kindly provided by Dr. Yamanaka), LIN28
(kindly provided by Dr. Yamanaka), and GFP (kindly provided
by Dr. Yamanaka) (see Note 3).
2. pVSV-G vector and GP-2 cells (Retroviral Gene Transfer and
Expression; TaKaRa, Shiga, Japan).
3. Opti-MEM I Reduced-Serum Medium (Invitrogen).
4. FuGENE 6 transfection reagent.
5. 0.45-μm pore-size cellulose acetate filter (Sartorius, Goettingen,
Germany).
6. Poly-L-lysine (Sigma).
2.3.4. Preparation 1. Sterilized forceps and scissors.

of Fetal Liver Cells 2. Collagenase solution: Dissolve Collagenase type I in DMEM
at 0.5% (w/v) concentration.
2.3.5. Retroviral Infection 1. Polybrene (Nacalai Tesque).

of Marmoset Cells 2. Mitomycin C-treated or irradiated MEF feeder cell plates.
2.3.6. Passage of iPSCs 1. Cell strainer, 100-μm nylon (BD Falcon).

2. Mitomycin C-treated or irradiated MEF feeder cell plates.
2.3.7. Storage 1. Cell Banker 2 (ZENOAQ, Koriyama, Fukushima, Japan).

of Established iPSCs 2. 2-ml plastic cryogenic vial (Iwaki).
2.4. Human iPSCs 1. Tissue culture plates and dishes: 100-mm (FPI, Kobe, Japan),
from Fibroblasts 6-, 24-, and 96-well (Nunc).
2.4.1. General Equipment 2. Conical tubes: 15- and 50-ml (Greiner).
3. Plastic disposable pipettes: 2-, 5-, 10-, 25-, and 50-ml (Greiner).
4. 0.22-μm bottle-top filter (Techno Plastic Products).
5. 0.22-μm pore size filter (Millipore).
6. 10-ml disposable syringe (Terumo).
7. Cryovial (Nunc).

2. 2.5% Trypsin.
3. 0.25% Trypsin/EDTA solution and 0.05% Trypsin/EDTA
solution.
4. Water (Sigma).
5. Gelatin-coated culture dishes: see Subheading 2.1.2, item 6.
6. mDMEM/10% FBS: see Subheading 2.1.2, item 4.
2.4.3. Preparation 1. Dermapunch (Maruho, Osaka, Japan).

and Culture of Human 2. Sterilized forceps and scissors.
Dermal Fibroblasts
3. Cell Banker 2 (ZENOAQ).
2.4.4. Lentivirus 1. pLenti6/UbC vector containing mouse Slc7a1 gene (Plasmid

Production 17224, Addgene).
2. 293FT cells (Invitrogen).
3. CalPhos Mammalian Transfection kit (TaKaRa).
4. Virapower Lentiviral expression system (Invitrogen).
5. Solution A: Dilute 3 μg of Virepower packaging mix (pLP1,
pLP2, and pLP/VSVG mixture) and 1 μg of pLenti6/UbC/
mSlc7a1 in 12.4 μl of 2 M Calcium Solution, and add up to
100 μl with sterile water.
6. Solution B: Transfer 100 μl of 2× HBS into a 60-mm dish.
7. 0.45-μm pore size cellulose acetate filter.
8. Blastocidin S hydrochloride: see Subheading 2.1.4, item 4.
2.4.5. Retrovirus 1. pMXs retrovial vectors containing the cDNAs of human OCT4
Production (Plasmid 17217), human SOX2 (Plasmid 17218), human
KLF4 (Plasmid 17219), and human C-MYC (Plasmid 17220)
(Cell biolabs, Inc., San Diego, CA, USA; http://www.cellbio-
labs.com/).
2. Plat-E packaging cells (available from Dr. Toshio Kitamura at
3. OPTI-MEM I.
4. FuGENE 6 transfection reagent.
5. Polybrene solution: see Subheading 2.1.4, item 9.
6. Puromycin: see Subheading 2.1.4, item 3.
7. Blastocidin S hydrochloride: see Subheading 2.1.4, item 4.
2.4.6. Induction of iPSCs 1. SNL medium: see Subheading 2.1.5, item 1.

from Human Fibroblasts 2. SNL cells: see Subheading 2.1.5, item 2.
3. 0.4 mg/ml mitomycin C: Dissolve 10 mg of mitomycin C in
25 ml of water. Filter through a 0.22-μm pore size filter, ali-
quot, and store at −20°C.
4. SNL feeder cells: Incubate SNL cells at 80–90% confluency with
12 μg/ml mitomycin C for 2 h and 15 min in 37°C, 5% CO2
incubator. Wash the cells with 4.5 ml of PBS twice, trypsinize,
and count the cell number. Plate the cells at 2.6 × 104 per cm2
onto gelatin-coated culture dishes; 1.5 × 106 cells/dish (100-
mm culture dish), 2.5 × 105 cells/well (6-well culture plate), and
5.2 × 104 cells/well (24-well plate) (see Note 4).
5. hESC medium: DMEM/F12 containing 20% KSR, 2 mM

L-glutamine, 0.1 mM non-essential amino acids (Sigma),
0.1 mM 2-mercaptoethanol (Sigma), and 50 U and 50 mg/ml
penicillin and streptomycin. Filter through a 0.22-μm bottle-
top filter and store at 4°C up to 2 weeks.
6. 50 μg/ml FGF-2: Dissolve 1 mg of FGF-2 (PeproTech, Rocky
Hill, NJ, USA) in 20 ml of hESC medium. Aliquot and store
at −20°C.
2.4.7. Picking, Expanding, 1. 10 mg/ml Collagenase IV: Dissolve 1 g of collagenase IV

Freezing, and Thawing (Invitrogen) in 100 ml of water, and filter through a 0.22-μm
Human iPSCs pore size filter. Aliquot and store at −20°C.
2. 0.1 M CaCl2: Dissolve 555 mg of CaCl2 in 50 ml of water, and
filter through a 0.22-μm pore size filter. Store at 4°C.
3. CTK solution: Mix 10 ml of 2.5% Trypsin, 10 ml of 10 mg/ml
collagenase IV, 1 ml of 100 mM CaCl2, and 20 ml of KSR with
59 ml of PBS (26). Aliquot and store at −20°C. Avoid repeated
freezing and thawing. Store at 4°C up to 1 week.
4. Cell scraper (Iwaki).
5. 10 mM Y-27632: Dissolve 5 mg of Y-27632 (Wako) in 1.48 ml
of sterile water. Aliquot and store at −20°C.
6. DAP213 solution: Mix 1.42 ml of DMSO (Sigma), 0.59 g of
acetamide (Sigma), and 2.2 ml of propylene glycol (Sigma)
with 6 ml of hESC medium. Filter through a 0.22-μm pore
size filter and store at −80°C.
3. Methods
3.1. Mouse iPSCs In most experiments of iPSC generation, reprogrammed cells have
from Fibroblasts been selected based on the expression of fluorescence protein or
drug-resistance genes driven by the promoter of pluripotency-
related genes such as Nanog and Oct4. Although this helps to select
highly reprogrammed cells, it is not always necessary to take advan-
tage of the system. For a wider usage of the iPSC technique, in this
part, we described the mouse iPSCs’ generation from embryonic
and adult fibroblasts without reporter-dependent selection.
3.1.1. Preparation 1. Euthanize female mice on the day 13.5 of pregnancy by cervi-
of Fibroblasts from cal dislocation (see Note 5). Wipe with 70% ethanol, and iso-
Mouse Embryos late uteri using sterilized forceps and scissors into 100-mm
culture dishes containing PBS. Separate the embryos from
their placenta and wash them with PBS twice. Remove the
embryo’s head, visceral tissues, and gonads.
2. Transfer the remaining bodies to a new 100-mm culture dish

containing PBS and mince them into small pieces. Transfer
into a 50-ml conical tube containing 0.25% Trypsin/EDTA
solution (3 ml per embryo) and incubate at 37°C for 20 min.
Then, add an additional 0.25% Trypsin/EDTA solution (3 ml
per embryo) and incubate at 37°C for 20 min. Invert the tube
gently several times and add an equal amount of mDMEM/10%
FBS (6 ml per embryo). Pipette up and down to dissociate the
tissues.
3. Centrifuge at 200 g for 5 min, discard the supernatant, and
resuspend the pellet in mDMEM/10% FBS. Count the cell
number and plate 1 × 107 cells per 100-mm gelatin-coated culture
dish containing 10 ml of mDMEM/10% FBS. Incubate at
37°C with 5% CO2 overnight (passage 1), and the next day
replace the medium to remove floating cells.
4. When the cells grow to confluency, split or freeze them at 1:4
dilution. Remove the medium, wash once with PBS, and
trypsinize with 1 ml of 0.25% Trypsin/EDTA at 37°C for
5 min. Then, add 9 ml of mDMEM/10% FBS and resuspend
by pipetting. For passage, split the cells to new 100-mm gelatin-
coated culture dishes at 1:4 dilution (passage 2) (see Note 6).
5. To prepare the freeze stocks, transfer the cell suspension to 15-ml
conical tubes and centrifuge at 200 g for 5 min. Discard the
supernatant and resuspend the cells with Recovery Cell Culture
Freezing Medium. Aliquot 1 ml of the cell suspension per freez-
ing vial. Keep the vials in a cell-freezing container at −80°C over-
night and then transfer them into a liquid nitrogen tank.
3.1.2. Preparation 1. Cut the tail from an adult mouse and wash with PBS (see
of Fibroblasts from Note 5). Incise using sterilized scissors, peel superficial dermis
Adult Mouse Tail by hand, and mince the remaining tail into 1-cm pieces with
scissors. Place two pieces per well of 6-well gelatin-coated
plates, add 2 ml of mDMEM/10% FBS, and incubate at 37°C
with 5% CO2 for 5 days.
2. Remove the tissues of tails and replace the medium with 2 ml of
fresh mDMEM/10% FBS. When they reach confluency, aspi-
rate the medium, wash twice with 2 ml of PBS, add 0.3 ml of
0.25% Trypsin/EDTA, and incubate at 37°C for 10 min. Add
2 ml of mDMEM/10% FBS, suspend the cells, and transfer to
a 15-ml conical tube. Centrifuge the cells at 200 g for 5 min.
3. Discard the supernatant, resuspend the cells with 10 ml of
mDMEM/10% FBS, and plate to a 100-mm gelatin-coated
culture dish (passage 2). When the cells become confluent,
trypsinize with 1 ml of 0.25% Trypsin/EDTA at 37°C for
5 min, and resuspend with 9 ml of mDMEM/10% FBS.
Passage to new 100-mm gelatin-coated culture dishes at 1:4

dilution (passage 3). These cells usually become confluent
within 3–4 days (see Note 6).
3.1.3. Retrovirus Production 1. Thaw a vial of Plat-E cells in 37°C water bath. Resuspend the
cells with 10 ml of mDMEM/10% FBS and transfer to a 100-
mm gelatin-coated culture dish. Incubate the cells in 37°C, 5%
CO2 incubator. From the next day onwards, cultivate the cells
in 10 ml of mDMEM/10% FBS supplemented with 1 μg/ml
puromycin and 10 μg/ml blastocidin S. Split the cells at 1:5
dilution when they reach confluency.
2. Twenty-four hours before transfection, aspirate the medium,
gently wash with PBS once, and add 1 ml of 0.05% Trypsin/
EDTA. After incubation at room temperature for 5 min, sus-
pend with 10 ml of mDMEM/10% FBS, and transfer to a
50-ml conical tube. Count the cell number and plate the cells
in mDMEM/10% FBS at 3.6 × 106 cells per 100-mm culture
dish, 1.5 × 106 cells per 60-mm culture dish, or 6 × 105 cells per
well of a 6-well culture plate. For the four iPSC factors to be
transduced, prepare five culture dishes to transfect the five plas-
mids pMXs-Oct4, Sox2, Klf4, c-Myc, and DsRed separately.
3. Transfer 0.3 ml of Opti-MEM I Reduced-Serum Medium to
1.5-ml plastic tubes. Add 27 μl of FuGENE 6 transfection
reagent, mix gently by tapping, and incubate at room tempera-
ture for 5 min. Then, add 9 μg of pMXs plasmid DNA, mix
gently by finger tapping, and incubate at room temperature for
15 min (see Note 7).
4. Add the DNA/FuGENE 6 mixture to the Plat-E cell culture
dishes dropwise and incubate at 37°C, 5% CO2 overnight.
Replace the medium with 10 ml of fresh mDMEM/10% FBS
and further incubate overnight.
5. Collect the supernatants from the Plat-E cell culture dishes
and filter them through a 0.45-μm cellulose acetate filter
(Fig. 1). Combine an equal volume of the virus supernatants
containing each factor. For the transduction of four iPSC fac-
tors, mix the supernatants of Oct4, Sox2, Klf4, c-Myc, and
DsRed at 1:1:1:1:4 ratio. For three iPSC factors without c-Myc,
mix the supernatants of Oct4, Sox2, Klf4, and DsRed at 1:1:1:3.
Add polybrene solution to the virus supernatant mixture at the
final concentration of 4 μg/ml and mix gently. Use immedi-
ately for transduction (see Note 8).
3.1.4. iPSCs’ Derivation 1. Twenty-four hours before retroviral gene transduction,

from Mouse Fibroblasts trypsinize MEF or TTF within passage 3, and plate 8 × 105 cells
per 100-mm gelatin-coated culture dishes. Prepare one extra
culture dish for transduction of DsRed in addition to those for
the iPSC factors. The next day, aspirate the medium and add
Fig. 1. Plat-E packaging cells after transfection of pMXs retrovirus plasmids. Phase and fluorescence images of Plat-E cells
just before collection of virus supernatants (Oct4 and DsRed). The Plat-E cells with pMXs-DsRed transfection show high
Red fluorescence when the virus is properly produced.
Fig. 2. Fibroblasts with successful gene transduction. Retroviral gene transduction can be monitored by red fluorescence
in mouse fibroblasts infected with the pMXs-DsRed retrovirus. The image was photographed after replating onto SNL
feeder cells.
the retrovirus supernatant mixture prepared at step 5 of

Subheading 3.1.3. Incubate the cells at 37°C, 5% CO2 over-
night and replace the medium with 10 ml of fresh mDMEM/10%
FBS. Two days later, exchange the medium with 10 ml of fresh
mDMEM/10% FBS again (see Note 9) (Fig. 2).
Fig. 3. Derivation of mouse fibroblast-derived iPSCs. (a) Morphology of mouse iPSC colony derived from fibroblasts (with
three iPSC factors: Oct4, Sox2, Klf4 ) just before picking. (b) Expansion culture of fibroblast-derived mouse iPSCs. The
image was photographed at passage 8 on gelatin-coated culture dish.
2. On the day 8 after transduction, trypsinize the cells with 0.25%

Trypsin/EDTA. Resuspend with 10 ml of mDMEM/10%
FBS and count the cell number. Replate them at a density of
0.5–5 × 104 cells (with four iPSC factors) or 3.5 × 105 cells (with
three iPSC factors) per 100-mm culture dishes with SNL feeder
cells (see Note 10). Incubate the cells at 37°C, 5% CO2 over-
night and replace the medium with 10 ml of mESC medium
the next day.
3. Exchange the medium with 10 ml of fresh mESC medium
every other day until the emerging iPSC colonies grow large
enough to be picked (Fig. 3a).
3.1.5. Picking and 1. Aliquot 20 μl of 0.25% Trypsin/EDTA per well of a 96-well

Expanding iPSC Colonies culture plate. Then, aspirate the mESC medium and wash the
cells with 10 ml of PBS once. Add 5 ml of PBS, pick colonies
with ESC-like morphology using a 10-μl pipette, and transfer
each colony to one well of the 96-well culture plate. When
picking iPSC colonies, choose the colonies without DsRed
fluorescence to select highly reprogrammed cells (see Note 11).
Incubate at 37°C for 15 min.
2. Add 180 μl of mESM medium to each well and pipette up and
down to dissociate the colony into single cells. Transfer each
cell suspension into one well of a 24-well plate with SNL feeder
cells and add 300 μl of mESC medium. Incubate at 37°C, 5%
CO2 until the cells grow to 50–60% confluency.
3. To expand the iPSCs, aspirate the medium, wash with PBS once,
and add 0.1 ml of 0.25% Trypsin/EDTA per well of a 24-well
culture plate. Incubate at 37°C for 10 min and add 0.4 ml of
mESC medium. Carefully pipette up and down to obtain a sin-
gle-cell suspension and transfer to a well of 6-well culture plates.
Add 1.5 ml of mESC medium and cultivate at 37°C, 5% CO2
until the cells reach 80–90% confluency (Fig. 3b).
3.1.6. Freezing 1. Aspirate the medium, wash with PBS once, and add 0.3 ml of
and Thawing iPSCs 0.25% Trypsin/EDTA per well of 6-well culture plates.
Incubate at 37°C for 10 min. Add 2 ml of mESC medium and
carefully pipette up and down to obtain single-cell suspension.
Transfer the cell suspension to a 15-ml conical tube.
2. Centrifuge the tube at 200 g for 5 min. Discard the superna-
tant and resuspend the cells with Recovery Cell Culture
Freezing Medium at 1–2 × 106 cells/ml. Aliquot 1 ml of the
cell suspension per freezing vial. Keep the vials in a cell-freez-
ing container at −80°C overnight and then transfer them into
a liquid nitrogen tank the next day.
3. To thaw the iPSC freeze stocks, warm the vials in 37°C water
bath until half of the ice crystals disappear. Transfer the cell
suspension into a 15-ml conical tube containing 9 ml of
mDMEM/10% FBS. Centrifuge at 200 g for 5 min and gently
resuspend the cells with 2 ml of mESC medium. Plate the cells
to a well of a 6-well culture plate with SNL feeder cells.
3.2. Mouse iPSCs MSCs are defined as plastic-adherent, fibroblast-like cells which
from Mesenchymal undergo sustained in vitro growth and can give rise to multiple
Stem Cells mesenchymal lineages (bone, adipose and cartilage tissue, etc.).
We previously established a method for isolating highly enriched
MSCs from adult murine bone marrow based on their expression
of PDGFRα and Sca-1 (27). The iPSCs generated from purified
MSCs (PαS) by Oct4, Sox2, and Klf4 seem to be the closest
equivalent to ESCs by global gene profile and germline transmis-
sion, compared with those from PDGFRα−/Sca-1− osteoprogeni-
tors and tail-tip fibroblasts (28). These results suggest that tissue
stem cells could be a promising cell source for producing high-
quality iPSCs.
3.2.1. Preparation of Bone 1. Dissect femurs and tibias from adult mice (3–20 mice) and
Marrow Cell Suspension remove residual tissues from the bones. Wash with PBS three
times.
2. Put the bones on a mortar and crush them with a pestle (see
Note 12). Wash the crushed bones several times with HBSS+
to remove the hematopoietic cells.
3. Incubate the bone fragments in 20 ml of enzymatic dissocia-
tion solution in 50-ml conical tube for 1 h at 37°C with shak-
ing (110 rpm/min). Filter the suspension through a cell
strainer (70-μm pore size), and collect the cells by centrifuga-
tion at 280 g for 7 min at 4°C.
4. Discard the supernatant. Resuspend the pellet with 1 ml of
sterile water for 5–10 s to burst red blood cells, and add 1 ml
of 2× PBS containing 4% FBS. At this step, cell debris can be
seen. Then, resuspend the cells in 10 ml of HBSS+. To remove
the debris, filter the cell suspension through a cell strainer

(70-μm pore size).
5. Collect the cells by centrifugation at 280 g for 7 min at 4°C.
Discard the supernatant, and resuspend the cells in 1 ml of
HBSS+.
3.2.2. Purification of MSCs 1. To compensate the fluorescence interference, transfer approxi-

from Bone Marrow Cell mately 1 × 105 cells into five 1.5-ml plastic tubes as “control
Suspension tubes.” Add a single fluorescence-conjugated antibody (PE,
APC, or FITC) to the “control tubes” one by one. The final
concentration of antibodies is 0.8 (anti-CD45), 1 (anti-TER119),
1.4 (anti-PDGFRα), and 2.5 μg/ml (anti-Sca-1). Prepare one
extra tube for an unstained (negative) control. Avoid light expo-
sure and incubate the tubes at 4°C for 30 min.
2. Add all fluorescence-conjugated antibodies to the remaining
cell suspension in 50-ml conical tube as “sample tube.” Avoid
light exposure and incubate the tube at 4°C for 30 min.
3. Add 500 μl and 9 ml of HBSS+ to the “control tubes” and
“sample tube,” respectively. Centrifuge the “control tubes” at
800 g for 3 min at 4°C. For the “sample tube,” centrifuge at
280 g for 7 min at 4°C. Discard the supernatant, and add 300–
500 μl and 5–9 ml of HBSS+ containing 2 μg/ml PI to the
“control tubes” and the “sample tube,” respectively.
4. For sorting cells, use a triplelaser cell sorter such as MoFlo
(Dako) or JSAN (Bay Bioscience), following to the instrument
calibration and standardization by the protocols established in
your laboratory. Compensate each laser using the “control
tubes,” measure PI fluorescence, and define a live cell gate
excluding PI-positive cells. Define additional gates to isolate
the cells positive for PDGFRα and Sca-1 but negative for
CD45 and TER119, according to the isotype control
fluorescence intensity (see Note 13). Routinely, 0.1–2% of
bone marrow cells are CD45−/TER119−, and 10–20% of
CD45−/TER119− cells are PDGFRα+/Sca-1+ (Fig. 4).
5. Collect the sorted PDGFRα+/Sca-1+/CD45−/TER119− cells
by centrifugation at 280 g for 7 min at 4°C. Discard the super-
natant and resuspend the cells with 1 ml of MSC medium.
Repeat this washing step again.
6. Plate the sorted cells onto a 100-mm tissue culture dish con-
taining 10 ml of MSC medium. Exchange the medium every
3 days until the cells reach confluency (Fig. 5a).
3.2.3. Retrovirus Production 1. Seed Plat-E cells at 8 × 106 cells per 100-mm dish.
2. On the next day, introduce 9 μg of pMX-based retroviral vec-
tors for DsRed, Oct4, Sox2, Klf4, and c-Myc individually into
a 104 b 104
103 103
CD45/TER119
Sca-1
102 102
101 101
100 100
0 1000 2000 3000 4000 100 101 102 103 104
FSC PDGFRα
Fig. 4. Isolation of mouse PαS MSCs from adult bone marrow. (a) Cell sorter profile of CD45−/TER119− non-blood cells
in whole bone marrow cells. In this experiment, 1.01% of cells are CD45−/TER119−. (b) Cell sorter profile of PDGFRα+/
Sca-1+ MSCs in CD45−/TER119− cells. PDGFRα+/Sca-1+ cells were separated after gating on CD45− and TER119−. In
this experiment, 11.3% of cells are PDGFRα+/Sca-1+. In toto, 0.01–0.4% of cells are usually isolated as PαS MSCs from
bone marrow.
Fig. 5. Derivation of mouse PαS-derived iPSCs. (a) Morphology of PαS MSCs purified from adult bone marrow. (b) Phase
and fluorescence images of PαS-derived iPSC colonies. When using transgenic mice with NanogGFP-IRES-Puro, fully repro-
grammed iPSCs are visualized by GFP fluorescence driven by the promoter of pluripotency marker gene Nanog.
separate dishes of Plat-E cells using 27 μl of FuGENE 6

transfection reagent.
3. After 24 h, replace the medium with 10 ml of mDMEM/10%
FBS and incubate overnight. Collect the virus-containing
supernatants from the Plat-E cultures and filter them through
0.45-μm cellulose acetate filters.
4. Make a mixture of equal volumes of supernatants containing
four or three (without c-Myc) iPSC factors and DsRed retrovi-
ruses. DsRed is used as a marker for infection efficiency and
transgene silencing. Add polybrene solution into the filtered
virus-containing supernatants at the final concentration of
4 μg/ml. Use immediately for transduction.
3.2.4. Induction of iPSCs 1. Twenty-four hours before retroviral gene transduction,

from PαS Cells trypsinize PαS cells to seed at 1 × 104 cells per 60-mm culture
dish covered with SNL feeder cells.
2. Remove the medium from PαS cell culture dishes and add the
virus/polybrene-containing medium. Incubate the cells for
24 h and replace the medium with 5 ml of MSC medium.
3. Two days after infection, exchange the medium with 5 ml of
mESC medium. The next day, plate 1 × 104 DsRed+-infected
cells per 100-mm culture dish covered with SNL feeder cells.
4. To select highly reprogrammed cells by NanogGFP-IRES-Puro, add
puromycin to the culture medium at the final concentration of
1.5 μg/ml after 3 weeks (four iPSC factors) or 4 weeks (three
iPSC factors: OCT4, SOX2, and KLF4) since the infection.
Change the medium every day until iPSC colonies grow large
enough to be picked (Fig. 5b).
3.3. Marmoset iPSCs In this chapter, we describe a protocol to establish common mar-
from Fetal Liver Cells moset iPSCs from fetal liver cells via retrovirus-mediated introduction
of six human transcription factors, i.e., OCT4, SOX2, KLF4,
C-MYC, NANOG, and LIN28 (24). We found that LIN28, in
addition to Yamanaka’s four transcription factors, improved the
efficiency of iPSCs’ establishment in marmosets. The availability of
marmosets, and their ease of breeding, may provide an alternative
to the use of traditional Old World nonhuman primates. In the
future, common marmosets and their iPSCs could provide a pow-
erful preclinical model for the study of regenerative medicine and
possibly increase interest in the field.
3.3.1. Virus Production Retroviruses carrying the transcription factors were produced using
the Retroviral Gene Transfer and Expression System according to
the manufacturer’s instructions.
1. Seed GP-2 cells at 3 × 106 cells per 100-mm poly-L-lysine-
coated dish 1 day prior to transfection (see Note 14).
2. Mix 27 μl of FuGENE 6 transfection reagent with 0.3 ml of
OPTI-MEM I in a 1.5-ml tube, and incubate at room tem-
perature for 5 min.
3. Combine 6 μg of each pMX vector (carrying human OCT4,
SOX2, KLF4, C-MYC, NANOG, LIN28, and GFP) and 6 μg
of pVSV-G vector with the FuGENE 6 and OPTI-MEM I
mixture. Mix gently, and incubate at room temperature for
15 min.
4. Add the DNA/FuGENE 6 complex to the GP-2 cell dish
culture in 10 ml of OPTI-MEM I, and incubate at 37°C, 5%
CO2 overnight. The next day, replace the medium contain-
ing the DNA/FuGENE 6 complex with 10 ml of cjD-
MEM/10% FBS.
5. At 48 and 72 h post transfection, collect the medium as a

virus-containing supernatant, and filter with a 0.45-μm pore-
size cellulose acetate filter. Aliquot and store the virus-contain-
ing supernatant at −80°C (see Note 15).
3.3.2. Preparation Common marmoset fetal liver cells were isolated from a miscarried
of Fetal Liver Cells female fetus.
1. Remove the fetus liver and mince (with sterilized scissors) on a
100-mm cell culture dish after washing twice with HBSS.
Add 5 ml of collagenase solution, and transfer the cell suspen-
sion to a 50-ml centrifuge tube. Incubate at 37°C for 30 min
with shaking.
2. Add 30 ml of cjDMEM/10% FBS, centrifuge at 190 g for
5 min, and discard the supernatant. Resuspend the cells with
10 ml of cjDMEM/10% FBS, plate onto 100-mm cell culture
dishes, and culture at 37°C with 5% CO2. Change the medium
every other day.
3.3.3. Retroviral Infection 1. Seed the liver cells at 1 × 106 cells per 100-mm cell culture dish
of Marmoset Cells 1 day prior to infection. The cells will reach 70–80% confluency
the following day.
2. Mix equal volumes of each virus-containing supernatant with
OCT4, SOX2, KLF4, C-MYC, NANOG, LIN28, and GFP.
The final volume of the mixture is 8–12 ml. Add polybrene
into the virus-containing mixture to the final concentration of
4 μg/ml.
3. Replace the culture medium with the virus-containing mix-
ture, and incubate the cells for a minimum of 4 h (maximum
overnight) at 37°C, 5% CO2.
4. After infection, replace the virus-containing mixture with cjD-
MEM/10% FBS (Fig. 6a) (see Note 16). Replace the medium
every other day.
Fig. 6. Derivation of marmoset iPSCs. (a) GFP fluorescence 2 days after viral infection. Approximately 33% of the visible
cells fluoresced. (b) Phase and fluorescence images of marmoset iPSCs emerged 3–5 weeks post infection with the six
iPSC factors. All iPSCs exhibited flat, packed, tight colony morphology, and a high nucleus-to-cytoplasm ratio. Fully repro-
grammed iPSCs are GFP negative under UV light because of the transgene silencing.
5. Seven days after transfection, harvest the cells by trypsinization

and plate onto MEF feeder cells at 1–2 × 105 cells per 100-mm
gelatin-coated dish. Replace the cjDMEM/10% FBS with
cjESC medium. Change the medium every other day.
3.3.4. Picking Colonies Three to 5 weeks after introducing the six transcription factors, sev-
eral colonies resembling ESCs will emerge (Fig. 6b) (see Note 17).
1. Aliquot 20 μl of cjESC medium (per well) into a 96-well plate.
Pick each colony in the 96-well plate using a 20-μl pipette, and
dissociate the colony to small clumps by repeated pipetting.
2. Transfer the cell suspension onto MEF feeder cells in gelatin-
coated 12-well plates and culture the cells in cjESC medium.
Change the medium every other day.
3.3.5. Passage of iPSCs Seven to 10 days after picking the colonies, the iPSC colonies
develop to approximately 100–200 μm in a diameter.
1. Aspirate the culture medium and wash the cells twice with
HBSS. Add 0.2 ml of Trypsin solution for the ESCs per well of
the 12-well plate, and incubate at 37°C for 5 min.
2. Add 1 ml of cjESC medium and remove colonies from the
feeder cells by repeated pipetting. Transfer the cell suspension
to a 15-ml centrifuge tube, centrifuge at 190 g for 5 min, and
discard the supernatant (see Note 18).
3. Dissociate the colonies by repeated pipetting to small clumps
of 20–30 cells. Replate on new MEF feeder cells in a 100-mm
gelatin-coated dish (see Note 19).
3.3.6. Storage 1. Aspirate culture medium and wash the cells twice with HBSS.
of Established iPSCs Add 2 ml of Trypsin solution for the ESCs to a 100-mm cell
culture dish, and incubate at 37°C for 5 min.
2. Add 10 ml of cjESC medium and remove colonies from the
feeder cells by repeated pipetting. Transfer the cell suspension
to a 15-ml centrifuge tube, centrifuge at 190 g for 5 min, and
discard the supernatant.
3. Add 3 ml of Cell Banker 2 and aliquot into 2-ml plastic cryo-
genic vials (see Note 20). Store the vials at −80°C.
3.4. Human iPSCs In our laboratory, we have generated human iPSCs by retroviral
from Fibroblasts transduction of four reprogramming factors (Oct4, Sox2, Klf4, and
c-Myc), which was initially introduced by Shinya Yamanaka in 2007
(29). A unique step of Yamanaka method is to introduce mouse
solute carrier family 7 member 1 (Slc7a1) gene, which encodes an
ecotropic retrovirus receptor, into human cells. Although there
are currently several strategies to deliver reprogramming factors
(e.g., various virus vectors, nonviral DNAs, and miRNA), we usually

utilize retrovirus vectors because of convenience and high efficiency.
Here, we described our protocol, especially focusing on the prob-
lems that we encountered.
3.4.1. Preparation 1. Obtain primary human fibroblasts from skin biopsy using a
of Culture Human Dermal 5-mm dermapunch (see Note 21). Place the biopsy specimen
Fibroblasts immediately in mDMEM/10% FBS on ice, and transport it to
the laboratory.
2. Transfer the biopsy sample to a 60-mm culture dish, and elimi-
nate the outer layer of the skin. Cut the inner skin into 1 mm
pieces using sterilized forceps and scissors. Place the four pieces
per 60-mm culture dish. Routinely, three to four dishes with
skin pieces can be prepared from a biopsy specimen.
3. When the pieces adhere to the culture dish, add 5 ml of
mDMEM/10% FBS into the dish. If some pieces do not
adhere, aspirate the medium and try this procedure again.
4. Incubate the cells in 37°C, 5% CO2 incubator and leave them
still for a week. When outgrowth of fibroblasts appears,
exchange the medium twice a week.
5. When the cells grow to 30–50% confluency, split them at 1:3
dilution. Aspirate the medium, wash twice with PBS, and
trypsinize with 0.5 ml of 0.05% Trypsin/EDTA at 37°C for
7 min. Add 3 ml of mDMEM/10% FBS and resuspend by
pipetting. Split the cells to new 60-mm culture dishes at 1:3
dilution (see Note 22).
6. Prepare the freeze stocks when the cells grow to 80% confluency.
Trypsinize with 1 ml of 0.05% Trypsin/EDTA at 37°C for
7 min. Add 6 ml of mDMEM/10% FBS and resuspend by
pipetting. Transfer the cell suspension to a 15-ml conical tube,
centrifuge at 160 g for 5 min, and discard the supernatant.
7. Resuspend the cells with Cell Banker 2 at 1 × 106 cells/ml
approximately. Aliquot 1 ml of the cell suspension per freezing
vial. Keep the vials in a freezing container at −80°C overnight,
and transfer them to the gas phase in a liquid nitrogen tank.
8. To thaw the cell stocks, warm the vials in 37°C water bath until
most (but not all) cells are thawed. Transfer the cells into a
15-ml conical tube containing 9 ml of mDMEM/10% FBS.
Centrifuge at 160 g for 5 min, discard the supernatant, and
resuspend the cells with 10 ml of mDMEM/10% FBS. Plate
the cells into a 100-mm culture dish. Change the medium
every other day.
3.4.2. Lentiviral Production 1. Thaw a vial of 293FT cells in 37°C water bath. Transfer the
cells to a 15-ml conical tube containing 9 ml of mDMEM/10%
FBS. Centrifuge at 160 g for 5 min and discard the supernatant.
Resuspend the cells with 10 ml of mDMEM/10% FBS, and

transfer to a 100-mm culture dish. Incubate the cells in 37°C,
5% CO2 incubator until the cells grow to 80–90% confluency.
Exchange the medium every other day.
2. When the cells grow to 70–80% confluency, trypsinize with
1 ml of 0.25% Trypsin/EDTA at 37°C for 3 min and resus-
pend with mDMEM/10% FBS. Split the cells at 1:3 to 1:5
dilution.
3. The day before transfection, plate the cells at 1 × 106 cells per
60-mm culture dish. The next day, prepare Solution A and B,
and drop Solution A onto Solution B dropwise. Incubate the
solution mixture at room temperature for 20 min.
4. Add the solution mixture to 293FT cell culture dish, and incu-
bate at 37°C, 5% CO2 overnight (see Note 23). Then, replace
the medium with 5 ml of mDMEM/10% FBS.
5. Twenty-four hours after the medium change, collect the super-
natant from the 293FT cell culture and filter it through a 0.45-
μm cellulose acetate filter. Store at −80°C.
3.4.3. Lentiviral 1. Twenty-four hours before transduction, plate the fibroblasts at

Transduction of Fibroblasts 3.2 × 105 cells per 60-mm culture dish. Incubate at 37°C, 5%
CO2 overnight.
2. Replace the medium with 5 ml of the lentivirus supernatant
supplemented with 4 μg/ml polybrene. Incubate at 37°C, 5%
CO2 overnight (see Note 24).
3. Twenty-four hours after transduction, aspirate the virus-
containing medium and add 5 ml of fresh mDMEM/10%
FBS.
4. When the cells grow to 70–80% confluency, passage the cells to
two new 100-mm culture dishes (see Note 25).
3.4.4. Retrovirus Production 1. The day before transfection, seed Plat-E cells at 3.6 × 106 cells
per 100-mm culture dish, and incubate at 37°C, 5% CO2 over-
night (see Note 26).
2. On the next day, mix 27 μl of FuGENE 6 transfection reagent
with 0.3 ml of OPTI-MEM I in a 1.5 ml tube (see Note 26).
Incubate at room temperature for 5 min.
3. Add 9 μg of pMXs vectors (encoding OCT4, SOX2, KLF4,
C-MYC, and GFP) one by one into the FuGENE 6/OPTI-
MEM I mixture. Mix gently and incubate at room temperature
for 15 min.
4. Add the DNA/FuGENE 6 complex dropwise into the Plat-E
cell culture dishes, and incubate at 37°C, 5% CO2 overnight
(see Note 27). After 24 h, replace the medium with 10 ml of
mDMEM/10% FBS and incubate further overnight.
5. On the next day, collect the supernatant from each Plat-E cell
culture, and filter through a 0.45-μm pore size cellulose ace-
tate filter. Add polybrene solution into the filtered virus-
containing medium at the final concentration of 4 μg/ml.
6. Make a mixture of equal volume of the supernatants contain-
ing each retrovirus (see Note 28).
3.4.5. Induction of iPSCs 1. The day before transduction, plate the fibroblasts expressing
from Human Fibroblasts mouse Slc7a1 which encodes an ecotropic retrovirus receptor
at 3.2 × 105 cells per 60-mm culture dish (Fig. 7a) (see Note 29).
Incubate at 37°C, 5% CO2 overnight.
2. Aspirate the medium and add 5 ml of retrovirus mixture pre-
pared at step 6 of Subheading 3.4.4. Incubate the cells at 37°C,
5% CO2 overnight, and replace the medium with mDMEM/10%
FBS. Exchange the medium every other day (Fig. 7b).
3. On the day 11 after infection, trypsinize the cells with 0.5 ml
of 0.05% Trypsin/EDTA at 37°C for 5 min. Resuspend with
4 ml of mDMEM/10% FBS, and count the cell number.
4. Seed 5 × 104 or 5 × 105 cells onto 100-mm culture dishes cov-
ered with SNL feeder cells containing 10 ml of mDMEM/10%
FBS. Incubate at 37°C, 5% CO2 overnight.
5. The next day, replace the medium with 10 ml of hESC medium.
Culture them in 37°C, 3% CO2 incubator. Exchange the
medium to 10 ml of hESC medium supplemented with 4 ng/
ml FGF-2 every other day, until the iPSC colonies become
large enough to be picked (see Note 30). Routinely, iPSC col-
onies are observed 2–3 weeks after the retroviral infection
(Fig. 8) (see Note 31).
Fig. 7. Retroviral transduction of human dermal fibroblasts. (a) Human fibroblasts before the retroviral infection. Replate the
cells at 3.2 × 105 cells on 60-mm culture dishes the day before infection. (b) Human fibroblasts 7 days after infection. The
infection efficiency can be evaluated by transduction of GFP-retrovirus.
Fig. 8. Human iPSCs derived from fibroblasts. At this experiment, 5 × 104 cells of fibroblasts
were replated onto SNL feeder cells in a 100-mm culture dish. ESC-like iPSC colonies
emerge by day 30 after retroviral infection.
3.4.6. Picking and 1. Aliquot 100 μl of hESC medium with FGF-2 per well of a
Expanding Human iPSCs 96-well culture plate. Pick iPSC colonies from the culture dish
under the stereomicroscope using a 20-μl pipette, and transfer
each colony to each well of the 96-well culture plate (see
Note 32). Pipette up and down to dissociate the colonies to
cell clumps composed of 20–30 cells (see Note 33).
2. Add 400 μl of hESC medium with FGF-2 per well, and trans-
fer the cell suspensions into a 24-well plate with SNL feeder
cells. Culture them in 37°C, 3% CO2 incubator until the cells
grow to 80–90% confluency.
3. To passage the iPSCs, aspirate the medium, wash with 0.5 ml
of PBS, and add 0.1 ml of CTK solution. Aspirate an excess of
CTK immediately, and incubate at 37°C for 5 min.
4. Add 0.5 ml of hESC medium with FGF-2 and transfer the cells
into a 1.5-ml plastic tube without pipetting. Centrifuge at
160 g for 5 min at room temperature and discard the superna-
tant. Add 1 ml of hESC medium with FGF-2 and pipette care-
fully to obtain cell clumps composed of 20–30 cells.
5. Transfer the cell suspension to a well of 6-well culture plates
with SNL feeder cells. Add 1 ml of hESC medium with FGF-2,
and incubate in 37°C, 3% CO2 incubator until cells grow to
80–90% confluency. Exchange the medium every day.
6. For further passages, aspirate the medium, wash with 2 ml/
well of PBS, and add 0.5 ml of CTK solution. Incubate at 37°C
for 2–5 min. Then, aspirate CTK solution and wash with 2 ml
of PBS twice.
7. Add 2 ml of hESC medium with FGF-2 and detach iPSCs by
using a cell scraper. Dissociate the iPSC colonies to cell clumps
composed of 20–30 cells by pipetting. Add 8 ml of hESC
medium with FGF-2, and plate the cells into a 100-mm culture
dish with SNL feeder cells. Culture in 37°C, 3% CO2 incubator
until the cells grow to 80–90% confluency again (see Note 34).
3.4.7. Freezing and To make the iPSC freeze stocks, prepare the cells which grow to
Thawing Human iPSCs 80–90% confluency. It is recommended to store the iPSCs at early
passages. We usually use Y-27632, a specific inhibitor for p160-
Rho-associated coiled-coil kinase (ROCK), to enhance a viability
of the frozen cells (30).
1. Aspirate the medium and wash the cells with 6 ml of PBS. Add
1 ml of CTK solution and aspirate the excess immediately.
Then, incubate at 37°C for 5 min.
2. Add 6 ml of hESC medium. Detach the iPSC colonies from
dish by using a cell scraper, and transfer the cell suspension to
two 15-ml conical tubes per 100-mm culture dish.
3. Centrifuge the cells at 160 g for 5 min. Remove the superna-
tant and resuspend the pellet with 0.2 ml of DAP213 solution
by pipetting (see Note 33). Transfer the cell suspension to
freezing vials. Put the vials quickly into liquid nitrogen (see
Note 35).
4. To thaw the freeze stocks, warm 10 ml of hESC medium in
37°C water bath. Add 0.8 ml of pre-warmed hESC medium
into each frozen viral, and thaw quickly by pipetting two to
three times.
5. Transfer the cell suspension to the 15-ml conical tube contain-
ing hESC medium. Centrifuge at 160 g for 5 min at room
temperature.
6. Aspirate the supernatant and add 4 ml of hESC medium sup-
plemented with 4 μl of 10 mM Y-27632 and FGF-2. Plate the
cells into 100-mm culture dishes with SNL feeder cells. Culture
them in 37°C, 3% CO2 incubator until the cells grow to 80–90%
confluency. Do not move the dish for the initial 48 h, and then
exchange the medium every day.
4. Notes
1. 2-Mercaptoethanol is toxic. Avoid inhalation, ingestion, or

contact with skin and eye. Use protective gloves and safety
glasses when handling.
2. Mitomycin C is toxic. Avoid inhalation, ingestion, or contact
with skin and eye. Use protective gloves and safety glasses
when handling.
3. Retroviral pMX vectors for human OCT4, SOX2, KLF4,

C-MYC, NANOG, and LIN28 were kindly provided by Dr.
Yamanaka (29). These vectors except NANOG and LIN28 are
available from Addgene (http://www.addgene.org/Shinya
Yamanaka).
4. Use SNL feeder cells within 3 days. Otherwise, the feeder cells
might detach from the culture dish during iPSCs’ induction
culture. It is possible to utilize frozen stocks of SNL feeder
cells kept at −80°C.
5. Experiments involving use of animals must be approved by the
international and institutional regulations. Technically, all the
procedures should be conducted aseptically.
6. We recommend using early passage fibroblasts (passage 3) as
donors for iPSC derivation because the prolonged culture
causes replicative senescence, which results in low efficiency of
the iPSC derivation.
7. Prepare one tube of DNA/FuGENE 6 mixture for each pMXs
plasmid. At this time, it is essential to prepare the proper con-
trol to confirm successful gene transduction. We normally uti-
lize pMXs retroviral vector of red fluorescence protein DsRed
to evaluate the transduction efficiency. Thus, in the case of four
iPSC factors’ transduction, a total five tubes of DNA/FuGENE
6 mixture are necessary for the production of Oct4, Sox2, Klf4,
c-Myc, and DsRed viruses.
8. The viral medium should be used soon after collection. Storing
of viral supernatant causes a significant reduction of transfec-
tion efficiency and consequently a lower number of iPSC colo-
nies. Because the retrovirus infects Plat-E packaging cells
themselves, the successful viral production can be convention-
ally evaluated by DsRed fluorescence in Plat-E cells transfected
with pMXs-DsRed (Fig. 1).
9. Check the transduction efficiency by monitoring DsRed
fluorescence. High efficiency of gene transduction is essential
for successful iPSC derivation. We routinely observe >80%
transfection efficiency.
10. The number of replated cells dramatically affects the frequency
of the iPSC derivation. For the transduction of four iPSC fac-
tors, replating too many cells results in overgrowth of trans-
formed cells and difficulty in isolation of highly reprogrammed
iPSCs. On the other hand, without c-Myc, the iPSCs’ frequency
becomes much lower so that all the cells can be replated.
Considering the number of iPSC colonies fluctuates among
experiments, it is better to validate the optimal cell number by
replating with a dilution series.
11. Highly reprogrammed iPSCs can be distinguished by a silenc-

ing of retroviral transgenes. By retroviral transduction of the
iPSC factors in combination with DsRed, we can visually
monitor the transgene silencing based on the fluorescence.
The iPSC colonies without DsRed fluorescence are most likely
to be highly reprogrammed iPSCs.
12. Crush the bones in pieces but do not grind them.
13. To adjust the isotype controls, arrange the preparatory experi-
ment cells similarly to the situation of the fluorescence com-
pensation using “control tubes.” Once the adjustment is done,
there is no need to check for the following experiments.
14. The cells reach 80–90% confluency the following day.
15. We can generate marmoset iPSCs using frozen viral stocks
stored for up to 1 month. Do not repeatedly freeze/thaw the
viral stocks to avoid reducing the viral titer.
16. Infection rates as assessed by GFP fluorescence should be more
than 30% (Fig. 6a).
17. This timing of iPSC colony appearance is consistent with the
study on rhesus monkey iPSCs (31). All colonies morphologi-
cally resembling marmoset ESCs (32) are GFP negative
(Fig. 6b). Although various colony types appear approximately
2 weeks post infection, these are not iPSC colonies. ESC-like,
clear-edged colonies should appear within 3–5 weeks post
infection.
18. When the differentiated cells occupy the majority of the dishes,
collect the undifferentiated colonies using a 20-μl pipette, or
remove the large clumps of differentiated cells by filtering the
cell suspension using a 100-μm nylon cell strainer.
19. Do not dissociate the cell clumps into single cells because too
much dissociation could trigger cell death. Cells are passaged
approximately every 5–7 days.
20. Do not dissociate the colonies to small clumps prior to
freezing.
21. For biopsy of human dermal fibroblasts, you must obtain a
proper informed consent from the donors.
22. Be careful not to plate the cells at low cell density to prevent
replicative senescence.
23. Transfection of the lentiviral vector should be done when
293FT cells grow to 70% confluency in a 100-mm culture
dish.
24. The overnight incubation with lentivirus is sometimes toxic to
fibroblasts. In that case, shorten the incubation duration down
to 5 h.
25. To evaluate the successful transduction, use a GFP-encoding

lentiviral vector as a control. Alternatively, it is possible to select
the infected cells by culturing with 10 μg/ml blastocidin S,
because pLenti6/UbC/mSlc7a1 vector carries a blastocidin-
resistance gene.
26. Prepare one cell culture dish and one 1.5-ml plastic tube per
plasmid. Five cell culture dishes and tubes are required for
transfection of OCT4, SOX2, KLF4, C-MYC, and GFP retro-
virus vectors.
27. Do not transfect more than two plasmids into a Plat-E cell cul-
ture dish. It causes a low-efficient generation of resultant iPSCs.
28. Use immediately for transduction. Do not freeze the virus
supernatants.
29. The efficiency of retroviral transduction markedly decreases
when using fibroblasts at older passages. It is recommended to
use passage 8–10 fibroblasts for iPSC induction.
30. Fuzzy-edged cell colonies appear approximately 2 weeks
after retroviral transduction, but they are not iPSC colo-
nies. Keep the culture until clear-edged, hESC-like colonies
are observed.
31. Check the emergence of iPSC colonies carefully, because the
timing differs in each experiment even when using the same lot
of fibroblasts.
32. We usually pick 20–30 iPSC colonies from a culture dish.
33. Do not dissociate iPSC colonies into single cells.
34. To keep undifferentiated iPSC culture, remove the differenti-
ated colonies by aspirating during passaging procedure. Transfer
only undifferentiated iPSC colonies to a new culture dish.
35. To ensure high cell viability, these procedures should be done
within 15 s.
Acknowledgements
This work was supported by grants from the Ministry of Education,

Culture, Sports, Science and Technology of Japan (MEXT); the
Ministry of Health, Labor, and Welfare; the Japan Society for the
Promotion of Science (JSPS); the National Institute of Biomedical
Innovation; the Project for Realization of Regenerative Medicine,
MEXT; the Funding Program for World-leading Innovative R&D
in Science and Technology (FIRST), JSPS; and Grant-in-Aid for
Young Scientists (B).
References
1. Stadtfeld M, Hochedlinger K (2010) Induced stem cells derived from adult hepatocytes. Mol
pluripotency: history, mechanisms, and appli- Reprod Dev 77:802–811
cations. Genes Dev 24:2239–2263 18. Park TS et al (2009) Derivation of primordial
2. Takahashi K, Yamanaka S (2006) Induction of germ cells from human embryonic and induced
pluripotent stem cells from mouse embryonic pluripotent stem cells is significantly improved
and adult fibroblast cultures by defined factors. by coculture with human fetal gonadal cells.
Cell 126:663–676 Stem Cells 27:783–795
3. Chamberlain SJ et al (2010) Induced pluripo- 19. Kim JB et al (2009) Oct4-induced pluripotency
tent stem cell models of the genomic imprinting in adult neural stem cells. Cell 136:411–419
disorders Angelman and Prader-Willi syndromes. 20. Honda A et al (2010) Generation of induced
Proc Natl Acad Sci U S A 107:17668–17673 pluripotent stem cells in rabbits: potential
4. Horii T et al (2008) Loss of genomic imprint- experimental models for human regenerative
ing in mouse parthenogenetic embryonic stem medicine. J Biol Chem 285:31362–31369
cells. Stem Cells 26:79–88 21. Shimada H et al (2010) Generation of canine
5. Do JT et al (2009) Generation of parthenoge- induced pluripotent stem cells by retroviral
netic induced pluripotent stem cells from par- transduction and chemical inhibitors. Mol
thenogenetic neural stem cells. Stem Cells Reprod Dev 77:2
27:2962–2968 22. Nagy K et al (2011) Induced pluripotent stem
6. Takahashi K (2010) Direct reprogramming cell lines derived from equine fibroblasts. Stem
101. Dev Growth Differ 52:319–333 Cell Rev 7(3):693–702
7. Humpherys D et al (2001) Epigenetic instability 23. Bao L et al (2011) Reprogramming of ovine
in ES cells and cloned mice. Science 293:95–97 adult fibroblasts to pluripotency via drug-
8. Stadtfeld M et al (2010) Aberrant silencing of inducible expression of defined factors. Cell
imprinted genes on chromosome 12qF1 in Res 21(4):600–8
mouse induced pluripotent stem cells. Nature 24. Tomioka I et al (2010) Generating induced
465:175–181 pluripotent stem cells from common marmo-
9. Liu L et al (2010) Activation of the imprinted set (Callithrix jacchus) fetal liver cells using
Dlk1-Dio3 region correlates with pluripotency defined factors, including Lin28. Genes Cells
levels of mouse stem cells. J Biol Chem 15:959–969
285:19483–19490 25. Sumer H et al (2011) NANOG is a key factor
10. Pick M et al (2009) Clone- and gene-specific for induction of pluripotency in bovine adult
aberrations of parental imprinting in human fibroblasts. J Anim Sci 89(9):2708–2716
induced pluripotent stem cells. Stem Cells 26. Fujioka T et al (2004) A simple and efficient
27:2686–2690 cryopreservation method for primate embry-
11. Toyooka Y et al (2003) Embryonic stem cells onic stem cells. Int J Dev Biol 48:1149–1154
can form germ cells in vitro. Proc Natl Acad 27. Morikawa S et al (2009) Prospective identification,
Sci U S A 100:11457–11462 isolation, and systemic transplantation of multipo-
12. Geijsen N et al (2004) Derivation of embry- tent mesenchymal stem cells in murine bone mar-
onic germ cells and male gametes from embry- row. J Exp Med 206:2483–2496
onic stem cells. Nature 427:148–154 28. Niibe K et al (2011) Purified mesenchymal
13. Hubner K et al (2003) Derivation of oocytes stem cells Are an efficient source for iPS cell
from mouse embryonic stem cells. Science induction. PLoS One 6:e17610
300:1251–1256 29. Takahashi K et al (2007) Induction of pluripo-
14. Eguizabal C et al (2009) Generation of pri- tent stem cells from adult human fibroblasts by
mordial germ cells from pluripotent stem cells. defined factors. Cell 131:861–872
Differentiation 78:116–123 30. Watanabe K et al (2007) A ROCK inhibitor
15. Hayashi K, Surani MA (2009) Self-renewing permits survival of dissociated human embry-
epiblast stem cells exhibit continual delinea- onic stem cells. Nat Biotechnol 25:681–686
tion of germ cells with epigenetic reprogram- 31. Liu H et al (2008) Generation of induced
ming in vitro. Development 136:3549–3556 pluripotent stem cells from adult rhesus mon-
16. Lavagnolli TM et al (2009) Presumptive germ key fibroblasts. Cell Stem Cell 3:587–590
cells derived from mouse pluripotent somatic 32. Sasaki E et al (2005) Establishment of novel
cell hybrids. Differentiation 78:124–130 embryonic stem cell lines derived from the
17. Imamura M et al (2010) Induction of primor- common marmoset (Callithrix jacchus). Stem
dial germ cells from mouse induced pluripotent Cells 23:1304–1313
Chapter 3
Generation of Trophoblast Stem Cells

Michael C. Golding
Abstract
The isolation and culture of both embryonic and extraembryonic stem cells provide an enormous opportunity
to study the molecular processes that establish and maintain lineage-specific, monoallelic patterns of gene
expression. This chapter describes the isolation an culture of trophectoderm stem cells from mouse
blastocyst stage embryos. Using this powerful in vitro system, scientists can now begin to tease apart the
epigenetic processes that result in placental patterns of imprinted gene expression and begin to better
understand the role these genes play in development and disease.
Key words: Placental stem cell, Trophectoderm, Extraembryonic lineage, TS cell
1. Introduction
Genomic imprinting is a specialized transcriptional regulatory

mechanism that restricts expression to the maternally or paternally
inherited allele (1). Misregulation of these lineage-specific patterns
of monoalleleic gene expression has been associated with numer-
ous developmental disorders and cancer (2, 3). Recently, we have
come to recognize the extreme importance of imprinted gene
expression to the proper development and function of the placenta,
through observed defects in the development of embryos produced
through assisted reproductive technologies (2, 4–10).
To better define these and other molecular events driving
mammalian embryogenesis, pluripotent stem cells from each of the
three distinct lineages present within the preimplantation blastocyst
have been derived (11–14). Embryonic (ES), trophectoderm (TS), and
extraembryonic endoderm (XEN) stem cells each possess the devel-
opmental potential of their founding lineages and exhibit distinct
patterns of imprinted X-inactivation and gene expression (14–17).
49
50 M.C. Golding
However, the molecular basis for the establishment and maintenance

of these differing monoallelic patterns of gene expression remains
poorly defined.
Trophectoderm stem cells represent a powerful system with
which to study the developmental origins of the extraembryonic
tissues that give rise to the placenta (13). In the analysis of patterns
of imprinted gene expression it is essential to be able to identify
parent-of-origin expression. To this end, genetic crosses between
distinct strains of mice allow the tracking and identification of
allele-specific patterns of gene expression through the identification
of single-nucleotide polymorphisms either through direct
sequencing or using restriction enzyme digestion (18). This chapter
describes the derivation of TS cells from embryos potentially
derived from these crosses.
2. Materials
2.1. Production 1. A source of mouse embryonic fibroblasts—either commercial

of Embryonic Feeders or primary cultures.
2. DMEM with 10% Fetal Bovine Serum and 1% Antibiotic/
Antimycotic.
3. Phosphate-buffered saline.
4. 0.1% (w/v) trypsin–1 mM EDTA.
5. Mitomycin C.
6. DMSO.
7. T175 Flasks or 15 cm tissue culture dishes.
8. Freezing vials.
9. −80 °C Freezer.
10. Liquid nitrogen tank or −160 °C freezer.
11. 37 °C Water bath.
2.2. Derivation 1. TS Cell Medium:

of Trophectoderm 500 ml RPMI (Sigma).
Stem Cells
6 ml Pen/strep (50 μg/ml each final concentration, Sigma).
6 ml 100 mM sodium pyruvate (Invitrogen 11360070, final
concentration 1 mM).
6 ml 10 mM B-mercaptoethanol (Sigma, final concentration
100 μM) (14.3 M BME use 35 μl in 50 ml to make 100×
stock).
1,000× FGF4 (R&D Systems).
1,000× FGF Basic (R&D Systems).
3 Generation of Trophoblast Stem Cells 51
1,000 × Heparin (1mg/ml).

6 ml 200 mM L-glutamine (Sigma, final concentration 2 mM).
FCS to final volume of 15% (Hyclone ES Serum, Fisher
Scientific).
2. Low-wall tissue culture dishes.
3. Mitomycin C-treated MEF feeders.
4. Source of Mouse blastocysts.
2.3. Culture of 1. Mitomycin C-treated MEFs.

Trophectoderm Stem 2. DMEM/10% FBS.
Cells in the Absence
3. 15 cm Dishes.
of Feeders
4. TS medium.
5. 0.45 μm Filter.
6. 10–15 ml Syringe.
7. 37 °C Water bath.
2.4. Freezing TS Cell 1. TS Cell Medium:

Cultures 500 ml RPMI (Sigma).
6 ml Pen/strep (50 μg/ml each final conc Sigma).
6 ml 100 mM sodium pyruvate (Invitrogen, final concentration
1 mM).
6 ml 10 mM B-mercaptoethanol (Sigma, final concentration
100 μM) (14.3 M BME use 35 μl in 50 ml to make 100×
stock).
1,000× FGF4 (R&D Systems).
1,000× FGF Basic (R&D Systems).
1,000× Heparin (1mg/ml).
6 ml 200 mM L-glutamine (Sigma, final conc 2 mM).
FCS to final volume of 15% (Hyclone ES Serum Fisher
Scientific).
2. Freezing Medium (50 ml).
25 ml FBS.
20 ml of TS cell Medium.
5 ml of DMSO.
3. Cryo-Vials.
4. −80 °C Freezer.
5. −160 °C Freezer or Liquid Nitrogen.
6. Cell Freezer.
52 M.C. Golding
3. Methods
3.1. Production of Trophectoderm Stem Cells require several soluble, secreted factors,
Embryonic Feeders: including Activin and TGF-Beta, in order to maintain an undif-
Mitomycin C ferentiated state (19). Moreover, TS cell lines grown on MEF-feeder
Treatment of Mouse layers are easier to maintain than those grown in MEF-conditioned
Embryonic Fibroblasts medium. Here we will describe the production of growth-arrested
mouse embryonic fibroblasts to be used as feeder layers in the
culture of trophectoderm stem cells. Using the powerful chemo-
therapeutic Mitomycin-C to irreversibly inhibit DNA replication,
treated MEFs can be plated and although their growth has been
arrested they still continue to secrete factors necessary for TS cell
maintenance. It is best to prepare multiple vials of feeder cells at
once to ensure uniformity.
1. Thaw a frozen vial of MEFs in a 37 °C water bath and transfer
entire contents into a 1.5-ml tube and centrifuge at 400 × g for
4 min.
2. Remove the supernatant and gently resuspend the cells in 1 ml
of DMEM/10% FBS.
3. Split cells (500 μl each) into two 15 cm dishes, each containing
25 ml DMEM/10% FBS.
4. Culture cells at 37 °C for 3–4 days or until cells reach ~90%
confluence. Do not let the cells become confluent.
5. Passage the cells by removing the growth medium and rinse
twice with 10 ml PBS per dish.
6. Add 2.5 ml 0.1% trypsin to each dish and incubate for ~2 min
at 37 °C. Tap each dish to dissociate the cell monolayer.
7. Add 10 ml of DMEM/10% FBS to the dish and gently pipette
to break cell aggregates.
8. Transfer cells to a 15 ml tube and centrifuge at 200 × g for
4 min. Resuspend cell pellet in 15 ml of DMEM/10% FBS.
9. Split (3 ml each) into five new 15 cm plates containing 22 ml
DMEM/10% FMS.
10. Culture cells at 37 °C for 3–4 days or until cells reach ~90%
confluence. Do not let the cells become confluent.
Caution—Mitomycin C is extremely toxic. Please exercise caution
when handling.
11. When cells are ready prepare 200 ml of DMEM/10%FBS con-
taining 10 μg/ml Mitomycin-C. Typically this compound is
sold in 2 mg aliquots. Resuspend the entire contents of the vial
in 4 ml of DMEM/10% FBS and add this to a final volume of
200 ml of DMEM/10% FBS. This will produce enough
Mitomycin-C medium to treat ten 15 cm plates. Add
Mitomycin-C medium and incubate cells for 2 h at 37 °C.
12. Remove the medium and rinse cells twice, with 20 ml of PBS.
13. Add 5 ml of Trypsin as described above and place in incubator
for ~2 min.
14. Tap each dish to dissociate the cell monolayer. Add 10 ml of
DMEM/10% FBS to the dish and gently pipette to break cell
aggregates.
15. Transfer the cells from each plate into a 15 ml tube and centri-
fuge at 200 × g for 4 min.
16. Resuspend cell pellet in 5 ml of Freezing Medium—60% FBS,
30% DMEM, and 10% DMSO—and aliquot into freezing vials.
Typically, one vial will contain enough cells to cover one 10 cm
dish or two 12-well plates.
17. Place vials in the −80 °C freezer overnight and transfer to liq-
uid nitrogen or a −160 °C freezer for long-term storage.
3.2. Derivation TS Cells are derived by plating blastocyst stage embryos on an

of Trophectoderm MEF-feeder layer and allowing outgrowths to form. When these
Stem Cells outgrowths are dissociated, culture of the derivative cells in TS
cell medium will promote the growth of trophectoderm and XEN
stem cell colonies. These colonies can be picked and stable TS cell
lines derived.
3.2.1. Preparation 1. FGF Basic (R&D Systems) and FGF4 (R&D Systems) need to
for TS Cell Medium be suspended in 1 ml of PBS/0.1% BSA (we use the NEB FBS
that comes with Restriction Enzymes and filter sterilize) and
make 50 μl aliquots and freeze at −80 °C.
2. 1000 × Heparin is made by diluting 1 mg/ml of Heparin
(Sigma catalogue # H3393) in PBS. Make 1 ml aliquots and
store at −80 °C.
3.2.2. Derivation 1. One or two days before day before blastocyst collection, plate
of Mouse TS Cells Mitomycin C-treated MEFs in low-wall 4-well plates in a final
volume of 0.5 ml DMEM/10%FBS per well.
2. On the day of blastocyst collection replace the DMEM/10%
FBS on the feeders with TS cell medium adding fresh FGF
basic and FGF4 to the medium.
3. Sacrifice mated females using methods approved by your
institution’s animal use and care committee at a time that will
allow the collection of late morula or early blastocyst stage
embryos (Day 3.0 to Day 3.5 days post coitus).
4. Isolate blastocysts by dissection of the uterine horns and
utilizing a 1 ml syringe, M2 medium, and a 26 gauge needle
to flush the embryos into a petri dish. Detailed protocols
describing mouse blastocyst collection have been described
elsewhere (20).
54 M.C. Golding
5. Wash the embryos through PBS–PVP and add a single embryo

to each well of the 4-well plate containing the feeder MEFs.
6. Return the plates to a standard tissue culture incubator (37 °C,
5% CO2) and allow blastocyst outgrowths to form.
7. After 2 days of culture replace medium with fresh TS cell
medium containing FGF basic and FGF4.
8. Plate a second group of mouse feeders on low-wall 4-well plates.
9. After a 4 or 5 days of growth blastocyst outgrowths should be
easily visible and have reached a size of greater than 750 μm.
Do not let outgrowths become too large as the efficiency of
stem cell isolation will rapidly begin to diminish (see Note 1).
10. Carefully aspirate the TS cell medium with a pipette and wash
with PBS. The outgrowths will be very loosely attached, so
take great care not to knock them loose.
11. To dissociate these structures add 0.1 ml 0.1% trypsin/1 mM
EDTA and incubate for 5 min at 37 °C.
12. After incubation, use a 200 μl pipette (set to a volume of 75 μl)
to dissociate the outgrowths by pipetting up and down.
13. Add 600 μl of Fesh TS cell medium containing fresh FGF basic
and FGF4 to the dissociated outgrowths and transfer to a well
in the new 4-well plates.
14. After 12 h, replace medium with fresh TS cell medium (+ heparin,
FGF basic, and FGF4).
15. Culture TS cells in the above medium in a standard tissue
culture incubator (37 °C, 5% CO2 incubator) replacing TS cell
medium every 2 days. After approximately 1 week, TS cell
colonies will begin to form. Often, XEN cells will also begin to
grow. These cells tend to grow as single cells in clumps that
begin to branch out. It is important to pick TS cell colonies
before the plate becomes too overgrown so as to minimize the
chance of XEN contamination (see Note 2) (Fig. 1).
16. Plate mouse feeders in either a flat-bottom 96-well plate or
24-well plate depending on the number of colonies.
17. Aspirate the medium from the 4-well plates and wash TS cell
colonies twice with PBS.
18. After the last wash, cover the cells with 150 μl of PBS.
19. Using a dissecting scope use a 20 μl pipette to add approxi-
mately 10 μl of 0.1% trypsin/1 mM EDTA directly onto the
target colony, then mechanically break the colony away from
the tissue culture surface, and pick up with the pipette.
20. Eject colony in a 0.5 ml tube on a 37 °C heating block.
21. After ~5 min add 100 μl of fresh TS cell medium containing
heparin, FGF basic, and FGF4.
Fig. 1. Light micrographs depicting sequential stages in the isolation and culture of trophectoderm stem cells. (a) Early
passage TS cells 4–6 days after dissociation of the initial embryonic body. Note the “smooth” cells along the expanding
margins and “rough” cells in the center. (b) TS cells after the third passage. Colonies at this point will be primarily composed
of “smooth” cells growing in individual colonies. (c) TS cells after ten passages in culture, growing on gelatin-coated plastic,
in conditioned medium. Cells grow in relatively homogenous colonies with actively proliferating, expanding margins of
smooth cells. The colonies shown here are confluent and need to be split within 12 h.
22. Pipette up and down to dissociate the cells and transfer to a

single well of the tissue culture plate containing the newly
plated feeders.
23. After 12–24 h replace medium with fresh.
24. Culture TS cells in the above medium in a standard tissue cul-
ture incubator (37 °C, 5% CO2 incubator) replacing TS cell
medium every 2 days.
25. TS cells may now be cultured in the above medium in a stan-
dard tissue culture incubator (37 °C, 5% CO2 incubator). Cells
are typically passaged (1:20) every 2–3 days. If they are split
1:20, they may become confluent in 2–3 days. TS cell media is
changed every second day; however, when they reach >60%
confluence, the media should be changed daily.
3.3. Maintenance and 1. To passage TS cells wash twice with PBS, and dissociate colo-
Passage of TS Cells nies with enough 0.1% trypsin/1 mM EDTA to cover the bot-
tom of the plate.
2. Cells should begin to lift off the plate after ~1–2 min at which
point colonies should be dissociated by gentle pipetting up and
down.
56 M.C. Golding
3. Add TS cell medium containing heparin, FGF basic, and FGF4

to stop the reaction. Split cells 1:20 and transfer into a new
well of feeder cells.
4. After 12 h replace the medium on the newly plated cells with
fresh TS cell medium containing heparin, FGF basic, and FGF4.
3.4. Culture and In certain experimental situations it is necessary to culture TS cells

Passage of TS Cells in in the absence of MEF-feeders. For example, when diagnosing
the Absence of Feeders imprinted gene expression it is best not to have feeder lines con-
tributing to the analysis. To this end, TS cells may be grown in
gelatin-treated plastic (see Note 3).
MEFs and differentiated trophoblast cells adhere to the tissue
culture dish more quickly than TS cells. This differential plating
time can be used to recover floating TS cells in the medium after
the MEFs and other cell types have adhered to the tissue culture
plastic. TS cells can be maintained in the absence of MMC-MEFs
in medium supplemented with 70% MEF-conditioned medium.
The example below is for a 100-mm cell culture dish. Adjust volumes
accordingly for different sizes of dishes or flasks.
3.4.1. Production of 1. Thaw a frozen vial of Mitomycin-C-treated MEFs in a 37 °C

MEF-Feeder-Conditioned water bath, transfer entire contents into a 1.5-ml tube, and
Medium centrifuge at 400 × g for 4 min.
2. Remove the supernatant and gently resuspend the cells in 1 ml
of DMEM/10% FBS.
3. Transfer cells into a 15 cm dish containing 25 ml DMEM/10%
FBS.
4. Culture cells at 37 °C for 24–48 h to let cells settle.
5. Replace medium with 25 ml TS cell medium that does not
contain heparin, FGF basic, and FGF4.
6. Incubate cells in a standard tissue culture incubator (37 °C, 5%
CO2 incubator) for 3 days.
7. On the third day pre-wet a 0.45 μm syringe filer with fresh TS
cell medium that does not contain FGF basic and FGF4.
8. Draw the medium on the 15 cm plate up into the 15 ml
syringe.
9. Filter medium into a 50 ml conical tube. It may take a few pulls
from each plate.
10. The medium is now feeder conditioned and may be stored at
−80 °C.
11. Replace aspirated medium with a second 25 ml volume of TS
cell medium that does not contain heparin, FGF basic, and
FGF4.
12. A second collection of filtered medium is possible by repeating
steps 5–10.
3.4.2. Culture of TS Cells 1. Grow TS cells on MEF-Feeders as described above.

in MEF-Feeder-Conditioned
2. Two to 3 h before passaging the cells, the tissue culture dishes
Medium
that will be used to culture the MEF-feeder-free TS cells need
to be treated with 1% Gelatin. Add enough 1% Gelatin in PBS
to completely cover the bottom of the tissue culture well and
place in a standard tissue culture incubator for 2–3 h.
3. Wash TS cells twice with PBS, and dissociate colonies with
enough 0.1% trypsin/1 mM EDTA to cover the bottom of the
plate.
down.
5. Add conditioned TS cell medium containing FGF basic and
FGF4 to stop the reaction.
6. Transfer the cells to a new tissue culture plate and allow feeder
cells to settle our for 1 h.
7. After 1 h a large proportion of the MEF-feeders will have
attached to the tissue culture dish. Pick up the remaining cells
and place in a suitable tube. Discard the tissue culture well/
dish with the attached feeders.
8. Remove the 1% gelatin solution from the tissue culture plates
and split the TS cells 1:7.5 or 1:10 and transfer into a well of
the gelatin-coated dish.
9. After 12 h replace the medium on the newly plated cells with
fresh conditioned TS cell medium containing heparin, FGF
basic, and FGF4.
10. Cells may be passaged in conditioned TS cell medium containing
FGF basic and FGF4 as described above.
3.5. Freezing TS 1. Prepare the TS Cell Freezing Medium by combining the

Cell Cultures components listed above. Place Medium on ice.
2. Wash confluent TS cell cultures twice with PBS, and dissociate
colonies with enough 0.1% trypsin/1 mM EDTA to cover the
bottom of the plate.
down.
4. Add TS cell medium containing FGF basic and FGF4 to stop
the reaction.
5. Remove an aliquot of cells to passage as necessary.
6. Transfer the remaining cells to a 1.5 ml tube and spin at
4,000 × g for 4 min.
7. Remove media and resuspend cells in 1 ml of TS cell freezing
medium.
58 M.C. Golding
8. Transfer cells to a 15 ml tube containing 4 ml of TS Cell

Freezing Medium.
9. Mix the cells by inversion and aliquot the cells in 1 ml volumes
into five cryovials.
10. Place cells in cell freezer and place in the −80 °C freezer
overnight.
11. The next day transfer frozen cells to either liquid nitrogen or
−160 °C freezer for long-term storage (see Note 4).
4. Notes
1. TS cells are a challenging cell type to derive and maintain in

culture. TS cells grow very slowly when plated at low density
but upon reaching a critical mass begin to grow very quickly.
This property is likely due to an as yet unidentified secreted
factor. Given these observations we have always found that
derivation of TS cells is more efficient when multiple blasto-
cysts are plated and dissociated in a single culture dish. Once
multiple TS cell colonies begin to emerge clonal populations
are picked.
2. When deriving TS cells, care must be taken not to allow XEN
stem cells to take over the culture dish. In contrast to the
smooth morphology of TS cell colonies, XEN cells grow in
clumps of individual cells but will quickly spread throughout
the dish.
3. TS cells typically take a long time to recover after being frozen.
Be sure to plate a large number of cells in a dish to ensure a
rapid recovery. Furthermore, after freezing a significant number
of cells will spontaneously begin to differentiate. Allow two to
three passages for the stem cell population to stabilize before
proceeding with your experiments.
4. Culture of TS cells in the absence of feeders is very challeng-
ing. A significant number of TS cell colonies will have subpopu-
lations that may differentiate when plated on plastic or glass.
Again, allow cells to passage two to three times in conditioned
medium before beginning experiments.
Acknowledgement
This work was supported by the NIH grant AA020129-02.

References
1. Verona RI, Mann MR, Bartolomei MS (2003) 11. Martin GR (1981) Isolation of a pluripotent
Genomic imprinting: intricacies of epigenetic cell line from early mouse embryos cultured in
regulation in clusters. Annu Rev Cell Dev Biol medium conditioned by teratocarcinoma stem
19:237–259 cells. Proc Natl Acad Sci U S A 78:7634–7638
2. Odom LN, Segars J (2010) Imprinting disor- 12. Nagy A, Rossant J, Nagy R, Abramow-Newerly
ders and assisted reproductive technology. W, Roder JC (1993) Derivation of completely
Curr Opin Endocrinol Diabetes Obes 17: cell culture-derived mice from early-passage
517–522 embryonic stem cells. Proc Natl Acad Sci U S A
3. Uribe-Lewis S, Woodfine K, Stojic L, Murrell 90:8424–8428
A (2011) Molecular mechanisms of genomic 13. Tanaka S, Kunath T, Hadjantonakis AK, Nagy
imprinting and clinical implications for cancer. A, Rossant J (1998) Promotion of trophoblast
Expert Rev Mol Med 13:e2 stem cell proliferation by FGF4. Science
4. Eggan K, Akutsu H, Hochedlinger K, Rideout 282:2072–2075
W, Yanagimachi R, Jaenisch R (2000) 14. Kunath T, Arnaud D, Uy GD, Okamoto I,
X-Chromosome inactivation in cloned mouse Chureau C et al (2005) Imprinted X-inactivation
embryos. Science 290:1578–1581 in extra-embryonic endoderm cell lines from
5. Hill JR, Burghardt RC, Jones K, Long CR, mouse blastocysts. Development 132:
Looney CR et al (2000) Evidence for placental 1649–1661
abnormality as the major cause of mortality in 15. Lewis A, Mitsuya K, Umlauf D, Smith P, Dean
first-trimester somatic cell cloned bovine W et al (2004) Imprinting on distal chromo-
fetuses. Biol Reprod 63:1787–1794 some 7 in the placenta involves repressive his-
6. Bourc’his D, Le Bourhis D, Patin D, Niveleau tone methylation independent of DNA
A, Comizzoli P et al (2001) Delayed and methylation. Nat Genet 36:1291–1295
incomplete reprogramming of chromosome 16. Terranova R, Yokobayashi S, Stadler MB, Otte
methylation patterns in bovine cloned embryos. AP, van Lohuizen M et al (2008) Polycomb
Curr Biol 11:1542–1546 group proteins Ezh2 and Rnf2 direct genomic
7. Xue F, Tian XC, Du F, Kubota C, Taneja M contraction and imprinted repression in early
et al (2002) Aberrant patterns of X chromo- mouse embryos. Dev Cell 15:668–679
some inactivation in bovine clones. Nat Genet 17. Latos PA, Stricker SH, Steenpass L, Pauler FM,
31:216–220 Huang R et al (2009) An in vitro ES cell imprint-
8. Santos F, Zakhartchenko V, Stojkovic M, Peters ing model shows that imprinted expression of
A, Jenuwein T et al (2003) Epigenetic marking the Igf2r gene arises from an allele-specific
correlates with developmental potential in expression bias. Development 136:437–448
cloned bovine preimplantation embryos. Curr 18. Market-Velker BA, Zhang L, Magri LS,
Biol 13:1116–1121 Bonvissuto AC, Mann MR (2010) Dual effects
9. Mann MR, Lee SS, Doherty AS, Verona RI, of superovulation: loss of maternal and paternal
Nolen LD et al (2004) Selective loss of imprint- imprinted methylation in a dose-dependent
ing in the placenta following preimplantation manner. Hum Mol Genet 19:36–51
development in culture. Development 19. Erlebacher A, Price KA, Glimcher LH (2004)
131:3727–3735 Maintenance of mouse trophoblast stem cell
10. Lin J, Shi L, Zhang M, Yang H, Qin Y et al proliferation by TGF-beta/activin. Dev Biol
(2011) Defects in trophoblast cell lineage 275:158–169
account for the impaired in vivo development 20. Nagy A (2003) Manipulating the mouse
of cloned embryos generated by somatic embryo: a laboratory manual. CSHL Press,
nuclear transfer. Cell Stem Cell 8:371–375 Cold Spring Harbor, NY
Chapter 4
Immunomagnetic Purification of Murine Primordial

Germ Cells
Emily Y. Smith and James L. Resnick
Abstract
Primordial germ cells (PGCs) play essential roles in both reproduction and development. In this chapter,
we describe a method used in our laboratory for the immunopurification of PGCs from the mouse embryo.
After dissection and disruption of the fetal gonad, PGCs are identified by a monoclonal antibody recogniz-
ing an epitope characteristic of pluripotent stem cells. After reaction with a paramagnetic bead-linked
secondary antibody, the cell mixture is applied to a strong magnetic field. PGCs are recovered by release
from the magnetic field. Purity is assessed by the alkaline phosphatase activity inherent to PGCs.
Key words: Primordial germ cells, Immunomagnetic purification
1. Introduction
Primordial germ cells (PGCs) are the embryonic progenitors of

eggs and sperm and are thus vital to both reproduction and
development. In addition to ensuring transgenerational continuity
of the germ line, murine PGCs implement mitotic amplification of
the germ lineage, colonization of the gonad, epigenetic reprogram-
ming of both imprinted and non-imprinted genes, and reactivation
of the X chromosome, and set the stage for entry into meiosis.
After a complex pattern of specification in the extraembryonic
mesoderm during gastrulation, PGCs perform these tasks while
migrating toward the genital ridge, the gonadal anlage. PGCs
undergo sexual differentiation shortly after arrival in the genital
ridge. In male embryos the PGCs arrest mitotically as prosper-
matogonia. In female embryos PGCs enter meiotic prophase (1).
Given their diverse roles and functions, it is not surprising
that investigators would seek to morphologically, physiologically,
61
62 E.Y. Smith and J.L. Resnick
and molecularly characterize PGCs. During the migratory phase

PGCs are not a uniform tissue, but instead are present either singly
or in small aggregates (2). This has naturally led to the development
of several methods for PGC isolation and purification. McLaren
and colleagues were among the first to purify PGCs. They developed
a method in which isolated genital ridges were mashed to release
the germ cells into the media. PGCs were then identified by their
unique morphology and isolated manually (3). Shortly thereafter,
De Felici and McLaren described a method for larger scale PGC
fractionation on Percoll gradients (4). Mayanagi et al. have more
recently described improvements in this method (5). Over the past
15 years, however, FACS sorting and immunomagnetic sorting
have become the most widely used methods. Both techniques yield
high purity with reasonable effort but require either special mouse
strains or immunological reagents. FACS sorting to purify PGCs
often takes advantage of monoclonal antibodies that recognize
SSEA-1, a PGC cell surface antigen (6, 7). The use of fluorescent
substrates of beta-galactosidase in combination with a mouse engi-
neered to express LacZ in PGCs has also been described (8).
Currently, however, the most widely used FACS method employs
sorting of PGCs engineered to express green fluorescent protein
(GFP) (9–11).
Our lab has had reliable success using immunomagnetic sorting
based on the SSEA1 surface antigen present on pluripotent mouse
cells. Compared to GFP-based FACS sorting, immunomagnetic
sorting has the disadvantages of requiring more “hands-on” effort
and is dependent upon both primary and secondary antibodies.
An advantage of immunomagnetic purification is that PGCs can be
obtained from mouse strains lacking the GFP expression marker.
We have found the method to be compatible with PGC culture
assays (12), RNA expression analysis, and DNA methylation
analysis. It can be used to isolate PGCs from embryos between
10.5 and 14.5 dpc. Beyond 14.5 dpc the expression of the SSEA-1
epitope is reduced (13).
The method used in our lab closely resembles that first
described by Pesce and DeFelice (14).
2. Materials
2.1. Dissection 1. Two jewelers forceps (such as Dumont #5).

Requirement 2. PBS (1× without Ca2+ or Mg2+).
3. Bacterial petri dishes.
4. Stereomicroscope.
4 Immunomagnetic Purification of Murine Primordial Germ Cells 63
2.2. Immunomagnetic 1. 0.05% Trypsin–EDTA. Warm to 37 °C before use. Store at

Purification 4 °C.
2. PBS–DNase buffer: 1× PBS, pH 7.2, 5 mM EDTA, 0.5% BSA,
20 μg/ml DNase. Prepare 5 ml per purification, making fresh
each time.
3. Equilibration buffer: 1× PBS, pH 7.2, 3% BSA. Prepare fresh,
500 μl for each purification.
4. TG-1 mAb or equivalent as hybridoma cell culture supernatant
(see Note 1).
5. MiniMACS starting kit (Miltenyi Biotech). Kit includes inte-
gral components for the purification procedure: Anti-Mouse
IgM MicroBeads, MiniMACS MS separation columns, MACS
MultiStand, and MiniMACS Separation Unit.
2.3. Alkaline 1. Cytospin.

Phosphatase Staining 2. CSA-100 silanated slides.
Components
3. Poly-D-lysine dissolved at 50 mg/ml in H2O.
4. 4% Paraformaldehyde.
5. Fast Red TR Salt (Sigma) dissolved at 1 mg/ml in H2O.
6. Naphthol AS-MX phosphate (Sigma).
3. Methods
3.1. Recovery of 1. Euthanize gravid female in accordance with institutional policy

Urogenital Ridges and recover reproductive tract bearing embryos. Submerge
uterine horns in PBS in bacterial petri dishes for further
dissection.
2. Working under the buffer and with the aid of a dissecting
microscope, separate embryos from decidual tissue.
3. Use a “scissor” action of two jewelers forceps to decapitate the
embryo at a position just posterior to the forelimb buds.
4. Remove the viscera. This is most easily accomplished as fol-
lows. With the ventral surface facing up steady the embryo by
inserting the forceps tips at the base of each hindlimb bud. The
viscera may then be removed using forceps held with the other
hand. The urogenital ridges remain attached to the dorsal body
wall.
5. Dissect the urogenital ridges away from the body wall.
6. Yield per column can be increased by teasing the genital ridge
away from the more dorsally located mesenephros with the aid
of a 28 gauge syringe needle, but for many applications we find
this to be unnecessary. If desired, 12.5 dpc or older male and
female genital ridges may be pooled separately based on their

morphological differences. We find it simplest to transfer geni-
tal ridges from the dissection medium to a microfuge tube
using a 1,000 μl pipette tip.
3.2. Immunomagnetic 1. Digest approximately 8–16 urogenital ridge pairs in 0.5 ml

Purification of PGCs trypsin–EDTA at 37 °C for 5 min in a microcentrifuge tube.
Use about 15 pairs for 10.5 dpc and 10 pairs 13.5. Do not
overload the column (see Note 2).
2. Triturate (see Note 3) and then centrifuge at 250 × g for 2 min
in a microcentrifuge. Carefully remove most of trypsin–EDTA
leaving tissue clumps behind in about 50–100 μl. Triturate
again.
3. Add 1 ml PBS–DNase buffer, triturate, and centrifuge.
4. Remove the majority of the PBS–DNase, leaving approximately
160 μl behind. Triturate thoroughly to generate a single cell
suspension.
5. Place on ice and add 40 μl TG-1 mAb. Incubate on ice with
shaking for 30 min.
6. Centrifuge, aspirate supernatant, and then triturate. Add 100 μl
PBS–DNase buffer and wash two additional times, triturating
each time.
7. After final wash, resuspend pellet in 180 μl ice-cold PBS–
DNase. Add 20 μl of Anti-Mouse IgM MicroBeads. Incubate
on ice for 30 min with shaking.
8. Place MiniMacs separation column in magnetic holder and
prewash the column with 500 μl equilibration buffer (see Note 4).
Take care to avoid bubble formation on the column. Allow
column to empty by gravity.
9. Apply the cell suspension and collect flow through into a 1.5 ml
microfuge tube while the column is in the magnet. Allow col-
umn to empty by gravity.
10. Reapply the flow through to the magnet two additional
times. The third flow through is the immunodepleted frac-
tion, consisting of the somatic cells of the urogenital ridge
(see Note 5).
11. With the column still in the magnet, apply 500 μl PBS–DNase
buffer and allow column to empty by gravity. Wash this way
three additional times.
12. Elute column to obtain purified PGCs. Remove column from
magnet, apply 500 μl PBS–DNase buffer, and allow column to
empty by gravity. Collect this purified fraction.
13. Apply 1 ml PBS–DNase buffer to the column and gently force
through using the plunger supplied with the column. Collect
4 Immunomagnetic Purification of Murine Primordial Germ Cells 65
and add to original purified fraction obtained in previous step.

The cells can be concentrated by centrifugation. If assessing
purity, remove 150 μl before centrifugation.
14. Follow alkaline phosphatase staining to assess purity.
3.3. Alkaline 1. Cytospin a 10% aliquot (150 μl) onto silanated slides for
Phosphatase Staining 10 min at 55 × g.
2. Fix in 4% paraformaldehyde for 10–20 min at room tempera-
ture. Wash by gently immersing the slide into water two to
four times.
3. Blot away excess water and overlay with Fast Red TR/Napthol
AS-MX. PGCs will stain red (see Note 6).
4. Notes
1. TG-1 is a mouse IgM that was originally intended to recognize

human thymocytes (15); however, it also recognizes SSEA-1
on the surface of mouse pluripotent teratocarcinoma cells and
PGCs. If this monoclonal is not available, substitute MC-480,
available from the Developmental Studies Hybridoma Bank.
2. Overloading the column will result in very slow column clearing
times and a reduction in the level of purity. As a general rule,
one column is sufficient for about fifteen 10.5 dpc genital
ridges or eight to ten 12.5–14.5 dpc genital ridges.
3. Trituration steps should be performed by 50–100 passages
through a 200 μl pipet tip. Be careful not to suck air into the
tip and thereby introduce bubbles into the solution.
4. We use two magnets and are able to run two samples (often
male and female) simultaneously. Samples may be processed in
sequence if only one magnet is available. Samples awaiting the
magnet should be held on ice after addition of the anti-IgM
microbeads.
5. We regularly observe a very small number of PGCs in the
immunodepleted fraction.
6. Add 40 μl of 1 mg/ml Fast Red TR to 1 ml of Napthol AS-MX
phosphate. Gently layer mixture over the deposited cells at
room temperature. The stain may be very slow under these
conditions, but may be sped up by placing on a slide warmer.
If stain still does not appear after 25 min, blot away old stain-
ing solution and apply fresh.
References
1. Ewen KA, Koopman P (2010) Mouse germ embryos using FACS-gal. Dev Biol 180(2):
cell development: from specification to sex 468–472. doi:S0012160696903206 [pii]
determination. Mol Cell Endocrinol 323(1): 9. Szabo PE, Hubner K, Scholer H, Mann JR
76–93. doi:S0303-7207(09)00621-2 [pii] (2002) Allele-specific expression of imprinted
10.1016/j.mce.2009.12.013 genes in mouse migratory primordial germ
2. Gomperts M, Garcia-Castro M, Wylie C, cells. Mech Dev 115(1–2):157–160.
Heasman J (1994) Interactions between primor- doi:S0925477302000874 [pii]
dial germ cells play a role in their migration in 10. Yoshimizu T, Sugiyama N, De Felice M, Yeom
mouse embryos. Development 120(1):135–141 YI, Ohbo K, Masuko K, Obinata M, Abe K,
3. Monk M, McLaren A (1981) X-chromosome Scholer HR, Matsui Y (1999) Germline-specific
activity in foetal germ cells of the mouse. J expression of the Oct-4/green fluorescent pro-
Embryol Exp Morphol 63:75–84 tein (GFP) transgene in mice. Dev Growth
4. De Felici M, McLaren A (1982) Isolation of Differ 41(6):675–684
mouse primordial germ cells. Exp Cell Res 11. Yeom YI, Fuhrmann G, Ovitt CE, Brehm A,
142(2):476–482. doi:0014-4827(82)90393-7 Ohbo K, Gross M, Hubner K, Scholer HR
[pii] (1996) Germline regulatory element of Oct-4
5. Mayanagi T, Kurosawa R, Ohnuma K, Ueyama specific for the totipotent cycle of embryonal
A, Ito K, Takahashi J (2003) Purification of cells. Development 122(3):881–894
mouse primordial germ cells by Nycodenz. 12. Cooke JE, Godin I, Ffrench-Constant C,
Reproduction 125(5):667–675 Heasman J, Wylie CC (1993) Culture and
6. McCarrey JR, Hsu KC, Eddy EM, Klevecz RR, manipulation of primordial germ cells. Methods
Bolen JL (1987) Isolation of viable mouse pri- Enzymol 225:37–58
mordial germ cells by antibody-directed flow 13. Donovan PJ, Stott D, Cairns LA, Heasman J,
sorting. J Exp Zool 242(1):107–111. Wylie CC (1986) Migratory and postmigratory
doi:10.1002/jez.1402420116 mouse primordial germ cells behave differently
7. Yamazaki Y, Mann MR, Lee SS, Marh J, in culture. Cell 44(6):831–838. doi:0092-
McCarrey JR, Yanagimachi R, Bartolomei MS 8674(86)90005-X [pii]
(2003) Reprogramming of primordial germ 14. Pesce M, De Felici M (1995) Purification of
cells begins before migration into the genital mouse primordial germ cells by MiniMACS
ridge, making these cells inadequate donors for magnetic separation system. Dev Biol
reproductive cloning. Proc Natl Acad Sci U S A 170(2):722–725. doi:S0012-1606(85)71250-
100(21):12207–12212. doi:10.1073/pnas.2035 X [pii] 10.1006/dbio.1995.1250
119100 2035119100 [pii] 15. Beverley PC, Linch D, Delia D (1980) Isolation
8. Abe K, Hashiyama M, Macgregor G, of human haematopoietic progenitor cells using
Yamamura K (1996) Purification of primor- monoclonal antibodies. Nature 287(5780):
dial germ cells from TNAPbeta-geo mouse 332–333
Part II
Identifying Imprinted Genes

Chapter 5
Whole Genome Methylation Profiling

by Immunoprecipitation of Methylated DNA
Andrew J. Sharp
Abstract
I provide a protocol for DNA methylation profiling based on immunoprecipitation of methylated DNA
using commercially available monoclonal antibodies that specifically recognize 5-methylcytosine.
Quantification of the level of enrichment of the resulting DNA enables DNA methylation to be assayed for
any genomic locus, including entire chromosomes or genomes if appropriate microarray or high-through-
put sequencing platforms are used. In previous studies (1, 2), I have used hybridization to oligonucleotide
arrays from Roche Nimblegen Inc, which allow any genomic region of interest to be interrogated, depen-
dent on the array design. For example, using modern tiling arrays comprising millions of oligonucleotide
probes, several complete human chromosomes can be assayed at densities of one probe per 100 bp or
greater, sufficient to yield high-quality data. However, other methods such as quantitative real-time PCR
or high-throughput sequencing can be used, giving either measurement of methylation at a single locus or
across the entire genome, respectively. While the data produced by single locus assays is relatively simple
to analyze and interpret, global assays such as microarrays or high-throughput sequencing require more
complex statistical approaches in order to effectively identify regions of differential methylation, and a brief
outline of some approaches is given.
Key words: Immunoprecipitation, DNA methylation, Imprinting
1. Introduction
One feature that has been associated with many imprinted genes is
the presence of parent-of-origin-specific Differentially Methylated
Regions (DMRs). Thus, the maternal and paternal genomes pos-
sess distinct epigenetic marks which distinguish them at imprinted
loci. Here we describe a DNA immunoprecipitation method to
perform comparative DNA methylation profiling between the two
parental genomes that can detect DMRs associated with imprinted
genes. This methodology takes advantage of the fact that patients
69
70 A.J. Sharp
with uniparental disomy contain chromosomes inherited from a

single parent, representing a system that allows the independent
study of a paternally or a maternally derived epigenome. Systematic
comparison of the two parental epigenomes in this way represents
a powerful method for detecting epigenetic differences between
the two parental epigenomes associated with imprinted loci (2).
This described protocol has been optimized for hybridization
of the immunoprecipitated DNA to Nimblegen microarrays.
However, the level of enrichment for methylated DNA can also be
assayed using other methodologies such as qPCR or high-throughput
sequencing. In this case, it may be appropriate to modify the
amount of starting DNA accordingly.
2. Materials
2.1. Methylated DNA 1. Antibody: Mouse monoclonal anti 5-methyl cytidine (Diagenode,
Immunoprecipitation Liege, Belgium).
Components 2. Beads: Protein A Agarose Beads (Invitrogen, Carlsbad, CA).
3. 5× IP buffer: 50 mM Sodium Phosphate (pH 7), 0.7 M NaCl,
0.25 % Triton X-100. Total volume 100 ml. Mix 50 ml 100 mM
Na-Phosphate (pH 7), 14 ml 5 M NaCl, 2.5 ml 10 % Triton
X-100, 33.5 ml distilled H2O.
4. Digestion buffer: 50 mM Tris–HCl (pH 8), 10 mM EDTA,
0.5 % SDS. Total volume 100 ml. Mix 5 ml 1 M Tris–HCl
(pH 8), 2 ml 0.5 M EDTA, 5 ml 10 % SDS, 88 ml distilled
H2O. Filter using a 0.2 μm filter and store at 4 °C.
5. Phosphate-Buffered Saline (PBS), pH 7. Store at 4 °C.
6. Proteinase K solution (10 mg/ml). Store at −20 °C.
7. 25:24:1 Phenol:chloroform:isoamyl alcohol. Store at 4 °C.
8. 1× TE Buffer (pH 8). Store at room temperature.
9. 24:1 Chloroform:isoamyl alcohol. Store at 4 °C.
3. Methods
3.1 Methylated DNA Due to the use of overnight incubations at two points, this
Immunoprecipitation protocol is best performed over a period of 3 days.
Day 1
1. Dilute 15 μg genomic DNA (see Note 1) in 440 μl sterile H2O
in a 1.5 ml screw-top tube (see Note 2).
5 Methylation Profiling by meDIP 71
2. Fragment DNA to a size range of approximately 200–800 bp.

Fragmentation can be achieved by sonication of DNA with a
Branson 450 sonifier using a standard tapered probe (see Note
3). Program the sonifier as follows: Time = 70 s; Amplitude = 10 %;
Pulse on 0.5 s; Pulse off 0.5 s. Suspend the tube of DNA in a
polystyrene float in ice/water bath during sonication to keep
the DNA solution cool, as significant heating of the solution
occurs during sonication which can cause denaturation of the
DNA and result in a nonrandom fragmentation pattern. Clean
the sonication probe with 70 % ethanol before and after frag-
menting each DNA sample to avoid contamination between
samples.
3. Check the size of the DNA fragments produced by sonication
by loading 15 μl of the sonicated DNA (equivalent to ~300 ng)
on 1.5 % agarose gel with a 100 bp ladder alongside. Most
fragments should be between 200 and 800 bp, with an average
size of ~500 bp (see Note 4).
4. Denature the DNA for 5 min at 95 °C in a hot block, and
immediately place the samples on ice.
5. Remove 75 μl (~1.5 μg) of each DNA sample into a new tube
labeled “sample name_sonicated” and store at +4 °C. This son-
icated DNA will be used later as the reference (input) DNA,
and will be precipitated on Day 3 of the protocol.
6. Add 88 μl of 5× IP buffer to the remaining 350 μl denatured/
sonicated DNA and mix by briefly vortexing.
7. Add 10 μg of 5meth-C antibody (1 μg/μl) to each tube of
DNA/1× IP buffer (see Note 1). Ensure that the lids are
screwed on each tube tightly, and secure in a rack.
8. Incubate the tubes of DNA/IP buffer/antibody at 4 °C with
gentle rotation/rocking overnight.
Day 2
1. Prechill a micro-centrifuge to 4 °C.
2. Resuspend the Protein A Agarose beads by gently shaking the
bottle. Remove 80 μl of beads per IP reaction into an Eppendorf
tube and place on ice.
3. Make 3 ml of PBS:0.1 % BSA (2,970 μl cold PBS + 30 μl
10 mg/ml BSA) per IP reaction. Mix by vortexing and chill on
ice.
4. Wash beads twice with 1 ml cold PBS:0.1 % BSA, as follows:
(i) Centrifuge the beads for 2 min at 3,824 ´ g, 4 °C.
(ii) Remove the majority of the supernatant from each tube
with a 1 ml pipette, taking care not to disturb the beads.
The remaining supernatant can be removed using a BD
72 A.J. Sharp
UltraFine needle and syringe, which has a needle bore

smaller than the beads, such that they cannot be pipetted.
(iii) Add 1 ml of PBS:0.1 % BSA and mix well by inversion.
(iv) Repeat steps (i) and (ii), and place the washed beads on
ice.
5. Make 1× IP Buffer by diluting one part 5× IP buffer with four
parts cold sterile water, and place on ice to chill.
6. Resuspend each aliquot of the washed agarose beads in 80 μl
of chilled 1× IP Buffer.
7. Add one tube of the agarose bead/1× IP Buffer slurry to each
tube of the DNA–antibody mixture. Flick each tube gently to
ensure that the agarose beads are fully resuspended in the 1×
IP Buffer before pipetting, as the beads settle in solution.
8. Ensure that the lids are screwed on tightly to each tube of
DNA/1× IP Buffer/antibody/agarose beads, secure in a rack,
and incubate the mix for 2 h at 4 °C with gentle rotation/
rocking.
9. Transfer the mix of DNA/1× IP Buffer/antibody/agarose
beads into a new labeled 1.5 ml screw-top tube to avoid car-
ryover of contaminating un-precipitated DNA from sides of
the first tube.
10. Wash beads twice with 1 ml cold 1× IP Buffer, as follows:
(i) Centrifuge the beads for 2 min at 3,824 ´ g, 4 °C.
(ii) Remove the majority of the supernatant from each tube
with a 1 ml pipette, taking care not to disturb the beads.
Remove the remaining supernatant using a BD UltraFine
needle and syringe.
(iii) Add 1 ml cold 1× IP Buffer and mix well by inversion.
(iv) Repeat steps (i) and (ii), and place the washed beads on
ice.
11. Resuspend the DNA/antibody/beads in 250 μl digestion
buffer.
12. Add 7 μl Proteinase K solution (10 mg/ml) to each tube,
ensure that the lids are screwed on tightly, and incubate over-
night at 55 °C with rotation.
Day 3
1. Add 250 μl of sterile H2O to each tube.
2. Working in a fume hood, add 500 μl of 25:24:1 phenol:
chloroform:isoamyl alcohol to each tube.
3. Ensure that lids are screwed on tightly and vortex each tube
thoroughly for ~30 s.
4. Centrifuge tubes for 3 min at 13,000 rpm.
5. Working in a fume hood, remove the majority of the upper

aqueous phase from each tube into a new labeled 2 ml screw-top
tube, taking care not to disturb the precipitates at the interface
or the lower organic phase.
6. To maximize the recovery of DNA, working in a fume hood
add 300 μl 1× TE Buffer to each tube containing the remaining
upper aqueous phase and phenol:chloroform.
7. Ensure that the lids are screwed on tightly, vortex thoroughly,
and centrifuge for 3 min at 13,000 rpm.
8. Working in a fume hood, remove the upper aqueous phase
(avoiding the interface/lower organic phase) and add it to
the first aliquot of the aqueous phase that was removed in step
5 into a labeled 2 ml screw-top tube.
9. Perform a second phenol:chloroform extraction on the DNA
solution by repeating steps 2–8 above to ensure removal of any
residual protein from the DNA.
10. Working in a fume hood, add an equal volume of 24:1
chloroform:isoamyl alcohol to each tube containing the DNA
solution.
11. Ensure that lids are screwed on tightly, and vortex thoroughly.
12. Centrifuge tubes for 3 min at 13,000 rpm.
13. For each sample, label two 1.5 ml Eppendorf tubes with “sam-
ple name_meDIP.” Remove the aqueous (upper) phase, avoid-
ing the interface and lower organic phase, and divide equally
between the two new labeled tubes.
14. Precipitate both the tubes of IP DNA and the 75 μl sonicated
input DNA from Day 1 as follows:
(i) Add 0.7 μl glycogen (20 mg/ml), 1 ml of ice-cold 100 %
ethanol, and 50 μl of 5 M NaCl to each tube. Mix by
vortexing.
(ii) Incubate tubes at −20 °C for 1 h.
(iii) Centrifuge tubes at 13,000 rpm for 15 min at 4 °C
(see Note 5).
(iv) Carefully remove the supernatant from each tube using a
pipette, taking care not to remove the DNA pellet.
15. Wash each DNA pellet by adding 300 μl cold 70 % ethanol,
mixing by briefly vortexing.
16. Centrifuge tubes for 5 min at 13,000 rpm, 4 °C.
17. Carefully remove all the supernatant, taking care not to remove
the DNA pellet.
18. Air-dry the pellets for ~30 min until all liquid has evaporated.
19. Resuspend each DNA pellet by adding 15 μl 1× TE buffer per
tube and incubating at 65 °C for 30 min.
74 A.J. Sharp
20. Vortex each tube well, pulse spin to collect all the droplets, and
combine together the two tubes of IP DNA per sample into a
single tube.
21. Measure the DNA concentration of each sample using a
Nanodrop spectrophotometer or a similar method. The
amount of IP DNA recovered is usually ~10–20 % of the amount
of input DNA.
3.2. Measurement Immunoprecipitated DNA and the corresponding input DNA are
of Enrichment for labeled with cy3 and cy5 fluorescent dyes, hybridized to tiling oligo-
Methylated DNA nucleotide arrays, scanned, and the images analyzed to extract log2
ratios representing the relative quantity of methylated:unmethylated
DNA at each probe locus. All steps are performed according to man-
ufacturer’s protocols. Alternatively, relative amounts of IP and input
DNA can be quantified by alternative technologies, such as real-time
PCR or high-throughput sequencing.
3.3. Processing of Rigorous analysis and interpretation of large datasets produced by

Microarray Data to hybridization of IP and input DNA to oligonucleotide arrays
Identify Regions of require some programming and/or statistical knowledge, such as
Differential Methylation use of the Bioconductor project (3). Below I outline a framework
Between Samples for the analysis of data from Nimblegen microarrays with a median
density of one probe per 100 bp across entire chromosomes that
can be used to identify regions of differential methylation between
samples. The exact thresholds and analysis approaches used in any
such global analysis will vary depending on the nature of the under-
lying data, specifics of the technical platform used, number and
nature of the samples assayed, underlying biological question, and
potentially many other factors specific to each experiment. As a
result, the outline below should be treated as a set of guidelines
that should be modified to suit each specific situation.
1. Due to technical variations between samples that may be caused
by a variety of factors, including (but not limited to) variable
efficiency of immunoprecipitation reactions, sample labeling,
or hybridization, it is usually necessary to perform normaliza-
tion across all samples within an experiment to try and remove
systematic sample-to-sample biases that might otherwise result
in significant artifacts when comparing different individuals.
A variety of normalization approaches are available, but I have
found quantile normalization (4) to be effective (Fig. 1).
Following quantile normalization, the mean and standard
deviation of the data distributions in each sample are identical,
allowing relatively unbiased comparison of log2 ratios across
different samples.
2. A small fraction of probes on any microarray perform poorly,
yielding highly unreliable data. These low-quality data points
may represent probes that are inherently unreliable due to their
sequence characteristics (5), or, for example, may be located
Fig. 1. Transformation of microarray hybridization data by quantile normalization allows unbiased comparison across arrays.
(Left panel) Density plot showing the varying distribution of raw log2 ratios in six individual hybridizations. Due to these differ-
ing distributions, comparisons across samples using these raw data would result in the detection of many differences that
are likely artifacts resulting from the inherently different underlying data distributions. (Middle panel) Raw data from six
individual hybridizations was transformed by quantile normalization to remove sample-specific biases resulting from differ-
ences in antibody enrichment, labeling, or hybridization. (Right panel) After quantile normalization, the six datasets show
identical distributions, allowing unbiased comparison across samples to identify differentially methylated regions.
on a section of the microarray surface containing hybridization

or scanning artifacts (e.g., dust). Given that with high-density
tiling arrays the probe spacing throughout the genome (~1 per
100 bp) is significantly smaller than the size of DNA fragments
being hybridized to the array (~500 bp), it is expected that
closely spaced probes will show somewhat correlated log2 val-
ues. Based on this expectation, low-quality data points can be
identified by implementing a sliding window analysis that
identifies probes that deviate significantly from the log2 value
of their immediate neighbors (outlier data points). I have
found that an effective approach is to use a sliding window to
identify all clusters of five consecutive probes which span a
physical distance of 1 kb or less. For each group of probes, if
the difference in log2 values between the central probe and the
median value of the five probes in that cluster exceeds the
interquartile range of log2 values on that entire chromosome,
it is flagged as an outlier. Rather than completely removing
outliers, which would result in loss of data, an alternative
approach is to replace them with the median log2 ratio of the
remaining four probes in the group. Based on these criteria,
the log2 values of approximately 2–4 % of probes per array are
replaced.
Overall, these normalization and filtering steps resulted in
significant noise reduction and improvements to data quality.
For example, in one prior study in which six samples were
tested in duplicate, the mean correlation between log2 ratios in
technical replicate hybridizations for the six individuals tested
increased from 0.83 in the raw data to 0.93 after quantile
normalization and outlier replacement (2) (Fig. 2).
76 A.J. Sharp
Fig. 2. Effects of outlier probe replacement on methylation profiles. The image shows a screenshot of probe log2 ratios in a
25 kb region of chromosome 15 from one array hybridization. The top track shows the raw data, while the lower panel
shows the same data after replacement of outlier probes (dotted ellipses). This step can significantly reduce noise caused
by poor-performing probes on the array.
3. Data can be further treated by application of a linear smoothing

function (6), which acts to reduce probe-to-probe variation.
4. Before performing any analysis to compare between samples to
detect regions of differential methylation, it is often useful to
remove probes that show very low variance between samples.
Probes that are inherently invariant have very low power to
detect differences between samples, and are not therefore use-
ful to include in downstream analysis. In fact, removal of invari-
ant probes from the dataset prior to performing formal
statistical testing can actually increase the overall statistical
power of an analysis by reducing the burden of multiple testing
correction that needs to be performed in any microarray analy-
sis. Generally a simple filter to remove any probe that shows
low variance across a population (e.g., standard deviation <0.2)
is appropriate.
5. To identify probes that show potential differential methylation
between samples, significance testing can be performed, using,
e.g., a Student’s t-test. Given that most microarray analyses
will involve large numbers of probes, it is also necessary to
apply a multiple testing correction to any p values generated in
these testes to avoid large numbers of false positive associations
(e.g., False discovery rate correction) (7).
6. Given that data from single probes is often unreliable and con-
sideration of single probe events will often result in a high
false positive rate, differentially methylated regions of higher
confidence can be identified by searching for clusters of >1
closely spaced probes that exceed a given threshold of significance.
In a previous study (2), use of a sliding window analysis to iden-
tify clusters of five probes separated by <1 kb which exceeded a
relatively low-stringency statistical threshold was successful in
identifying regions of differential methylation (the central probe
shows an FDR-adjusted p < 0.1, and at least two of the four
neighboring probes show an unadjusted p < 0.05).
4. Notes
1. The amount of starting DNA can be varied depending on the

amount of immunoprecipitated material that is required for
later quantification by array hybridization, sequencing, or qPCR.
However in this case it is necessary to also vary the amount of
antibody used in step 7 of Day 1 of the protocol in order to
retain the same 3:2 ratio of DNA:antibody. As a rule, generally
the final yield of the immunoprecipitation reaction is approxi-
mately 10–15 % of the starting amount.
2. The use of screw-top tubes is strongly recommended where
indicated, as there are multiple steps in this protocol where
standard snap-top Eppendorf tubes can leak, resulting in the
loss of DNA and the release of organic solvents.
3. Other random DNA fragmentation methods can be used, such
as Covaris DNA shearing or chemical cleavage methods.
4. In some cases, after sonication the DNA sample may have a
larger size range than anticipated. In this case, additional frag-
mentation can be performed to reduce the size to an accept-
able range (200–500 bp).
5. The final DNA pellets are generally very small, and are some-
times barely visible to the naked eye. It is recommended to rotate
each Eppendorf tube so that the hinges are on the outer edge of
the centrifuge rotor. If this is done, even if the precipitated DNA
pellet is invisible after centrifugation, its location in the tube will
always be below the hinge, making it easier to subsequently
remove the supernatant without losing the DNA pellet.
Acknowledgements
This work was supported by funding from the Fondation Jerome

LeJeune.
References
1. Sharp AJ, Stathaki E, Migliavacca E, Antonarakis SE (2010) Methylation profiling
Brahmachary M, Montgomery S, Dupre Y, in cases with uniparental disomy identifies
Antonarakis SE (2011) DNA methylation novel differentially methylated regions on
profiles of human active and inactive X chro- chromosome 15. Genome Res 20:1271–1278
mosomes. Genome Res 21(10):1592–1600 3. Gentleman RC, Carey VJ, Bates DM, Bolstad
2. Sharp AJ, Migliavacca E, Dupre Y, Stathaki E, B, Dettling M, Dudoit S, Ellis B, Gautier L, Ge
Sailani MR, Mackay D, Robinson DO, Cobellis Y, Gentry J, Hornik K, Hothorn T, Huber W,
G, Cobellis L, Brunner H, Steiner B, Iacus S, Irizarry R, Leisch F, Li C, Maechler M,
78 A.J. Sharp
Rossini AJ, Sawitzki G, Smith C, Smyth G, of DNA copy-number. Hum Mol Genet 16:
Tierney L, Yang JY, Zhang J (2004) 2770–2779
Bioconductor: open software development for 6. Pelizzola M, Koga Y, Urban AE, Krauthammer M,
computational biology and bioinformatics. Weissman S, Halaban R, Molinaro AM
Genome Biol 5:R80 (2008) MEDME: an experimental and analyti-
4. Bolstad BM, Irizarry RA, Astrand M, Speed TP cal methodology for the estimation of DNA
(2003) A comparison of normalization meth- methylation levels based on microarray
ods for high density oligonucleotide array data derived MeDIP-enrichment. Genome Res 18:
based on bias and variance. Bioinformatics 1652–1659
19:185–193 7. Benjamini Y, Hochberg Y (1995) Controlling
5. Sharp AJ, Itsara A, Cheng Z, Alkan C, the false discovery rate: a practical and power-
Schwartz S, Eichler EE (2007) Optimal design ful approach to multiple testing. J R Stat Soc B
of oligonucleotide microarrays for measurement 57:289–300
Chapter 6
Identification of Imprinted Loci by Transcriptome

Sequencing
Tomas Babak
Abstract
Enabled by high-throughput technologies that are capable of generating millions of sequencing reads,
transcriptome sequencing is emerging as an important approach for mapping allelic imbalance (AI), where
transcription is biased toward one allele in a diploid system. AI is identified by counting sequencing reads
that map to genomic regions containing heterozygous SNPs, where the base identity of the SNP is used
to distinguish allelic origin. Genomic imprinting is a special case of AI where bias is toward parental sex
and can be identified by transcriptome sequencing of systems that represent reciprocally inherited loci. The
focus of this protocol is on experimental design, analysis, and interpretation of genomic imprint discovery
using whole transcriptome sequencing.
Key words: Imprinting, Transcriptome sequencing, Allelic imbalance
1. Introduction
While a comprehensive map of imprinted loci in all cell/tissue

types of all mammals facilitates the evolutionary characterization of
genomic imprinting, inherent challenges of traditional discovery
approaches have mostly limited their application to developing
mouse stages. Classical genetic screens based on uniparental diso-
mies and reciprocal translocations (1, 2) and genetic mapping of
parent-of-origin phenotypes in humans (e.g., ref. 3) revealed the
initial imprinted loci. Most imprinted genes emerged from fine
mapping of these initial regions, typically by RFLP analysis of cDNA.
Microarray profiling of embryos with uniparental disomies or
entire genomes (4, 5) extended the map, but embryonic lethality
induced by these genetic perturbations has limited their discovery
potential. A two-dimensional RFLP approach (6) and genotyping
microarrays (7) have been applied to imprint discovery in adult/
wild-type tissues but require extensive analysis to rule out signal
79
80 T. Babak
from noise. Mapping imprinting by transcriptome sequencing,

which has only recently become possible, is advantageous in that it
does not require a priori knowledge of the expected genomic local-
ization or phenotype, and can be applied to any diploid progeny of
genetically diverged parents.
NextGen sequencing (NGS) has rapidly changed the landscape
of nucleic acid-driven research. In addition to standard sequence
determination, the ability to generate millions of sequencing reads
has also enabled quantification of input abundance by simply
counting the reads. RNA-Seq (8, 9), the most commonly utilized
form of transcriptome sequencing, is based on random priming of
polyA + purified RNA to generate cDNA, which is again random
primed to generate double-stranded DNA that is then sequenced.
Biological inference typically involves mapping of the reads to a
reference genome and quantifying events (e.g., splicing or gene
expression) by counting. Similarly, when a mapped sequencing
read overlaps a heterozygous SNP, allelic origin of that read can be
discerned using the identity of the polymorphic base. AI can be
inferred by comparing the two sums of allelic reads that map to
that SNP. Since AI is widespread in mammalian cells (10) and
genomic imprinting is likely to be causal for only a small propor-
tion (11), demonstrating AI in a reciprocally inherited fashion is
imperative for imprinting discovery. The most straightforward
experimental design consists of two crosses of inbred strains, where
each parental sex is represented by both genetic backgrounds. In
this simple system each progeny is heterozygous at all SNPs and
both maternal and paternal copies of each allele are represented
(Fig. 1). RNA-Seq is then applied to both crosses and genomic
imprinting distinguished as AI that is consistently biased toward
parental sex (vs. AI biased toward same parental strain which is
much more common). Although the classical definition of genomic
imprinting is based on monoallelic expression, it has been expanded
Inbred Strain 1
Inbred Strain 2
F1 RNAseq F1
Fig. 1. Schematic of reciprocal cross. Inbred parental strains are crossed in reciprocal to produce F1 progeny that are
sequenced and analyzed in pairs.
6 Identification of Imprinted Loci by Transcriptome Sequencing 81
to include incomplete parent-of-origin expression biases (11) and

enables identification of cell/tissue-specific imprinting that is
diluted by tissue heterogeneity.
Several groups have successfully used transcriptome sequencing
for imprint discovery (12–14) and the approach has outstanding
promise, but unresolved challenges exist. For example, all pub-
lished studies using RNA-Seq to map AI have thus far not addressed
systematic bias. It has been shown that AI reproduces very well
across technical and biological replicates when considering the
same SNP (15) but little has been done to assess concordance
across different SNPs. Systematic priming biases induced by SNPs
could lead to incorrect AI calls. One would expect strong agreement
on AI among two SNPs within the same exon, for example, and
these could be used to empirically estimate systematic bias. Until
AI is robustly modeled, mock reciprocal crosses (i.e., biological
replicates) can be used to gauge false-discovery and I expand on
this below. A second challenge is application outside the somewhat
artificial setting of inbred mouse crosses. Transcriptome sequenc-
ing has not yet been applied for imprint discovery in species where
inbred strains or controlled crosses are not available (such as
humans). Two experimental designs could be employed to make
this feasible. The first is a controlled screen in a family where par-
ents and offspring are genotyped; AI is determined at all heterozy-
gous SNPs in offspring, and parent-of-origin inheritance determined
from phased haplotypes. A large family with distantly related parents
would be ideal. The second approach is to screen an outbred popu-
lation and identify imprinting as AI with no sequence dependence
(i.e., AI is always observed but biased for either allele with equal
likelihood) since consistent bias toward one allele suggests a genetic
mechanism and this is generally the case (10).
The focus of this protocol is on identification of genomic
imprinting in reciprocally crossed F1 mouse strains using standard
RNA-Seq with emphasis on analysis practices. The assumption is
that the reader has access to total RNA from reciprocally crossed
mouse tissues and a reasonable (>5 million) map of SNPs for these
strains. The scope of the approach could be expanded to any diploid
species and additional recommendations for application outside
inbred mouse strains are also included.
2. Materials
2.1. Constructed Construction of RNA-Seq libraries was first described in yeast and
RNA-Seq Libraries mouse (8, 16) and is now available in kit format from several
manufacturers. In my experience the standard RNA-Seq kit sold by
Illumina works very well and the end result is a library of high
complexity (measured by the proportion of sequencing reads that
82 T. Babak
align to unique genomic locations). The TruSeq RNA kit (Illumina),

which is designed for higher sample throughput, works well and
I have seen great data from as little as 100 ng total RNA input. A few
practical changes involving deoxy-uridine triphosphate (dUTP)
and uracil-N-glycosylase (UNG) (17, 18) effectively introduce
strand specificity and are recommended since imprinted antisense
transcription is known to exist. Library complexity and even cover-
age are essential for measuring AI and some of the low-input kits
suffer in this regard because they have multiple series of amplification.
NSR-seq (not-so-random primer sequencing) (19) is one of the
earlier approaches that was applied for identifying AI (12), and has
the advantage of capturing non-polyadenylated transcripts and is
also strand specific, but personal experience and a recent evaluation
(17) have revealed undesirable evenness of coverage (i.e., coverage
is “spiky”). In summary, any approach that quantitatively captures
input transcript abundance and yields a library of high complexity
will work and I have seen excellent data suitable for mapping AI
from libraries made with mRNAseq/TruSeq RNA kits purchased
from Illumina modified with dUTP/UNG treatment (18) to
achieve strand specificity.
2.2. NextGen 454, SOLiD, and Illumina are currently the major suppliers of
Sequencing Capacity NGS sequencers. Any of these platforms and likely many other
emerging platforms will work, although Illumina and SOLiD are
currently the only commercially available RNA-Seq platforms for
generating tens to hundreds of millions of reads. Overall sequenc-
ing depth is dependent on the length and number of sequencing
reads and the heterozygous SNP density of the system. Methods
exist to estimate the minimum required sequencing (20) and more
will always improve sensitivity. In practice, 4 Gb of single-end
RNA-Seq data from reciprocally crossed C57BlxCAST samples
(i.e., 8 Gb total data, 4 Gb from each cross) is sufficient to
confidently identify >90% of previously validated imprints in that
tissue. 2 Gb will result in slightly lower performance (70–80% sen-
sitivity at the same detection threshold) and even 1 Gb will yield
acceptable results (~60% sensitivity). The ideal read length is a
trade-off between molecular complexity (long reads and PE reads
limit the number of molecules represented in the library) and
sequencing of SNPs. The ideal read length would on average capture
1 SNP/read and can be estimated using a published model (20).
Considering practical challenges I recommend using single-end
75–100 bp reads. Paired-end (PE) data improves mapping perfor-
mance but only marginally. With a mean RNA-Seq insert size of
~200 bp, the 3¢ ends of pairs can overlap which leads to diminish-
ing returns. Reads shorter than 50 bp are not recommended since
this will lead to significant mapping bias (see Note 9).
2.3. Computational 1. Access to Linux/Unix working environment with at least 6 Gb

Resources RAM.
2. Installation of Novoalign v2.07.11 or newer (21) or equivalent
short read sequence aligner. V2.07.11 has capability of report-
ing mismatches to masked bases (Ns).
3. Reference genome. Most sequenced mammalian genomes can
be downloaded from UCSC (22). If the genome for the spe-
cies is not available, reads could be assembled into transcript
models that then serve as a reference, but this is beyond the
scope of this protocol.
4. Map of SNPs. 15 mouse strains were recently sequenced by the
Sanger Institute and SNP maps are available for download
(23). If working with a system that has a reference genome
but where SNPs are unknown, genotyping arrays or genome
sequencing can be used to map heterozygous SNPs. In humans
additional SNPs can be imputed and phased using MaCH (24)
to improve sensitivity of AI mapping. SNPs can also be inferred
from the RNA-Seq data. This does not work well for mapping
AI since discovery favors SNPs biased in expression toward the
non-reference allele and thus the resulting AI profile becomes
artificially skewed toward non-reference alleles. However, the
approach can be effective for imprint discovery by calling SNPs
on pooled (in equal amount) sequencing data from the recip-
rocal crosses where imprinted SNPs are supported by near
50:50 proportions. SAMtools (25), GATK (26), or soapSNP
(27) can all be used to identify SNPs from mapped RNA-Seq
data and are comparable in performance.
5. Perl/Python installation or equivalent for custom manipula-
tion of data.
6. Matlab, R, Excel, or equivalent for visualizing results.
3. Methods
1. Generate a masked version of genome where known SNPs are

replaced with Ns (see Note 1).
2. Index genome by running Novoindex (default options and -k
14 -s 3) on a single fasta file of masked genome (see Note 2).
3. Align fastq-formatted raw reads files (paired-end or single-end)
against masked genome using Novoalign (default options and
-a -o IUBMatch -r None, and -i 0 1000 if aligning PE reads)
(see Note 3). A summary of the alignment approach is shown
in Fig. 2.
84 T. Babak
FA S TQ file
N ovoalign/S im ilar
S plice Junctions G enom e Transcripts
C ore
O ptional M atch N o m atch
Unique match
R edundant N o m atch
to genome
Alignment
Allelic Counts Allelic Counts
Summary
(at each SNP) (summed/locus)
File
SNP
Fig. 2. Alignment, SNP-identification, AI-quantification pipeline. Alignment is accomplished with an independent algorithm
(e.g., Novoalign (21)) against the genome, and optionally splice junctions and full-length transcripts. Unique matches (in
the genome) are retained and used for SNP prediction and quantification of ASE.
4. If improved alignment sensitivity is desired, also align reads to

splice junctions and full-length transcripts (especially useful if
aligning paired-end reads) using Novoalign (same settings as
above except -r All 50) (see Note 4). Convert transcript align-
ment coordinates to genomic coordinates using the transcript
genomic coordinates (see Note 5).
5. Discard redundant alignments (where reads map to more than
one genomic location) and generate report files that store
alignment coordinates and genomic mismatches for each read
(see Note 6).
6. For each SNP, tally the number of reads that support the refer-
ence and alternate bases (see Note 6).
7. At each SNP, let A represent the number of reference-specific
reads and B the number of alternate allele-specific reads.
Quantify the degree of AI as A/(A + B). The probability of AI
can be estimated using the cumulative binomial distribution.
This can also be done in Excel where binomial-p (probability
of no AI) = binomdist(min(A, B), A + B, 0.5, 1). In Matlab the
binomcdf function from the statistics package can be called.
The same principles can also be used on all allele-specific reads

summed across a transcript (i.e., sum over all SNPs within the
transcript) (see Note 7).
8. Genomic imprinting requires AI to be measured in tissue-matched
reciprocally crossed samples. If s1 = sample 1 and s2 = reciprocal
sample, Genomic imprinting may exist if (AIs1 > 0.5 and AIs2 < 0.5)
or if (AIs1 < 0.5 and AIs2 > 0.5), i.e., reciprocal bias exists. The
probability of imprinting can be estimated as the less significant
binomial estimate of the two samples (see Note 8).
9. Select a suitable threshold of significance for calling imprint-
ing by using a mock reciprocal cross as a negative control
(see Note 9).
4. Notes
1. SNP maps can be downloaded from Sanger (23) and masking

greatly reduces alignment biases (28).
2. This step creates an .idx file that is used as input for genome
alignment. Junction and transcript indices (step 4) can be made
with the same settings.
3. Over a dozen short-read alignment algorithms are currently
available. BWA (29), SOAP2 (30), and Bowtie (31) are based
on the Burrows–Wheeler Transform (BWT) algorithm and
are by far the fastest aligners with sensitivity and specificity
comparable or better than most. However, in testing these
and eight other popular aligners on simulated single-end and
paired-end data (with imputed mismatches representative of
quality scores and expected variation), Novoalign (21)
attained the highest sensitivity (7–8% higher sensitivity than
BWT approaches with avg. alignment rate of 87% vs. 79–80%)
and comparable specificity (<0.1% erroneous alignments) to
all aligners. Not surprisingly, the AI profile was less biased
toward reference alleles than for other approaches tested
owing to a better ability to align over SNPs. -a will trim
adapter sequences, -o IUBMatch will report N > (ACGT)
base changes, -r None will not report reads that align in more
than one genomic regions, and -i 0 1000 will allow pairs to
match up to 1,000 bp apart.
4. Extensive custom scripting will be required to perform this
step and there is more than one way to compile a reference
transcriptome. I made a splice junction coordinate file from all
possible exon skipping events (up to two exons skipped) from
RefSeq, ENSEMBL, UCSC known gene, and Genbank mRNA
BED files downloaded from UCSC (22). I then sorted to
86 T. Babak
remove duplicate entries (sort -k 6,6 -k 1,1 -k 2,2n -k 3,3n -u

unsorted_with_6_columns.bed > sorted_unique_junctions.
bed) and retrieved the fasta equivalent from UCSC Table
Browser and indexed using Novoindex (-k 14 -s 3). Aligning
paired-end reads to transcripts will considerably improve the
number of reads that align as pairs. I again recommend RefSeq,
ENSEMBL, UCSC KG, and Genbank mRNAs as a compre-
hensive transcript set. It is important to allow reporting of all
matches since -r None will ignore matches to multiple isoforms
which will be most of them (i.e., use -r All 50). Redundant
filtering (step 5) done in genomic space removes truly redun-
dant matches. BED files for junctions and transcripts can be
used to convert alignments back into genomic coordinates.
5. If mapping paired-end reads, a paired match takes precedence
over single matches (i.e., if maps as a pair once take that align-
ment and disregard all others). At this point reads that do not
contain N > (ACGT) changes can be discarded if further SNP
discovery will not be done.
6. Reads mapping to opposite strands should be tallied indepen-
dently if strand-specific RNA-Seq was used (i.e., each SNP may
have up to two sets of counts).
7. The cumulative binomial distribution models the maximum
number of successes in a sequence of independent binary
events, each of which yields success with some probability. For
example, the chance of getting three or fewer heads when
flipping a fair coin ten times is 17.2%. Summing reads across
SNPs violates the binomial assumption when a single sequenc-
ing read spans more than one SNP since it expects all counts to
be independent. Ideally, a read (whether single-end or paired-
end) should only be counted once. An ad hoc approach to
ensure that this is the case is to only consider SNPs that are
further apart than the read length (fragment length if using
paired-reads). In practice, the extent of systematic error in
measuring AI with RNA-Seq contributes significantly more
uncertainty in the binomial calculation than violating counting
independence as described. Negative controls are imperative
for estimating false-discovery (see step 9).
8. A suitable threshold for making an imprint call depends on the
extent of acceptable false-discovery (i.e., proportion of calls
that are not truly imprinted; see step 9) and will vary from
sample to sample and with the selected RNA-Seq protocol.
In practice, a binomial p-value of 0.001 results in a false-
discovery rate (FDR) of ~10% using standard Illumina
mRNAseq.
9. For this control to be valid, samples need to be prepared com-
pletely in parallel, they must be sequenced to equivalent depths,
and all must pass quality control criteria. The FDR can be
estimated by plotting the number of imprinted sites as a

function of binomial-p cutoff from data generated from bio-
logical replicates. Since there is no genuine reciprocal inheri-
tance of any allele in this scheme, all calls are false-positives and
their rate will translate to a genuine cross if all samples are
sequenced to an equal depth. Random removal of reads should
be done to ensure that all samples have the same number of
input reads. A similar plot for a genuine reciprocal cross can be
used to estimate sensitivity (number of known imprinted sites
detected) and by combining the data into a plot of FDR vs.
sensitivity a useful threshold for making imprinting calls can be
selected. The same criteria can be applied to AI inferred from
reads summed across SNPs in the same transcript.
References
1. Cattanach BM, Kirk M (1985) Differential 8. Mortazavi A, Williams BA, McCue K, Schaeffer
activity of maternally and paternally derived L, Wold B (2008) Mapping and quantifying
chromosome regions in mice. Nature 315: mammalian transcriptomes by RNA-Seq. Nat
496–498 Methods 5:621–628
2. Surani MA, Reik W, Allen ND (1988) 9. http://www.illumina.com
Transgenes as molecular probes for genomic 10. Pickrell JK, Marioni JC, Pai AA, Degner JF,
imprinting. Trends Genet 4:59–62 Engelhardt BE, Nkadori E, Veyrieras JB,
3. Nicholls RD, Knoll JH, Butler MG, Karam S, Stephens M, Gilad Y, Pritchard JK (2010)
Lalande M (1989) Genetic imprinting sug- Understanding mechanisms underlying human
gested by maternal heterodisomy in nondele- gene expression variation with RNA sequenc-
tion Prader-Willi syndrome. Nature 342: ing. Nature 464:768–772
281–285 11. Morison IM, Ramsay JP, Spencer HG (2005)
4. Choi JD, Underkoffler LA, Collins JN, A census of mammalian imprinting. Trends
Marchegiani SM, Terry NA, Beechey CV, Genet 21:457–465
Oakey RJ (2001) Microarray expression 12. Babak T, Deveale B, Armour C, Raymond C,
profiling of tissues from mice with uniparental Cleary MA, van der Kooy D, Johnson JM, Lim
duplications of chromosomes 7 and 11 to iden- LP (2008) Global survey of genomic imprint-
tify imprinted genes. Mamm Genome 12: ing by transcriptome sequencing. Curr Biol
758–764 18:1735–1741
5. Mizuno Y, Sotomaru Y, Katsuzawa Y, Kono T, 13. Gregg C, Zhang J, Weissbourd B, Luo S,
Meguro M, Oshimura M, Kawai J, Tomaru Y, Schroth GP, Haig D, Dulac C (2010) High-
Kiyosawa H, Nikaido I, Amanuma H, resolution analysis of parent-of-origin allelic
Hayashizaki Y, Okazaki Y (2002) Asb4, Ata3, expression in the mouse brain. Science (New
and Dcn are novel imprinted genes identified York, NY) 329:643–648
by high-throughput screening using RIKEN 14. Wang X, Sun Q, McGrath SD, Mardis ER,
cDNA microarray. Biochem Biophys Res Soloway PD, Clark AG (2008) Transcriptome-
Commun 290:1499–1505 wide identification of novel imprinted genes in
6. Plass C, Shibata H, Kalcheva I, Mullins L, neonatal mouse brain. PLoS One 3:e3839
Kotelevtseva N, Mullins J, Kato R, Sasaki H, 15. Babak T, Garrett-Engele P, Armour CD,
Hirotsune S, Okazaki Y, Held WA, Hayashizaki Raymond CK, Keller MP, Chen R, Rohl CA,
Y, Chapman VM (1996) Identification of Grf1 Johnson JM, Attie AD, Fraser HB, Schadt EE
on mouse chromosome 9 as an imprinted gene (2010) Genetic validation of whole-transcrip-
by RLGS-M. Nat Genet 14:106–109 tome sequencing for mapping expression
7. Morcos L, Ge B, Koka V, Lam KC, Pokholok affected by cis-regulatory variation. BMC
DK, Gunderson KL, Montpetit A, Verlaan DJ, Genomics 11:473
Pastinen T (2011) Genome-wide assessment of 16. Nagalakshmi U, Wang Z, Waern K, Shou C,
imprinted expression in human cells. Genome Raha D, Gerstein M, Snyder M (2008) The
Biol 12:R25 transcriptional landscape of the yeast genome
88 T. Babak
defined by RNA sequencing. Science (New data to estimate haplotypes and unobserved
York, NY) 320:1344–1349 genotypes. Genet Epidemiol 34:816–834
17. Levin JZ, Yassour M, Adiconis X, Nusbaum C, 25. Li H, Handsaker B, Wysoker A, Fennell T,
Thompson DA, Friedman N, Gnirke A, Regev Ruan J, Homer N, Marth G, Abecasis G,
A (2010) Comprehensive comparative analysis Durbin R (2009) The Sequence Alignment/
of strand-specific RNA sequencing methods. Map format and SAMtools. Bioinformatics
Nat Methods 7:709–715 (Oxford, England) 25:2078–2079
18. Parkhomchuk D, Borodina T, Amstislavskiy V, 26. McKenna A, Hanna M, Banks E, Sivachenko
Banaru M, Hallen L, Krobitsch S, Lehrach H, A, Cibulskis K, Kernytsky A, Garimella K,
Soldatov A (2009) Transcriptome analysis by Altshuler D, Gabriel S, Daly M, DePristo MA
strand-specific sequencing of complementary (2010) The genome analysis toolkit: a
DNA. Nucleic Acids Res 37:e123 MapReduce framework for analyzing next-gen-
19. Armour CD, Castle JC, Chen R, Babak T, eration DNA sequencing data. Genome Res
Loerch P, Jackson S, Shah JK, Dey J, Rohl CA, 20:1297–1303
Johnson JM, Raymond CK (2009) Digital 27. Li R, Li Y, Fang X, Yang H, Wang J, Kristiansen
transcriptome profiling using selective hexamer K, Wang J (2009) SNP detection for massively
priming for cDNA synthesis. Nat Methods parallel whole-genome resequencing. Genome
6:647–649 Res 19:1124–1132
20. Fontanillas P, Landry CR, Wittkopp PJ, Russ 28. Degner JF, Marioni JC, Pai AA, Pickrell JK,
C, Gruber JD, Nusbaum C, Hartl DL (2010) Nkadori E, Gilad Y, Pritchard JK (2009) Effect
Key considerations for measuring allelic expres- of read-mapping biases on detecting allele-
sion on a genomic scale using high-throughput specific expression from RNA-sequencing
sequencing. Mol Ecol 19(Suppl 1):212–227 data. Bioinformatics (Oxford, England) 25:
21. http://www.novocraft.com/main/index.php 3207–3212
22. Rhead B, Karolchik D, Kuhn RM, Hinrichs AS, 29. Li H, Durbin R (2009) Fast and accurate short
Zweig AS, Fujita PA, Diekhans M, Smith KE, read alignment with Burrows-Wheeler trans-
Rosenbloom KR, Raney BJ, Pohl A, Pheasant form. Bioinformatics (Oxford, England)
M, Meyer LR, Learned K, Hsu F, Hillman- 25:1754–1760
Jackson J, Harte RA, Giardine B, Dreszer TR, 30. Li R, Yu C, Li Y, Lam TW, Yiu SM, Kristiansen
Clawson H, Barber GP, Haussler D, Kent WJ K, Wang J (2009) SOAP2: an improved ultra-
(2010) The UCSC Genome Browser database: fast tool for short read alignment. Bioinformatics
update. Nucleic Acids Res 38:D613–D619 (Oxford, England) 25:1966–1967
23. http://www.sanger.ac.uk/resources/mouse/ 31. Langmead B, Trapnell C, Pop M, Salzberg SL
genomes/ (2009) Ultrafast and memory-efficient align-
24. Li Y, Willer CJ, Ding J, Scheet P, Abecasis GR ment of short DNA sequences to the human
(2010) MaCH: using sequence and genotype genome. Genome Biol 10:R25
Chapter 7
Data Mining as a Discovery Tool for Imprinted Genes

Chelsea Brideau and Paul Soloway
Abstract
This chapter serves as an introduction to the collection of genome-wide sequence and epigenomic data, as
well as the use of these data in training generalized linear models (glm) to predicted imprinted status. This
is meant to be an introduction to the method, so only the most straightforward examples will be covered.
For instance, the examples given below refer to 11 classes of genomic regions (the entire gene body,
introns, exons, 5¢ UTR, 3¢ UTR, and 1, 10, and 100 kb upstream and downstream of each gene). One
could also build models based on combinations of these regions. Likewise, models could be built on com-
binations of epigenetic features, or on combinations of both genomic regions and epigenetic features.
This chapter relies heavily on computational methods, including basic programming. However, this
chapter is not meant to be an introduction to programming. Throughout the chapter, the reader will be
provided with example code in the Perl programming language.
Key words: Epigenetics, Epigenomics, Imprinting, Data mining, Bioinformatics, Generalized linear
model
1. Introduction
Genomic imprinting refers to genes that are expressed from one of

the two parental alleles in a parent-of-origin-specific manner. Until
recently, about 100 mouse imprinted genes had been identified, with
many more genes predicted to be imprinted (http://igc.otago.ac.nz/
home.html) (1, 2). However, application of new methods, such as
whole transcriptome sequencing and computational prediction,
has identified additional imprinted genes (3–7).
The identification of novel imprinted genes has become
increasingly important with the realization that imprinting
defects are associated with a variety of complex disorders, such as
obesity, diabetes, and schizophrenia (8–11). Given the impor-
tance that imprinted genes play in human health, several studies
have attempted genome-wide identification of imprinted genes
89
90 C. Brideau and P. Soloway
(1–6, 12–24). These have done so mainly using experimental

methods, with some success.
The first studies identified loci with allele-specific DNA methy-
lation, one hallmark of imprinted genes, applying methods such as
Restriction Landmark Genome Scanning (RLGS) to DNAs from
progeny of interstrain reciprocal crosses (18, 25–28). In this
method, DNA is cut with a methylation-sensitive restriction
enzyme, followed by radioactive end-labeling. Then, the radioac-
tive DNA fragments are digested with a second restriction enzyme
and run on the first dimension of a two-dimensional agarose gel.
The DNA fragments are then digested in the gel with a third
restriction enzyme, and the second dimension of the two-dimen-
sional gel is run. After exposing the gel to film, a pattern of spots is
visible and can be compared between reciprocal crosses to deter-
mine whether there are any potential differences in methylation
between the two alleles. One of the major drawbacks of the RLGS
method was the low throughput, as the genomic region associated
with each spot needed to be cloned and identified. This issue has
been resolved, since all spots have been cloned and identified.
However, this is a low-resolution method, as there are a limited
number of sites that can be queried using this method. Reduced
representational bisulfite sequencing (29) or whole genome
bisulfite sequencing (30) of DNAs from progeny of reciprocal
crosses can also be used to comprehensively identify sites of allele-
specific DNA methylation.
In the past decade, methods for genome-wide identification of
imprinted genes have increased dramatically in terms of through-
put. One large-scale study identified candidate imprinted transcripts
in the mouse genome by expression profiling of cDNA clones.
cDNA microarrays were used to detect differential expression by
comparing mRNA levels in the P9.5 gynogenetic and androge-
netic mouse embryos (1). Of the ~28,000 FANTOM2 transcripts
analyzed, ~2,000 were identified as imprinted candidates.
Interestingly, 39 of the 2,000 transcripts mapped to known
imprinted regions of the mouse genome, while 56 were ncRNAs,
and 159 were antisense transcripts. Experimental validation of two
transcripts located in the Prader–Willi syndrome region identified
these transcripts as imprinted, indicating that allele-specific array-
based methods are useful for large-scale identification of novel
imprinted genes.
Additionally, four recent papers have successfully identified
novel imprinted genes using massively parallel sequencing
approaches (3–6). In each case, RNA and cDNA were prepared
from reciprocal F1 mouse tissues and the cDNA subjected to mas-
sively parallel sequencing. By using polymorphic strains for the
reciprocal crosses, SNPs in the sequenced material can be used to
identify the expressed alleles, as well as those genes that express
only, or predominantly, one allele. Use of reciprocal crosses is
7 Data Mining for Imprinted Genes 91
important, as it allows one to distinguish strain-specific expression

effects from parent-of-origin expression effects. (For a more detailed
description of F1 hybrid studies, please refer to Chapter 6.) The
first of the four studies used neonatal mouse brain and successfully
identified three novel imprinted genes in this tissue. Imprinting of
each gene was confirmed by Sanger and pyrosequencing of PCR
products spanning allele-specific SNPs (3). The second used e9.5
mouse embryos from reciprocal F1 mouse crosses and identified six
novel imprinted genes (4). This study also suggests that many ncR-
NAs are subject to imprinted expression, as more than half of all
imprinted single-nucleotide polymorphisms did not overlap previ-
ously discovered imprinted transcripts and a large fraction of these
represent novel ncRNAs within known imprinted loci. The two
most recent studies examined patterns of imprinted expression in
brains of adult mice from reciprocal crosses (5, 6). The authors
found elevated expression from the maternal X chromosome, indi-
cating a bias in X-chromosome activation. Furthermore, over
1,300 candidate autosomal imprinted genes were identified. Two of
the candidate genes were examined further and found to be imprinted
in female, but not in male, adult mouse brain. These studies dem-
onstrate the feasibility of unbiased, transcriptome-wide analysis for
the identification of novel imprinted genes.
A third method has also been used in imprinted gene identifica-
tion: computational prediction. This method will be discussed in
detail throughout the remainder of the chapter. Briefly, this method
involves identifying features of interest (for example, DNA sequence
features, transcription factor or chromatin-remodeling protein-bind-
ing sites, or epigenetic status), demonstrating enrichment for those
features at known imprinted loci, identifying loci of unknown imprint-
ing status that carry those features, and experimentally testing
imprinted status. Characteristic epigenetic features have been
identified at gene regulatory elements of both nonimprinted and
imprinted genes (31–33). Furthermore, epigenetic mechanisms are
known to regulate genomic imprinting at several well-studied
imprinted loci (34–51). With the application of genome-wide
sequencing technologies to chromatin immunoprecipitation experi-
ments (52), epigenomic data sets have become widely available for a
variety of epigenetic marks, allowing the importance of epigenetic
marks in the control of imprinted expression to be used as a tool to
predict which additional genes in a given genome may be imprinted.
In this method, species-specific epigenomic data on a variety of
features of interest are collected and, using training sets of genes, a
computer learning approach is used to identify those features of
interest, which are most important for the prediction of imprinted
status. Once trained models have been created, they are used to
search genome-wide for predicted novel imprinted genes. By iden-
tifying the epigenetic features that serve as the strongest predictors
of imprinting, this approach can also identify those epigenetic
mechanisms that are most likely to control imprinted states. This is

something that identification of imprinted genes by transcriptome
sequencing cannot do. Genome-wide identification of novel
imprinted genes based on sequence features alone was pioneered in
a series of two studies, which used a two-tiered machine-learning
program to predict novel mouse and human imprinted genes
genome-wide (2, 16). The first tier used a training set of known
imprinted genes and presumed non-imprinted control genes to
train the prediction program based on data on a variety of sequence
features, but focusing on repetitive elements and transcription fac-
tor binding sites. The second tier was where the resulting model was
run on the genome-wide data to predict novel imprinted genes.
Although they did not experimentally verify any candidate imprinted
genes in the mouse genome, they predicted a total of 600 imprinted
mouse genes. A similar approach was used for the human genome
and successfully verified two new imprinted genes on a chromo-
some that was not previously known to contain any imprinted genes.
However, as a cautionary note, the imprinting status of genes was
not verified using reciprocal F1 crosses, so false positives due to
genetic background, but not parent of origin expression bias, can-
not be ruled out. Subsequent studies using computational methods
have resulted in experimental validation of twelve imprinted genes
in the mouse genome, in addition to the two candidate genes from
analysis of the human genome mentioned above (2, 7, 16, 20, 24).
As data become available that describe placement of additional
epigenetic marks in other tissues, or other features altogether, such
as sequences that physically interact, these methods can comple-
ment experimental methods such as transcriptome sequencing to
identify imprinted genes and to provide insights into mechanisms
controlling imprinting (3, 4, 53).
2. Materials
2.1. Hardware A computer connected to the Internet.

A multi-CPU cluster will be helpful, although not strictly necessary.
For some steps dealing with large data sets (e.g., ChIP-Seq data
sets), processing on a laptop or a desktop may not be possible.
2.2. Software Perl: http://www.perl.org/get.html.

A text editing program, such as Notepad++: http://notepad-plus-
plus.org/download.
[R]: http://www.r-project.org/.
Microsoft Excel, or similar spreadsheet program.
Microsoft Word, or similar word processing program.
Unzipping program capable of handling tar.gz files.
3. Methods
3.1. Data Mining: In this section, you will extract genomic regions you wish to include
Extracting Genomic in any analysis planned. This is done using UCSC and Galaxy and
Regions of Interest saving .txt files containing those genomic regions.
1. Direct your Web browser of choice to the UCSC Genome
Browser Web site: http://genome.ucsc.edu/cgi-bin/hgGate-
way (54). From the menu at the top of the page, click on the
“Tables” link. Once the page has loaded, select the relevant
“clade” and “genome” from the drop-down list next to each.
Next to the “assembly” option, the most recent assembly will
be sufficient for most purposes.
2. To download genomic coordinates for all known genes, select
“Genes and Gene Prediction Tracks” from the “group” menu,
select the desired track from the “track” menu (see Note 1),
ensure that “knownGene” is selected next to the “table”
option, and that “selected fields from primary and related
tables” is selected next to the “output format” option. To
download coordinates for all known genes, leave everything
else as is. However, to download coordinates for only a subset
of all known genes, click on either the “paste list” or the
“upload list” button next to “identifiers (names/accessions).”
Then, if you have selected “paste,” paste the names of the
genes you wish to work with into the box and press the “sub-
mit” button. If you have selected “upload,” click “browse,”
select the appropriate file from your computer, and then press
the “submit” button. Once you have done this, enter the file
name to which you will save your downloaded genomic coor-
dinates in the box next to “output file” (e.g., All Gene
Coordinates.xls). Then, press the “get output” button. You
will now be taken to a different Webpage where you will be
given options to select. This page will be divided into separate
sections. In the very top section, click the “select all” button.
Then search (control + F) for “Gene Symbol” and make sure
that the box next to this option is selected. Finally, scroll back
up to the top of the page and press the “Get output” button in
the very top section of the page. Once the file has downloaded,
check that you can open it using Excel, or a similar spreadsheet
management program. You will notice that the first column
does not contain conventional gene names. If you would like,
you can replace this column with the last column, which will
contain more familiar looking gene names. Make sure that you
have saved any changes.
3. Next, filter this file to remove all duplicate entries. Under the
data tab, select “Remove Duplicates.” In the pop-up menu,
unselect all except “Chrom,” “txStart,” and “txEnd,” which
should be in columns B, D, and E. Click OK. Once your

duplicates have been removed, open a new Excel file, and name
it accordingly (e.g., Genes Filtered). Copy the data for “chrom”
from column B of your filtered file into the first column of
your new file. Also copy the “txStart” and “txEnd” data from
columns D and E of your filtered file into the second and third
columns of your new file. Delete the header row, by right-
clicking on the “1” to the left of the header row and selecting
delete. Save this file with an appropriate name (e.g., Genes.txt,
see Note 2) and close this file.
4. To calculate genomic coordinates for upstream genic regions,
first sort your “Genes Filtered” Excel spreadsheet by the data
in the “strand” column. To do this, select “sort” from the
“data” menu. Make sure to indicate that your data contain
headers and that you want to sort by the “strand” column,
which should be column “C.” Then, select sort ascending (see
Note 3). You should now have all of the genes on the “−”
strand at the top of your Excel sheet and all of the genes on the
“+” strand at the bottom of your Excel sheet. Now, insert two
columns by right-clicking on the “C” above “strand.” Select
“Insert” from the menu that pops up. Repeat to add a second
column. You should now have two empty columns between
“chrom” and “strand.” Label your new columns with an appro-
priate name. For this example, we will be calculating interval
100 kb upstream of your transcription start site for all genes, so
we will label column C as “100 kb up Start” and column D as
“100 kb up End.” To perform this calculation for all genes on
the “−” strand, type =G2 + 100,000 into cell D2 and press
enter. Copy the formula from cell D2. Then, go to the last cell
in column D that still has a “−” in the “strand” column. Select
this cell and paste the formula. Now, go back to the top of your
sheet and select cell D2. The cell should be outlined with a
thick black border and a small square should be apparent at the
lower right-hand corner of the cell. Position your mouse cur-
sor over this square. A cross should form. When you see the
cross form, double click on the cross with the left-hand mouse
button. The formula should fill in for all of the genes on the
“−” strand. To calculate the interval 100 kb upstream of your
transcription start site for all genes on the “+” strand, locate
the first cell under column C that has a “−” in the “strand”
column. This cell should be empty. Note the row number you
are on and then type =F followed immediately by the row
number you are on and −100,000 (e.g., =F22485−100,000).
Press enter. Select this cell and double click on the cross with
the left-hand mouse button. The formula should fill in for all
of the genes on the “+” strand. Now, you should have two
columns partially filled with numbers. Next, type =G2 into cell
C2. Select this cell and double click on the cross with the
left-hand mouse button. Then, find the first empty cell in

column E and note the row number you are on. Into this
empty cell, type =F, followed immediately by the row number
you are on (e.g., =F22485). Press enter. Select this cell and
double click on the cross with the left-hand mouse button. You
should now have two columns filled with numbers. Open a
new Excel file and name it accordingly (e.g., 100kbUP.txt).
Highlight the data in columns B, C, and D, copy, and paste
these into your new spreadsheet using the “paste special” func-
tion (see Note 4). Delete the header row, by right-clicking on
the “1” to the left of the header row and selecting delete. Save
as a .txt file and close the file. Repeat this procedure for each
upstream interval you wish to examine, but change the addi-
tion and subtraction of 100,000 accordingly (e.g., 10,000 and
1,000 for 10 kb and 1 kb upstream, respectively).
5. To calculate genomic coordinates for downstream genic regions,
the procedure is very similar, but with a couple of subtle, but
important, differences. Label your newly emptied columns,
which previously held the genomic coordinates for upstream
genic regions with an appropriate name. For this example, we
will be calculating interval 100 kb downstream of your tran-
scription start site for all genes, so we will label column C as
“100 kb dn Start” and column D as “100 kb dn End.” To
perform this calculation for all genes on the “−” strand, type
=F2−100,000 into cell D2 and press enter. Copy the formula
from cell D2. Then, go to the last cell in column D that still
has a “−” in the “strand” column. Select this cell and paste the
formula. Now, go back to the top of your sheet and select cell
D2. The cell should be outlined with a thick black border and
a small square should be apparent at the lower right-hand corner
of the cell. Position your mouse cursor over this square. A cross
should form. When you see the cross form, double click on the
cross with the left-hand mouse button. The formula should fill
in for all of the genes on the “−” strand. To calculate the inter-
val 100 kb upstream of your transcription start site for all genes
on the “+” strand, locate the first cell under column C that has
a “−” in the “strand” column. This cell should be empty. Note
the row number you are on and then type =G followed imme-
diately by the row number you are on and +100,000 (e.g.,
=G22485 + 100,000). Press enter. Select this cell and double
click on the cross with the left-hand mouse button. The formula
should fill in for all of the genes on the “+” strand. Now, you
should have two columns partially filled with numbers. Next,
type =I2 into cell C2. Select this cell and double click on the
cross with the left-hand mouse button. Then, find the first
empty cell in column E and note the row number you are on.
Into this empty cell, type =H, followed immediately by the row
number you are on (e.g., H22485). Press enter. Select this cell
and double click on the cross with the left-hand mouse button.
You should now have two columns filled with numbers. Open
a new Excel file and name it accordingly (e.g., 100kbDN.txt).
Highlight the data in columns B, C, and D, copy, and paste
these into your new spreadsheet using the “paste special” func-
tion (see Note 4). Delete the header row, by right-clicking on
the “1” to the left of the header row and selecting delete. Save
as a .txt file and close the file. Repeat this procedure for each
downstream interval you wish to examine, but change the
addition and subtraction of 100,000 accordingly (e.g., 10,000
and 1,000 for 10 kb and 1 kb downstream, respectively).
6. Download and install Perl to C:\Perl from http://www.perl.
org/get.html.
7. Download and install a text editing program (e.g., Notepad++,
gedit, Aquamacs, etc.).
8. To download genomic coordinates for exons, go back to the
Table browser at the UCSC Genome Browser Web site. Make
sure that the correct “clade,” “genome,” and “assembly” are
still selected. Select “Genes and Gene Prediction Tracks” from
the “group” menu and select the desired track from the “track”
menu, as above in step 2. However, this time, next to “output
format,” select “BED—Browser Extensible Data” and tick the
box next to “Send to Galaxy.” Click “get output.” On the next
screen, under “Create one BED record per,” make sure that
“Coding Exons” is ticked. Click “send query to Galaxy.” Your
browser should be redirected to the Galaxy Web site and your
data will appear under “History” on the right-hand side of the
browser screen. In the left-hand “Tools” menu, click on “Text
Manipulation,” and then “Cut columns from a table.” Now,
cut all columns except those containing “Chrom,” “Start,”
and “End.” To do this, type the columns you wish to keep into
the box next to “Cut columns” (e.g., c1, c2, c3). You can view
the existing column order by clicking on the data set in the
“History” pane on the right-hand side of the screen. The
option under “Delimited by” should set to “Tab.” Press exe-
cute and wait for your job to finish. When it has finished run-
ning, click on the name of the job. This should expand the file
window and you should see several icons at the top right of the
file window. Click on the pencil icon to “edit attributes.” Once
the new window has opened, scroll down to the “change data
type” heading and select txt from the drop-down menu. Click
the save button. This will change the format of the file. Once
the file has finished updating, click on the name of the job to
expand the file window, if needed, and save the file by clicking
on the disk icon. Save the file under an appropriate name
(“ExonsAll.txt”). Now, move this file to the same directory
where you have installed Perl (see Note 5).
9. Now, open Notepad++ or a similar text editing program, and

copy or type the text below into a new file (see Note 6).
#!/usr/local/bin/perl -w
# removes duplicates
my $usage = ‘duplicates.pl
Removes duplicate entries from a file.
USAGE:
./duplicates.pl input.txt output.txt’;
# the next 2 lines tell the program that the user will enter the
input and output files to use
my $input = shift @ARGV || die “$usage\n”;
my $output = shift @ARGV || die “$usage\n”;
open (INPUT, “<$input”)
or die “can’t open INPUT FILE”; #opens the input file or dies
trying
open (OUT, “ > $output”)
or die “can’t open OUTPUT FILE”; #opens the output file or
dies trying
print “Running…\n”;
my @INPUT = <INPUT>;
chomp @INPUT;
my %hash = map {$_, 1} @INPUT;
# enters data into a hash
my @unique = keys %hash;
foreach (@unique) {
print (OUT $_, “\n”);
}
print “Done!\n”;
close (INPUT); #closes input file
close (OUT); #closes output file
exit 0; #closes program
Type or paste the text above into a new file and name this file
Duplicates. To save as a Perl file, select “Perl source file (*.pl,
*.pm, *.plx)” from the drop-down menu next to “Save as
type.” Make sure that the file is saved in the same directory as
Perl. You will also need to check that the file called ExonsAll.
txt, which you created in Subheading 3.1, step 8 above, is
saved in this directory as a .txt file. Open Command Prompt,
or similar command line program. In Windows, Command
Prompt can be found by following Start ->All Programs
->Accessories ->Command Prompt. Change directory from
the current directory to the folder where you have installed

Perl (see Note 5). To change a directory, type cd followed by
the new directory (e.g., cd C:\Perl). Your Command Prompt
should now read C:\Perl>, or something very similar. Type perl
Duplicates.pl ExonsAll.txt Exons.txt and press enter. This will
create a file called “Exons.txt,” which contains a list of all the
genomic coordinates of all exons, filtered for duplicates.
10. To download genomic coordinates for introns, go back to the
Table browser at the UCSC Genome Browser Web site. Make
sure that the correct “clade,” “genome,” and “assembly” are
still selected. Select “Genes and Gene Prediction Tracks” from
the “group” menu and select the desired track from the “track”
menu, as above in step 2. However, this time, next to “output
format,” select “BED—Browser Extensible Data” and tick the
box next to “Send to Galaxy.” Click “get output.” On the next
screen, under “Create one BED record per,” make sure that
“Introns” is ticked. Click “send query to Galaxy.” Your browser
should be redirected to the Galaxy Web site and your data will
appear under “History” on the right-hand side of the browser
screen. In the left-hand “Tools” menu, click on “Text
Manipulation,” and then “Cut columns from a table.” Now,
cut all columns except those containing “Chrom,” “Start,”
and “End.” To do this, type the columns you wish to keep into
the box next to “Cut columns” (e.g., c1, c2, c3). Press execute
and wait for your job to finish. When it has finished running,
click on the name of the job. This should expand the file win-
dow and you should see several icons at the top right of the file
window. Click on the pencil icon to “edit attributes.” Once the
new window has opened, scroll down to the “change data
type” heading and select txt from the drop-down menu. Click
the save button. This will change the format of the file. Once
the file has finished updating, click on the name of the job to
expand the file window, if needed, and save the file by clicking
on the disk icon. Save the file under an appropriate name
(“IntronsAll.txt”). Now, move this file to the same directory
where you have installed Perl (see Note 5).
11. Open Command Prompt, or similar command line program.
In Windows, Command Prompt can be found by following
Start ->All Programs ->Accessories ->Command Prompt.
Change directory from the current directory to the folder
where you have installed Perl. To change a directory, type cd
followed by the new directory (e.g., cd C:\Perl); see Note 5.
Your Command Prompt should now read C:\Perl>, or some-
thing very similar. Type perl Duplicates.pl IntronsAll.txt
Introns.txt and press enter. This will create a file called “Introns.
txt,” which contains a list of all the genomic coordinates of all
introns, filtered for duplicates.
12. To download genomic coordinates for 5¢ UTRs, go back to

the Table browser at the UCSC Genome Browser Web site.
Make sure that the correct “clade,” “genome,” and “assem-
bly” are still selected. Select “Genes and Gene Prediction
Tracks” from the “group” menu and select the desired track
from the “track” menu, as above in step 2. However, this time,
next to “output format,” select “BED—Browser Extensible
Data” and tick the box next to “Send to Galaxy.” Click “get
output.” On the next screen, under “Create one BED record
per,” make sure that “5¢ UTRs” is ticked. Click “send query to
Galaxy.” Your browser should be redirected to the Galaxy Web
site and your data will appear under “History” on the right-
hand side of the browser screen. In the left-hand “Tools”
menu, click on “Text Manipulation,” and then “Cut columns
from a table.” Now, cut all columns except those containing
“Chrom,” “Start,” and “End.” To do this, type the columns
you wish to keep into the box next to “Cut columns” (e.g., c1,
c2, c3). Press execute and wait for your job to finish. When it
has finished running, click on the name of the job. This should
expand the file window and you should see several icons at the
top right of the file window. Click on the pencil icon to “edit
attributes.” Once the new window has opened, scroll down to
the “change data type” heading and select txt from the drop-
down menu. Click the save button. This will change the for-
mat of the file. Once the file has finished updating, click on the
name of the job to expand the file window, if needed, and save
the file by clicking on the disk icon. Save the file under an
appropriate name (“5UTRAll.txt”). Now, move this file to the
same directory where you have installed Perl (see Note 5).
Change directory from the current directory to the folder where
you have installed Perl (see Note 5). To change a directory,
type cd followed by the new directory (e.g., cd C:\Perl). Your
Command Prompt should now read C:\Perl>, or something
very similar. Type perl Duplicates.pl 5UTRAll.txt 5UTR.txt
and press enter. This will create a file called “5UTR.txt,” which
contains a list of all the genomic coordinates of all introns,
filtered for duplicates.
14. To download genomic coordinates for 3¢ UTRs, go back to
Make sure that the correct “clade,” “genome,” and “assem-
bly” are still selected. Select “Genes and Gene Prediction
Tracks” from the “group” menu and select the desired track
from the “track” menu, as above in step 2. However, this time,
next to “output format,” select “BED—Browser Extensible
Data” and tick the box next to “Send to Galaxy.” Click “get
output.” On the next screen, under “Create one BED record

per,” make sure that “3¢ UTRs” is ticked. Click “send query to
Galaxy.” Your browser should be redirected to the Galaxy Web
site and your data will appear under “History” on the right-
hand side of the browser screen. In the left-hand “Tools”
menu, click on “Text Manipulation,” and then “Cut columns
from a table.” Now, cut all columns except those containing
“Chrom,” “Start,” and “End.” To do this, type the columns
you wish to keep into the box next to “Cut columns” (e.g., c1,
c2, c3). Press execute and wait for your job to finish. When it
has finished running, click on the name of the job. This should
expand the file window and you should see several icons at the
top right of the file window. Click on the pencil icon to “edit
attributes.” Once the new window has opened, scroll down to
the “change data type” heading and select txt from the drop-
down menu. Click the save button. This will change the for-
mat of the file. Once the file has finished updating, click on the
name of the job to expand the file window, if needed, and save
the file by clicking on the disk icon. Save the file under an
appropriate name (“3UTRAll.txt”). Now, move this file to the
same directory where you have installed Perl (see Note 5).
Change directory from the current directory to the folder where
you have installed Perl (see Note 5). To change a directory, type
cd followed by the new directory (e.g., cd C:\Perl). Your
Command Prompt should now read C:\Perl>, or something
very similar. Type perl Duplicates.pl 3UTRAll.txt 3UTR.txt
and press enter. This will create a file called “3UTR.txt,” which
contains a list of all the genomic coordinates of all introns,
filtered for duplicates.
3.2. Data Mining: In this section, you will identify locations of additional features you
Extracting Additional wish to correlate with imprinting status. The examples below cover
Features miRNAs, CpG islands, G-quartets, CTCF sites, a variety of epige-
netic features, and GC percent. Identifying miRNAs and CpG
islands requires the use of UCSC and Galaxy, in a way that is very
similar to what was already described for extracting the genic
regions. For additional features, other databases are used (insula-
torDB, Broad, Quadruplex). You may want to consider other fea-
tures (for example, transcription factor binding sites) and additional
databases exist where those locations can be captured (http://ecr-
base.dcode.org/).
1. To download genomic coordinates for CpG islands, go back to
Make sure that the correct “clade,” “genome,” and “assembly”
are still selected. Select “Expression and Regulation” from the
“group” menu and select the “CpG Islands” track from the
“track” menu. However, this time, next to “output format,”
select “all fields from selected table” and tick the box next to
“Send to Galaxy.” Click “get output.” On the next screen,
click “send query to Galaxy.” Your browser should be redi-
rected to the Galaxy Web site and your data will appear under
“History” on the right-hand side of the browser screen. In the
left-hand “Tools” menu, click on “Text Manipulation,” and
then “Cut columns from a table.” Now, cut all columns except
those containing “Chrom,” “ChromStart,” and “ChromEnd.”
To do this, type the columns you wish to keep into the box
next to “Cut columns” (e.g., c2, c3, c4). Press execute and
wait for your job to finish. When it has finished running, click
on the name of the job. This should expand the file window
and allow you to save the file by clicking on the disk icon.
Make sure that you can open your file in Excel. Delete the
header row, by right-clicking on the “1” to the left of the
header row and selecting delete. Name the file accordingly
(e.g., CpG.txt) and close the file.
2. To download genomic coordinates for micro-RNA (miRNA)
clusters, go back to the Table browser at the UCSC Genome
Browser Web site. Make sure that the correct “clade,”
“genome,” and “assembly” are still selected. Select “Genes
and Gene Prediction Tracks” from the “group” menu and
select the “miRNA” track from the “track” menu. However,
this time, next to “output format,” select “all fields from
selected table” and tick the box next to “Send to Galaxy.” Click
“get output.” On the next screen, click “send query to Galaxy.”
Your browser should be redirected to the Galaxy Web site and
your data will appear under “History” on the right-hand side
of the browser screen. In the left-hand “Tools” menu, click on
“Text Manipulation,” and then “Cut columns from a table.”
Now, cut all columns except those containing “Chrom,”
“ChromStart,” and “ChromEnd.” To do this, type the col-
umns you wish to keep into the box next to “Cut columns”
(e.g., c2, c3, c4). Press execute and wait for your job to finish.
When it has finished running, click on the name of the job.
This should expand the file window and allow you to save the
file by clicking on the disk icon. Open the file with Excel.
Delete the header row, by right-clicking on the “1” to the left
of the header row and selecting delete. Then, save the file as a
.txt file, making sure to give it an appropriate name (e.g.,
miRNA.txt), and close the file.
3. To obtain genomic coordinates for CTCF sites for human or
mouse, direct your Web browser to the Insulator Database Web
site at http://insulatordb.uthsc.edu/help.php#download (55).
On this page, you will have two choices of download: experi-
mentally verified CTCF binding sites and computationally
predicted CTCF binding sites. Once you have downloaded the

file(s), open it first in Word or a similar text editing program
and save as a .txt (Plain Text File) file; see Note 2. When
prompted in the next Window, leave all options as they are,
except choose “CR only” from the drop-down menu next to
“End lines with:”. Open the saved .txt file with Excel. Sort the
file based on species. To do this, click on the “Data” tab and
select “Sort.” Sort by the “Species” column, which should be
column B. Make sure to indicate that your data contain head-
ers and click OK. Now, erase any data that do not apply to the
species in which you are interested. The easiest way to do this
is to search (press control and f at the same time) for your spe-
cies name. This should take you to the first row containing
data applying to your species of interest. Delete all rows above
this, except for the header, by clicking on the row number to
the left of the first row you wish to delete, scrolling to the top
of the page, and holding down shift while clicking on the last
row you wish to delete. Then, right-click within the high-
lighted area and select “delete.” To delete any unwanted spe-
cies following your species of interest, find the last row
containing data you wish to keep, and delete click on the row
number to the left of the first row you wish to delete, scroll to
the bottom of the page, and hold down shift while clicking on
the last row you wish to delete.
4. Then, delete all columns, except the “Chromosome Location”
file. To do this, click on the letter above each column you wish
to eliminate, and hold down the control key while clicking.
Once you have selected all of the columns you wish to delete,
right-click and select delete. You should be left with a single
column labeled “Chromosome Location.” Click on the A
above “Chromosome Location” and select “Text-to-Columns”
from the “Data” tab. Make sure that “Delimited” is selected
and then click “Next.” In the next window, deselect “Tab” and
select “Other.” In the box next to “Other,” type a colon (:)
and click “Finish.” Part of the data formerly contained in col-
umn A will have moved to column B. Click on the B above
your new column and select “Text-to-Columns” from the
“Data” tab. Make sure that “Delimited” is selected and then
click “Next.” In the next window, deselect “Tab” and select
“Other.” In the box next to “Other,” type a dash (-) and click
“Finish.” You should now have your chromosomal locations
split between three columns. Delete the header row, by right-
clicking on the “1” to the left of the header row and selecting
delete. Save your file under an appropriate name (e.g.,
CTCFcompOLD.txt or CTCFexpOLD.txt) and close the file.
5. The data from the Insulator Database are mapped to the hg18
assembly for human, the mm8 assembly for mouse, the rn3
assembly for rat, and the galGal2 assembly for chicken. If this
is not the option you have been selecting next to the “assembly”
option on the UCSC Genome Browser, you will need to con-
vert the genomic coordinates from February 2006, mm8 to
the assembly you are working with. To convert CTCF site
genomic coordinates from one genome assembly to another,
direct your Web browser to the Galaxy Web site at http://
main.g2.bx.psu.edu/ (56, 57). Upload your file, making sure
to choose the version without column names, by clicking on
“Get Data” and selecting “Upload File” from the left-hand
menu. Click “Browse” and select the file you wish to upload.
At the top of the window will be a drop-down menu under
“File Format.” Select “bed” from this menu. Above the
“Execute” button, there should be a drop-down menu which
allows you to select the genome. Make sure to select both the
correct species and genome assembly (hg18 assembly for
human, the mm8 assembly for mouse, the rn3 assembly for rat
and the galGal2 assembly for chicken) from this menu. Then,
click “Execute.” When your file has finished loading to Galaxy,
click on “Lift Over” and select “Convert genome coordinates”
from the menu on the left. Select your newly uploaded file
from the drop-down menu under “Convert coordinates of”
and select the genome assembly you wish to convert to from
the drop-down menu under “To:”. Click execute. Two new
files will appear in the menu to the right. Once the jobs have
finished running, click on the header of the file with “MAPPED
COORDINATES” in the name. Click on the disk icon to save
the file. Locate the file and open it with Excel. Then, save the
file as a .txt file, making sure to give it an appropriate name
(e.g., CTCFcomp.txt or CTCFexp.txt).
6. To obtain genomic coordinates for histone modification data,
direct your Web browser to ftp://ftp.broad.mit.edu/pub/
papers/chipseq/ at the Broad Institute. The data are arranged
by first author and publication date of each of the papers in
which the data are published. Select the data set you are most
interested in by double-clicking on the relevant folder. In each
folder is a file called “Readme.txt.” If you click on this file, you
will be able to read a key indicating the type of data found in
each subfolder. The subfolder of most interest is likely to be
the “Alignments” folder, which contains both the sequences
and the coordinates of uniquely aligned ChIP-Seq reads.
However, some data sets contain files called “WindowIntervals.
tar.gz” and “HMMIntervals.tar.gz,” which may be of interest
as well. These, respectively, contain intervals enriched for cer-
tain histone modifications inferred by fixed-size windows and
intervals Enriched for certain histone modifications inferred by
a Hidden Markov Model (HMM). Within the “Alignments”
subfolder, you are able to choose those histone ChIP-Seq data
sets which most interest you. All of the downloadable files are
in tar.gz format, and will require an unzipping program capable

of handling this type of file. Also, keep in mind that the genomic
coordinates for the histone-modification data might be from a
different assembly than the genomic coordinates you are using.
The ReadMe file will tell you which assembly was used, and
you can use Galaxy, as in step 14 above, to convert coordinates
between different assemblies; see Note 7.
7. To obtain genomic coordinates for predicted G-quartet sites in
your genome of interest, direct your Web browser to http://
www.quadruplex.org/?view=quadbaseDownload and click on
the link for your organism and genome assembly of interest
(58). If your particular genome assembly is not available, it is
possible to convert coordinates between genome assemblies
using Galaxy, so choose a different assembly, and make a note
of which version you have selected. Click on the “All Files” link
to download all predicted G-quartet sites in your genome of
interest. Once the file has downloaded, rename it appropriately
(e.g., GQraw) and upload the file to Galaxy by clicking on
correct species and genome assembly. Now, cut all columns
except the first three, which contain information regarding
genomic location. In the left-hand “Tools” menu, click on
“Text Manipulation,” and then “Cut columns from a table.”
Type the columns you wish to keep into the box next to “Cut
columns” (e.g., c1, c2, c3). Press execute and wait for your job
to finish. When it has finished running, click on the name of
the job. This should expand the file window and allow you to
save the file by clicking on the disk icon. Locate the file and
open it with Excel. Delete the header row, by right-clicking on
the “1” to the left of the header row and selecting delete. Then,
save the file as a .txt file, making sure to give it an appropriate
name (e.g., GQs.txt), and close the file. Genomic coordinates
for predicted G-quartet sites in your genome of interest can
also be obtained in the method described in Note 8.
8. To calculate the GC% of each gene body, introns, exons, 5¢
UTRs, 3¢ UTRs, and any upstream and downstream regions,
direct your Web browser to the Galaxy Web site at http://
main.g2.bx.psu.edu/. Upload your file containing the filtered
genomic coordinates for each of your regions of interest, mak-
ing sure to choose the version without column names, by click-
ing on “Get Data” and selecting “Upload File” from the
left-hand menu. Click “Browse” and select the file you wish to
upload. At the top of the window will be a drop-down menu
under “File Format.” Select “bed” from this menu. Above the
correct species and genome assembly. Once your file has loaded
to Galaxy, select “Fetch Sequences” and “Extract Genomic
DNA” from the “Tools” menu on the left-hand side of the
page. Make sure that the file you just uploaded is selected in
the drop-down menu under “Fetch sequences corresponding
to Query:” and change “Output” data type to “Interval.”
Click “Execute.” When it has finished running, click on the
name of the job. This should expand the file window and allow
you to save the file by clicking on the disk icon. Locate the file
and open it with Excel. Delete the header row, by right-clicking
on the “1” to the left of the header row and selecting delete.
Then, save the files as a .txt file, making sure to give it an
appropriate name (e.g., CGintron.txt), and close the file. Do
this for each of the genomic regions you are interested in.
9. Now, open Notepad++ or a similar text editing program, and
copy or type the text below into a new file (see Note 6).
# calculate GC%
my $usage = ‘gc-count.pl
Compute GC content in a set of sequences.
USAGE:
./gc-count.pl input.txt output.txt
‘;
my $input = shift @ARGV || die “$usage\n”;
open (INPUT, “<$input”)
or die “can’t open INPUT FILE”; #opens the input file or dies
trying
open (GCOUT, “ > $output”)
or die “can’t open OUTPUT FILE”; #opens the output file or
dies trying
while (<INPUT>) {#tells the program what to do while the input
file is open
chomp; #removes any new line symbols from the end of each
line
(@INPUT) = split/\t/; #splits each line on tabs
$SEQ = $INPUT[3]; #defines the variable $SEQ
my @seqarray = split ‘’, $SEQ; #splits the variable $SEQ at each

character
my $GC = 0; # counter for G’s and C’s
foreach my $char (@seqarray) {#tells the program what to do
with each character
if ($char = ~ m/[GgCc]/) {#regular expression to search for
matches to any of G g C or c
$GC++; # increment the GC counter
}
}
$percentGC = (($GC/(length($SEQ)-1))*100); #calculates
GC%
print (GCOUT $GC, “\t”, length($SEQ)-1, “\t”, $percentGC,
“\n”); #prints GC%
}
close (INPUT); #closes input file
close (GCOUT); #closes outout file
gc-count. To save as a Perl file, select “Perl source file (*.pl,
Perl and click “Save.” Open Command Prompt, or similar
command line program. Command Prompt can be found by
following Start ->All Programs ->Accessories ->Command
Prompt. Change directory from the current directory to the
folder where you have installed Perl (see Note 5). To change a
directory, type cd followed by the new directory (e.g., cd C:\
Perl). Your Command Prompt should now read C:\Perl>, or
something very similar. Type perl gc-count.pl followed by your
input and output file names (e.g., perl gc-count.pl CGintron.
bed CGintronOUT.txt); see Note 9. Repeat this process for
every region you are interested in (gene body, introns, exons,
5¢ UTRs, 3¢ UTRs, and any upstream and downstream regions).
Make sure to change the name of the input and output file
names each time.
3.3. Data Mining: 1. To obtain genomic coordinates of all known imprinted genes,
Identifying Known direct your Web browser to the Otago Catalogue of Imprinted
Imprinted Genes for Genes: http://igc.otago.ac.nz/home.html. Click on the
Model Training “Summary Tables” link. Select your organism of interest from
the drop-down menu next to “Taxon” and select “Imprinted
Genes” from the drop-down menu next to “Category.” Click
“Search.” Copy and paste the resulting list into Excel and save
under an appropriate name (e.g., Allimprinted.xls). Remove

the extra data from the “Gene” column by clicking on the C
above the column. Next, select “Text-to-Columns” from the
“data” tab. Make sure that “delimited” is selected and click
“Next.” In the new window, tick both “comma” and “space”
and click “Finish.” Copy the gene names from the “Gene”
column. Direct your Web browser of choice to the Table
browser at the UCSC Genome Browser Web site. Make sure to
select the relevant “clade,” “genome,” and “assembly.” Then,
click on the “paste list” button next to “identifiers names/
associations:”. A new window will open. Paste the names of
the genes you have copied from your Excel sheet into the box
and press the “submit” button. Once you have done this,
choose the “selected fields from primary and related tables”
option from the drop-down list next to “output format.” Enter
the file name to which you will save your downloaded genomic
coordinates in the box next to “output file” (e.g.,
KnownImprinted.txt). Then, press the “get output” button.
In the next window, tick the boxes next to “chrom,” “txStart,”
and “txEnd.” Once the file has downloaded, check that you
can open it using Excel, or a similar spreadsheet management
program. Next, filter this file to remove all duplicate entries.
Under the data tab, select “Remove Duplicates.” In the pop-
up menu, unselect all boxes. Click OK. Delete the header row,
by right-clicking on the “1” to the left of the header row and
selecting delete. Save and close the file.
2. To obtain a list of whether each gene in your filtered list of all
known genes (e.g., Genes.txt) is imprinted or not, install Perl
from http://www.perl.org/get.html by following the instruc-
tions on the Webpage. Open Notepad++ or a similar text editing
program, and copy the text below into a new file:
# Determines whether each known gene is imprinted.
my $usage = ‘Imprinted.pl
Determines whether each known gene in your filtered list of all
known genes is imprinted.
USAGE:
./Imprinted.pl within.txt find.txt output.txt
‘;
my $within = shift @ARGV || die “$usage\n”;
my $find = shift @ARGV || die “$usage\n”;
open (WITHIN, “<$within”)
or die “can’t open file to search within”; #opens the first input
file or dies trying
or die “can’t open OUTPUT file”; #opens the output file or dies
trying
while (<WITHIN>) {#tells the program what to do while the
input file is open
chomp; #removes any new line symbols from the end of each line
(@WITHIN) = split/\t/; # splits each line on tabs
# the next 3 lines define variables
$chrI = $WITHIN [0];
$startI = $WITHIN [1];
$endI = $WITHIN [2];
my $NUM = 0; #sets the variable $NUM equal to 0
open (FIND, “<$find”)
or die “can’t open file to search with”; #opens the second input
file or dies trying
foreach $chrI (@WITHIN){#loops through 1st file
while (<FIND>) {#loops through 2nd file
(@FIND) = split/\t/; # splits each line on tabs
$chrS = $FIND [0];
$startS = $FIND [1];
$endS = $FIND [2];
if ((($chrI = $chrS) && ($startI<= $startS) && ($endI > =
$endS))) {#finds matches
$NUM++; #increases count if match is found
}
}
}
#the lines below tell the program to print a 0 if no matches were
found and a 1 if any #match was found.
if ($NUM > 0){
print (OUT $NUM/$NUM, “\n”);
}else{
print (OUT 0, “\n”);
}
}
close (WITHIN); #closes first input file
close (FIND); #closes second input file
Imprinted. To save as a Perl file, select “Perl source file (*.pl,
Perl and click “Save.” Open Command Prompt, or similar
command line program. In Windows, Command Prompt can
be found by following Start ->All Programs ->Accessories
->Command Prompt. Change directory from the current direc-
tory to the folder where you have installed Perl (see Note 5).
To change a directory, type cd followed by the new directory
(e.g., cd C:\Perl). Your Command Prompt should now read
C:\Perl>, or something very similar. Type perl Imprinted.pl
followed by your input and output file names (e.g., perl
Imprinted.pl Genes.txt KnownImprinted.txt Imprinted.txt,
see Note 10). Now you will have a list containing a column of
1’s and 0’s. A value of 1 means that the gene is imprinted and a
value of 0 means that the gene is not imprinted. This file will
not make much sense on its own, but it will once we combine it
with another file, which we will do in Subheading 3.4, step 3.
3.4. Data Now that you have collected data on each of your features of inter-
Manipulation: est, you need to count the number of times each of those features
Counting Occurrences occurs within each of your genomic regions of interest, for every
of Features of Interest gene in the genome. To do this, you will use some basic Perl
scripts.
1. To tabulate the number of times each of the features you have
collected occurs within each of these regions of interest:
Introns, Exons, 5¢ UTRs, or 3¢ UTRs, open Notepad++, or a
similar text editing program, and copy the text below into a
new file (see Note 6):
# tabulate number of occurrences of one file within another file
my $usage = ‘countIN.pl
Tabulates the number of occurrences of one file within another
file. Meant to be a 2 round counting program.
E.g. the first round fiunds the exons which contain CpG
islands. The 2nd round will find the exons within each known
gene in the genome, effectively counting the number of CpG
islands within exons, but classified by gene.
USAGE:
./countIN.pl within.txt find.txt output.txt
‘;
file or dies trying
trying
while (<WITHIN>) {#tells the program what to do while the first
file is open
(@WITHIN) = split/\t/; #splits each line on tabs
my $NUM = 0;
file or dies trying
(@FIND) = split/\t/; #splits each line on tabs
$chrS = $FIND [0];
$endS = $FIND [2];
print (OUT $chrS, “\t”, $startS, “\t”, $endS, “\n”); #prints
matches if found
}
}
}
}
countIN. To save as a Perl file, select “Perl source file (*.pl,
Perl and click “Save.” You will use a different process for the
remaining regions; make sure to use this program only for
introns, exons, 5¢ UTRs, and 3¢ UTRs. The aim of this pro-
gram is to identify which introns, exons, 5¢ UTRs, or 3¢ UTRs
contain any of your marks of interest. The second program,
which we will run below, will identify the gene to which each
of the introns, exons, 5¢ UTRs, and 3¢ UTRs belongs. To run
the countIN program, open Command Prompt, or similar
command line program. In Windows, Command Prompt can
be found by following Start ->All Programs ->Accessories
C:\Perl>, or something very similar. Type perl count.pl fol-
lowed by your first input file containing one of the following
genomic intervals, introns, exons, 5¢ UTRs, or 3¢ UTRs, then
your second input file containing the marks of interest you
have collected (e.g., your list of CpG Islands, miRNA clusters,
CTCF binding sites, etc.), and, finally, the output file (e.g.,
perl countIN.pl Exons.txt miRNA.txt miRNAexonsOUT_1.
txt). For very large data sets, like ChIP-Seq data sets, it may
not be possible to perform this analysis on a laptop or a desk-
top computer. In this case, it may be useful to have access to a
multi-CPU cluster. See Notes 11 and 12. Repeat this process
for each of the following regions you are interested in: introns,
exons, 5¢ UTRs, or 3¢ UTRs. Make sure to change the name of
both the input file as well as the output file, each time.
2. Once you have run the above program on introns, exons, 5¢
UTRs, and 3¢ UTRs, you will want to run the program below
which will both tabulate the number of times each of the fea-
tures you have collected occurs within each of these regions of
interest: gene body, introns, exons, 5¢ UTRs, 3¢ UTRs, and any

upstream and downstream regions. And, importantly, this pro-
gram will also determine in which gene each of the counts
occurs. To run this program, open Notepad++ or a similar
text editing program, and copy the text below into a new file
(see Note 6):
my $usage = ‘count.pl
Tabulates the number of occurrences of one file within another
file. E.g. number of CpG Islands found within each known
gene in the genome.
USAGE:
./count.pl within.txt find.txt output.txt
‘;
file or dies trying
trying
while (<WITHIN>) {#tells the program what to do while the first
file is open
(@WITHIN) = split/\t/; #splits each line on tabs
my $NUM = 0;
file or dies trying

(@FIND) = split/\t/; #splits each line on tabs
$chrS = $FIND [0];
$endS = $FIND [2];
$NUM++; #increases count if match is found
}
}
}
print (OUT $NUM, “\n”);
}
count. To save as a Perl file, select “Perl source file (*.pl, *.pm,
*.plx)” from the drop-down menu next to “Save as type.”
Make sure that the file is saved in the same directory as Perl
and click “Save.” Open Command Prompt, or similar com-
mand line program. In Windows, Command Prompt can be
found by following Start ->All Programs ->Accessories
C:\Perl>, or something very similar. Type perl count.pl fol-
lowed by your first input file containing the genomic intervals
you wish to examine (e.g., your known gene locations or
upstream/downstream genomic intervals), your second input
file containing the marks of interest you have collected (e.g.,
your list of CpG Islands, miRNA clusters, CTCF binding sites,
etc.), and output file names (e.g., perl count.pl Genes.txt
miRNA.txt miRNAgenesOUT.txt). Important: For introns,
exons, 3¢ UTRs, and 5¢ UTRs, you will follow a slightly different
format. First of all, the first input file will remain as Genes.txt
independent of whether you are analyzing introns, exons, 5¢
UTRs, or 3¢ UTRs. Second, the second input file will be different
than the example given above. In these cases only, you will
want to use the output file from Subheading 3.4, step 1, above
(e.g., miRNAexonsOUT_1.txt). Therefore, the full command
for these regions will look something like perl count.pl Genes.
txt miRNAexonsOUT_1.txt miRNAgenesOUT.txt. See Notes
12 and 13. Repeat this process for every region you are inter-
ested in (gene body, introns, exons, 5¢ UTRs, 3¢ UTRs, and
any upstream and downstream regions). Make sure to check
that you have entered the correct name for both the input files
as well as the output file, each time.
3. Next, we need to get all of the data you have collected into a
format that we can use for model training. To do this, open
your file containing your filtered list of all known genes (e.g.,
Genes.txt), as well as the file containing the information regarding
whether each known gene is imprinted or not (e.g., Imprinted.
txt, from Subheading 3.3 above). The file containing the infor-
mation regarding whether each known gene is imprinted or
not should contain a single column of 1’s and 0’s. The first
empty column in your file containing your filtered list of all
known genes should be column D. Copy the column of num-
bers you have in the file containing the information regarding
whether each known gene is imprinted (e.g., Imprinted.txt)
and paste it into column D of your file containing the genomic
coordinates of all know genes (e.g., Genes.txt). Now, open the
file containing data on one of your features of interest within
that genomic region (e.g., miRNAgenesOUT.txt). The only
thing present in this file should be a column of numbers. The
next empty column in your file containing the filtered list of all
known genes should be column E. Copy the column of num-
bers into column E of the file containing your filtered list of all
known genes. Repeat this copy and paste process for each of
the features of interest you have collected data for. Once you
have done this, save your file with an appropriate name, and as
a .csv file (e.g., Genes.csv, see Note 14). Then, repeat this
entire process for each of the genomic intervals you are inter-
ested in examining, making sure to add your features of inter-
est to the columns in the same order as you did in this first file.
For example, if in your first file, you copied GC% to column E,
miRNA clusters to column F, and predicted CTCF binding
sites to column G, make sure you copy GC% to column E,
miRNA clusters to column F, and predicted CTCF binding
sites to column G in all of the other files you create for introns,
exons, 5¢ UTRs, 3¢ UTRs, and upstream and downstream
genomic regions. Please note that you will reuse the file con-
taining the information regarding whether each known gene is
imprinted (e.g., Imprinted.txt). Therefore, each of the new
files you create will contain the exact same information in col-
umn D. Save each file with an appropriate name as a .csv file.
3.5. Data In this section, you will need to create two separate data sets. One
Manipulation: Creating data set will be used to train your models. This data set will contain
Data Sets for Model a mix of known imprinted genes and non-imprinted control genes.
Training and Model The program will be told which genes are imprinted and which are
Testing not. This will allow the computer to identify those features which
distinguish imprinted genes from non-imprinted genes. The second
data set, which you will use for model testing will also contain a mix-
ture of imprinted and non-imprinted genes. The difference here is
that the computer will not be told which genes are imprinted and
which are not, although you will have a record of this information.
You will be able to use the results of this analysis to determine the
specificity and sensitivity of your trained models. This is a very impor-
tant step, as it will allow you to evaluate the performance of your
models before running your programs on the genome-wide data.
Again, you will use some basic Perl scripts to manipulate your data.
1. To obtain genomic coordinates of a random mix of known
imprinted genes and non-imprinted control genes to use in the
future as training and test data sets, paste or type the following
program into Notepad++ (see Note 6):
#!/usr/bin/perl -w
open (OUT, “ > Random Numbers.txt”) or die “can’t open
output file”; #opens output file or dies trying
$i = 0; #sets variable $i equal to 0
while ($i<100,000) {#tells the program what to do while $i is less
than 100,000
my $numbers = 100000; #sets the variable $numbers to 100000
my $random_number = int(rand($numbers)); #tells the pro-
gram to generate a random number
print (OUT $random_number, “\n”); #prints the random
number
$i++; #increases the value of $i by 1
}
close (OUT); #closes the program
exit 0; #exits the program
Save the file as Numbers. To save as a Perl file, select “Perl
source file (*.pl, *.pm, *.plx)” from the drop-down menu next
to “Save as type.” Make sure that the file is saved in the same
directory as Perl. Open Command Prompt, or similar com-
mand line program. In Windows, Command Prompt can be
found by following Start ->All Programs ->Accessories
->Command Prompt. Change directory from the current
directory to the folder where you have installed Perl (see Note
5). To change a directory, type cd followed by the new direc-
tory (e.g., cd C:\Perl). Your Command Prompt should now
read C:\Perl>, or something very similar. Type perl Numbers.

pl and press enter. This will create a file called “Random
Numbers.txt.”
2. Open a file containing data for one of your genomic regions of
interest (e.g., Genes.csv). Do not save any of the following
changes to this file. It is important to keep all rows in the exact
same order between all of your files, because eventually all of
the data will be put together into one file. To avoid saving any
changes accidentally, open the file and immediately save it
under a different name (e.g., GenesCOPY.csv). Sort your
spreadsheet in descending order, based on the column that
contains information on whether each gene is imprinted. This
should be column D. Once you have done this, those genes
with a 1 will be on top. Copy the genes with a 1 to a new file.
Open the file named “Random Numbers.txt.” You should see
a single column of numbers. Paste the column of numbers
from “Random Numbers.txt” into the first empty column of
your new file. Delete any numbers that extend past the rows
containing your data. To do this, click on the first cell which
contains a random number beyond your rows of data. While
holding down the shift key, double click on the bottom bound-
ary of the cell. You should now have highlighted to the bottom
of the column of random numbers. Right-click and select
delete. Now, sort your spreadsheet in ascending order. To do
this, select “sort” from the “data” menu. Copy approximately
three-quarters of the rows from this sheet to a new Excel
spreadsheet. This will become part of the data set you will use
to train your computational prediction models. It is important
to reserve a subset of genes that were not used for training so
that you can test how well your trained models perform before
running them on your genome-wide data. The three-quarter
to one-quarter split is somewhat arbitrary, and can therefore be
changed to suit your needs, but it is important to strike a bal-
ance between having more genes in the training set (which
gives better opportunities for training) and having an indepen-
dent set of genes for testing (which is essential). Now, delete
the column containing the random numbers and save your
new sheet with an appropriate name (e.g., Genes Training.
csv). Copy the remaining one-quarter of your known imprinted
genes to a new Excel spreadsheet. This will become part of the
data set you will use to test the performance of your computa-
tional prediction models. Then, delete the column containing
the random numbers and save your new sheet with an appro-
priate name (e.g., Genes Test.csv). Now, go back to your
GenesCOPY.csv file and copy the genes with a 0 to a new file.
Again, paste the column of numbers from “Random Numbers.
txt” into the first empty column of your new file. Delete any
numbers that extend past the rows containing your data. To do
this, click on the first cell which contains a random number

beyond you rows of data. While holding down the shift key,
double click on the bottom boundary of the cell. You should
now have highlighted to the bottom of the column of random
numbers. Right-click and select delete. Sort your spreadsheet
in ascending order. To do this, select “sort” from the “data”
menu. Copy the first three-quarters of the rows from this sheet,
add it to the bottom of the Excel spreadsheet containing your
list of training genes (e.g., Genes Training.csv), delete the col-
umn containing the random numbers, and save the file. Copy
the remaining one-quarter of your known imprinted genes,
add it to the bottom of the Excel spreadsheet containing your
list of test genes (e.g., Genes Test.csv), delete the column con-
taining the random numbers, and save the file. If your original
file is still open (e.g., Genes.csv), close your original file con-
taining data for one of your genomic regions of interest genes
without saving any changes. Repeat this process for each
genomic regions you are interested in (e.g., gene body, introns,
exons, 5¢ UTRs, 3¢ UTRs, and any upstream or downstream
genomic intervals). Now, you have collected all of the epige-
nomic and sequence data you need to start model training.
3.6. Data Analysis: This step uses R to determine which of your features of interest are
Calculation of correlated with known imprinted genes.
Correlation
1. Download and install [R] from http://www.r-project.org/ (59).
Coefficients
2. Often, you will want to know how well each feature you have
collected data on correlates with Imprinted status in each of
the genomic intervals. The examples discussed below deal only
with the most straightforward correlation calculations. For
instance, the examples given below refer to 11 classes of
genomic regions (the entire gene body, introns, exons, 5¢
UTR, 3¢ UTR, and 1, 10, and 100 kb upstream and down-
stream of each gene). One could also calculate correlation with
imprinted status based on combinations of these regions.
Correlation could also be calculated for combinations of epige-
netic features, or combinations of both genomic regions and
epigenetic features, with imprinted status. To determine the
degree of correlation between a given feature in a given
genomic region and imprinting, you can calculate both a cor-
relation coefficient and a corresponding p-value using R. First,
open R. Go to File, and select New Script. Then, paste or type
the text below into the script editing box that opens up (the
example shown is for the gene body, see Note 6):
setwd(“c:/Perl”) # where I want my working directory. Tells the
#program where to look #for files
dataG<- read.csv(“Genes.csv”, header = TRUE)[1:xxx,] #tells
the #program which file to load
t((cor(dataG$Imprinted, dataG[,4:yyy]))) #calculates correlation

#coefficients and prints #them in an easy to use format
Before running this script, you will need to make a few changes.
You may need to change the first line, where you set your work-
ing directory. To do this, see Note 15. At the end of the second
line, where it reads [1:xxx,] the xxx needs to be changed to a
number. This number will depend on how many rows are there
in your input file (e.g., Genes.csv). Once you have determined
how many rows are there in this file, simply substitute this
number for xxx. Likewise, at the end of the script, where it
reads [,4:yyy], the yyy needs to be changed to a number. This
number depends on how many features of interest you are
working with. The simplest way to find this number is to type
setwd(“c:/Perl”) # where I want my working directory. Tells the
#program where to look for files
dataG<- read.csv(“Genes.csv”, header = TRUE)[1:xxx,] #tells
length(dataG)
into your R script editing box, and then run the script. To run
a script in Windows, highlight all of the text in your script and
press both control and the r key at the same time. A number
should appear on the screen, and this will tell you how many
columns are there in your file. You can also open the file in
Excel and manually count the number of columns. Replace yyy
with this number. Then, run the script. To run a script in
Windows, highlight all of the text in your script and press both
control and the r key at the same time. To determine how to
run in a script using other operating systems, right click within
the script editing box. When the script has finished running,
you should see two columns, one with the names of all of your
features of interest, and the next with numbers, which are your
correlation coefficients. Copy the columns and open a new
Excel file. Label the first column “Features” and the second
column “Genes,” or whatever genomic interval you are exam-
ining. Click on the cell below “Features” and then select “Paste
Special” by right-clicking. In the box that pops up, select
“Text” and click OK. When you repeat this process, you can
delete the additional “Features” columns leaving only the cor-
relation coefficients.
3. To calculate p-values for your correlation coefficients, paste or
type the text below into your script editing box in R (the exam-
ple shown is a continuation of the example in step 1 and deals
with the gene body, see Note 6). The script below is taken from
http://tolstoy.newcastle.edu.au/R/help/01c/2272.html.
cors<-cor(dataG[,4:yyy]) #saves columns 4-yyy in a variable
#called cors
#the lines below calculate the correlation coefficient p-values

cor.prob<- function(X, dfr = nrow(X) - 2) {
R<- cor(X)
above<-row(R)<col(R)
r2<- R[above]^2
Fstat<-r2 * dfr/(1 - r2)
R[above]<-1 - pf(Fstat, 1, dfr)
R}
probs<-cor.prob(dataG[,4:33]) #stores correlation coefficient
p-#values for columns 4-yyy in a variable named probs
as.matrix(probs[1,]) #prints the correlation coefficient p-values
#in an easily useable format
Do not forget to change xxx to the number of lines contained
in your training file. And, do not forget to change yyy to the
number of columns contained in your file. Run the script.
When the script has finished running, you should see two col-
umns, one with the names of all of your features of interest,
and the next with numbers, which are your correlation
coefficients. Copy the columns and open a new Excel file. To
do this, label the first column “Features” and the second col-
umn “Genes,” or whatever genomic interval you are examin-
ing. Click on the cell below “Features,” and then select “Paste
Special” by right-clicking. In the box that pops up, select
“Text” and click OK. When you repeat this process, you can
delete the additional “Features” columns, leaving only the
p-values of the correlation coefficients.
4. Repeat steps 1 and 2 for all of your genomic regions of interest
(e.g., gene body, introns, exons, 5¢ UTRs, 3¢ UTRs, and any
upstream or downstream genomic intervals). For each run,
you will need to make a few changes. First, do not forget to
change xxx to the number of lines contained in your input file.
Second, do not forget to change yyy to the number of columns
contained in your file. Also, in the script for step 1, you will
need to change the input file (e.g., Genes.csv to 100kbUP.csv),
as well as the variable called “dataG.” If you wanted to calcu-
late correlation coefficients for the region 100 kb upstream of
all known genes, this might be changed to data100u. For
example (the example shown is for the region 100 kb upstream
of each gene, see Note 6):
setwd(“c:/Perl”) #tells the program where to look for files
data100u<- read.csv(“100kbUP.csv”, header = TRUE)
[1:xxx,] #tells the program which file to load
t((cor(data100u$Imprinted, data100u[,4:yyy]))) #calculates
#correlation coefficients and prints them in an easy to use
format
In the script for step 2, you will have to make some similar
changes. Again, you will need to change the variable called
“dataG.” Do not forget to change xxx to the number of lines
contained in your training file. And, do not forget to change
yyy to the number of columns contained in your file. Finally,
you may need to change the first line where you set your work-
ing directory. To do this, see Note 15. Using the same exam-
ple, if you wanted to calculate p-values for the correlation
coefficients for the region 100 kb upstream of all known genes,
this might be changed to data100u. For example:
cors<-cor(data100u[,4:yyy]) #saves columns 4-yyy in a variable
#called cors
#the lines below calculate the correlation coefficient p-values
cor.prob<- function(X, dfr = nrow(X) - 2) {
R<- cor(X)
above<-row(R)<col(R)
r2<- R[above]^2
Fstat<-r2 * dfr/(1 - r2)
R[above]<-1 - pf(Fstat, 1, dfr)
R}
probs<-cor.prob(data100u[,4:yyy]) #stores correlation coefficient
#p-values for columns 4-#yyy in a variable named probs
as.matrix(probs[1,]) #prints the correlation coefficient p-values
#in an easily useable format
3.7. Model Training In this step, you will use R to train computational prediction mod-
els for each of your genomic regions of interest. At this point, the
computer is told which genes within the data set are imprinted and
which genes are not. This allows the computer to pick out features
that are important for distinguishing imprinted from non-imprinted
genes. Again, only the most straightforward examples of model
training will be discussed below. For instance, the following exam-
ples train models using 11 classes of genomic regions (the entire
gene body, introns, exons, 5¢ UTR, 3¢ UTR, and 1, 10, and 100 kb
upstream and downstream of each gene). One could also build
models using combinations of these genomic regions. Furthermore,
models could be built using combinations of epigenetic features,
or combinations of both genomic regions and epigenetic features.
1. The next step is to train the model using the training set files
you created in Subheading 3.5 above. To do this, paste or type
the text below into your R script editing box (the example
shown is for the region 100 kb upstream of each gene, see
Note 6):
setwd(“c:/Perl”) # where I want my working directory
Data100u<- read.csv(“Genes Training.csv”, header = TRUE)

[1:xxx,],] #tells the program which file to load
Variables100u<- data100[,5:yyy] #stores data contained in
#columns 4-yyy in a variable named Variables100u
#the following lines specify your models
Base100u<- glm(data100$Imprinted ~ 1, family = binomial)
Full100u<- glm(data100u$Imprinted ~ variables100u[,1] +
variables100u[,2] + variables100u[,3] + variables
100u[,4] + variables100u[,5] + variables100u[,6] + variables
100u[,28] + variables100u[,29], family = binomial)
#removes unnecessary variables from model
Res100u < - step(full100u, scope = list(upper = 100u, lower = ~1),
direction = “both”, trace = FALSE)
#prints a summary of which variables are included in the model
#and the p-value of each variable
summary(res100u)
in your training file. And, do not forget to change yyy to the
number of columns contained in your file. Finally, you may
need to change the first line where you set your working direc-
tory. To do this, see Note 15. Also, you will notice that
variables100u[,x] range from 1 through 29 in the example
above. If you have more features of interest or fewer features of
interest, you will need to adjust the number of variables100u
entries. For example, if you have 8 features of interest, you will
have variables100u 1 through 8. Likewise, if you have 40 fea-
tures of interest, you will have variables100u 1 through 40. To
determine how many features of interest you have, type
length(data100) into your R script editing box and run the
script. The output will be the number of columns in your input
file. However, remember that the first three columns contain
information about Chromosome, Start, and End locations.
Also, the fourth column contains information regarding
whether each gene is imprinted. Therefore, the number of fea-
tures of interest will be the number returned by length(data100)
minus 4. Once you have made the necessary adjustments, run
the script. When the script has finished running, you will have
a trained model for this genomic region.
2. Repeat step 4 for all of your genomic regions of interest (e.g.,

gene body, introns, exons, 5¢ UTRs, 3¢ UTRs, and any upstream
or downstream genomic intervals). For each run, you will need
to make a few alterations. To change between regions, alter the
text in bold accordingly. For example, to move from 100 kb up
to 3¢ UTR (see Note 6):
data3utr < - read.csv(“3UTR Training.csv”, header = TRUE)
[1:xxx,] #tells the program which file to load
variables3utr < - data3utr[,5:yyy] #stores data contained in
#columns 4-yyy in a variable #named Variables100u
#the following lines are specify your models
base3utr < - glm(data3utr$Imprinted ~ 1, family = binomial)
full3utr < - glm(data3utr$Imprinted ~ variables3utr[,1] +
variables3utr[,2] + variables3utr[,3] + variables3utr[,4] +
variables3utr[,29], family = binomial)
#removes unnecessary variables from model
res3utr < - step(full3utr, scope = list(upper = full3utr,
lower = ~1), direction = “both”, trace = FALSE)
#prints a summary of which variables are included in the model
#and the p-value of each variable
summary(res3utr)
Also, do not forget to change xxx to the number of lines con-
tained in your training file. Likewise, do not forget to change
yyy to the number of columns contained in your file. And, you
may need to change the first line where you set your working
directory. To do this, see Note 15. Finally, do not forget to
adjust the number of variables3utr to match the number of
features of interest you have.
3.8. Model Testing Here you will use R and the second data set of genes, which were
not used for training your models, to test how well your trained
models perform before running them on your genome-wide data.
1. Once your models have been trained, you will want to test how
well the trained models perform in predicting imprinted status.
To do this, paste or type the following script into your R script

editing box (the example shown is for introns, see Note 6):
dataI < - read.csv(“Introns Test.csv”, header = TRUE)[1:xxx,]
#tells the program which file to load
predictedI < -predict.glm(resI,newdata = dataI,type = “response”)
#predicts whether each #gene in your test file is imprinted
and #saves the information in the variable predictedI
names(which(predictedI > 0.8)) #prints the index of each gene
#predicted to be imptinted at 80% confidence
in your test file. Also, you may need to change the first line
where you set your working directory. To do this, see Note 15.
Run the script. In this example, we are testing how well the
model for introns performs. The output from this script will
tell you the index of any genes that are predicted as imprinted.
To determine which gene the index corresponds to, simply
find that row number in your test file. The gene name can be
obtained by searching the UCSC Genome Browser using the
genomic coordinates of the gene. At this point, you will want
to record whether each gene in your test set was called as
imprinted or not imprinted, and whether that call was accu-
rate. If the results are not satisfactory, you can change the strin-
gency of the calls. To do this, locate the line above which reads
names(which(predictedI > 0.8)). To make calls more stringent,
increase the number at the end of the line from 0.8. To make
the calls less stringent, decrease the number at the end of the
line from 0.8. Repeat this process for each genomic region you
are interested in. For each run, you will need to make a few
alterations. To change between regions, alter the text in bold
accordingly. For example, to move from Introns to Exons (see
Note 6):
dataE < - read.csv(“Exons Test.csv”, header = TRUE)[1:xxx,]
#tells #the program which file to load
predictedE < -predict.glm(resE, newdata = dataE,
type = “response”) #predicts whether #each gene in your test
file is imprinted and #saves the information in the variable
#predictedE
names(which(predictedE > 0.8)) #prints the index of each gene
#predicted to be imptinted at 80% confidence
If your models are not performing as well as you had hoped,
you may want to try adding additional epigenetic or sequence
features of interest. You can also try increasing the number of
genes in your training set, if you have enough genes to do so.
If not, you can increase the ratio of non-imprinted to imprinted

genes in your training set. Finally, you can adjust the confidence
level (in the line predictedE > 0.8) to see whether a different
cutoff gives better results for your particular data set.
3.9. Genome-Wide Here you will use R to run your trained and tested computational
Prediction of prediction models on the genome-wide data you have gathered on
Imprinted Status each of your features of interest across each of your genomic
regions of interest.
1. Once you have an idea of how well your models are perform-
ing, you can then run your models on your genome-wide data
sets to predict which genes in your genome of interest might
be imprinted. To run your models on your genome-wide data,
paste or type the following script into your R script editing box
(the example shown is for exons, see Note 6):
dataE < - read.csv(“Exons.csv”, header = TRUE)[1:xxx,] #tells
predictedE < -predict.glm(resE, newdata = dataE, type = “res-
ponse”) #predicts whether #each gene in your test file is imprinted
and #saves the information in the variable predictedE
ImpE < -names(which(predictedE > 0.8)) #saves the index of
each gene #predicted to be imprinted at 80% confidence to a
variable #called ImpE
ImpE #prints the data contained in ImpE to the screen
in your test file. Also, you may need to change the first line
where you set your working directory. To do this, see Note 15.
Run the script. The output from this script will tell you the
index of any genes that are predicted as imprinted. Remember,
you can change the stringency of the calls by locating the line
above which reads ImpE<-names(which(predictedI > 0.8)). To
make calls more stringent, increase the number at the end of
the line from 0.8. To make the calls less stringent, decrease the
number at the end of the line from 0.8. There may be several
hundred genes that are predicted as imprinted in each genomic
region, so you may want to wait until the next step to identify
individual genes. However, if you would like to determine
which gene the index corresponds to, simply find that row
number in your test file. The gene name can be obtained by
searching the UCSC Genome Browser using the genomic
coordinates of the gene. At this point, you will want to record
whether each gene in your test set was called as imprinted or
not imprinted. Repeat this process for each genomic region
you are interested in. For each run, you will need to make a
few alterations. To change between regions, alter the text in
bold accordingly. For example, to move from Introns to Exons

(see Note 6):
dataI < - read.csv(“Introns.csv”, header = TRUE)[1:xxx,]
#tells #the program which file to load
predictedI < -predict.glm(resI, newdata = dataI, type = “res-
ponse”) #predicts whether #each gene in your test file is imprinted
and #saves the information in the variable predictedI
ImpI < -names(which(predictedI > 0.8)) #saves the index of
each gene predicted to be imprinted at 80% confidence to a
variable called ImpI
ImpI #prints the data contained in ImpI to the screen
3.10. Compiling a In this step, you will use R to compile a list of all predicted imprinted
Candidate Gene List genes, genome-wide.
1. Now you are ready to compile a candidate list of imprinted
genes.
In the following example,
imp100u corresponds to 100 kb upstream.
imp5utr corresponds to 5¢ UTRs.
impG corresponds to the gene body.
impE corresponds to exons.
impI corresponds to introns.
imp3utr corresponds to 3¢ UTRs.
imp1d corresponds to 1 kb downstream.
To compile a single list of all genes predicted as imprinted
across all genomic regions of interest, paste or type this script
into your R script editing box (refer to the key above to deter-
mine which genomic interval corresponds to which shorthand
notation, see Note 6):
imp100u < -which(predicted100 > 0.8) #saves the genes which
are #predicted as imprinted at 80% confidence using data
from the #region 100 kb upstream of all genes in a variable
called imp100u
imp10u < -which(predicted10 > 0.8) #saves the genes which are #pre-
dicted as imprinted at 80% confidence using data from the
#region 10 kb upstream of all genes in a variable called imp10u
imp1u < -which(predicted1 > 0.8) #saves the genes which are
predicted #as imprinted at 80% confidence using data from the
region 10 kb #upstream of all genes in a variable called imp1u
imp5utr < -which(predicted5 > 0.8) #saves the genes which are
#predicted as imprinted at 80% confidence using data from the
#5¢UTRs of all genes in a variable called imp5utr
impG < -which(predictedG > 0.8) #saves the genes which are pre-
dicted #as imprinted at 80% confidence using data from within
each gene #body in a variable called impG
impE < -which(predictedE > 0.8) #saves the genes which are pre-
dicted #as imprinted at 80% confidence using data from the exons
of all #genes in a variable called impE
impI < -which(predictedI > 0.8) #saves the genes which are pre-
dicted #as imprinted at 80% confidence using data from the
introns of all #genes in a variable called impI
imp3utr < -which(predicted3utr > 0.8) #saves the genes which
are #predicted as imprinted at 80% confidence using data from
the #3¢UTRs of all genes in a variable called imp3utr
imp1d < -which(predicted1d > 0.8) #saves the genes which are
predicted #as imprinted at 80% confidence using data from the
region 1 kb #downstream of all genes in a variable called imp1d
imp10d < -which(predicted10d > 0.8) #saves the genes which are
#predicted as imprinted at 80% confidence using data from the
#region 10 kb downstream of all genes in a variable called imp10d
imp100d < -which(predicted100d > 0.8) #saves the genes which
are #predicted as imprinted at 80% confidence using data from
the #region 100 kb downstream of all genes in a variable #called
imp100d
impALL < -c(imp100u, imp10u, imp1u, imp5utr, impG, impE,
impI, imp3utr, imp1d, imp10d, imp100d) #stores all predicted
imprinted #genes from all of the regions above in a single vari-
able called #impALL
TableImpALL < -table(impALL) #changes the way the data in
impALL are #stored
names(which(TableImpALL > 8) #prints to the screen those genes
which #are predicted as imprinted by 8 or more of the genomic
regions of #interest
Do not forget that you may need to change the first line where
you set your working directory. To do this, see Note 15. If you
do not have one of the genomic regions included above, or if
you have a genomic region in addition to those listed above,
simply add or delete these regions as necessary. The output
from this script will list those genes that are predicted as
imprinted in 8 or more of your genomic regions of interest. To
change this threshold, simply replace the 8 in this line
names(which(TableImpALL > 8) with a more appropriate

number. The output from this script will tell you the index of
those genes that are predicted as imprinted by the number of
models you have specified (e.g., 8).
2. To determine which gene the index corresponds to, transfer
the index number of those genes that are predicted as imprinted
to a new Excel file, with one index number on each line of
column one. Save this file in tab-delimited text form (.txt) and
call it something appropriate (e.g., List.txt). Also, make sure
that you have saved this file in the same directory where you
have installed Perl (see Note 5). Then, paste or type the fol-
lowing program into Notepad++ (see Note 6):
my $usage = ‘GeneNames.pl
Extracts information from one file using search criteria form
another file.
USAGE:
./GeneNames.pl list.txt genes.txt output.txt
‘;
my $list = shift @ARGV || die “$usage\n”;
my $genes = shift @ARGV || die “$usage\n”;
open (LIST, “ < $list”)
file or dies trying
trying
while (<LIST>) {#tells the program what to do while the first file
is open
(@LIST) = split/\t/; # splits each line on tabs
$line = $LIST [0]; #defines a variable
open (GENES, “ < $genes”)
file or dies trying
while (<GENES>) {#loops through 2nd file
if ($. == $line) {
print (OUT $_, “\n”); #prints matches if found
}
}
close (GENES);
}
close (LIST); #closes first input file
close (GENES); #closes second input file
Save the file as GeneNames. To save as a Perl file, select “Perl
source file (*.pl, *.pm, *.plx)” from the drop-down menu next
to “Save as type.” Make sure that the file is saved in the same
directory as Perl. You will also need to check that the file called
Genes.txt, which you created in Subheading 3.1, step 3 above,
is saved in this directory as a .txt file. Open Command Prompt,
or similar command line program. In Windows, Command
Prompt can be found by following Start ->All Programs
->Accessories ->Command Prompt. Change directory from
the current directory to the folder where you have installed
Perl (see Note 5). To change a directory, type cd followed by
the new directory (e.g., cd C:\Perl). Your Command Prompt
should now read C:\Perl>, or something very similar. Type perl
GeneNames.pl List.txt Genes.txt GNOUT.txt and press enter.
This will create a file called “GNOUT.txt.” Depending on
what you decided to do in Subheading 3.1, step 2 above, either
your first or your last column should contain gene names.
3. Once you have your candidate list of imprinted genes, it is
essential to determine whether any are actually imprinted using
traditional methods of analysis.
4. Notes
1. The choice of which track to use depends on your particular

situation and the difference is that UCSC Genes contain gene
entries from several sources, including RefSeq, UniProt,
GenBank, CCDS, and comparative genomics. Other data sets,
for example the RefSeq genes data set, contain genes from a
single source, such as the NCBI RNA reference sequences
collection, only.
2. To save a file as a text (.txt) file, choose the option of Text (Tab
Delimited) from the drop-down menu next to “Save as type.”
3. Depending on your version of Excel, “sort ascending” may be
different. In some versions it may be represented as an A on
top of a Z, both of which are positioned next to an arrow
pointing down, and in others it may be called “sort A to Z.”
4. For your paste to be effective, you will have to paste values only.
To do this, use the “paste special” function and select values.
5. To determine where Perl is installed, follow Start ->Search and
search for “Perl” among “All files and Folders.” Once the
search has finished running, look for a folder named “Perl.”
If there is more than one folder named “Perl,” look for the one
that says C:\ next to it. Now, pull up your Command Prompt
window next to your search window so that you can see both
at the same time. Type cd into Command Prompt, but do not
press Enter. Click on the “Perl C:\” and drag and drop the
folder into your Command Prompt window. Command
Prompt should now read C:\Perl>. Press Enter. You have
successfully changed your directory.
6. The # symbol in Perl or R scripts indicates an author’s com-
ment and therefore is not run by the program. In the scripts
provided, I have tried to give a brief summary of what each line
in the script does, and these explanations can be found follow-
ing the # symbol.
7. Often, equivalent releases of the same genome assembly are
identified by different names. For a list of equivalent genome
assemblies, see http://genome.ucsc.edu/FAQ/FAQreleases.
html.
8. To obtain genomic coordinates for predicted G-quartet sites,
download the QUADPARSER program by directing your Web
browser here: http://www.quadruplex.org/?view=quadparser.
Choose the correct operating system for your computer and
follow the prompts to install the program. Download your
genome of interest from the UCSC Genome Browser Web site
at http://hgdownload.cse.ucsc.edu/downloads.html. Click
on the link for your organism of interest. On the next page,
find the assembly you are interested in and click on the link for
“Full data set.” Download “chromFa.tar.gz,” and unzip the
files to the same directory you have installed QUADPARSER.
Open CommandPrompt. Change directory to the location
where QUADPARSER is installed. To run the program, type
quadparser –n, followed by your input and output file names
(e.g., quadparser –n Chr1.txt Chr1GQ.txt) and press enter. If
you run into problems, QUADPARSER has a help page avail-
able at http://www.quadruplex.org/?view=quadparser_
instructions. A help function is also available by entering
quadparser –h into CommandPrompt and pressing Enter.

Once you have output files (e.g., Chr1GQ), for each chromo-
some, upload them all to Galaxy, one at a time by clicking on
correct species and genome assembly. Once all of your chro-
mosome files have been uploaded, select “Text Manipulation”
from the left-hand menu and click on “Concatenate queries.”
Select your first file from the drop-down menu under
“Concatenate query.” Then, click on the “Add new querys”
button above the “execute” button and select your second file
from the drop-down menu. Repeat this process until you have
added all of your chromosome files and press “Execute.” Once
your files have concatenated, cut all columns except the first
three, which contain information regarding genomic location.
In the left-hand “Tools” menu, click on “Text Manipulation,”
and then “Cut columns from a table.” Type the columns you
wish to keep into the box next to “Cut columns” (e.g., c1, c2,
c3). Press execute and wait for your job to finish. When it has
finished running, click on the name of the job. This should
expand the file window and allow you to save the file by click-
ing on the disk icon. Locate the file and open it with Excel.
Delete the header row, by right-clicking on the “1” to the left
of the header row and selecting delete. Then, save the file as a
.txt file, making sure to give it an appropriate name (e.g., GQs.
txt), and close the file.
9. When using this program, the input file name is the file Perl is
going to read DNA sequences from in order to calculate GC%
of each region you are interested in. You downloaded these
files in Subheading 3.2, step 8, and saved each file (the exam-
ple file name given was CGintron) from Galaxy. The file for-
mat is .bed, so the full name of you input file will have .bed at
the end (e.g., CGintron.bed). Your output file is the file to
which Perl will write the calculated GC%. Name this some-
thing appropriate (e.g., CGintronOUT) and use .txt as the
file type, as these can be easily opened in Excel (e.g.,
CGintronOUT.txt).
10. When using this program, the first input file name is the file
Perl is going to loop through line by line. This first input file
will contain your list of all known genes (e.g., Genes.txt). It is
very important that this is your first input file, because you
want your output to be in the same order as in your file con-
taining the list of all know genes (e.g., Genes.txt). Your second
input file name is the file containing the information Perl will
try to match within your first file. This second file will be your
list of all known imprinted genes (e.g., KnownImprinted.txt).
Your output file is the file to which Perl will write the informa-
tion regarding whether each gene is imprinted. Name this
something appropriate (e.g., Imprinted) and use .txt as the file
type, as these can be easily opened in Excel (e.g., Imprinted.
txt). As the information regarding whether each gene is
imprinted will not vary between genomic regions, this program
needs to be run only once.
will contain one of these lists: your list of introns, exons, 5¢
UTRs, or 3¢ UTRs. It is very important that one of these four
files is your first input file for two reasons. The first is because
you want your output to be in the same order as it is in these
files. The second reason is because this program is meant to be
run before the program called count (Subheading 3.4, step 2).
This program identifies those introns, exons, 5¢ UTRs, or 3¢
UTRs that contain each of your features of interest, and reports
these data in a format that the count program can use. The
count program, in turn, identifies which genes the introns,
exons, 5¢ UTRs, or 3¢ UTRs belong to and tallies the number
of occurrences of your features of interest. Your second input
file name is the file containing the information Perl will try to
match within your first file. This second file will contain data
on your features of interest (e.g., GC%, miRNA clusters, CTCF
binding sites, histone modification data, etc.). Your output file
is the file to which Perl will write the information regarding
whether each gene is imprinted. Name this something appro-
priate (e.g., miRNAexonsOUT_1) and use .txt as the file type,
as these can be easily opened in Excel (e.g., miRNAexon-
sOUT_1.txt). The _1 at the end of the file name is added
because the files we create here will be sent through the count
program and we will want to distinguish the output files from
the two programs (e.g., miRNAexonsOUT_1.txt versus miR-
NAexonsOUT.txt).
12. The files containing the histone ChipSeq data are very large. In
fact, some may contain millions of data points. Therefore, run-
ning programs that utilize these files may take quite a long
time on standard desktop computers. To speed up the process,
you may want to look into using a cluster, if one is available for
you to use.
will contain one of these lists: your list of all known genes, or
any upstream or downstream genomic regions you are inter-
ested in. It is very important that one of these types of files is
your first input file, because you want your output to be in the
same order as it is in these files. It is also important that you do

not use your list of introns, exons, 5¢ UTRs, or 3¢ UTRs as
your first input file. This will be explained further in a few sen-
tences. Your second input file name is the file containing the
information Perl will try to match within your first file. For
your list of all known genes, and any upstream or downstream
genomic regions, the second file will contain data on your fea-
tures of interest (e.g., GC%, miRNA clusters, CTCF binding
sites, histone modification data, etc.). However, for introns,
exons, 5¢ UTRs, and 3¢ UTRs, this formula differs. In these
cases only, you will be using the output file from Subheading 3.4,
step 1 (e.g., miRNAexonsOUT_1.txt). These files contain the
data in a form that the count program can use effectively. Your
output file is the file to which Perl will write the information
regarding whether each gene is imprinted. Name this some-
thing appropriate (e.g., miRNAgenesOUT) and use .txt as the
file type, as these can be easily opened in Excel (e.g., miR-
NAgenesOUT.txt).
14. To save a file as a comma-delimited file (.csv), choose the
option of CSV (Comma Delimited) from the drop-down menu
next to “Save as type.”
15. Depending on where you have installed Perl, the line where
you set your working directory may have to be altered. To
determine where Perl is installed, see Note 5 above.
References
1. Nikaido I, Saito C, Mizuno Y et al (2003) 8. Crespi B (2008) Genomic imprinting in the
Discovery of imprinted transcripts in the mouse development and evolution of psychotic spec-
transcriptome using large-scale expression trum conditions. Biol Rev Camb Philos Soc
profiling. Genome Res 13:1402–1409 83:441–493
2. Luedi PP, Hartemink AJ, Jirtle RL (2005) 9. Mackay DJ, Callaway JL, Marks SM et al
Genome-wide prediction of imprinted murine (2008) Hypomethylation of multiple imprinted
genes. Genome Res 15:875–884 loci in individuals with transient neonatal dia-
3. Wang X, Sun Q, McGrath SD et al (2008) betes is associated with mutations in ZFP57.
Transcriptome-wide identification of novel Nat Genet 40:949–951
imprinted genes in neonatal mouse brain. PLoS 10. Shao WJ, Tao LY, Gao C et al (2008) Alterations
One 3:e3839 in methylation and expression levels of
4. Babak T, Deveale B, Armour C et al (2008) imprinted genes H19 and Igf2 in the fetuses of
Global survey of genomic imprinting by tran- diabetic mice. Comp Med 58:341–346
scriptome sequencing. Curr Biol 18:1735–1741 11. Xie T, Chen M, Gavrilova O, Lai EW, Liu J,
5. Gregg C, Zhang J, Butler JE et al (2010) Sex- Weinstein LS (2008) Severe obesity and insulin
specific parent-of-origin allelic expression in resistance due to deletion of the maternal Gsa
the mouse brain. Science 329:682–685 allele is reversed by paternal deletion of the Gsa
6. Gregg C, Zhang J, Weissbourd B et al (2010) imprint control region. Endocrinology 149:
High-resolution analysis of parent-of-origin 2443–2450
allelic expression in the mouse brain. Science 12. Hatada I, Sugama T, Mukai T (1993) A new
329:643–648 imprinted gene cloned by a methylation-sensi-
7. Brideau CM, Eilertson KE, Hagarman JA et al tive genome scanning method. Nucleic Acids
(2010) Successful computational prediction of Res 21:5577–5582
novel imprinted genes from epigenomic fea- 13. Hayashizaki Y, Shibata H, Hirotsune S et al
tures. Mol Cell Biol 30:3357–3370 (1994) Identification of an imprinted U2af
binding protein related sequence on mouse 27. Hayward BE, Kamiya M, Strain L et al (1998)
chromosome 11 using the RLGS method. Nat The human GNAS1 gene is imprinted and
Genet 6:33–40 encodes distinct paternally and biallelically
14. Kaneko-Ishino T, Kuroiwa Y, Miyoshi N et al expressed G proteins. Proc Natl Acad Sci U S A
(1995) Peg1/Mest imprinted gene on chro- 95:10038–10043
mosome 6 identified by cDNA subtraction 28. Kamiya M, Judson H, Okazaki Y et al (2000)
hybridization. Nat Genet 11:52–59 The cell cycle control gene ZAC/PLAGL1 is
15. Kuroiwa Y, Kaneko-Ishino T, Kagitani F et al imprinted–a strong candidate gene for transient
(1996) Peg3 imprinted gene on proximal chro- neonatal diabetes. Hum Mol Genet 9:453–460
mosome 7 encodes for a zinc finger protein. 29. Meissner A, Gnirke A, Bell GW et al (2005)
Nat Genet 12:186–190 Reduced representation bisulfite sequencing
16. Luedi PP, Dietrich FS, Weidman JR et al (2007) for comparative high-resolution DNA methyla-
Computational and experimental identification tion analysis. Nucleic Acids Res 33:5868–5877
of novel human imprinted genes. Genome Res 30. Lister R, O’Malley RC, Tonti-Filippini J et al
17:1723–1730 (2008) Highly integrated single-base resolu-
17. Maeda N, Hayashizaki Y (2006) Genome-wide tion maps of the epigenome in Arabidopsis.
survey of imprinted genes. Cytogenet Genome Cell 133:523–536
Res 113:144–152 31. Guenther MG, Levine SS, Boyer LA et al
18. Plass C, Shibata H, Kalcheva I et al (1996) (2007) A chromatin landmark and transcrip-
Identification of Grf1 on mouse chromosome tion initiation at most promoters in human
9 as an imprinted gene by RLGS-M. Nat Genet cells. Cell 130:77–88
14:106–109 32. Heintzman ND, Stuart RK, Hon G et al (2007)
19. Pollard KS, Serre D, Wang X et al (2008) A Distinct and predictive chromatin signatures of
genome-wide approach to identifying novel- transcriptional promoters and enhancers in the
imprinted genes. Hum Genet 122:625–634 human genome. Nat Genet 39:311–318
20. Ruf N, Bähring S, Galetzka D et al (2007) 33. Wen B, Wu H, Bjornsson H et al (2008)
Sequence-based bioinformatic prediction and Overlapping euchromatin/heterochromatin-
QUASEP identify genomic imprinting of the associated marks are enriched in imprinted
KCNK9 potassium channel gene in mouse and gene regions and predict allele-specific
human. Hum Mol Genet 16:2591–2599 modification. Genome Res 18:1806–1813
21. Schulz R, Menheniott TR, Woodfine K et al 34. Li E, Beard C, Jaenisch R (1993) Role for
(2006) Chromosome-wide identification of DNA methylation in genomic imprinting.
novel imprinted genes using microarrays and Nature 66:362–365
uniparental disomies. Nucleic Acids Res 34:E88 35. Wu MY, Tsai TF, Beaudet AL (2006) Deficiency
22. Smith RJ, Dean W, Konfortova G et al (2003) of Rbbp1/Arid4a and Rbbp1l1/Arid4b alters
Identification of novel imprinted genes in a epigenetic modifications and suppresses an
genome-wide screen for maternal methylation. imprinting defect in the PWS/AS domain.
Genome Res 13:558–569 Genes Dev 20:2859–2870
23. Wolf JB, Cheverud JM, Roseman C et al (2008) 36. Delaval K, Govin J, Cerqueira F et al (2007)
Genome-wide analysis reveals a complex pat- Differential histone modifications mark mouse
tern of genomic imprinting in mice. PLoS imprinting control regions during spermato-
Genet 4:e1000091 genesis. EMBO J 26:720–729
24. Wood AJ, Roberts RG, Monk D et al (2007) A 37. Lindroth AM, Park YJ, McLean CM et al
screen for retrotransposed imprinted genes (2008) Antagonism between DNA and H3K27
reveals an association between X chromosome Methylation at the Imprinted Rasgrf1 Locus.
homology and maternal germ-line methyla- PLoS Genet 4:e1000145
tion. PLoS Genet 3:e20 38. Mikkelsen TS, Hanna J, Zhang X et al (2008)
25. Shibata H, Hirotsune S, Okazaki Y et al (1994) Dissecting direct reprogramming through inte-
Genetic mapping and systematic screening of grative genomic analysis. Nature 454:49–55
mouse endogenously imprinted loci detected 39. Nagano T, Mitchell JA, Sanz LA et al (2008)
with restriction landmark genome scanning The Air noncoding RNA epigenetically silences
method (RLGS). Mamm Genome 5:797–800 transcription by targeting G9a to chromatin.
26. Shibata H, Yoshino K, Muramatsu M et al Science 322:1717–1720
(1995) The use of restriction landmark genomic 40. Bell AC, Felsenfeld G (2000) Methylation of a
scanning to scan the mouse genome for endog- CTCF-dependent boundary controls imprinted
enous loci with imprinted patterns of methyla- expression of the Igf2 gene. Nature 405:
tion. Electrophoresis 16:210–217 482–485
41. Hark AT, Schoenherr CJ, Katz DJ et al (2000) 50. Zhao Z, Tavoosidana G, Sjolinder M et al
CTCF mediates methylation-sensitive enhancer- (2006) Circular chromosome conformation
blocking activity at the H19/Igf2 locus. Nature capture (4 C) uncovers extensive networks of
405:486–489 epigenetically regulated intra- and interchromo-
42. Hikichi T, Kohda T, Kaneko-Ishino T et al somal interactions. Nat Genet 38:1341–1347
(2003) Imprinting regulation of the murine 51. Qiu X, Vu TH, Lu Q et al (2008) A complex
Meg1/Grb10 and human GRB10 genes; roles deoxyribonucleic acid looping configuration
of brain-specific promoters and mouse-specific associated with the silencing of the maternal
CTCF-binding sites. Nucleic Acids Res 31: Igf2 allele. Mol Endocrinol 22:1476–1488
1398–1406 52. Barski A, Cuddapah S, Cui K et al (2007) High-
43. Kanduri C, Pant V, Loukinov D et al (2000) resolution profiling of histone methylations in
Functional association of CTCF with the insu- the human genome. Cell 129:823–837
lator upstream of the H19 gene is parent of 53. Lieberman-Aiden E, van Berkum NL, Williams
origin-specific and methylation-sensitive. Curr L et al (2009) Comprehensive mapping of long-
Biol 10:853–856 range interactions reveals folding principles of
44. Takada S, Paulsen M, Tevendale M et al (2002) the human genome. Science 326:289–293
Epigenetic analysis of the Dlk1-Gtl2 imprinted 54. Rhead B, Karolchik D, Kuhn RM et al (2010)
domain on mouse chromosome 12: implica- The UCSC Genome Browser database: update
tions for imprinting control from comparison 2010. Nucleic Acids Res 38:D613–D619
with Igf2-H19. Hum Mol Genet 11:77–86 55. Bao L, Zhou M, Cui Y (2008) CTCFBSDB: a
45. Yoon BJ, Herman H, Hu B et al (2005) Rasgrf1 CTCF-binding site database for characteriza-
Imprinting is regulated by a CTCF-dependent tion of vertebrate genomic insulators. Nucleic
methylation-sensitive enhancer blocker. Mol Acids Res 36:D83–D87
Cell Biol 25:11184–11190 56. Blankenberg D, Von Kuster G, Coraor N, et al
46. LaSalle JM, Lalande M (1996) Homologous (2010) Galaxy: a web-based genome analysis
association of oppositely imprinted chromo- tool for experimentalists. Curr Prot in Mol
somal domains. Science 272:725–728 Biol, Chapter 19, Unit 19.10.1–21
47. Murrell A, Heeson S, Reik W (2004) Interaction 57. Goecks J, Nekrutenko A, Taylor J et al (2010)
between differentially methylated regions par- Galaxy: a comprehensive approach for support-
titions the imprinted genes Igf2 and H19 into ing accessible, reproducible, and transparent
parent-specific chromatin loops. Nat Genet computational research in the life sciences.
36:889–893 Genome Biol 25:R86
48. Kato Y, Sasaki H (2005) Imprinting and loop- 58. Huppert JL, Balasubramanian S (2005)
ing: epigenetic marks control interactions Prevalence of quadruplexes in the human
between regulatory elements. Bioessays 27:1–4 genome. Nucleic Acids Res 33:2908–2916
49. Ling JQ, Li T, Hu JF et al (2006) CTCF 59. R Development Core Team (2008) R: A lan-
mediates interchromosomal colocalization guage and environment for statistical computing.
between Igf2/H19 and Wsb1/Nf1. Science R Foundation for Statistical Computing, Vienna,
312:269–272 Austria. URL http://www.R-project.org
Part III
Identifying the Regulatory Features of Imprinted Domains

Chapter 8
Engineering of Large Deletions and Duplications In Vivo

Louis Lefebvre
Abstract
Gene targeting in embryonic stem (ES) cells coupled with the site-specific Cre/loxP recombination system
offers unique opportunities to identify and analyze the roles of cis-acting sequences in the regulation of
imprinted gene expression. Although several different approaches have been described to engineer large
chromosomal rearrangements in ES cells, these strategies can be labor-intensive and often require several
subcloning of the original stem cells, therefore limiting the chances of obtaining germ line transmission of
the mutation introduced. Here we describe an alternative approach which is based on in vivo recombina-
tion, therefore limiting the number of steps performed in ES cells and allowing to take advantage of the
growing number of loxP insertional mutations already available in transgenic mice.
Key words: Gene targeting, Cre/loxP recombination, Embryonic stem cells, Targeted meiotic
recombination, TAMERE
1. Introduction
The Cre recombinase of bacteriophage P1 can catalyze site-specific

recombinations between loxP site in mammalian cells (1). Each
loxP site is a short 34-bp sequence consisting of a unique core of
8 bp, flanked by 13-bp inverted repeats: ATAACTTCGTATAat
gtatgcTATACGAAGTTAT (2). It is the relative orientation of the
core element which will determine the outcome of the recombi-
nation event: when present on the same chromosome in cis, loxP
sites inserted in the same orientation will give rise to a deletion of
the intervening sequences, whereas loxP sites in opposite orienta-
tions will generate inversion of the loxP-flanked region (3). These
basic properties of the Cre/loxP system have been exploited in a
number of different targeted and random approaches to generate
large sets of deletions in ES cells (4–6). The main disadvantages of
these approaches is that they typically require three subsequent
subcloning of ES cells: targeting of the two loxP sites defining the
breakpoints of the rearrangement, followed by Cre-mediated
137
138 L. Lefebvre
M P
loxP S
loxP
MI
Sycp1-Cre
MII
Del Dp
Fig. 1. Targeted meiotic recombination (TAMERE). Targeted loxP site insertions are generated
at nonallelic positions on the same chromosome, and in the same orientation relative to
the centromere (circle at top of acrocentric chromosome). Triple transgenic males carry-
ing the loxP site insertions on the maternal (M, black) and paternal (P, gray) homologues
of the targeted chromosome and the Sycp1-Cre transgene will undergo a Cre-mediated
recombination during chromosome pairing at meiosis I (MI). After meiosis II (MII) mature
gametes carrying the deletion of the intervening sequence (Del) or its duplication (Dp) will
be recovered in mature sperm and these new mutations can be established in a mouse
line by direct breeding of these “trans-loxing” males.
recombination via transient production of Cre from an expression

vector. Depending on the frequency at which the recombinants
are recovered, such approaches often rely on strong genetic selec-
tion conferred by the reconstitution of a functional selectable
marker at the recombination breakpoint.
An alternative approach overcoming some of these limitations
has been developed to obtain recombination between loxP sites
targeted at nonallelic positions on the two parental homologues
in vivo (7). Termed TAMERE, for targeted meiotic recombina-
tion, this strategy takes advantage of transgenic lines expressing
Cre in germ cells. In its original form, TAMERE was based on the
Sycp1-Cre transgene, producing Cre recombinase in primary sper-
matocytes, at a stage when loxP sites on homologous chromosomes
are brought in close proximity upon pairing at meiosis I. In males
carrying the Cre transgene as well as loxP sites on the two parental
homologues (termed trans-loxing males), Cre can catalyze the
recombination between the two loxP site in trans and produce the
expected deletion and duplication in germ cells (Fig. 1). These
new mutations are independently recovered in the progeny of
trans-loxing males without selection and frequencies of 0.1–20%
have been reported (7–10).
8 Engineering of Large Deletions and Duplications In Vivo 139
This chapter describes the steps required to introduce loxP site

insertions in the mouse genome by homologous recombination in
ES cells and the generation of novel duplication and deletion alleles
in vivo.
2. Materials
All ES cell work should be performed in a dedicated tissue culture

facility equipped with humidified incubators (37 °C, 5% CO2), a
laminar flow cabinet, tabletop centrifuge to spin down cells, an
inverted microscope for cell observation, and a stereomicroscope
with light base for colony picking. An electroporator for mammalian
cells (e.g., Bio-Rad GenePulser Xcell with CE Module 165-2661)
with electroporation cuvettes (0.4 cm electrode gap; e.g., VWR
89047-210, Bio-Rad 165-2088) is also required.
2.1. MEF and ES Cell 1. Mouse embryonic fibroblast (MEFs) resistant to neomycin,
Culture hygromycin, and puromycin can be prepared from transgenic
mouse lines (e.g., DR4; JAX #003208 (11)) following estab-
lished protocols (12) or purchased (e.g., Stemcell Technologies,
http://www.stemcell.com/).
2. MEF medium (MEFM): Dulbecco’s modified Eagle’s medium
(DMEM) supplemented with 10% fetal bovine serum (FBS) (see
Note 1). Some authors also add 2 mM L-glutamine (Invitrogen
25030-081) and antibiotics (100 units penicillin/100 μg
streptomycin; Invitrogen 15140-122).
3. ES medium (ESM) 500-ml bottle: 400 ml DMEM high glu-
cose (4.5 g/l; e.g., Millipore SLM-120-B), 2 mM L-glutamine
(5 ml of 200 mM stock; Invitrogen 25030-081), 0.1 mM non-
essential amino acids (5 ml of 10 mM stock; Invitrogen 11140-
050), 0.1 mM 2-mercaptoethanol (5 ml of 10 mM stock;
Sigma M7522; 70 μl of 14.3 M stock in 100 ml PBS), 1 mM
sodium pyruvate (5 ml of 100 mM stock; Invitrogen 11360-
070), 100 units penicillin/100 μg streptomycin (5 ml of
10,000 U/10,000 μg stock; Invitrogen 15140-122), 15% of
FBS (75 ml; ES qualified, e.g., Hyclone SH30070, StemCell
Technologies 06952) (see Note 2), and leukemia inhibitory
factor (LIF) at 10 μg/l (e.g., StemCell Technologies 02740,
Santa Cruz Biotech. sc-4989) (see Note 3).
4. 0.25% Trypsin: Add 10 ml of 2.5% trypsin (Invitrogen 15090-
046) to 90 ml of autoclaved EDTA in Hank’s buffered saline:
0.35 g NaHCO3, 0.4 g KCl, 0.06 g KH2PO4, 0.01 g phenol
red, 1.0 g glucose, 8.0 g NaCl, 0.09 g Na2HPO4⋅7H2O, and
0.2 g EDTA to 900 ml with distilled water.
140 L. Lefebvre
5. PBS (Ca2+ and Mg2+ free): 137 mM NaCl (8 g/l), 2.7 mM KCl
(0.2 g/l), 8.1 mM Na2HPO4⋅2H2O (1.44 g/l), and 1.76 mM
KH2PO4 (0.2 g/l) in distilled water. Adjust pH to 7.2 with
HCl and bring to 1 l with distilled water.
6. Mitomycin C: Prepare 100× (1 mg/ml) stock by dissolving
2 mg of Mitomycin C (Sigma M0503) in 2 ml PBS. Store at
−20 °C in 100 μl aliquots.
7. Gelatin: 0.1% in water (0.5 g in 500 ml), autoclaved. Store at
4 °C. To prepare gelatinized plates, add enough 0.1% gelatin
to cover the surface of the plate (4 ml per 10 mm plates, 100 μl
per well of 96-well plate), let sit briefly, aspirate gelatin solu-
tion, and let dry for 5 min.
8. Puromycin: Prepare 1 mg/ml stock (500×) by dissolving
10 mg of puromycin (Sigma P8833) in 10 ml of sterile distilled
water. Store at −20 °C in 1-ml aliquots. For R1 ES cells, we use
puromycin at a final concentration of 2 μg/ml (1 ml of 500×
stock per 500 ml bottle of ESM). Puromycin is also available as
a 10 mg/ml (5,000×) solution (Invitrogen A11138-02).
9. Geneticin (G418): Prepare 100 mg/ml stock (500×) by dis-
solving 5 g of G418 (e.g., Gibco 11811, Sigma G5013) in
50 ml of sterile distilled water. Store at −20 °C in 1-ml ali-
quots. For R1 ES cells, we use G418 at a final concentration of
200 μg/ml (1 ml of 500× stock per 500 ml bottle of ESM).
2.2. TAMERE 1. PCR primers to genotype the Cre mice. Forward: ATGTCC
AATTTACTGACCGTAC. Reverse: GTTTCACTGGTTATG
CGGCG. These primers amplify the first 356 bp of the Cre
coding region.
3. Methods
3.1. Gene Targeting: 1. Using recombinant DNA or recombineering (13) techniques,

Electroporation build a targeting vector to generate each loxP site insertions.
The general features of such vectors are shown in Fig. 2. The
arms of homology can be subcloned from an available genomic
clone, amplified from genomic DNA isogenic to the ES cell
used (see Note 4), or recovered from genomic BAC clones by
gap repair (14).
2. Linearize 50 μg of the targeting vector using a restriction
enzyme with a single cut site, positioned outside of the arms of
homology. Heat inactivate the enzyme, phenol extract the digest
and recover the linear DNA by ethanol precipitation. Wash the
pellet with a large volume of 70% ethanol, dry the pellet, and
resuspend in 25 μl of sterile distilled water (see Note 5).
Fig. 2. Targeting vector for loxP site insertion. Each of the chromosomal breakpoints of the
TAMERE rearrangements (deletion and duplication) is defined by a loxP site insertion on
the desired chromosome. Any existing loxP site insertion already available in transgenic
mouse can be used as one of the breakpoint. Additional loxP site insertions can be gener-
ated by targeting in ES cells. A typical targeting vector will contain a selectable marker
cassette (e.g., PGK-neo-pA) inserted in a continuous fragment of genomic DNA of 6–8 kb.
This cassette can be flanked by loxP sites, as shown here. This will allow to monitor exci-
sion of the cassette in vivo and will eliminate residual foreign genetic elements at the
rearrangement breakpoint. Enrichment for targeted events can also be provided by
the addition of a negative selection cassette (e.g., diphtheria toxin-A chain, DTA) outside
the smallest arm of homology.
Confirm the completion of digestion and estimate the final

concentration by agarose gel electrophoresis.
3. Thaw a vial of MEFs by heating rapidly in your hand or in
your 37 °C incubator (see Note 6). Using a 1-ml pipette gen-
tly collect MEFs from the vial and add drop by drop to a 14-ml
tube containing 10 ml of MEFM. Gently invert the tube and
pellet MEFs by centrifugation at 300 × g for 5 min. Aspirate
supernatant and resuspend the MEF pellet in MEFM to reach
a cell density of 2 × 105 cells/ml. Plate 10 ml on a 100 mm
dish (see Note 7).
4. Grow MEFs until they reach confluence (2–3 days) and split
cells onto two 150 mm dishes (~1:5 passage) as follows.
Aspirate MEFM and rinse 100 mm dish twice with PBS. Add
trypsin to cover the surface of the dish (~2 ml) and incubate
2–3 min at 37 °C. Recover MEFs in a 14-ml tube by adding
6 ml of MEFM to the dish to inactivate the trypsin and gently
pipetting up and down to generate a single cell suspension.
Pellet MEFs, aspirate supernatant, and resuspend cells in 4 ml
MEFM. Add 2 ml of MEFs to each of two 150 mm dishes
containing 23 ml of MEFM. Rock plates back and forth, right
and left and incubate at 37 °C, 5% CO2.
5. Grow MEFs until they reach confluence (2–3 days). Feeders
are prepared by mitotically inactivating MEFs with mitomycin
C. Aspirate MEFM and add 10 ml MEFM per plate. Add mito-
mycin C to 10 μg/ml (100 μl aliquot of 1 mg/ml stock), swirl
plates, and incubate for 2 h at 37 °C (see Note 8).
6. Aspirate MEFM and rinse plates twice with PBS. Trypsinize as
described in step 4, and resuspend combined feeder pellet in
MEFM to bring to 2 × 105 cells/ml (~100 ml from two
150 mm plates).
142 L. Lefebvre
7. Seed feeders on gelatinized plates as follows: One 60 mm plate

(5 ml), four 100 mm plates (10 ml each), two 96-well plates
(200 μl per well), using a multichannel pipette (see Note 9).
8. Thaw a vial of germ line-competent ES cells as described in
step 3, using ESM. Resuspend the ES cell pellet in ESM to
reach a cell density of 106 cells/ml. Seed 1 ml on the 60 mm
feeder plate containing 4 ml of ESM. Gently rock the plate and
incubate at 37 °C, 5% CO2.
9. Grow ES cells for 48 h, changing medium every day. Trypsinize
as described in step 4, using ESM (see Note 10). Resuspend pellet
in 4 ml of ESM and seed 2 ml of ES cell suspension per 100 mm
feeder plate containing 8 ml of ESM. This is a ~1:5 passage.
Gently rock the plate and incubate at 37 °C, 5% CO2.
10. Grow ES cells for 48 h, changing medium every day.
11. The day of the electroporation, change medium in the morn-
ing. In the afternoon, trypsinize ES cells as above and resus-
pend the combined ES cell pellets in 2 ml of cold PBS. Count
cells from a 1:10 dilution in PBS using a hemocytometer. For
each electroporation, pellet 5.6 × 106 cells and resuspend in
0.8 ml of cold PBS (7 × 106 cells/ml) (see Note 11).
12. Put electroporation cuvettes (0.4 cm gap size) on ice and set
the electroporator for one pulse of 250 V at 500 μF.
13. Using a 1-ml pipette transfer 0.8 ml of ES cells to each cuvette
and add the DNA (20–40 μg of linear targeting vector).
Gently mix.
14. Deliver the electroporation pulse (time constant should be of
1.7–1.9 ms) and let the cuvette sit on ice for 30 min.
15. Transfer the cells to a 14-ml tube containing 5 ml of ESM and
seed two 100 mm feeder plates containing 8 ml of ESM with
2.8 ml of the ES cell suspension. Incubate at 37 °C, 5% CO2.
16. The next day, change the medium and start drug selection.
Use puromycin at 2.0 μg/ml or G418 at 200 μg/ml (see
Note 12).
17. Change the medium every day. Colonies big enough for
picking should appear 8–10 days after electroporation (see
Note 13).
3.2. Gene Targeting: Although the description of colony picking, analysis and expansion
Growth and Analysis is beyond the scope of this chapter, as well as the formation of
of ES Clones in germ line chimeras, the reader can refer to detailed description of
96-Well Plates these steps in the literature (12, 15).
3.3. TAMERE The generation of new deletion and duplication alleles in vivo by
TAMERE requires three simple breeding steps involving mice car-
rying the Sycp1-Cre transgene (see Note 14) as well as the two
Fig. 3. Generation of new deletion (Del) and duplication (Dp) alleles in vivo by TAMERE.
In the breeding scheme shown here the two nonallelic loxP site insertions are denoted
loxA and loxB. The number of loxP sites present at each allele is referred to as 1lox and
2lox for single loxP site insertions and floxed alleles, respectively. If one or both alleles
already carry a single loxP site, the same steps are required. (a) First, mice carrying the
loxA allele are crossed with Sycp1-Cre transgenics. The reciprocal cross can also be done.
The goal is to obtain a male carrying the loxA allele and the Cre transgene. (b) In this male,
Cre is active in germ cells so transmission of 2loxA will delete the selectable marker at
high frequency. Consequently, close to half of the progeny from this hemizygous male
should contain the 1loxA allele and simple PCR genotyping distinguishing 1loxA from
2loxA can be used to confirm efficient germ line Cre activity. By breeding this male with
females carrying the 2loxB allele (hemizygotes, or homozygotes), trans-loxing males car-
rying the two loxP site insertions (compound hemizygotes) and the Cre transgene can be
obtained. (c) When bred to wild-type females (WT), trans-allelic recombination will occur
in the germ cells of trans-loxing males. PCR genotyping of the progeny can be used to
identify hemizygotes for each of the possible alleles. Segregation of the Cre transgene is
not shown here. Note that for all genotypes, the maternal allele is shown first.
nonallelic loxP site insertions on the same chromosome (Fig. 3).

For simplicity, the two loxP site insertion alleles are referred to as
loxA and loxB, and the number of loxP site present at each allele (1
or 2, for flowed alleles), as 1lox and 2lox. All animals are genotyped
by genomic PCR on ear punch lysates. Segregation of the Sycp1-
Cre transgene is monitored using the Cre primers given in
Subheading 2.
1. Cross 2loxA carrying mice with Sycp1-Cre transgenics to obtain
double transgenic males (Fig. 3a; see Note 15). Since Cre is
only active in germ cells in the Sycp1-Cre line (16), these ani-
mals are mosaic and only germ cells should contain the 1loxA
allele (see Notes 16 and 17).
2. Cross loxA Sycp1-Cre double transgenic males to females carrying
the second loxP site insertion, loxB, to obtain triple transgenic,
144 L. Lefebvre
trans-loxing males (Fig. 3b). In these male progeny the 2loxA

allele should have been efficiently converted to 1loxA during
germ line transmission.
3. Cross the trans-loxing males carrying both loxP site insertions
and the Cre transgene to wild type inbred or outbred females
and genotype progeny at weaning. Most progeny should be
hemizygous for either of the loxP insertion (1loxA and 1loxB).
PCR reactions should also be established to specifically identify
the two recombination products, the deletion and duplication
alleles. Unless they are associated with lethal phenotypes in
heterozygotes, they should be recovered at approximately the
same frequency (see Note 18).
4. Notes
1. The quality of FBS is less critical for MEFs and we use regular
(cheaper) tissue culture grade FBS for all MEF and feeder
growth.
2. ES cells are particularly affected by the quality of the serum used.
For small scale experiments, it is preferable to spend the money
on ES qualified serum (pretested). Otherwise, different batches
of serum can be tested for maintenance of good ES cell mor-
phology; order the optimum serum in large quantities (15).
3. Recombinant LIF can also be produced in bacteria as a GST
fusion.
4. In the case of ES cell lines established from F1 embryos, pure
genomic DNA from one of the two parental strains should be
used for PCR amplification.
5. Presence of excess salt in the DNA solution can cause arcing
during electroporation. Consequently the 70% ethanol washes
are critical.
6. To limit cell death, it is important to thaw the frozen vial
rapidly and immediately add its content in culture media to
dilute the freezing medium which contains DMSO.
7. Although this should not be required, using gelatinized plates
can increase the adherence of MEFs and feeders.
8. Alternatively, MEFs can be mitotically inactivated by expo-
sure to 6,000–10,000 rads in a gamma cell irradiator after
trypsinization.
9. The 60 mm feeder plate can be used within 2 h, but ideally the
feeders are allowed to settle overnight.
10. Since ES cells form tight colonies, increase the trypsin incuba-
tion to 4–5 min at 37 °C. Pipette the ES cell suspension several
times once ESM has been added to the plate, to disaggregate

cell clumps.
11. Alternatively, commercial electroporation buffers can be used
(e.g., Millipore/Chemicon ES-003-D).
12. Each drug selection has its own kinetics and cell death is not
expected on the same days. Puromycin is fat-acting and sensi-
tive cells with be killed within 2 days of drug selection. For
G418, this will take 4–5 days.
13. Colonies large enough for picking can clearly be seen with the
naked eye when the plate is observed from below.
14. Although all the published reports of TAMERE have relied on
the Sycp1-Cre transgenic line, there is no reason to believe that
other lines expressing Cre in the germ line could not also be
suitable for this approach. In fact, trans recombination can also
be obtained using a ubiquitous CMV-Cre line (L.L. unpub-
lished observation).
15. If only one of the two loxP insertions is present in mice as a
floxed allele (2lox), it is advantageous to introduce it first on
the Cre background, as conversion of this allele from 2lox to
1lox can be used to monitor the efficiency of the germ line Cre
in the second generation.
16. We have previously demonstrated this conditional germ cell-
specific deletion by Southern blot analysis of genomic DNA
isolated from liver (all 2loxA), or testis and sperm (mixture of
2loxA and 1loxA). This analysis also revealed that in some
Sycp1-Cre male, ubiquitous excision is observed in all tissue
analyzed, suggesting leakiness of the Cre transgene and recom-
bination early in development (9).
17. A previous study has reported the inactivation of loxP site by
DNA methylation following Sycp1-Cre mediated recombination
in the male germ line (17). However, these observations have
not been confirmed by others, and we have not observed such
loss of excision in second and even third generation recombi-
nation events in Sycp1-Cre males.
18. The frequency of TAMERE appears quite variable from one
locus to another and might also depend on the size of the
rearrangements expected. In four published studies using
TAMERE, each recombinant allele was recovered in 0.1–19%
of the progeny genotyped (7–10).
Acknowledgment
This work was supported by CIHR grant MOP-82863 and a

Canada Research Chair to LL.
146 L. Lefebvre
References
1. Sauer B, Henderson N (1988) Site-specific Beckwith-Wiedemann region. Hum Mol Genet

DNA recombination in mammalian cells by the 18:4255–4267
Cre recombinase of bacteriophage P1. Proc 10. Tarchini B, Huynh TH, Cox GA, Duboule D
Natl Acad Sci U S A 85:5166–5170 (2005) HoxD cluster scanning deletions iden-
2. Nagy A (2000) Cre recombinase: the universal tify multiple defects leading to paralysis in the
reagent for genome tailoring. Genesis 26:99–109 mouse mutant Ironside. Genes Dev 19:
3. Branda CS, Dymecki SM (2004) Talking about 2862–2876
a revolution: the impact of site-specific recom- 11. Tucker KL, Wang Y, Dausman J, Jaenisch R
binases on genetic analyses in mice. Dev Cell 6: (1997) A transgenic mouse strain expressing
7–28 four drug-selectable marker genes. Nucleic
4. Bilodeau M, Girard S, Hébert J, Sauvageau G Acids Res 25:3745–3746
(2007) A retroviral strategy that efficiently cre- 12. Nagy A, Gertsenstein M, Vintersten K,
ates chromosomal deletions in mammalian Behringer R (2003) Manipulating the mouse
cells. Nat Methods 4:263–268 embryo: a laboratory manual. Cold Spring
5. Ramírez-Solis R, Liu P, Bradley A (1995) Harbor Laboratory Press, Cold Spring
Chromosome engineering in mice. Nature 378: Harbor, NY
720–724 13. Sharan SK, Thomason LC, Kuznetsov SG,
6. Zheng B, Sage M, Sheppeard EA, Jurecic V, Court DL (2009) Recombineering: a homolo-
Bradley A (2000) Engineering mouse chromo- gous recombination-based method of genetic
somes with Cre-loxP: range, efficiency, and engineering. Nat Protoc 4:206–223
somatic applications. Mol Cell Biol 20: 14. Copeland NG, Jenkins NA, Court DL (2001)
648–655 Recombineering: a powerful new tool for
7. Hérault Y, Rassoulzadegan M, Cuzin F, mouse functional genomics. Nat Rev Genet
Duboule D (1998) Engineering chromosomes 2:769–779
in mice through targeted meiotic recombina- 15. Wurst W, Joyner A (1993) Production of
tion (TAMERE). Nat Genet 20:381–384 targeted embryonic stem cell clones. In: Gene
8. Kmita M, Fraudeau N, Hérault Y, Duboule D targeting: a practical approach
(2002) Serial deletions and duplications sug- 16. Vidal F, Sage J, Cuzin F, Rassoulzadegan M
gest a mechanism for the collinearity of Hoxd (1998) Cre expression in primary spermato-
genes in limbs. Nature 420:145–150 cytes: a tool for genetic engineering of the
9. Lefebvre L, Mar L, Bogutz A, Oh-McGinnis R, germ line. Mol Reprod Dev 51:274–280
Mandegar MA, Paderova J, Gertsenstein M, 17. Rassoulzadegan M, Magliano M, Cuzin F
Squire JA, Nagy A (2009) The interval (2002) Transvection effects involving DNA
between Ins2 and Ascl2 is dispensable for methylation during meiosis in the mouse.
imprinting centre function in the murine EMBO J 21:440–450
Part IV
Epigenetics of Imprinted Regions

Chapter 9
Methylated DNA Immunoprecipitation (MeDIP)

from Low Amounts of Cells
Julie Borgel, Sylvain Guibert, and Michael Weber
Abstract
Methylated DNA immunoprecipitation (MeDIP) is an immunocapturing approach for unbiased enrichment
of DNA that is methylated on cytosines. The principle is that genomic DNA is randomly sheared by soni-
cation and immunoprecipitated with an antibody that specifically recognizes 5-methylcytidine (5mC),
which can be combined with PCR or high-throughput analysis (microarrays, deep sequencing). The
MeDIP technique has been originally used to generate DNA methylation profiles on a genome scale in
mammals and plants. Here we provide an optimized version of the MeDIP protocol suitable for low
amounts of DNA, which can be used to study DNA methylation in cellular populations available in small
quantities.
Key words: DNA methylation, MeDIP, Cytosine, CpG, Profiling, Epigenomics, Microarrays, Deep
sequencing
1. Introduction
DNA methylation occurs on the carbon 5 of cytosines and plays

essential roles in genome regulation in a variety of organisms and
in both normal and disease contexts (1). To better understand the
role of this epigenetic mark, several strategies have been developed
to assess the distribution of cytosine methylation at a genome-wide
scale (2). Some of these technologies use methylation-sensitive
(e.g., HpaII) or methylation-specific (e.g., McrBC) restriction
enzymes, with the caveat that they are biased towards specific
restriction motifs. Other methods combine sodium bisulfite con-
version and deep sequencing, which offers a powerful readout at a
single-nucleotide resolution but requires large sequencing efforts
when applied genome-wide (3). Alternative strategies use affinity
purification of methylated DNA that can be coupled to microarray
hybridization or deep sequencing. These are based on the use of
149
150 J. Borgel et al.
methyl-binding protein domains (MBD) that recognize methylated

DNA, or in the case of MeDIP on the use of antibodies that
specifically recognizes 5-methylcytidine (5mC). These affinity
methods provide valuable tools for a rapid and unbiased profiling
of DNA methylation at more limited costs.
The principle of MeDIP is that genomic DNA is randomly
sonicated and immunoprecipitated with a monoclonal antibody
directed against 5mC (4). The methylated fraction of the genome
can be analyzed at a single-gene resolution by conventional PCR
and real-time PCR, or on a genome-wide scale by microarray
hybridization or deep sequencing. It is however important to keep
in mind that enrichment-based methods also have certain limita-
tions. First, they offer an incomplete resolution (defined by the
size of sonicated fragments), and for this reason bisulfite sequenc-
ing still remains the method of choice when detailed methylation
information at single nucleotide resolution is required. Second,
there is a confounding effect of the DNA sequence because the
methylation enrichment also depends on the local CpG concentra-
tion. Indeed, low MeDIP enrichments can indicate either an unm-
ethylated state or the absence of sufficient CpG targets in very
CpG-poor regions of the genome. This effect can be corrected by
applying bioinformatics normalization to obtain absolute cytosine
methylation levels with a relatively good accuracy (5–7). As a con-
sequence, it also appears that the accuracy of MeDIP measure-
ments decreases in regions that are very CpG-poor. The classical
MeDIP protocol was originally designed to work with relatively
large amounts of DNA (at least 2 μg) (8). Here we describe an
optimized protocol that can be used to immunoprecipitate methy-
lated DNA from as low as 20,000 cells (9) (Fig. 1).
2. Materials
2.1. Isolation 1. Eppendorf LoBind® and standard 1.5 ml microtubes.

of Genomic DNA 2. Lysis buffer: 20 mM Tris pH 8.0, 4 mM EDTA, 20 mM NaCl,
2% SDS.
3. Proteinase K, 10 mg/ml stock. Store at −20 °C.
4. Dry heating block for 1.5 ml microtubes.
5. PCI (phenol–chloroform–isoamyl alcohol 25:24:1).
6. Linear polyacrylamide (LPA), 5 mg/ml stock, used as a
coprecipitant.
7. Refrigerated microcentrifuge.
8. Qubit® Fluorometer (Invitrogen) for quantification of low
amounts of nucleic acids.
9 MeDIP From Low Amounts of Cells 151
Fig. 1. Principle of MeDIP (methylated DNA immunoprecipitation). Genomic DNA is randomly sheared by sonication and
immunoprecipitated with an antibody that recognizes 5-methylcytidine (5mC Ab). A portion of the sonicated DNA is left
untreated and serves as input control. When MeDIP is performed on low amounts of starting DNA, a whole genome
amplification (WGA) step is performed on the input and methylated DNA. Enrichments in the methylated fraction can be
measured at a single gene resolution by real-time PCR, or on a global scale by microarray hybridization and deep sequenc-
ing. The deep sequencing image capture is reprinted by permission from Macmillan Publishers Ltd: Nature Biotechnology
28:1097-105, © 2010 -12).
2.2. Sonication 1. Eppendorf LoBind® 1.5 ml microtubes.

2. Diagenode Bioruptor® sonicator (standard model), with an
automated cooling system that allows for continuous cooling
of the water bath.
3. Equipment for small size agarose gel electrophoresis.
2.3. Immuno- 1. Eppendorf LoBind® 2 and 1.5 ml microtubes, and standard

precipitation of 1.5 ml microtubes.
Methylated DNA 2. Magnetic rack for microtubes, used for recovering the mag-
netic beads.
3. Dry heating block for 2 ml microtubes, with shaking.
4. IP buffer 10×: 100 mM Na-phosphate pH 7.0, 1.4 M NaCl,
0.5% Triton X-100. Store at room temperature.
5. 1 M Na-phosphate pH 7.0 buffer: Mix 39 ml 2 M monobasic

sodium phosphate (NaH2PO4) (276 g/l), 61 ml 2 M dibasic
sodium phosphate (Na2HPO4) (284 g/l), and 100 ml H2O.
6. IP buffer 1×: Dilute 1 ml IP buffer 10× in 9 ml H2O. Store at
room temperature.
7. Mouse anti 5-methylcytidine monoclonal antibody, clone 33D3,
available at a 1 mg/ml concentration from various suppliers
such as Eurogentec or AbD Serotec. Other mouse monoclonal
antibodies such as the ones developed by Diagenode work
equally well. Store the antibody as 5 μl aliquots at −20 °C.
8. Vortex Genie 2 shaker with a platform for microtubes, placed
at room temperature.
9. Overhead rotator for microtubes, placed in a 4 °C cold room.
10. Magnetic beads: Dynabeads M-280 Sheep anti-mouse IgG
(Invitrogen).
11. PBS-BSA 0.05%: Mix 9.5 ml PBS with 0.5 ml BSA at 10 mg/
ml concentration.
12. Proteinase K digestion buffer: 50 mM Tris pH 8.0, 10 mM
EDTA, 0.5% SDS.
13. Proteinase K, 10 mg/ml stock. Store at −20 °C.
14. PCI (phenol–chloroform–isoamyl alcohol 25:24:1).
15. Linear polyacrylamide (LPA), 5 mg/ml stock, used as a copre-
cipitant.
16. Refrigerated microcentrifuge.
2.4. Amplification 1. Genomeplex® complete whole genome amplification kit WGA2

and Analysis (Sigma-Aldrich).
2. Real-time PCR reagents and apparatus.
3. Methods
3.1. Isolation This protocol is suitable for isolating genomic DNA from 20,000
of Genomic DNA to 200,000 mammalian cells. If extracting DNA from higher num-
ber of cells, please refer to the standard MeDIP protocol (8). The
use of LoBind microtubes in the initial step allows to minimize the
loss of DNA during the procedure.
1. Resuspend the cells in a LoBind 1.5 ml microtube in 300 μl
lysis buffer containing 20 μl proteinase K (10 mg/ml stock)
(see Note 1).
2. Incubate at 55 °C in the dry heating block for 3 h.
Fig. 2. Example of sonicated DNA migrating on a 1% agarose gel and stained with ethidium
bromide. In this experiment, we sonicated 1 μg mouse genomic DNA in a volume of
150 μl H2O and loaded 100 ng on the agarose gel. The numbers above the gel indicate the
number of 30 s sonication pulses, which shows that 12 pulses leads to an optimal sonica-
tion under these conditions. Ideally, sheared DNA fragment should have an average size
of 400 bp and be no longer than 1,000 bp.
3. Extract with 1 volume PCI. Transfer the upper phase in a new

standard microtube (see Note 2).
4. Precipitate the DNA with 3 volumes (900 μl) ethanol containing
300 mM NaCl. Add 1 μl LPA if the amount of cells is <100,000.
Store at −20 °C overnight.
5. Centrifuge for 40 min at full speed at 4 °C.
6. Wash the pellet with 500 μl ethanol 70% and centrifuge for
20 min at full speed at 4 °C.
7. Resuspend the pellet in 30 μl H2O.
8. Quantify the amount of DNA with the Qubit fluorometer
(see Note 3).
3.2. Sonication Genomic DNA is randomly sheared by sonication to generate frag-

ments between 200 and 1,000 bp. This step is crucial because the
size of the DNA fragments will determine the resolution of the
MeDIP assay.
1. Dilute 100–1,000 ng of genomic DNA in 150 μl H2O in a
1.5 ml LoBind microtube.
2. Sonicate 12 times for 30 s (with 30 s intervals) with the
Bioruptor (see Note 4).
3. If possible, verify the efficiency of the sonication by loading at
least 50 ng on a 1% agarose gel. Ideally, the sheared fragments
should have an average size of 400 bp and be no larger than
1,000 bp (Fig. 2) (see Note 5).
3.3. Immuno- The sonicated DNA is then immunoprecipitated with a monoclonal

precipitation of antibody directed against 5-methylcytidine (5mC). Importantly, a
Methylated DNA portion of the sonicated DNA (at least 10 ng) should be left
untreated to serve as input control. We describe here on optimized
immunoprecipitation protocol for 75–100 ng sonicated DNA. For
immunoprecipitation of larger amounts of DNA, you can refer to
the standard MeDIP protocol (8).
1. Dilute 75–100 ng sonicated DNA in 135 μl H2O in a 2 ml
LoBind microtube (see Note 6).
2. Denature at 95 °C for 10 min in the dry heating block, and
immediately cool on ice.
3. Add 15 μl IP buffer 10×.
4. Add 1/5 μl 5mC antibody (see Note 7).
5. Incubate for 2 h at 4 °C on the overhead rotator (see Note 8).
6. Prewash 2 μl magnetic beads with 500 μl PBS-BSA 0.05% in a
standard 1.5 ml microtube. Incubate 5 min at room tempera-
ture with vortexing on the vortex Genie 2 (see Note 9).
7. Collect the magnetic beads on the magnetic rack and repeat
the washing step with 500 μl PBS-BSA 0.05%.
8. Collect the magnetic beads on the magnetic rack and wash briefly
with 500 μl IP buffer 1× to eliminate the traces of PBS-BSA.
9. Collect the magnetic beads on the magnetic rack and resus-
pend in 2 μl IP buffer 1× (see Note 9).
10. Transfer the magnetic beads to the sample.
11. Incubate for 2 h at 4 °C on the overhead rotator (see Note 10).
12. Collect the magnetic beads on the magnetic rack and wash
with 700 μl IP buffer 1× by incubating 10 min at room tem-
perature with vortexing on the vortex Genie 2.
13. Repeat the washing step with 700 μl IP buffer 1× twice.
14. Collect the magnetic beads on the magnetic rack and resus-
pend in 250 μl proteinase K digestion buffer.
15. Add 5 μl proteinase K and incubate 30 min at 50 °C in the dry
heating block with shaking (see Note 11). Transfer the sample
in a 1.5 ml LoBind microtube.
16. Extract with one volume (250 μl) PCI. Transfer the upper
phase in a new standard 1.5 ml microtube (see Note 2).
17. Precipitate the DNA with 3 volumes (750 μl) ethanol contain-
ing 300 mM NaCl and 1 μl LPA. Store at −20 °C overnight.
18. Centrifuge for 40 min at full speed at 4 °C.
19. Wash the pellet with 500 μl ethanol 70% and centrifuge for
20 min at full speed at 4 °C.
20. Resuspend the pellet in 10 μl H2O and store at −20 °C.
3.4. Amplification Enrichments in the MeDIP fraction can be measured by real-time

and Analysis PCR or microarray hybridization. If the MeDIP is performed with
small amounts of starting material, it is necessary to perform a
nonspecific amplification to increase the amount of DNA for down-
stream analyses, with the caveat that it might introduce amplification
biases. We routinely use the Genomeplex® complete whole genome
amplification kit WGA2 (Sigma-Aldrich), following the manufac-
turer’s protocol on 10 ng of input DNA and the entire MeDIP
product. This amplification step does not alter the MeDIP profiles
at most targets, however we and others experienced that it can
introduce amplification biases in the MeDIP enrichments at cer-
tain targets, especially the ones that are CpG-rich (9, 10). The
MeDIP procedure can be validated after the whole genome
amplification by performing real-time PCR in the input and MeDIP
fraction on endogenous methylated and unmethylated controls.
Real-time PCR can be performed on 10 ng of input and MeDIP
DNA. Methylated controls are typically imprinting control regions
or promoters of germline-specific genes that appear methylated in
most somatic cells, whereas unmethylated negative controls are
CpG island promoters that remain constitutively unmethylated in
most cells or regions that contain very few CpGs. Table 1 gives a
number of primers for real-time PCR in methylated and unmethy-
lated controls that can be used to validate the MeDIP in human
and mouse. Enrichments in the MeDIP fraction are calculated rela-
tive to one unmethylated negative control with the following for-
mula: enrichment = (IPtarget/INtarget)/(IPnc/INnc), with IPtarget and
INtarget representing the amount of the target sequence in the
MeDIP and input fraction, and IPnc and INnc representing the amount
of the unmethylated negative control sequence in the MeDIP and
input fraction. An example of typical enrichment profile obtained
by MeDIP on small amounts of human or mouse genomic DNA is
given in Fig. 3. Keep in mind to interpret the MeDIP results with
caution because apparent low MeDIP enrichments can reflect in
some cases an absence of sufficient CpGs in the target or a bias
introduced during the whole genome amplification. For these rea-
sons, we highly recommend to complement the MeDIP results
with bisulfite sequencing at selected targets whenever possible. For
genome-wide analyses, WGA-amplified input and MeDIP fractions
can be differentially labeled with Cy3 and Cy5 and cohybridized to
high-density oligonucleotides microarrays. It is also possible to
couple the MeDIP procedure with deep sequencing; however, it is
important to keep in mind that the MeDIP generates single-
stranded DNA that cannot directly be used to generate libraries for
deep sequencing. This can be circumvented by ligating the library
adapter oligos between the sonication and the immunoprecipita-
tion step (5, 7, 11, 12).
Table 1
Primer sequences for real-time PCR validation of MeDIP in somatic cells
Organism Gene name Comment Sequence
Mouse Dpep3 Methylated in Forward: GCAGGTTACCCACAGAGACG

somatic cells Reverse: GTGACCAAGACTGAGCACCA
Mouse Prss21 Methylated in Forward: CAAGACGTTGGTGCCACTG
somatic cells Reverse: CACTGCCCCCAGTCTCAC
Mouse H19 ICR Partially methylated Forward: GCATGGTCCTCAAATTCTGCA
in somatic cells Reverse: GCATCTGAACGCCCCAATTA
Mouse IGd Control sequence Forward: CCCTCTGGCCCTGAATTTAT
with very few Reverse: CACCCAGCAATGCTTCAGT
CpGs
Mouse Tbx15 Unmethylated CpG Forward: TCCCCCTTCTCTTGTGTCAG
island Reverse: CGGAAGCAAGTCTCAGATCC
Human TSH2B Methylated in Forward: CAGACATCTCCTCGCATCAA
somatic cells Reverse: GGAGGATGAAAGATGCGGTA
Human BRDT Methylated in Forward: CCCTTTGGCCTTACCAACTT
somatic cells Reverse: GCCCTCCCTTGAAGAAAAAC
Human IG5 Control sequence Forward: GACCATGTCCAGGCAAAAGT
with very few Reverse: AGGCTCCTACAGACGTGGAA
CpGs
Human UBE2B Unmethylated CpG Forward: CTCAGGGGTGGATTGTTGAC
island Reverse: TGTGGATTCAAAGACCACGA
4. Notes
1. Due to the viscosity of the solution after cell lysis, we recommend

to add the proteinase K to the lysis buffer before mixing it with
the cells.
2. We recommend not using a LoBind microtube for the ethanol
precipitation step because we experienced that it hinders the
recovery of the pellet.
3. Typically, up to 100 ng genomic DNA can be recovered from
20,000 diploid mammalian cells.
4. Because the sonication efficiency varies with DNA quality,
quantity and sonicator settings, it is highly recommended to
first verify the efficiency of the sonication on nonprecious
DNA. For this, sonicate nonprecious DNA in the same condi-
tions and verify the size of the sheared DNA by loading at least
50 ng sonicated DNA on a 1% agarose gel. If the sonication
needs to be optimized for <50 ng, you can sonicate several
tubes in parallel and then check the size of the pooled DNA.
Fig. 3. Examples of MeDIP enrichment profiles measured by real-time PCR. MeDIP was performed with 200 ng sonicated
DNA from mouse E9.5 embryos (a) or human primary fibroblasts (b), followed by whole genome amplification. The graphs
show the enrichment in the MeDIP versus input fraction of methylated sequences over unmethylated negative controls.
Values are normalized with the formula (IPtarget/INtarget)/(IPnc/INnc) to the unmethylated negative controls (nc) IGd (mouse) or
UBE2B (human), whose ratios are set to 1 (see Note 12).
To ensure a better consistency in the sonication, we also rec-

ommend always filling up the sonicator tube holder with empty
microtubes containing 150 μl H2O.
5. If sonicating small amounts of DNA, it is possible to monitor
the efficiency of the sonication by sonicating in parallel an
equal amount of nonprecious DNA and loading it on a 1%
agarose gel.
6. We use 2 ml instead of 1.5 ml microtubes because it allows for
a better mixing of small volumes.
7. For a better reproducibility, we suggest to first dilute 1 μl 5mC
antibody in 4 μl IP buffer 1×, and then add 1 μl of the dilution
to the sample. This amount of antibody has been optimized for
somatic cells with a standard methylation level. It can be adjusted
for cells with an unusual hyper- or hypomethylation state.
8. Verify that the sample is properly mixed in the microtube.
Alternatively, you can also perform gentle horizontal shaking
on a vortex Genie 2 with a platform for microtubes placed in a
4 °C cold room.
9. If performing several MeDIPs in parallel, you can wash all the
beads together. For a better consistency, we also recommend to
wash more beads than necessary for the experiment. For
instance in the case of 5 MeDIPs, wash 15 μl of beads, resus-
pend in 15 μL IP buffer 1×, and use 2 μl per MeDIP reaction.
10. Verify that the sample is properly mixed in the microtube.

Alternatively, you can also perform gentle horizontal shaking
on a vortex Genie 2 with a platform for microtubes placed in a
4 °C cold room. In that case, the horizontal shaking should be
strong enough to prevent the sedimentation of the magnetic
beads.
11. The shaking speed must be sufficient to prevent the sedimenta-
tion of the magnetic beads. We routinely use 900 rpm.
12. Enrichment values of methylated over unmethylated controls
in the MeDIP fraction can sometimes vary drastically from 100
to 10,000 between experiments, which reflects the stochastic
nature of the MeDIP and WGA procedures.
Acknowledgments
This work was supported by the CEFIC Long Research Initiative

(LRI-EMSG49-CNRS-08).
References
1. Meissner A (2010) Epigenetic modifications 7. Chavez L, Jozefczuk J, Grimm C et al (2010)

in pluripotent and differentiated cells. Nat Computational analysis of genome-wide DNA
Biotechnol 28:1079–1088 methylation during the differentiation of
2. Laird PW (2010) Principles and challenges of human embryonic stem cells along the endo-
genomewide DNA methylation analysis. Nat dermal lineage. Genome Res 20:1441–1450
Rev Genet 11:191–203 8. Mohn F, Weber M, Schubeler D et al (2009)
3. Lister R, Pelizzola M, Dowen RH et al (2009) Methylated DNA immunoprecipitation
Human DNA methylomes at base resolution (MeDIP). Methods Mol Biol 507:55–64
show widespread epigenomic differences. 9. Borgel J, Guibert S, Li Y et al (2010) Targets
Nature 462:315–322 and dynamics of promoter DNA methylation
4. Weber M, Davies JJ, Wittig D et al (2005) during early mouse development. Nat Genet
Chromosome-wide and promoter-specific anal- 42:1093–1100
yses identify sites of differential DNA methyla- 10. Jia J, Pekowska A, Jaeger S et al (2010)
tion in normal and transformed human cells. Assessing the efficiency and significance of
Nat Genet 37:853–862 methylated DNA immunoprecipitation
5. Down TA, Rakyan VK, Turner DJ et al (MeDIP) assays in using in vitro methylated
(2008) A Bayesian deconvolution strategy genomic DNA. BMC Res Notes 3:240
for immunoprecipitation-based DNA 11. Li N, Ye M, Li Y et al (2010) Whole genome
methylome analysis. Nat Biotechnol 26: DNA methylation analysis based on high
779–785 throughput sequencing technology. Methods
6. Pelizzola M, Koga Y, Urban AE et al (2008) 52:203–212
MEDME: an experimental and analytical 12. Harris RA, Wang T, Coarfa C et al (2010)
methodology for the estimation of DNA Comparison of sequencing-based methods to
methylation levels based on microarray profile DNA methylation and identification
derived MeDIP-enrichment. Genome Res 18: of monoallelic epigenetic modifications. Nat
1652–1659 Biotechnol 28:1097–1105
Chapter 10
Chromatin Immunoprecipitation to Characterize

the Epigenetic Profiles of Imprinted Domains
Purnima Singh and Piroska E. Szabó
Abstract
Imprinted genes are marked by parental allele specific DNA methylation and histone modifications which
regulate their monoallelic expression. Chromatin immunoprecipitation (ChIP) is the technique of choice
to characterize the histones associated with either maternal or paternal chromosomes. To study allele-
specific chromatin composition at imprinted regions, the method has to be efficient to work on limiting
amount of starting material, and specific enough to recognize one of the parental alleles. We optimized the
commonly used ChIP technique for efficient recovery of one parental allele from small number of cells.
We provide examples to show that this ChIP protocol can specifically distinguish between parental alleles
in mouse embryo fibroblasts carrying maternal and paternal duplication of mouse distal Chr7 and also in
normal mouse embryo fibroblasts carrying single nucleotide polymorphism at imprinted regions.
Key words: Genomic imprinting, Monoallelic, Differentially methylated regions, Chromatin immu-
noprecipitation, Histone modifications
1. Introduction
Genomic imprinting is a form of epigenetic regulation whereby the

allele inherited from either the mother or father is functional
(1, 2). Imprinted genes often occur in clusters and are often associ-
ated with differentially methylated regions (DMRs). DMRs are
classified as imprinting control regions (ICRs) when they deter-
mine the parental allele-specific expression of the imprinted genes
in their respective domain. The promoters of imprinted genes and
also the DMRs and ICRs exhibit parental allele specific DNA
methylation and post translational histone modifications. Chromatin
immunoprecipitation (ChIP) (3) is performed to determine allele
specific histone modifications at the maternal or paternal allele
159
160 P. Singh and P.E. Szabó
at a DNA sequence in vivo. Histones are cross-linked to DNA

with formaldehyde followed by shearing the DNA into smaller
fragments by sonication. Antibodies against specific histones are
used to precipitate the complex. The cross-linking is reversed, and
the bound proteins are digested using proteinase K to release the
DNA fragments which can be analyzed for enrichment using locus
specific PCR primers (4, 5).
One challenge in analyzing allele-specific epigenetic features is
to differentiate between the parental alleles. One possible approach
is to use cells with uniparental duplications, where two copies of a
chromosome segment are inherited from one parent and no copies
from the other parent. Chromatin from primary mouse embryo
fibroblast (MEFs) carrying maternal and paternal duplication of
distal Chr7, MatDup.dist7 and PatDup.dist7, can be used to
characterize maternal and paternal allele specific histone
modifications, respectively at the H19/Igf2 and Cdkn1c/Kcnq1
domains (1, 2), because these are both located along the dupli-
cated region of Chr7 (6, 7). The H19-Igf2 region is controlled by
an ICR, which is methylated in the paternal chromosome but is
associated with the insulator protein CTCF in the maternal chro-
mosome. The Cdkn1c/Kcnq1 domain is controlled by the KvDMR1
that carries DNA methylation in the maternal allele and func-
tions as the promoter for the noncoding regulatory RNA, Kcnq1ot1
in the paternal allele. Another possible approach for allele-specific
chromatin analysis utilizes DNA sequence differences, such as sin-
gle nucleotide polymorphisms (SNPs) between the mother and
father to distinguish the parental alleles in normal cells (8, 9).
The other challenge in analyzing imprinted genes is, that the
starting material can be limiting. General ChIP protocols suggest
using 25 μg of chromatin or 107 cells per reaction, which is hard to
obtain when only small embryos or small populations of purified
embryonic cells are available. The following ChIP protocol is rou-
tinely performed in our laboratory using 4 μg of chromatin or
400,000 cells. Using this optimized technique we demonstrate
here that an active histone modification mark, H3K4me2, is associ-
ated with the unmethylated maternal and paternal allele of H19/
Igf2 ICR and KvDMR1, respectively and a repressive mark,
H3K9me3, is associated with the reciprocal, methylated alleles at
these DMRs. This ChIP protocol has been used in combination
with tiling microarrays to reveal chromosome-wide allele-specific
features along distal Chr7 and distal Chr15 (7). It was also used in
combination with multiplex single nucleotide primer extension
(SNuPE) assays in the sequenom-allelotyping platform to generate
histone modification profiles across multiple DMRs in the mouse
genome (8, 9). We also provide an even smaller scale version (*),
which gives reproducibly allele-specific results using only 100,000
purified mouse embryonic germ cells (10).
10 ChIP to Analyze Imprinted Domains 161
2. Materials
All buffers are prepared in tissue culture grade water, filter-sterilized,

and stored at appropriate temperature.
1. DMEM (Dulbecco’s modified eagle medium).
2. FBS (Fetal bovine serum).
3. Formaldehyde 37%.
4. PBS (phosphate buffered saline without calcium and magnesium).
5. Complete protease inhibitor cocktail tablets, (Roche applied
science) 50×, one tablet dissolved in 1 ml water.
6. ChIP Dilution Buffer: 0.01% SDS, 1.1% Triton X-100, 1.2 mM
EDTA, 16.7 mM Tris–HCl pH 8.1, and 167 mM NaCl. Store
at RT. Add 4 μl/ml of 50× complete protease inhibitor cock-
tail before use.
7. ChIP Lysis Buffer: 1% SDS, 10 mM EDTA, and 50 mM Tris–
HCl, pH 8.1. Store at RT. Add complete protease inhibitor
cocktail to 1× concentration before use.
8. Preblocked A/G Beads: 2 ml A/G Beads (Santa Cruz#
sc-2003), 20 μl herring sperm DNA (Invitrogen# 15634-
017), 56 μl 50× complete protease inhibitor cocktail, 280 μl
10 mg/ml BSA (PENTEX), and 2 ml ChIP dilution Buffer.
Preblock the beads by gentle rotation for 2 h in the cold room
in 15 ml falcon tube. Spin at 250 × g for 5 min. Discard super-
natant and resuspend the beads in 2 ml of ChIP dilution buffer
supplemented with 40 μl of 50× complete protease inhibitor
cocktail. Store at 4 °C for up to a month (see Notes 1–3).
9. Buffer A (Low-salt Buffer): 0.1% SDS, 1.0% Triton X-100, 2.0 mM
EDTA, 20 mM Tris–HCl, pH 8.1, 150 mM NaCl. Store at 4 °C.
10. Buffer B (High-salt Buffer): 0.1% SDS, 1.0% Triton X-100,
2.0 mM EDTA, 20 mM Tris–HCl, pH 8.1, 500 mM NaCl.
Store at 4 °C.
11. Buffer C (LiCl Buffer): 0.25 M LiCl, 1.0% Igepal-CA630 (also
known as NP-40), 1.0% Sodium deoxycholate, 1.0 mM EDTA,
10 mM Tris–HCl, pH 8.1. Store at 4 °C.
12. ChIP Elution Buffer: 1% SDS, 0.1 M NaHCO3.
13. Proteinase K.
14. TE Buffer: 1.0 mM EDTA, 10 mM Tris–HCl, pH 8.0.
15. 3 M Sodium Acetate (pH 5.5).
16. Phenol (equilibrated to pH 8.0).
17. Chloroform.
18. Linear polyacrylamide (11).
19. Ethanol.
3. Methods
3.1. Cross-linking 1. Grow primary mouse embryo fibroblasts (MEFs) close to

and Cell Harvesting confluency in a 150 mm dish containing 25 ml of culture
medium (DMEM supplemented with 6% fetal bovine serum)
to obtain about 2.5 × 107 cells per plate. Prepare chromatin in
two ways, using N-ChIP or X-ChIP conditions, to test out
new antibodies. *Trypsinized and flow-sorted cells, such as
primordial germ cells, are cross-linked in suspension (see Notes
4 and 5).
3.1.1. Preparation 1. Add formaldehyde directly into the media to a final concentra-
of N-Link Chromatin tion of 1% (675 μl of 37% formaldehyde per 25 ml medium)
inside a fume hood. It is best to tilt the plate and quickly add
formaldehyde by touching the side of the plate where medium
level is highest and without delay mixing up the medium by
swirling. *For formaldehyde cross-linking of 100K cells, first
pellet cells at 728 ´ g for 10 min in an eppendorf tube, then
remove medium and add 500 μl of PBS with 27 μl of 37%
formaldehyde. Suspend cells and mix by pipetting.
2. Incubate the plate at room temperature with gentle agitation
for exactly 2 min on a horizontal shaking platform. *Tubes are
placed on a rotating platform for 2 min.
3. Stop the cross-linking reaction by adding 2.6 ml of 1.25 M
glycine solution, mix well immediately. *For 100K cells, add
25 μl of 1.25 M glycine and spin at 728 ´ g for 10 min.
4. Remove media inside a fume hood by pouring it from the plate
into a glass beaker and blotting the last drop off on a paper
towel. Wash cells twice with 20 ml ice-cold PBS. *Remove liq-
uid from 100K pellet by pipetting. To wash 100K cells, sus-
pend the cell pellet in 700 μl of cold PBS and spin.
5. Aspirate PBS completely after the second wash. Place cell cul-
ture dish or *eppendorf tube on ice.
6. Scrape cells off the plate in 5 ml ice-cold PBS with a plastic
scraper and collect into a 15 ml falcon tube by pipetting.
7. Pellet the cells by centrifugation for 5 min at 250 × g at 4 °C.
Aspirate supernatant without touching the cell pellet, keep cell
pellet on ice. *For 100K cells, aspirate the last drop using a
pulled pasteur pipette attached to a mouthpiece under a dis-
secting microscope so as not to dislodge the cell pellet.
3.1.2. Preparation 1. X-ChIP is the same as N-ChIP, except for step 2.

of X-Link Chromatin 2. Incubate at room temperature for exactly 10 min.
3.2. Sonication 1. Resuspend pelleted cells in 750 μl ChIP Lysis buffer containing
complete protease inhibitors. *Add 100 μl of lysis buffer to
100K cells, mix, snap-freeze in liquid N2, and store at −80 °C.
These small samples are sonicated on the day of use.
2. Sonication is carried out in 1.7 ml eppendorf tubes using
Branson Sonifier cell Disruptor-350 with a micro tip. Pulse
three to four times for 10 s at 40% duty cycle and 4 output
control. Keep samples on ice at all times. In our experience,
N-ChIP and X-ChIP samples need to be sonicated three and
four times, respectively.
3. If using Diagenode Bioruptor UCD-200, sonicate at high “H”
setting with 15 s “ON” 1 min “OFF” for 5 min. Repeat six
times with 1 min on ice between each cycle. Diagenode Bioruptor
is preferred for sonicating 100K cells in 100 μl volume to pre-
vent frothing and loss of chromatin on the sonicator tip.
4. Centrifuge the sheared chromatin at 15,710 ´ g for 5 min at
4 °C. Transfer the supernatant containing the chromatin to a
fresh tube. Store chromatin on ice in cold room during
quantification/before freezing at −80 °C or use directly for
ChIP.
3.3. Determination 1. To determine the concentration of DNA in the sonicated chro-

of Chromatin DNA matin, take a small aliquot and reverse-cross-link it. For a 25 μl
Concentration and chromatin preparation, add 25 μl of ChIP lysis buffer, add
Fragment Size NaCl to a final concentration of 0.3 M (3 μl of 5 M NaCl) and
incubate at 65 °C for 4 h.
2. Add 1 μl each of 0.5 M EDTA, 1 M Tris–HCl, pH 6.5, and
20 mg/ml proteinase K. Incubate at 55 °C for 1 h.
3. Bring the volume to 100 μl by adding 50 μl of TE (pH 8.0).
Extract with phenol–chloroform. Add 50 μl of phenol and mix
vigorously. Add 50 μl of chloroform and mix well. Centrifuge
at 15,710 ´ g for 10 min. Collect the top, aqueous phase and
also the interphase at this point. Add 100 μl of chloroform and
mix well. Spin at 15,710 ´ g. Collect only the top aqueous
phase this time and transfer it to a fresh tube.
4. Precipitate DNA: add 3 μl of linear polyacrylamide and 1/10
volume of 3 M sodium acetate, pH 5.5 and mix. Add 2.5 vol-
umes of chilled 100% ethanol, mix, and freeze sample at −80 °C for
30 min. To collect DNA pellet, centrifuge sample at 15,710 ´ g
for 15 min. Wash DNA pellet with chilled 70% ethanol.
Resuspend DNA in 25 μl Tris-EDTA buffer (TE), pH 8.0.
5. Check DNA concentration using Nanodrop (see Note 6).
a b c
kb
1.5
1.0
0.5
0.4
0.3
0.2
0.1
M 1 2 M 1 2 M 1 2
Fig. 1. Testing the efficiency of chromatin sonication. Agarose gel electrophoresis of sonicated DNA to determine the frag-
ment size of the reverse cross-linked chromatin (a) Ideal DNA sonication is shown with fragment size ranging up to 1,200 bp
with median fragment length of ~500 bp. The small fraction <200 bp on this gel is RNA and can be eliminated by RNAse
treatment. (b) Undersonication. High molecular weight unfragmented DNA can be seen on top of the gel. (c) Oversonication.
The majority of fragments are below 400 bp. (lane 1, M: 1 kb-ladder, lane 2, MatDup.dist7, and lane 3, PatDup.dist7 DNA).
6. Load 1 μg of DNA on a 1.5% agarose gel and run it in 1×TAE

(Tris-Acetate-EDTA) buffer to check the size of chromatin
fragments. They should range from 0.2 to 1.5 kb. (Fig. 1)
(see Note 7).
7. Dilute chromatin preparations (which were stored on ice to
this point) to 0.4 μg/μl concentration with ChIP lysis buffer
containing 1× complete protease inhibitors. You will use 10 μl
(4 μg nucleic acid equivalent chromatin) per ChIP reaction. At
this point, chromatin preparations can be used right away or
100 μl aliquots can be snap-frozen in liquid nitrogen for long-
term storage at −80 °C.
3.4. Immuno- 1. Quickly defrost frozen sheared chromatin sample on the day of
precipitation ChIP. *100K cell aliquots can be defrosted, combined and
sonicated in a larger batch on the day of ChIP. Sonication
efficiency of small aliquots is similar to larger samples. This can
be optimized and tested beforehand.
2. Take out 10 μl sheared chromatin for “input.”
3. Use 10 μl of chromatin for each IP. Add 90 μl ChIP Lysis
Buffer with 1× complete protease inhibitors to the 10 μl chro-
matin fraction in an eppendorf tube. Dilute tenfold with ChIP
Dilution Buffer containing 4 μl/ml 50× complete. *100K cells
had been already suspended in 100 μl ChIP Dilution Buffer
containing 4 μl/ml of 50× complete. Dilute these tenfold with
ChIP dilution buffer containing 4 μl/ml 50× complete.
4. Preclear chromatin with preblocked Protein A/G agarose beads.

Add 50 μl Protein A/G agarose into each reaction. Rotate at
4 °C for 1 h. Pellet beads by centrifugation for 5 min at 250 × g.
Transfer supernatant chromatin to a new tube (see Note 8).
5. Add 4 μg of the antibody to the supernatant and incubate
overnight at 4 °C on a rotating platform. Set up one sample
with normal rabbit IgG antibody as negative control (see Notes
9–11).
6. To collect the antibody–chromatin complexes, the next morn-
ing add 50 μl Protein A/G agarose beads to each IP. Rotate for
2 h at 4 °C.
7. To collect beads centrifuge for 1 min at 92 ´ g and discard
supernatant.
8. Wash agarose beads on a rotating platform for 5 min with 1 ml
of the ice-cold buffers A, B and C. Collect beads by centrifuga-
tion after each wash at 1,000 rpm for 1 min at 4 °C.
(a) Buffer A with 2 μl/ml 50× complete in cold room.
(b) Buffer B with 2 μl/ml 50× complete in cold room.
(c) Buffer C with 2 μl/ml 50× complete in cold room.
(d) TE Buffer with 2 μl/ml 50× complete at room tempera-
ture (repeat this wash two more times)
9. Remove as much buffer as possible without disturbing the
beads (see Notes 12–14).
3.5. Elution, Reverse 1. Add 150 μl freshly made ChIP Elution buffer to the washed
Cross-linking, and beads (see Note 15). Incubate for 15 min at room temperature
Proteinase K with gentle agitation.
Treatment 2. Pellet beads by centrifugation at 1,487 ´ g for 2 min. Transfer
the ChIP eluate supernatant to a new tube.
3. Repeat elution. Combine the second eluate with the first one.
4. Reverse-cross-link the eluted DNA by adding 18 μl 5 M NaCl
per 300 μl eluate followed by incubation for 4 h at 65 °C (see
Note 16).
(Remember to reverse-cross-link the reserved “input” DNA
samples also at this point)
5. Add 6 μl each of 0.5 M EDTA, 1 M Tris–HCl, pH 6.5, and
20 mg/ml proteinase K. Incubate for 1 h at 55 °C (see
Note 17).
3.6. DNA Isolation 1. Add 1.5 ml of Binding buffer QG (QIAquick Kit, QIAGEN).
and Purification 2. Purify DNA according to the manufacturer’s instruction. Wash
with PE four times(see Note 18).
3. Recover DNA with 100 μl warm elution buffer. Store at
−20 °C. Samples are ready for subsequent qPCR assays or
sequenom allelotyping (Figs. 2 and 3) (see Notes 19–23).
a Normal MatDup.dist7 PatDup.dist7

7 7 15 7
7 7 15 7
Mat
Snrpn
T9H T9H Pat
H19/Igf2 ICR
KvDMR1
MatDup.dist7 MatDup.dist7
b PatDup.dist7
c
PatDup.dist7
1400 6000
1200
5000
Copy number precipitated
Copy number precipitated
1000
4000
800
3000
600
2000
400
200 1000
0 0
H19/Igf2 ICR KvDMR Snrpn C-myc
Fig. 2. Detecting allele-specific H3K4me2 enrichment at imprinted regions using cells with uniparental duplications of
distal Chr7. (a) MatDup.dist7 and PatDup.dist7 MEFs carry two maternal (black ) or two paternal (grey ) copies of chromo-
some 7 regions, located distally to the T9H translocation breakpoint. These cells allow the analysis of allele-specific
marks at the H19/Igf2 ICR and at the KvDMR1 along the maternally or paternally duplicated distal chromosome 7 region.
Parental allele-specific methylation and hypomethylation of the DMRs is shown by closed and open lollipops, respectively.
(b) Real-time PCR was used to quantify an active chromatin mark, H3K4me2, levels at specific loci in MatDup.dist7 and
PatDup.dist7 MEFs at two imprinted regions. The paternally methylated H19/Igf2 ICR shows H3K4me2 in the MatDup.
dist7 MEFs, whereas the maternally methylated KvDMR1 exhibits H3K4me2 enrichment in PatDup.dist7 MEFs. (c) Control
regions. The control Snrpn DMR is located outside of the duplicated chromosome region, therefore the allele-specificity
of positive H3K4me2 enrichment cannot be discerned at this locus. The control c-myc promoter is constitutively active
and enriched in H3K4me2.
3.7 Multiplex 1. Dissolve all the primers and probes (Table 1) in TE (pH 8.0) to
Quantitative Real-Time a final concentration of 100 μM. The oligo tubes from IDT are
PCR for Allele Specific spun briefly before dissolving, as the pellet may dislodge during
Histone Modifications shipment. Mix equal volumes of the upper and lower (U + L)
primers. These can be stored at −20 °C (see Note 24).
2. Prepare MIQ 5× Buffer by mixing equal volumes of 10× iTaq
Buffer and 5 mM MgCl2 from the iTaq DNA Polymerase kit
(Bio-Rad #170-8875).
Table 1
Primers for multiplex real-time PCR
DMR Probe sequence and dye PCR primers
H19-Igf2 ACATTCACACGAGCATCCA CACTTACACCCAGGACTCAAAGG

ICR-FAM GGAGGC
FAM GCGTATAAACCCCACAACTGATTC
KvDMR1 CCGCAGTGGCTCCGTATTCGTT CGGCTGGGCTCCATCTTC
TEX CGACCTCGGGGCTCAAAG
Snrpn promoter CATGCGTCCCAGGCAATGGCTGC TCCTTTTGGTAGCTGCCTTTTGG
TAMRA CCGCAATGGCTCAGGTTTGTC
c-myc promoter CTGCCTCGCTCCACACAA AGATAACTCATTCGTTCGTCCTTCC
TACGCCA
Cy5 TGTGTTCTTGCCCTGCGTATATC
3. For amplification standards use a sonicated mouse genomic

DNA dilution series: 10 ng/μl, 1 ng/μl, 0.1 ng/μl, 0.01 ng/μl,
0.001 ng/μl, and a no-template control. In the iQ5 real-time
plate setup menu, define units as copy numbers (10,000;
1,000; 100; 10; 1; and 0 copies, respectively). This calculation
considers that one diploid mouse genome equals 6 pg of
genomic DNA. Set up the multiplex PCR reaction as follows:
X1 (ml)
MIQ 5× Buffer 5.0

dNTP mix (25 mM each) 0.45
iTaq (Biorad) 5 U/μl 0.325
Primer (U + L) (4) 0.2 Each pair (up to 5 total pairs)
Probes (4) 0.075 Each probe (up to 5 total
probes)
DNA 3 μl of ChIP eluate or 30 ng input
or standard genomic DNA
H2O Up to 25 μl
The following PCR parameters are used to run the reactions in

Bio-Rad iQ5-Thermal Cycler:
95 °C 3 min
40 cycles
95 °C 30 s
55 °C 45 s
To illustrate the applicability of this ChIP protocol to allele-

specific analysis, we precipitated chromatin from MatDup.dist7
and PatDup.dist7 MEFs using the H3K4me2 antibody and
amplified the H19/Igf2 ICR, the KvDMR, as well-as the con-
trol Snrpn promoter and c-myc promoter regions. Four sets of
primers and probes (Table 1) were used in a multiplex real-
time PCR reaction (Fig. 2). We also performed ChIP with
H3K4me2 and H3K9me3 antibodies using chromatin from
normal 129XJF1 and JF1X129 MEFs and submitted aliquots
of the ChIP elution for multiplex sequenom allelotyping as we
have done earlier (8, 9). The percent H3K4me2 and H3K9me3
enrichment in the total immunoprecipitation was measured
utilizing SNPs between the parental alleles (Fig. 3a, b). The
PAT
a b MAT
H3K4me2 H3K9me3
120% 120%
% allele in total ChIP
100% 100%
80% 80%
60% 60%
40% 40%
20% 20%
0% 0%
129 X JF1 JF1 X 129 129 X JF1 JF1 X 129 129 X JF1 JF1 X 129 129 X JF1 JF1 X 129
KvDMR H19/Igf2 ICR KvDMR H19/Igf2 ICR
c 129XCS 100 K cells

100%
90% PAT (CS)
% allele in total ChIP
80%
MAT (129)
70%
60%
50%
40%
30%
20%
10%
0%
H3K4me2 H3K9me3
H19/Igf2 ICR
Fig. 3. Detecting allele-specific chromatin in normal cells. Chromatin was prepared from 129XJF1 (129 mother and JF1
father) and JF1X129 (JF1 mother and 129 father) MEFs and was subjected to ChIP using (a) H3K4me2 and (b) H3K9me3
antibodies. Sequenom allelotyping assays were used to measure the percent maternal and paternal component in the total
ChIP DNA at the KvDMR1 and at the H19/Igf2 ICR. H3K4me2 is enriched in the unmethylated paternal allele (PAT) at the
KvDMR1 and maternal allele (MAT) at the H19/Igf2 ICR. H3K9me3 shows enrichment at the reciprocal, methylated, alleles.
(c) ChIP using 100,000 MEFs. These MEFs were obtained by mating a 129 mother to a CAST/Ei (CS) father. Sequenom
allelotyping shows correct allele specific enrichment in 129XCS MEFs for H3K4me2 and H3K9me3 at the H19/Igf2 ICR
similar to ChIP from 4 μg chromatin.
results for the H3K4me2 antibody were in agreement with the

results of the MatDup.dist7-PatDup.dist7 experiment, show-
ing active chromatin-specific H3K4me2 enrichment in the
unmethylated DMR alleles. The repressing mark, H3K9me3,
however, was present in the CpG-methylated alleles at both
DMRs. The ChIP protocol using 100K 129XCS MEFs also
revealed the correct allele specific bias (Fig. 3c). These experi-
ments demonstrate that our optimized ChIP protocol efficiently
and quantitatively recovers the correct parental alleles of
imprinted regions from chromatin even when the cell numbers
are limited.
4. Notes
1. Protein A/G agarose beads are preferred over either Protein A

or G because they bind nearly all isotypes.
2. The protein A/G beads are preblocked to reduce nonspecific
binding.
3. Gently swirl the beads to uniformly resuspend them before
taking an aliquot. Vigorous shaking should be avoided, because
the beads may stick to the side of tube and dry. Also, vortexing
or high speed centrifugation may break the beads.
4. For most histone modifications N-linked chromatin works well.
For certain nonhistone proteins like CTCF insulator and tran-
scription factors X-linking is preferred. The optimal cross-linking
strength should be experimentally determined for each anti-
body by ChIP, real-time PCR and allele-specific protocols.
5. Mouse embryos/organs can be used for chromatin prepara-
tion. The embryo/organ is first suspended in PBS and cross-
linked in suspension (9).
6. Keep in mind that RNAse step is not necessary in this DNA
preparation. Therefore, the DNA will have some contaminat-
ing RNA (4 μg nucleic acid equivalent chromatin has less than
4 μg DNA; 100K fetal germ cells contain 600 ng DNA and
~1 μg total nucleic acids).
7. A <1.2 kb smear on the gel is considered optimal shearing with
~500 bp as median fragment size. It is critical to obtain opti-
mal size distribution of sonicated chromatin as fragment size
>1.2 kb may result in high background from neighboring
chromosomal regions. Fragment size <200 bp may result in
inadequate labeling yields for downstream ChIP-chip experi-
ments. Fragmentation needs to be optimized with each sonica-
tion device and cell type being used for the experiment.
Table 2
Volumes for preclearing the chromatin
Preblocked
Chromatin Lysis buffer Dilution buffer A/G beads
Sonicated chromatin from plate 10 μl (~4 μg) 90 μl 900 μl 50 μl

Chromatin from 100K cells 100 μl (~1 μg) 0 μl 900 μl 50 μl
8. The preclearing step inhibits nonspecific DNA and protein

binding to Protein A/G beads and thereby reduces the back-
ground. Preclearing can be done in large amount for multiple
ChIPs by multiplying the volumes in Table 2.
9. Set up your first ever ChIP using 4 μg DNA equivalent chro-
matin and well-characterized antibodies, such as those in
Fig. 3, which give reliable allele-specific enrichment at DMRs.
10. It is a good practice to test new batches of commercial antibod-
ies for specificity in a dot blot before using them in ChIP (9).
11. A1:1 (4 μg antibody and 4 μg chromatin) ratio usually works
for histone modifications. However, it is best to determine
optimal concentrations for each antibody. Polyclonal antibod-
ies give better results than monoclonal ones in ChIP.
12. Washing buffers can be prepared beforehand and stored at
4 °C. Add 50× complete protease inhibitor just before use.
13. 1 ml pipette tips can be used for washes. Gently dispense wash-
ing buffers from the side of the tubes. Avoid touching the
beads when taking out supernatant after washes.
14. Remove buffer leaving ~50 μl behind, spin again, and use a
100 μl pipette tip to take out as much buffer as possible.
15. Prepare elution buffer not more than 10 min before use.
16. Reverse cross-linking removes the protein and allows purifying
genomic DNA. Proteinase K treatment digests proteins in the
DNA-protein mix.
17. 65 °C and 55 °C temperatures should be maintained during
reverse cross-linking and proteinase K treatments for best
results. Rotating of the samples is not required.
18. ChIP protocols suggest stopping/freezing after reverse cross-
linking or proteinase K treatment and continuing the process
the next day. In our hands finishing all the steps on the second
day gives the best results.
19. If washes with PE buffer are done less than four times, SDS
from the elution buffer may remain and you may see SDS crys-
tals in the eluate.
20. Repeat freeze-thawing or long-term storage of ChIP-
precipitated DNA should be avoided.
21. The ChIP-precipitated DNA can be extracted using phenol–
chloroform and precipitated using linear polyacrylamide or
glycogen (small scale ChIP). However, if ChIP DNA is to be
amplified for ChIP-on chip, it is best to use Qiagen columns
for purification, which removes excess salts.
22. The copy number of immunprecipitated DNA should be higher
for specific antibodies than for nonspecific IgG. Using real-
time PCR we generally measure less than 10 copies for
nonspecific IgG and above 20 copies, up to 10,000 copies for
specific antibodies from a 3 μl ChIP eluate (Fig. 2) (9). The
downstream allele-specific measurement is more accurate with
higher copy numbers for the specific region in the ChIP
elution.
23. ChIPs and downstream allele-specific allelotyping assays should
be done in duplicates or triplicates. Small standard deviation in
these tests (Fig. 3) indicates that the precipitation worked and
there is real enrichment with a specific antibody at a specific
site. High standard deviation is a warning signal suggesting
that the PCR randomly amplified background precipitation.
24. Reciprocal mouse crosses substantiate the allele-specific
findings (Fig. 3).
25. Use the Beacon Designer software for four-color real-time
qPCR probe and primer design for other regions or consult
our paper for more DMR sets (9). Multiplexing in the real-
time PCR provides internal controls, saves time and impor-
tantly, saves the majority of the ChIP elution for downstream
processes, such as sequenom-allelotyping or amplification for
microarray hybridization.
Acknowledgments
We thank Jeff Mann for the MatDup.dist7 and PatDup.dist7

MEFs, Li Han for her work on the initial phases of this project, and
Diana Tran for her comments on the manuscript. This work was
supported by a Public Health Service grant (GM064378) from the
National Institute of General Medicine to P.E.S.
References
1. Ideraabdullah FY, Vigneau S, Bartolomei MS 7. Singh P, Wu X, Lee DH et al (2011)
(2008) Genomic imprinting mechanisms in Chromosome-wide analysis of parental allele-
mammals. Mutat Res 647:77–85 specific chromatin and DNA methylation. Mol
2. Koerner MV, Barlow DP (2010) Genomic Cell Biol 31:1757–1770
imprinting—an epigenetic gene-regulatory 8. Singh P, Han L, Rivas GE et al (2010) Allele-
model. Curr Opin Genet Dev 20:164–170 specific H3K79 Di- versus trimethylation
3. Solomon MJ, Larsen PL, Varshavsky A (1988) distinguishes opposite parental alleles at
Mapping protein-DNA interactions in vivo with imprinted regions. Mol Cell Biol 30:
formaldehyde: evidence that histone H4 is retained 2693–2707
on a highly transcribed gene. Cell 53:937–947 9. Singh P, Cho J, Tsai SY, Rivas GE, Larson GP,
4. Orlando V, Strutt H, Paro R (1997) Analysis of Szabo PE (2010) Coordinated allele-specific
chromatin structure by in vivo formaldehyde histone acetylation at the differentially methy-
cross-linking. Methods 11:205–214 lated regions of imprinted genes. Nucleic Acids
5. Hebbes TR, Thorne AW, Crane-Robinson C Res 38:7974–7990
(1988) A direct link between core histone 10. Lee DH, Singh P, Tsai SY et al (2010) CTCF-
acetylation and transcriptionally active chroma- dependent chromatin bias constitutes transient
tin. EMBO J 7:1395–1402 epigenetic memory of the mother at the H19-
6. McLaughlin KJ, Szabo P, Haegel H, Mann JR Igf2 imprinting control region in prosper-
(1996) Mouse embryos with paternal duplica- matogonia. PLoS Genet 6:e1001224
tion of an imprinted chromosome 7 region die 11. Gaillard C, Strauss F (1990) Ethanol precipita-
at midgestation and lack placental spongiotro- tion of DNA with linear polyacrylamide as car-
phoblast. Development 122:265–270 rier. Nucleic Acids Res 18:378
Chapter 11
Quantitative Chromosome Conformation Capture

Raffaella Nativio, Yoko Ito, and Adele Murrell
Abstract
It is becoming increasingly apparent that chromatin is not randomly folded into the nucleus, but instead
is highly organized into specific conformations within the nucleus. One consequence of such higher order
structure is that chromatin looping can bring together genomic elements which are separated by several
hundreds of kilobases, such as enhancers and promoters, and functionally facilitate their interaction. The
Chromosome Conformation Capture (3C) assay is a powerful technique to detect looping structures and
assess the probability of interaction between distant genomic elements (1–3). Here we describe the 3C
methodology, its power, and limitations, together with the controls and normalization steps required for
an accurate analysis.
Key words: 3C: Chromosome Conformation Capture, Cross-link, Digestion, Intramolecular ligation,
3C product, 3C template, PCR standard template, Association frequency
1. Introduction
The 3C technique detects the proximal association frequencies

between distant genomic elements. If these elements represent
regulatory regions, it is possible that they functionally interact. The
3C procedure consists of five experimental steps (2) (Fig. 1). The
first step involves cross-linking of DNA and protein complexes.
Formaldehyde cross-linking is commonly used, and such fixation
enables a snapshot of interactions between DNA and proteins. The
second step involves digestion of the cross-linked chromatin com-
plexes with a restriction enzyme that cuts DNA in the regions
under 3C analysis. This step is followed by ligation of the restricted
chromatin at low DNA concentration so as to favor intramolecular
ligation. The chromatin complexes are then reverse cross-linked
and the DNA is purified. The purified DNA (3C template) con-
tains religated sites in the genome that are enriched because of
their proximity through looping interactions as well as religated
173
174 R. Nativio et al.
3C association
a
Crosslink
b
a
Digestion
b
a
Ligation
b
Reverse crosslink
a b
Detection by qPCR
a b
Fig. 1. 3C procedure. Schematic representation of a 3C assay. Light grey and dark grey
boxes represent two interacting genomic elements, a and b, that are separated by a long
intervening region (black curved line). The 3C assay starts with a cross-linking step using
formaldehyde to capture protein–protein and protein–DNA interactions. A second step
consists of enzymatic digestion with a restriction enzyme known to recognize sites in the
investigated regions. In the third step, the cross-linked complex is religated under condi-
tions that favor intramolecular ligation. Lastly, the cross-links are reversed by heat treat-
ment; the DNA is then purified and the resulting 3C product is detected by qPCR.
sites at lower abundance that represent regions associated at lower

frequency. A final PCR step with two primers that amplify across
the ligated site is used to quantitatively detect the 3C product.
Generally, when examining a specific locus for looping interactions,
a specified region is chosen as an anchor and the anchor PCR
primer is tested for associations between other locations where bait
primers (in the same orientation as the anchor primer) are placed.
A crucial interpretation of the 3C assay depends on how the
association frequency is calculated and the ability to distinguish
between a specific interaction between two genomic sequences and
the background of random ligations due to the flexibility of the
11 Quantitative Chromosome Conformation Capture 175
chromatin fiber. Thus the detection of an association between two

elements does not necessarily mean that they are associated in vivo,
since the flexibility of the chromatin fiber enables random colli-
sions between sites on the same chromosome. However, such ran-
dom ligation events are inversely proportional to the distance
between two sites and can be plotted as a random ligation curve,
while specific associations can be detected as “spikes” above the
random ligation curve (2). In a simpler definition, two sites are
considered associated when their association frequency value is
higher than the one measured in the intervening sites, given that
these are not engaged in similar associations. Therefore, it is impor-
tant when assessing the association between two elements to first
establish the baseline of random interactions.
Furthermore, it is important to bear in mind that associations
detected by 3C need to be genetically or biochemically verified
prior to any conclusions regarding a functional association.
Depending on the restriction enzymes used, the 3C assay still has
low resolution and often multiple adjacent restriction sites within a
5 kb stretch of DNA will associate with similar frequencies to a
distant restriction site used as an anchor. Therefore the 3C signals
indicate proximities to associations, rather than pinpointing the
exact sequences involved in the association.
Since 3C captures all associations taking place in a cell popula-
tion at a given time, it is possible that the associations detected
between different elements do not all take place at the same time
and in the same cell. For example, it has been shown that the asso-
ciations of a particular site with four other sites take place preferen-
tially in a pair-wise manner when analyzed by single cell imaging
(4). The heterogeneity of cell populations can further decrease the
sensitivity in the detection of an association frequency, especially
when the associations are specific to a particular cell condition. In
order to detect the association, cells need to be selected and
enriched for the appropriate state. This means that low-frequency
associations can still be specific and biologically significant although
they are not detectable as spikes over other associations.
Once all the 3C controls have been considered, the 3C analysis
is able to provide important insights into the higher order organi-
zation of the analyzed locus.
2. Materials
2.1. Reagents 1. Phosphate-Buffered Saline (PBS).

and Buffers 2. 37% formaldehyde (Sigma).
3. Lysis buffer: 50 mM Tris–HCl pH 8, 1% SDS, 10 mM
EDTA.
4. Protease Inhibitor Cocktail (PI) (Sigma).

5. Triton-X100 (Sigma).
6. Digestion buffer (New England Biolabs, NEB).
7. Restriction enzyme (NEB).
8. T4 ligase buffer (NEB).
9. T4 ligase (NEB).
10. Proteinase K (Sigma).
11. Phenol.
12. Chloroform.
13. 100% Ethanol.
14. Glycogen (Invitrogen).
15. 3 M Sodium Acetate.
16. qPCR primers (Sigma).
17. Picogreen (Invitrogen).
18. 1× Sybr-green master mix (Power SYBR, ABI).
19. 384 well real-time machine (7900HT Fast Real time PCR sys-
tem, ABI).
20. Commercially obtained genomic DNA (Becton Dickinson).
2.2. Equipment 1. Optical microscope for qualitative and quantitative assessment

of nuclei.
2. Infinite M200 Tecan System for Picogreen-based quantification
of the 3C products.
3. DNA purification columns (DNA Clean α Concentrator™ 25 Kit,
Zymo Research) for purification of the PCR standard amplicons.
4. 2 ml Eppendorf tubes.
5. 15 ml falcon tubes.
6. Table microcentrifuge.
7. Macrocentrifuge.
3. Methods
3.1. Cell Cross-Linking 3C is a multistep process, the main steps of which are cross-linking
and Nuclei Preparation of cells and nuclei preparation; chromatin digestion and religation;
DNA template preparation; and identification and analysis of
chromatin interactions.
1. Grow adherent cells in a 15 cm dish at semi-confluence.
2. Remove the culture medium and wash the cells with 10 ml of
ice-cold PBS.
3. Cross-link the cells with 20 ml of 1% formaldehyde (diluted in

PBS) at 37 °C for 10 min (see Notes 1 and 2).
4. Remove the formaldehyde and wash the cells three times with
10 ml of ice-cold PBS. Leave 2 ml of PBS from the last wash
into the dish to scrape the cells off the dish.
5. Scrape the cells off with a spatula and transfer into a 2 ml
Eppendorf tube.
6. Pellet the cells by centrifugation into a table microcentrifuge at
100 × g for 5 min.
7. Remove the PBS from the tube and add 1 ml of ice-cold lysis
buffer (to release the nuclei from the cytoplasm) supplemented
with fresh Proteinase Inhibitor Cocktail and allow the cells to
lyse for 10 min on ice (see Note 3).
8. Confirm that lysis is complete by examining 10 μl aliquot of
the lysate microscopically. Lysis is complete when nuclei can be
seen under the microscope.
9. After lysis is complete pellet the nuclei by centrifugation at
900 × g for 5 min.
10. Remove carefully the lysis buffer and resuspend the nuclei in
200 μl of 1× digestion buffer (see Note 4).
11. Count the nuclei under the microscope and resuspend 1.5 × 106
nuclei in 300 μl of 1× digestion buffer. Nuclei can be stored at
−80 °C at this stage.
3.2. Chromatin 1. Add 1.8% Triton-X to sequester the remaining traces of SDS
Digestion and and put the nuclei to shake at 37 °C for 1 h on a shaker.
Religation 2. Put aside 30 μl of the nuclei suspension to use as the “undi-
gested control” later (freeze sample at −20°C and purify
together with the 3C sample) (see Note 7).
3. Digest the remainder of the nuclei with 1,000 U of the chosen
restriction enzyme and incubate at 37 °C on a shaker for 8 h or
overnight (see Notes 5 and 6).
4. After digestion put aside a 30 μl aliquot of the reaction to use
as the “digestion control” to measure the percent of digestion
(freeze sample at −20 °C and purify together with the 3C sam-
ple) (see Note 7). Only templates with digestion >70% can be
used to proceed to the next steps (see Note 8).
5. Dilute the digested chromatin to 2.5 ng/μl by using 1× of T4
ligase buffer and add 3,200 U of T4 ligase (high concentra-
tion) in a total volume of 1.5 ml in order to favor intramolecu-
lar ligation. Leave ligation to proceed for 8 h or overnight (see
Notes 9 and 10).
6. Digest the chromatin with 1,000 U of a second enzyme (high
concentration) which recognizes sites different from the first
enzyme and that digests outside of the regions that are assessed
for interaction (see Notes 11 and 12).
7. Reverse cross-link the DNA–protein complexes by treatment
with 100 μg/ml of Proteinase K for 4 h or overnight at
65 °C.
3.3. DNA Template 1. Extract the relegated DNA with phenol and chloroform treat-
Purification ment. Add 1 volume of phenol, mix, and centrifuge samples at
14,000 rpm for 10 min (see Note 13).
2. Collect the aqueous phase in a new tube and repeat the phenol
treatment.
3. Transfer the aqueous phase to a new tube, add 1 volume of
chloroform, mix, and centrifuge samples at 14,000 rpm for
10 min.
4. Place the aqueous phase in a new tube, add 1 volume of H2O,
2 volumes of 100% ethanol, 1/10 volume of 3 M Sodium
Acetate, and 1/200 volume of 20 μg/μl of glycogen.
5. Allow the DNA to precipitate for 2 h at −80 °C or overnight at
−20 °C.
6. Wash the DNA with 1 ml of 70% of ethanol and centrifuge at
14,000 rpm for 5 min.
7. Repeat the 70% ethanol wash.
8. Dry the pellet at 37 °C and resuspend the DNA into 250 μl of
H2O (see Note 14).
3.4. Identification and Chromatin interactions are measured by determining the ligation
Analysis of Chromatin frequencies between nonadjacent restriction sites. The quantification
Interactions is done by qPCR followed by normalization for differences in
genomic copy number and digestion–ligation efficiency in the 3C
template. Considerations of 3C primer design, production of stan-
dard curves for 3C targets, and quantitative 3C analysis are
described.
1. Use a primer design tool (e.g., Primer Express Software V3.0,
ABI) to design a series of primers with similar melting tem-
peratures and PCR efficiencies (see Note 15) that bind next to
the restriction sites analyzed by 3C. For locus-wide analysis of
the association of a particular site, a common primer (anchor
primer) adjacent to this site can be used in combination with
specific primers for each of the sites tested for an association.
Since two sites are scored as being associated when their asso-
ciation frequency is higher than those of intervening sites, it is
important to design primers also for testing the intervening
region. It is also important to test genome-wide associations in
a reciprocal manner by setting the anchor primer in one of the
elements previously tested for association. This is a control to
ensure the validity of the detected associations.
The PCR efficiency of the 3C primers is assessed by real-time

PCR on a PCR standard template.
2. Prepare a PCR standard template that contains equimolar
amounts of all possible religation products from the region of
interest. A PCR standard template can be easily made by diges-
tion and religation of small genomes. However, in the case of
humans and mice, where genomic size is large, direct use of
genomic DNA would produce too many combinations of reli-
gated products, making the detection by PCR of a specific 3C
product difficult. In this case a PCR standard template can be
prepared by PCR amplification of the genomic regions span-
ning each restriction site analyzed by 3C. The resulting ampli-
cons are purified by column and quantified using a Picogreen
assay (follow manual instructions). Equimolar amounts of
amplicons are mixed together, digested with the same enzyme
used in the 3C experiment, religated, phenol/chloroform
extracted, ethanol precipitated, and dissolved in H2O.
3. Quantification of the 3C product is done by qPCR and is based
on standard curves for each primer set. Make standard curves
using ¼ serial dilutions of the PCR standard template. The 3C
values are calculated with the parameters of the standard curves
for each 3C primer set as follows: 3C value = 10(Ct − i)/s, where
Ct represents the cycle threshold of the PCR reaction and i
and s represent the intercept and the slope of the standard
curve, respectively.
4. The qPCR conditions are as follows: use 1.5 μl of the 3C or
PCR standard template in a qPCR reaction. The PCR mixture
contains 200 nM of FW and REV primers and 1× Sybr-green
master mix made up to a final volume of 12.5 μl with H2O.
Use a 96- or 384-well real-time machine for the PCR
amplification. Carry out the qPCR as follows: after an initial
2-min preincubation step at 50 °C and 10 min at 95 °C, run
40 amplification cycles, each consisting of 95 °C for 15 s and
60 °C for 1 min. Obtain the melting curve by a 15-min incu-
bation step at 95 °C followed by a 15-min step at 60 °C.
5. When comparing different 3C biological replicates or biologi-
cal conditions it is important to be able to normalize the 3C
products for differences in genomic copy number and diges-
tion–ligation efficiency of the 3C template (Fig. 3). Detect the
genomic copy number of the 3C template by qPCR amplification
of a region that does not contain any of the restriction sites
recognized by the enzyme. The genomic copy number is also
used to ensure that the amount of the 3C template is within
the range of the standard curve for any given 3C product. The
genomic copy number of the 3C template used in each qPCR
reaction is 10,000 copies that correspond to approximately
70 ng of DNA.
6. Detect the digestion–ligation efficiency of the 3C template by

measuring the religation frequency of two adjacent sites. The
religation between two adjacent sites produces a circular mol-
ecule of DNA; therefore the circularization frequency of this
DNA fragment is representative of the efficiency of both diges-
tion and religation. The circularization frequency is measured
by qPCR amplification across the religated site. As example,
the circularization of the i–j fragment is shown in Fig. 2c.
a b c
3C association genomic copy number control digestion-ligation efficiency control
a
Crosslinking
b
j
i
a
Digestion
i j
Ligation
b
i-j
Reverse crosslinking
a b
Detection by qPCR
i-j
c d e
d AF(a–b)= c / (d x e)
Fig. 2. Normalization for genomic copy number and digestion–ligation efficiency. Schematic on how to normalize a 3C
product for genomic copy number and digestion–ligation efficiency. (a) The first column represents the 3C procedure to
detect the interaction between two genomic elements, a and b (a is indicated as purple box; b is indicated by a pink box),
which are distant along the primary genomic sequence (black curved line indicated genomic separation). (b) The second
column shows the qPCR product that is used to normalize for genomic copy number. This product derives from amplification
of a genomic region that lacks Bam HI sites and therefore is not digested when treated with the enzyme. (c) The third
column shows the qPCR product that is used to normalize for digestion–ligation efficiency. Normalization for digestion–
ligation efficiency is done using the circularization frequency of two adjacent sites (i–j). (d) The formula to calculate the
Association Frequency between two elements AF(a–b) is the 3C association value divided by the circularization frequency
value and genomic copies of the 3C template: AF(a–b) = c/(d × e). The application of this normalization allows the analysis
of the same interactions among biological replicates or different biological conditions.
7. The Association Frequency between two elements a and b

(AFa–b) normalized for genomic copy number and digestion–
ligation efficiency is calculated as follows:
AFa − b = c / (d × e ),
where c represents the 3C product derived by amplification of

the religated sites a and b, d represents the genomic copy
number of the 3C template, and e the circularization frequency
of two adjacent sites (i and j) (Fig. 2a–c).
Once the association frequency values of a particular element
with other sites are normalized for genomic copy number and
digestion–ligation efficiency, they can be plotted on a graph
and a chromatin looping model can be drawn to get insights
into the function of higher order chromatin organization at
the locus (Fig. 3).
4. Notes
1. Formaldehyde cross-linking is used to capture associations

between distant genomic elements by cross-linking protein–
protein and protein–DNA molecules by their amino and imino
groups. Cross-linking is preferably done on cells in order to
capture the interactions in the natural state of a cell. If cross-
linking is done on nuclei it is possible that some interactions
may have been lost during the procedure of nuclei preparation.
The amount, time, and temperature at which cross-linking is
carried out can affect the efficiency of cross-linking. Also, cell
size is important since it affects the formaldehyde concentra-
tion in the nucleus. The detection of stable associations may
require less cross-linking than weak or less frequent associa-
tions. If cross-linking is not efficient, the chromatin interac-
tions are lost during the following 3C steps and therefore they
cannot be detected.
2. If cells grow in suspension, transfer them in a 50 ml falcon
tube; wash once with 20 ml ice-cold PBS and cross-link with
1% formaldehyde at 37 °C for 10 min on a rotor. Pellet cells
in a centrifuge at 100 × g for 5 min and remove the formalde-
hyde. Wash cells three times with 20 ml of ice-cold PBS.
Resuspend cells in 2 ml of PBS left from the last wash and
transfer them into a 2 ml falcon tube. Proceed with the lysis as
for adherent cells.
3. Cell size is one of the factors to consider when lysing cells.
Larger cells are more difficult to lyse, especially when cross-
linked. The epithelial breast cell line HB2 and the fibroblast
cell line HS27 cells have high cytoplasmic content and therefore
Fig. 3. 3C data and chromatin looping model. Example of a 3C experiment results: the
association frequency values of two different anchor primers within a specific locus
are shown in a graph. (a) Schematic representation of the 350 kb genomic region
including three different genomic elements (different shaded boxes). Vertical lines indi-
cate the position of the Restriction Sites (RSs) recognized by the selected restriction
they require strong lysis with 1% of SDS. Cells with high

cytoplasm content such as fibroblasts require to be passed a
few times through a 27 G syringe needle. However, cells of
smaller size such as the breast cancer cell line SUM 159 and
CAL51 can be lysed by using only 0.5% of SDS. The process of
lyses and nuclei preparation can be checked by inspecting the
lysate under the microscope. In most 3C procedures, the lysis
step is followed by a nuclear permeabilization step but in this
case it is not necessary since the SDS also permeabilizes the
nuclear membrane.
4. Nuclei can easily clump. Avoid nuclei clumping by tapping the
pellet nuclei before addition of digestion buffer and add the
digestion buffer in aliquots of 100 μl per time and gently
pipette up and down each time you add the digestion buffer.
5. The choice of the restriction enzyme is important in the design
of the 3C assay. This is because association between two ele-
ments is measured by the association frequency of the restric-
tion sites within or close to the elements under study. If the
enzyme does not cut within or in the proximity of the two ele-
ments that are to be tested for association, then no association
can be studied. Restriction sites have to be present also in the
region between the two elements since an association between
two elements is significant only when it is detected at a higher
frequency than that measured with sites in the intervening
region. Enzymes that digest very frequently (four base pair
cutters) increase the resolution of 3C and are therefore used to
analyze associations between elements in a small region while
enzymes that digest less frequently (six base pair cutters) can
be used to analyze elements separated by long distances. In this
case, a limitation of the 3C assay is that, independently of how
many restriction sites there are in between two associated ele-
ments, it is not possible to assess the association between two
elements closer than 30 kb. This is because the association
between the two sites would be masked by random collisions
of the chromatin fiber due to its flexibility within the nuclear
space. The assessment of associations with different restriction
Fig. 3. (continued) enzyme and letters point out the RSs analyzed by 3C (3C RS) in the
graphs and histograms below. Thick vertical lines in panels b and c link the positions of
the restriction sites used as anchors in each panel with the overview of the locus in (a).
The X-axis is labeled according to genomic position and position 0 is arbitrarily fixed
42 kb upstream of the element a. (b) The anchor restriction site is a. A peak of association
is detected with the h and q elements. (c) Associations detected with the anchor site in q.
An association is present with the a and h elements. The specificity of the interaction is
confirmed by the detection of the a–h interaction with both the anchor primers. (d) A
model of the chromatin looping interactions at the locus is based on the 3C analysis
reported in b and c. The three elements a, h, and q contact each other at the base of the
chromatin loops.
Fig. 4. Digestion efficiency of the 3C template. The assessment of the digestion efficiency of the 3C template at a specific
restriction site (RS) is schematically represented. (a) The cross-linked chromatin is split into two equal aliquots, (b) one is
digested with the selected restriction enzyme, Bam HI (Bam HI +) in this example, and the other is put aside and used as
the undigested control (Bam HI −). (c) Primers binding adjacent to the RS are used to amplify by qPCR the undigested sites
both in the Bam HI-treated and Bam HI-untreated samples. The products resulting from these amplifications (d) are used
to calculate the percent of digestion at the selected site. (e) The formula to calculate the percent of Digestion of the RS is
%D(RS) = (1-(a /b))´100, where a is the qPCR product from amplification across the RS in the Bam HI-treated sample and
b derives from amplification in the Bam HI-untreated sample.
enzymes gives more confidence in the interpretation of the

3C results.
6. Add the enzyme in aliquots of three and let the sample digest
for 2 h before adding the following aliquot. After the addition
of the last aliquot of enzyme, the digestion can be left to
proceed overnight.
7. The digested and undigested fractions are purified similarly to

the 3C template: phenol and chloroform extraction, ethanol
precipitation, and resuspension in H2O (80 μl).
8. An important qualitative control is the assessment of digestion
efficiency at the restriction sites analyzed by 3C. Poor enzy-
matic digestion results in low religation and subsequently low
amplification of the 3C product representing the association
between two sites. Moreover the digestion efficiency has to be
similar between different sites analyzed by 3C; otherwise the
associations between these different sites cannot be compared.
The digestion efficiency of the 3C template is assessed by qPCR
across each restriction site as shown in Fig. 4. The percent
digestion is determined by comparing qPCR amplification
between digested and undigested fractions present in same
genomic copy number.
9. High dilution of chromatin favors intramolecular ligation.
Intramolecular ligation is required to religate the digested
chromatin fragments that have been captured in the same
cross-linked complex. The starting chromatin concentration is
estimated by counting nuclei before digestion and by knowing
that 1 × 106 nuclei contain 7.2 μg of DNA.
10. More ligations can be carried out at this point and samples can
be pooled for DNA extraction.
11. The secondary digestion step is required to avoid the forma-
tion of long DNA templates formed during the ligation steps as
they can negatively affect the efficiency of the PCR reaction.
12. Add the enzyme in aliquots of three, similarly to the treatment
with the first enzyme.
13. Before phenol treatment, the religated samples from the same
condition can be pooled in a 15 ml falcon tube.
14. If the DNA has been precipitated in several Eppendorf tubes,
aliquot the 250 μl of H2O among these tubes and recollect the
dissolved DNA into a single tube.
15. It is possible to compare interaction between different sites only
when the corresponding primer sets have similar efficiency.
References
1. Splinter E, Grosveld F, de Laat W (2004) 3C 3. Miele A, Gheldof N, Tabuchi TM, Dostie J,
technology: analyzing the spatial organization Dekker J (2006) Mapping chromatin interac-
of genomic loci in vivo. Methods Enzymol tions by chromosome conformation capture.
375:493–507 Curr Protoc Mol Biol Chapter 21, Unit 21.11
2. Dekker J (2006) The three ‘C’s of chromo- 4. Chang HY, Cuvier O, Dekker J (2009) Gene
some conformation capture: controls, controls, dates, parties and galas. Symposium on
controls. Nat Methods 3:17–21 Chromatin Dynamics and Higher Order
Organization. EMBO Rep 10:689–693
Chapter 12
Genome-Wide Analysis of DNA Methylation in Low Cell

Numbers by Reduced Representation Bisulfite Sequencing
Sébastien A. Smallwood and Gavin Kelsey
Abstract
Development of high-throughput sequencing technologies now enables genome-wide analysis of DNA
methylation of mammalian cells and tissues. Here, we present a protocol for Reduced Representation
Bisulfite Sequencing (RRBS) applicable to low amounts of starting material (from 200 to 5,000 cells).
RRBS is a cost-effective and powerful technique offering the advantages of absolute DNA methylation
quantification and single nucleotide resolution while covering mainly CpG islands. Typically one sequenc-
ing experiment using the Illumina Genome Analyser IIx platform provides information on the DNA
methylation status of more than half of the CpG islands of the mouse genome.
Key words: DNA methylation, CpG islands, Bisulfite sequencing, Reduced representation bisulfite
sequencing, High-throughput sequencing, Genomic imprinting, Illumina genome analyser, Bismark
1. Introduction
DNA methylation is an epigenetic modification of the mammalian

genome that plays a critical role in regulating gene expression and
thereby the definition of specific cellular identity. Notably DNA
methylation at CpG islands (CGIs), which are associated with the
majority of gene promoters, plays a crucial role during embryonic
development, cellular differentiation, and genomic imprinting (1–
3). The development of deep-sequencing technologies is having a
major impact on epigenetic studies in general and on DNA methy-
lation more specifically. It is now possible to comprehend DNA
methylation dynamics in a genome-wide manner, for all mamma-
lian organisms with a reference genome. Several techniques to
study DNA methylation now exist, each with their strengths and
weaknesses depending on the biological question asked and the
resources available in each laboratory (4, 5).
187
188 S.A. Smallwood and G. Kelsey
DNA methylation can be profiled genome-wide by sequencing

after methylated cytosine (5mC) enrichment either using a specific
anti-5mC antibody (MeDIP-Seq) or protein domain (methyl-binding
domain; MBD-Seq) (6, 7). This type of approach is cost-effective,
as one sequencing run provides a good coverage of the entire mouse
or human genome. It also provides information on both single-
copy loci and repetitive elements and MeDIP-Seq allows as well the
distinction between 5mC and 5-hydroxymethyl-cytosine (5hmC)
(6). However, these 5mC enrichment approaches do not offer
single-nucleotide resolution as they are based on enrichment of
fragmented DNA (typically 300 bp length), and more importantly,
5mC methylation quantification is relative. In addition, precipita-
tion-based methods entail a nonspecific background contribution
that sets a minimum limit of input material below which signal-to-
noise ratios become unacceptable. Alternatively, genome-wide
DNA methylation profiles can be studied by bisulfite sequencing
(BS-Seq or shotgun BS-Seq) approaches (8). In this case, treatment
of DNA with sodium bisulfite converts unmethylated Cytosines
into Uracils (Thymines after PCR amplification) while 5mCs are
protected. Therefore BS-Seq provides single-nucleotide resolution
and absolute quantification of DNA methylation. On the downside,
this technique does not allow the distinction between 5mC and
5hmC, yet. In addition, to obtain coverage of the entire genome a
large number of sequencing runs is required and therefore this
approach is expensive with current technology.
Alternative techniques based on BS-Seq exist, and here we
describe in detail a protocol for Reduced Representation Bisulfite
Sequencing (RRBS). RRBS combines the advantages of BS-Seq in
providing accurate DNA methylation level of individual cytosines,
and enrichment in CGIs (4, 9–11). While this technique was origi-
nally performed using relatively large amounts of genomic DNA
(1 μg), thus limiting its application, it is now possible to perform
RRBS with 10–300 ng genomic DNA (12) or even lower (11)
(this protocol). One sequencing run on the Illumina Genome
Analyser IIx platform typically gives reliable information on ~1
million individual CpGs, ~65 % of them overlapping the CGIs
identified by CAP-Seq (13). Since the main biological targets of
5mC are gene promoters, the majority of which reside in CGIs,
RRBS is a powerful, informative, flexible, and yet relatively inex-
pensive technique.
We present here a RRBS protocol suitable for 200–5,000
somatic cells (6 pg DNA per cell). We detail DNA purification,
sequencing library generation, quality control, and basic bioinfor-
matic analysis. Using this protocol, we successfully generated and
sequenced libraries from as low as 500 pg of genomic DNA (11).
How does RRBS work? First, genomic DNA is digested with a
restriction endonuclease, in this case MspI (C^CGG), which cuts
more frequently within CGIs, thereby providing the basis for CGI
12 Low Cell Analysis of Methylation by RRBS 189
enrichment based on appropriate size selection (in frequency,

smaller fragments correspond more to CGIs). After MspI diges-
tion, DNA fragments are repaired and 5mC Illumina sequencing
adapters are ligated, followed by bisulfite conversion. Then after a
first PCR amplification, size selection on agarose gel is performed
to recover the smaller fragment fraction. This is followed by a sec-
ond PCR amplification and purification. The protocol presented
here can be easily performed, in 2–3 days, and is particularly
efficient as MspI digestion, End-Repair/A-tailing, adapter ligation,
bisulfite conversion can be performed within the same PCR tube,
with no intermediate purification steps, thus limiting DNA loss.
2. Materials
1. QIAamp DNA Micro Kit (QIAGEN).

2. EB buffer (10 mM Tris–HCl, pH 8.5).
3. Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen).
4. Fluorometer.
5. PCR tubes with individual lids.
6. Water (UltraPure DNase/RNase-Free; Invitrogen).
7. PCR cycler.
8. MspI (Fermentas).
9. Klenow Fragment exo- (Fermentas).
10. PCR-Grade dNTPs: dATP, dCTP, dGTP, dTTP (Roche
Applied Science).
11. Nucleotide End-Repair Mix (1 mM dATP; 0.1 mM dGTP;
0.1 mM dCTP).
12. T4 DNA Ligase HC (30 u/μl) (Fermentas).
13. ATP 50 mM (Fermentas).
14. Illumina 5mC Adapters. We are using PE adapters, which pro-
vide the flexibility for single-run or paired-end run. Adapters
must have 5mC to conserve their sequence upon bisulfite treat-
ment. 5mC adapters can be bought directly from Illumina (Early
Access Methylation Adapter Oligos, Illumina). Alternatively,
5mC adapters oligos can be synthesized by another company
such as Sigma-Aldrich and annealing done separately.
15. Imprint DNA Modification Kit (Sigma-Aldrich).
16. Ethanol 100 %.
17. Pfu Turbo Cx Hotstart DNA polymerase (Agilent).
18. Illumina Library PCR primers PE1.0 and PE2.0.
19. Agarose (molecular biology grade).
20. TBE buffer (Tris/Borate/EDTA buffer).

21. Ethidium bromide.
22. 100 and 50 bp DNA ladders (no ready-mix).
23. 6× Loading dye.
24. UV Transilluminator.
25. Scalpel blades.
26. QIAquick Gel Extraction kit (QIAGEN).
27. Platinum Pfx DNA Polymerase (Invitrogen).
28. Agencourt AMPure XP (Beckman Coulter Genomics).
29. Bioanalyser and Illumina Sequencing platform (Genome
Analyser GIIx).
3. Methods
This protocol describes how to generate RRBS libraries starting

from 200 to 5,000 cells corresponding to ~500 pg to 20 ng
genomic DNA after purification. Here, we refer to starting mate-
rial >5 ng (L.A. for Low Amount) or <5 ng (V.L.A. for Very Low
Amount). It is recommended to perform a negative control (no
DNA) in parallel of your samples. For V.L.A. a positive control
(20 ng of genomic DNA) can also be performed. Bench, pipettes
and racks used should be thoroughly cleaned with DNA decon-
tamination solution. It is recommended to use low DNA binding
plastic ware.
3.1. DNA Purification 1. Use the QIAamp DNA microkit according to the manufac-
turer’s protocol for tissue. Do not use carrier RNA. Elute the
DNA using 22 μl of warm EB buffer (see Note 1).
2. Quantify DNA using Quant-iT PicoGreen kit according to the
manufacturer’s instructions. Use 2 μl of eluted DNA per repli-
cate and perform duplicates. In our hands, this technique is
accurate for as little as 100 pg/μl (see Note 2).
3.2. MspI Digestion In a PCR tube add the following:

– Genomic DNA
– 0.9 μl of MspI
– 1.8 μl of Tango 10× buffer
– H2O to 18 μl final volume
Incubate for 3 h at 37 °C, followed by enzyme heat inactivation at
80 °C for 20 min.
Proceed to next step or pause (see Note 3).
3.3. End Repair/A- In the same PCR tube (Subheading 3.2), directly add:
Tailing – 1 μl of Klenow Fragment exo-
– 0.8 μl of nucleotide end-repair mix
Incubate for 40 min at 37 °C, followed by enzyme heat inactiva-
tion at 75 °C for 15 min.
Proceed to next step or pause.
3.4. Adapter Ligation In the same PCR tube (Subheading 3.3), directly add:
– 1 μl of HC T4 DNA ligase
– 1 μl of 5mC sequencing adapters (For L.A.:1.5 μM; for V.L.A.:
0.5 μM) (see Note 4)
– 0.5 μl of ATP (50 mM)
– 2 μl of H2O
Incubate overnight at 4 °C, followed by enzyme heat inactivation
at 65 °C for 20 min (see Note 5).
3.5. Bisulfite Take 24 μl of the previous reaction mix (Subheading 3.4) and use
Modification directly the Imprint DNA Modification Kit two-step modification
procedure, with the following modifications (see Note 6):
– For L.A.: once DNA modification mix is added, incubate at
65 °C for 90 min, 99 °C for 2 min, 65 °C for 30 min.
– For V.L.A.: once DNA modification mix is added, incubate at
65 °C for 90 min.
Perform purification steps according to the manufacturer’s proto-
col, and elute DNA in 22 μl (2 × 11 μl) of warm EB buffer.
3.6. First Amplification To the BS converted DNA, set up a 25 μl PCR reaction by

Step adding:
– 20.6 μl of bisulfite treated DNA
– 0.4 μl of Pfu Turbo Cx Hotstart DNA polymerase (see
Note 7)
– 2.5 μl of Pfu Turbo Cx Hotstart 10× buffer
– 0.5 μl of dNTP mix
– 0.5 μl of primers PE1.0
– 0.5 μl of primers PE2.0
Perform six PCR cycles (95 °C-2 min; 95 °C-20 s, 65 °C-30 s,
72 °C-30 s) (see Note 8).
3.7. Fragment Size Prepare a 1.7 % agarose gel with Ethidium Bromide (see Note 9).
Selection Prepare 50 and 100 bp ladder in Pfu Turbo Cx Hotstart 10×
buffer in a 25 μl volume for each sample (see Note 10). Load
directly the product from the first amplification step
(Subheading 3.6) on the gel using 6× loading dye. Load 50 bp
ladder/sample/100 bp ladder and repeat for each sample,
leaving a few wells empty in between to avoid cross-contami-
nation. Run at 80 mV for ~45 min. Place the gel still in its
plastic case on a UV Transilluminator, and using the DNA lad-
ders as a guide, cut between 150 and 400 bp.
Proceed to Gel Extraction with Qiagen QIAquick following the
manufacturer’s protocol (see Note 11).
Elute in 40 or 44 μl of warm EB buffer (see Note 12).
3.8. Second Set up a 50 μl PCR reaction by adding:

Amplification Step – 36.2 μl of size selected DNA from the previous step
and Purification
– 0.8 μl of Platinum Pfx DNA Polymerase (see Note 13)
– 5 μl of Platinum Pfx 10× buffer
– 2 μl of MgSO4 (50 mM)
– 2 μl of dNTPs
– 2 μl of primer PE1.0
– 2 μl of primer PE2.0
Do 12 Cycles (94 °C 2 min; 94 °C-20 s, 65 °C-30 s, 68 °C-30 s)
(see Note 14).
The 50 μl library is purified using Agencourt AMPure XP mag-
netic beads according to the manufacturer’s protocol, and elu-
tion is done in 20 μl EB buffer (see Note 15).
3.9. QC Analysis Bioanalyser using DNA High Sensitivity chips is performed with
and General 1 μl of purified library. Bioanalyser is important to assess the overall
Troubleshooting quality of the library. The following points need to be examined.
Control samples (no DNA) should be clear: presence of DNA
within 150–400 bp size range is a sign of contamination, and a
new library should be prepared.
The size distribution of the MspI fragments is not uniform,
and some peaks are visible, mainly corresponding to MspI frag-
ments for interspersed repeats (Fig. 1). While the overall intensity
of the peaks might vary slightly between samples, their size is none-
theless constant between different experiments. The absence or
shift in size of one or more peaks is an indication of a suboptimal
library (Fig. 1). Absence of peaks can be the result of inaccurate
size selection on the agarose gel; the library can still be sequenced
but CGI coverage will be reduced. Shift in size can be the result of
strong PCR bias and clonality, and therefore, it is not recom-
mended to sequence this library.
Fig. 1. Quality assessment of RRBS libraries using Bioanalyser. (a) Typical Bioanalyser profile of a good quality RRBS library.
The library corresponds to size fragments in the 150–400 bp size range, with a characteristic pattern of peaks correspond-
ing to MspI fragments from interspersed repeats, and limited adapter contamination (117 bp). (b) Example of Bioanalyser
profile of a low quality RRBS library. Peaks are present at the right sizes; however, strong adapter contamination is also
present (117 bp). This library can still be sequenced but sequence complexity may be reduced. Alternatively a second gel
extraction can be performed to remove adapter dimers (see Subheading 3.9). (c) Example of Bioanalyser profile of a poor
quality RRBS library. There is obvious bias in peak distribution (strong 150 bp peak and absence of 273 bp peak) and sub-
stantial adapter contamination (118 bp). It is not recommended to sequence this library. These libraries were generated from
mouse DNA. RRBS libraries from DNA of other species will have a different pattern of peaks from repetitive sequences.
Bioanalyser analysis allows the assessment of adapter dimer

contamination in the library, which appears as a ~117 bp peak.
Presence of adapter dimers can result from poor size selection, or
indication that the amount of genomic DNA used was overesti-
mated. In the presence of a small quantity of dimers, libraries can
still be sequenced but sequence complexity may be reduced. If
adapter dimers represent the majority of the library in terms of DNA
concentration, it is recommended to generate a new library. If this
is not feasible, a second gel extraction can be performed to remove
adapter dimers, followed by a couple of extra PCR cycles to com-
pensate for the loss of material during this process (see Note 14).
Libraries should be purified again and bioanalyser analysis repeated.
3.10. Library Bioanalyser analysis also informs on library concentration (3 μl of

Quantification and 10 nM libraries are normally required for Illumina sequencing). If
Sequencing concentration is too low, a few additional PCR cycles can be per-
formed (see Note 14). It is also recommended to determine library
concentration by qPCR using as standards Illumina libraries of
known concentration (14). Alternatively, Quant-iT PicoGreen
dsDNA Assay can also be performed. We usually perform sequencing
on the Genome Analyser GAIIx platform with single read analysis
and 40 bp read length. RRBS library sequencing can be done with
around ~250,000–280,000 clusters per tile. This usually gives ~30
millions sequences after Bareback processing (15) resulting in ~15–18
millions aligned sequences by Bismark (16) (see below).
3.11. Data Analysis For data analysis, a recent PC or Mac is needed and sufficient;
however, 8 Gb or more of RAM memory can be useful (e.g., 64 bit
version of Windows). Specific software is required for alignment of
RBBS/BS-Seq reads. We are currently using Bismark (16) that can
perform both read mapping and DNA methylation calls. In addi-
tion Bismark can discriminate between DNA methylation in a CpG
context or another context (CHG, CHH). DNA methylation call
visualization and analysis can be made in a general program used
for high-throughput sequencing analysis. We are using Seqmonk
(www.bioinformatics.bbsrc.ac.uk/projects/seqmonk/).
Here is a short summary of the steps required for CGI DNA
methylation analysis in Seqmonk. First data need to be imported,
in the form of a list of cytosines, their position and their methyla-
tion status (“+” strand corresponds to methylated cytosine while
“−” strands correspond to unmethylated cytosines). Probes corre-
sponding to individual cytosines are generated using Data/Define
probes/Contig probe generator. Then probes are quantified by
number of reads to assess the number of time each individual cyto-
sine has been sequenced (Data/Quantification/Read count). Then
filter out the probes with less than five reads depth (Filtering/
Filter on value/Individual probes) (this filter gives sufficiently
accurate CGI methylation; however, a filter of ten reads can also be
used). Once the filtering is done, perform a new probe quantification
to give methylation values in percentage: Data/Quantification/

Difference quantification/Forward only/As a percentage of/All
reads. It is possible to get a CGI methylation value by averaging
methylation of individual CpGs within the CGI. To do so, after
importing the CGI dataset (13) in Seqmonk as an annotation track
(File/Import Annotation), perform a report (Reports/Create fea-
ture report/annotate CGIs with any overlapping probes/do not
exclude features with no probes). The report (tabulated text file)
can be imported in Excel where the table function and filters asso-
ciated can be used for analysis. CGIs can be filtered by the number
of probes associated (which corresponds to the number of infor-
mative CpGs per CGIs). We generally consider only CGIs for
which at least 10 % of CpGs are informative (i.e., a minimum of
five reads), but this number can be increased to 20 or 30 % for
more stringent analysis.
4. Notes
1. Obtaining good quality genomic DNA is key to successful

RRBS. This can be challenging when starting with very low
numbers of cells. Phenol–chloroform purification of DNA is
an option and the use of glycogen (20 μg) as carrier in ethanol
precipitation does not interfere with the subsequent enzymatic
steps. However, we obtained more reproducible results using
the QIAamp DNA microkit. Even with fewer than 1,000 cells,
recovery is ~50–70 % of the theoretical amount of DNA (6 pg
of DNA per somatic cell).
2. For very low amounts of DNA PicoGreen can be optional, and
RRBS performed without quantification, based simply on the
expected yield of DNA from the initial number of cells. In this
case, elute DNA in 20 μl of warm EB buffer.
3. Enzymatic reaction can be paused at 4 °C for up to 16 h. For
longer times, freeze the reaction mix at −20 °C.
4. It is important to perform initial tests with different adapter
concentrations (using 20 ng genomic DNA), as annealing
efficiency of oligos might vary from stock to stock. Using the
optional step described in Note 11, if after gel analysis a library
is not or only weakly visible, increase adapter concentration; if
only or large amounts of adapter dimers are observed, reduce
adapter concentration.
5. For L.A. conditions, adapter ligation can also be performed at
22 °C for 90 min with good results. Overnight incubation at
4 °C is always recommended for V.L.A. conditions.
6. The main advantage of this bisulfite modification kit versus
others is low DNA degradation and therefore better DNA
recovery for low amount of starting material. On the down-

side the conversion rate of unmethylated cytosine is not opti-
mal. In our experience it is typically ~95 % (for CGI DNA
methylation analysis this does not introduce significant/detect-
able artifacts). It is possible to use carrier DNA at this stage
(12), but this has not been tested in our laboratory.
7. The use of Pfu Turbo Cx Hotstart DNA polymerase is critical at
this step, in our experience. This enzyme is resistant to uracil stall-
ing and therefore both methylated and unmethylated fragments
are amplified with similar efficiencies. The use of another poly-
merase could result in an overestimation of DNA methylation.
8. This two-step PCR strategy is designed to increase the amount
of DNA for gel extraction, which is the major limiting step of
our protocol (where most material may be lost). Our tests have
shown that this does not introduce major biases in sequence
amplification.
9. Prepare a gel using a volume of agarose and combs just sufficient
to allow loading of 30 μl volumes. Using combs too big or too
much agarose will result in a heavier agarose slice after size selec-
tion, resulting in lower recovery during QIAquick gel extraction.
10. It is important to use DNA ladders prepared in Pfu Turbo Cx
Hotstart buffer, otherwise differences of migration between
ladders and samples might occur, resulting in nonoptimal size
selection.
11. Once cut, the gel slices should not weigh more than 300 mg
or the DNA recovery might be low. If necessary, adapt comb
size and migration time appropriately.
12. It is possible at this stage to perform an optional step to vali-
date all previous steps, without the need to proceed to final
library generation and purification with AMPure XP beads.
For this, set up a 2 × 20 μl PCR (with Platinum Pfx DNA
Polymerase, and conditions described in Subheading 3.8) with
2 μl of size selected fragments. Amplify for 20 and 25 cycles for
L.A. or 25 and 30 cycles for V.L.A. Load directly on a 3 %
agarose gel. Assess size selection accuracy and the presence of
adapter dimers. If this optional test is not performed the elu-
tion volume used in the QIAquick column can be reduced to
40 μl. Alternatively, for initial optimisation of adapter concen-
tration, set up 4 × 20μl PCRs with 10 μl of size selected frag-
ments: amplify for 15, 20, 25, and 30 cycles.
13. Platinum Pfx DNA Polymerase produces better yields than Pfu
Turbo Cx Hotstart polymerase. At this stage uracil stalling due
to bisulfite conversion is no longer a problem.
14. Twelve cycles of PCR are generally enough to obtain “sequen-
cable” libraries for both L.A. and V.L.A. conditions. In the
case of low library concentration, and if no original sample is
left, extra PCR cycles can be performed without affecting

significantly the overall quality (up to four extra cycles).
Purification using beads and Bioanalyser analysis should be
repeated. More than 22 cycles in total and the overall DNA
methylation values might be overestimated, and more than 26
cycles in total and major artifacts will appear.
15. As an alternative to AMPure XP magnetic beads, Qiagen
Minielute columns can be used. However, percentage of recov-
ery and overall quality of the library will be lower, especially if
adapter dimers are present.
Acknowledgments
This work was supported by the Biotechnology and Biological

Sciences Research Council, the Medical Research Council, and the
Centre for Trophoblast Research.
References
1. Smallwood SA, Kelsey G (2012) De novo 9. Meissner A, Mikkelsen TS, Gu H et al (2008)
DNA methylation: a germ cell perspective. Genome-scale DNA methylation maps of pluri-
Trends Genet 28:33–42. doi:10.1016/j. potent and differentiated cells. Nature 454:
tig.2011.09.004 766–770
2. Feng S, Jacobsen SE, Reik W (2010) Epigenetic 10. Gu H, Bock C, Mikkelsen TS et al (2010)
reprogramming in plant and animal develop- Genome-scale DNA methylation mapping of
ment. Science 330:622–627 clinical samples at single-nucleotide resolution.
3. Ferguson-Smith AC (2011) Genomic imprint- Nat Methods 7:133–136
ing: the emergence of an epigenetic paradigm. 11. Smallwood SA, Tomizawa S-I, Krueger F et al
Nat Rev Genet 12:565–575 (2011) Dynamic CpG island methylation land-
4. Harris RA, Wang T, Coarfa C et al (2010) scape in oocytes and preimplantation embryos.
Comparison of sequencing-based methods to Nat Genet 43:811–814
profile DNA methylation and identification of 12. Gu H, Smith ZD, Bock C et al (2011)
monoallelic epigenetic modifications. Nature Preparation of reduced representation bisulfite
Biotechnol 28:1097–1105 sequencing libraries for genome-scale DNA
5. Bock C, Tomazou EM, Brinkman AB et al methylation profiling. Nat Protoc 6:468–481
(2010) Quantitative comparison of genome- 13. Illingworth RS, Gruenewald-Schneider U,
wide DNA methylation mapping technologies. Webb S et al (2010) Orphan CpG Islands iden-
Nat Biotechnol 28:1106–1114 tify numerous conserved promoters in the
6. Ficz G, Branco MR, Seisenberger S et al (2011) mammalian genome. PLoS Genet 6:e1001134
Dynamic regulation of 5-hydroxymethylcyto- 14. Quail MA, Kozarewa I, Smith F et al (2008)
sine in mouse ES cells and during differentia- PERSPECTIVE A large genome center’s
tion. Nature 473:398–402 improvements to the Illumina sequencing sys-
7. Serre D, Lee BH, Ting AH (2010) MBD- tem. Nat Methods 5:1005–1010
isolated Genome Sequencing provides a high- 15. Krueger F, Andrews SR, Osborne CS (2011)
throughput and comprehensive survey of DNA Large scale loss of data in low-diversity illumina
methylation in the human genome. Nucleic sequencing libraries can be recovered by
Acids Res 38:391–399 deferred cluster calling. PLoS One 6:e16607
8. Lister R, Pelizzola M, Dowen RH et al (2009) 16. Krueger F, Andrews SR (2011) Bismark: a
Human DNA methylomes at base resolution flexible aligner and methylation caller for
show widespread epigenomic differences. Bisulfite-Seq applications. Bioinformatics 27:
Nature 462:315–322 1571–1572
Part V
Analysis of Imprinted Expression

Chapter 13
Isolation of RNA and DNA from Single Preimplantation

Embryos and a Small Number of Mammalian Oocytes
for Imprinting Studies
Sarah Rose Huffman, Md Almamun, and Rocío Melissa Rivera
Abstract
Researchers whose experimental models are mammalian oocytes and preimplantation embryos are often
limited by the yield of nucleic acids that can be isolated from such a small sample size. In addition, the
limited number of cells from these types of samples makes the simultaneous recovery of RNA and DNA
very difficult and often sample pooling is necessary to increase nucleic acid yield. Here we report a simple
set of procedures using commercially available kits that results in consistent yield and quality of nucleic
acids. After sample lysis, RNA is isolated and converted to a reusable cDNA library. Following RNA isola-
tion, DNA is precipitated, isolated, and bisulfite converted for DNA methylation studies. Our results
demonstrate the feasibility of isolating RNA and DNA from a small number of cells with repeatability of
results.
Key words: Mammalian embryo, Blastocyst, Oocyte, RNA isolation, DNA isolation, Bisulfite
mutagenesis
1. Introduction
One of the challenging tasks of researchers who use models that

contain small numbers of cells, such as mammalian preimplanta-
tion embryos, is the ability to isolate both DNA and RNA from
them. Depending on the model organism these may be hard to
obtain. For example, under natural cyclic conditions, a mouse
could produce ~6–15 embryos, while a cow would only produce
one or two embryos. The difficulties are further exacerbated by the
developmental stage (e.g., 2-cell vs. blastocyst) at which embryos
are used. Mouse preimplantation-stage embryos range in number
of cells from 1 to approximately 80 (1). In addition, a blastocyst-
stage mouse embryo will contain ~2 ng of total RNA and ~500 pg
201
202 S.R. Huffman et al.
of DNA. Therefore, often times several embryos need to be pooled

in order to have the necessary yield of nucleic acids to perform
experiments. While conclusions made from such an experimental
design are valid, they can only pertain to groups of embryos or
even groups of female donors.
RNA and DNA isolation from individual samples is essential in
the case where allelic determinations of gene expression and DNA
methylation are sought after when doing genomic imprinting stud-
ies. Genomic imprinting is an epigenetic modification that results
in parent-specific gene expression (2). Imprinted genes are found
in clusters (3) and are often regulated by a differentially methylated
region known as an imprinting control region (ICR). We (4) and
others (5) have demonstrated that manipulation of embryos dur-
ing the preimplantation stage results in loss-of-imprinting during
mid-gestation. In those studies a correlation was made between
biallelic expression of imprinted genes and loss of methylation at
ICRs. Similar studies are not very feasible using single preimplan-
tation embryos due to their paucity of cells. Here we provide the
procedures we use to isolate RNA and DNA from single embryos
and small groups of oocytes. The nucleic acids can be then used for
downstream expression and methylation analyses with good and
consistent results.
Caveat: The protocol presented in this chapter uses oligo dT
magnetic beads which only allows for the hybridization of poly-
adenylated transcripts. However, we have been able to consis-
tently isolate one nonpolyadenylated transcript that contains long
stretches (>20) of adenosines within the transcriptional unit.
2. Materials
1. TE (Tris EDTA)—1 mM EDTA and 10 mM Tris–HCl in

RNAse/DNAse free water. Autoclave.
2. TIT (TE, IGEPAL, Tween 20)—equal parts of TE, 10%
IGEPAL, and 10% Tween 20. Autoclave.
3. Commercially available Kits.
2.1. RNA Isolation 1. Dynabeads mRNA DIRECT kit (Invitrogen).

(a) Lysis buffer/binding buffer (100 mM Tris–HCl, pH 7.5,
M EDTA, pH 8, 1% LiDS, 5 mM dithiothreitol (DTT)).
(b) Washing Buffer A—(10 mM Tris–HCl, pH 7.5, 0.15 M
LiCl, 1 mM EDTA, 0.1% LiDS).
(c) Washing Buffer B—(10 mM Tris–HCl, pH 7.5, 0.15 M
LiCl, 1 mM EDTA).
Alternatively, Oligo d(T)25 magnetic beads (NEB)
13 RNA and DNA Isolation from Single Blastocysts 203
(a) Lysis/Binding Buffer (20 mM Tris–HCl (pH 7.5), 500 mM

LiCl, 0.5% LiDS, 1 mM EDTA, 5 mM DTT).
(b) Wash Buffer I (20 mM Tris–HCl (pH 7.5), 500 mM LiCl,
0.1% LiDS, 1 mM EDTA, 5 mM DTT).
(c) Wash Buffer II (20 mM Tris–HCl (pH 7.5), 500 mM
LiCl, 1 mM EDTA).
2. Other materials—1.5 ml microcentrifuge magnet for affinity
procedures (available from several vendors).
2.2. DNA Bisulfite 1. Sigma Imprint DNA Modification Kit.

Conversion Kit
3. Methods
Foreword: There are a few procedures we utilize which have

increased the yield of nucleic acids obtained from small samples.
These are:
1. When possible, centrifuge tubes between steps to collect drop-
lets from the lid and sides of tubes.
2. Use the smallest pipet tip possible (e.g., when pipetting <10 μl
use a 10 μl tip instead of 20 μl) and pipet slowly.
3. Perform all procedures (from RNA isolation to bisulfite muta-
genesis) in one day (see Note 1).
3.1. Sample Collection 1. Prepare the necessary number of 1.5 ml RNAse/DNAse free
microcentrifuge tube containing 100 μl of lysis/binding buffer.
2. Single embryos or groups of oocytes are placed in the lysis-
buffer containing tubes (see Note 2).
3. Centrifuge the tubes and proceed to RNA isolation or store at
−80.0 °C.
3.2. mRNA Isolation 1. All procedures are done at room temperature.

by Affinity Purification 2. Remove the bottle containing the oligo dT-bound magnetic
beads and gently resuspend the beads by swirling and inverting
the bottle.
3. Pipet out 10 μl of beads for every sample plus one for a blank
control and transfer to an RNase/DNase free 1.5 microcentri-
fuge tube. For example, if analyzing five embryos then pipet
out 60 μl of bead solution.
4. Centrifuge briefly to collect all beads at the bottom of the tube.
5. Place tube on a microcentrifuge tube holder magnet for at least
1 min or until the fluid is perfectly clear.
6. Remove the supernatant using a 100–200 RNase/DNase free

filtered pipet tip while tube is in the magnet (see Note 3).
7. Add 100 μl lysis buffer to the beads and rinse by vortexing
continuously for 5 min (Note 4).
8. Place tubes on the magnet until the supernatant looks clear.
9. Pipet out the supernatant and remove the tube from the magnet.
10. Resuspend beads in lysis buffer. The amount of lysis buffer
added should equal the original volume of bead solution
removed from the bead-stock bottle included with the kit (i.e.,
60 μl).
11. Thaw embryo lysate and centrifuge briefly.
12. Add 10 μl of bead suspension to each of the samples to be
analyzed.
13. The remaining beads will serve as a blank control. Add 100 μl
of lysis buffer to the blank control tube.
14. Place the microcentrifuge tubes containing embryo lysates and
beads on a vortexer for 5 min (medium to low speed) to allow
RNA to anneal to the beads (see Note 5).
15. Place the tubes on a tube rack and incubate for 5 min.
16. Briefly centrifuge tubes and place on magnet for 3–5 min until
the supernatant is completely clear (i.e., beads are all collected
on the back of the tube).
17. Remove supernatant and transfer to a new 1.5 ml microcentri-
fuge tube and save until use (see Note 6; see Subheading 3.3).
18. Wash mRNA/beads complexes by vortexing continuously for
5 min with 100 μl Washing Buffer A.
19. Briefly centrifuge tubes and place on the magnet for 3–5 min.
20. Remove supernatant and repeat wash with Buffer A (steps
15–17).
21. Wash mRNA/beads complexes by vortexing continuously for
5 min with 100 μl of Washing Buffer B.
22. Briefly centrifuge tubes and place on the magnet for 5 min.
23. Repeat wash with Buffer B (steps 18–20; see Note 7)
24. Add Buffer B a third time and resuspend the mRNA/beads
complexes by gentle pipetting.
25. Transfer all of the suspension to a 0.2 ml PCR tube.
26. Place tubes on magnet and leave undisturbed while preparing
the cDNA master mix (cDNAMM).
Preparation of a reusable cDNA library
27. Prepare cDNAMM using your standard procedures (at 10.1 μl/
sample).
28. Completely remove Buffer B from the 0.2 ml PCR tubes (see
Note 8).
29. Add 10 μl of cDNAMM and gently centrifuge.
30. Incubate at 42.0 °C for 1 h in a rotating hybridization oven to
keep the beads in suspension (see Note 9).
31. Place PCR tubes containing the bead/RNA/cDNA mix in a
PCR machine heated to 95.0 °C for 1 min (see Note 10).
32. Remove the supernatant of the first tube as quickly as possible
using a 10 μl pipet tip.
33. Discard the supernatant containing tip.
34. Repeat steps 31–33 until all tubes are processed.
35. Resuspend the magnetic bead-bound cDNA in 50 μl TIT buf-
fer and store at 4.0 °C (see Note 11).
Second strand PCR
36. Create a PCR program to generate a second strand from the
bead/cDNA complex with the following settings; denatur-
ation step 94–95.0 °C/15–20 s, annealing temperature of the
primer 15 s, extension 15–30 s (based on length of amplicon
and optimal temperature of the polymerase), and a final dena-
turation step 94–95.0 °C for 2 min.
37. Prepare a standard PCR master mix (PCRMM; 20–25 μl/
sample).
38. Place a new 1.5 ml tube on a magnet (having at least two mag-
nets simplifies this step) and a new PCR tube on a plastic rack
for every sample to be analyzed.
39. Add 1.5 μl of nuclease-free water to the 1.5 ml tubes.
40. Add 10–15 μl PCRMM to the new PCR tube and set aside.
41. Briefly centrifuge the PCR tubes containing the magnetic
bead-attached cDNA and place on magnet and completely
remove TNT.
42. Gently resuspend beads in 10 μl PCRMM. Place tubes on PCR
machine and run the second strand cycle.
43. Remove tubes immediately upon cycle completion and place
on magnet.
44. Remove supernatant as quickly as possible and add to the bot-
tom of the magnet-held 1.5 ml tube.
45. Immediately, transfer the second strand from the 1.5 ml tube
to the PCR tube containing the PCRMM.
46. The PCR reaction is ready for amplification.
47. Wash the 1.5 ml tube with 50 μl TNT to recover any beads
inadvertently transferred during the second strand recovery
Fig. 1. Amplification of imprinted genes using morula- and early blastocyst-stage mouse
embryos. Each lane contains the amplicon from an individual embryo. mRNA from
C57BL/6 (B6) × B6(CAST7)(C7) F1 hybrid embryos was extracted by affinity purification using
oligo dT magnetic beads. Left panel—the bands are for the imprinted RNA Kcnq1ot1 (not
a polyadenylated transcript). Right panel—the bands are for the imprinted RNA H19 (poly-
adenylated transcript). Samples were restricted with a restriction enzyme that differenti-
ates between the parental alleles to determine the allele the RNA was transcribed form. In
the gels the top band represents one of the alleles and bottom bands are the second
(restricted) allele. When the RNA is expressed from only one of the alleles it is referred to
as monoallelic expression. In the case where RNA is transcribed from both parental alleles
the expression is said to be biallelic. Amplicons were resolved in a 7% polyacrylamide gel.
(b) and (c) represent strain controls for expression and restriction enzyme digestion.
and return to the original tube containing the sample’s cDNA/

bead complex.
48. Store at 4.0 °C.
Figure 1 shows examples of imprinted gene expression analysis
of single mouse preimplantation embryos.
3.3. DNA Isolation 1. Add 2.5 volumes of ice cold absolute ethanol to the tube con-
taining the supernatant from step 17 of Subheading 2.1.
2. Gently mix and incubate at −20 °C for 1 h (see Note 12).
3. Centrifuge at full speed (14,000 rpm/~18,000 × g) for
15 min.
4. Carefully remove half of the supernatant without disturbing
the pellet (see Notes 13 and 14).
5. Centrifuge again at full speed for 5 min.
6. Place the pipet tip on the opposite side of the tube where the
pellet is and remove the supernatant
7. Air dry the DNA pellet for ~5 min (see Note 15).
8. Resuspend the DNA in 21 μl nuclease free water.
9. If the isolated DNA is intended for methylation studies, pro-
ceed to the bisulfite mutagenesis procedure immediately.
Fig. 2. Sequencing of bisulfite converted DNA of a single blastocyst-stage B6xC7 F1 hybrid embryo. DNA was isolated,
bisulfite converted, and amplified by nested PCR using bisulfite converted DNA-specific primers. The amplicons were
cloned and sequenced. The region amplified is the H19/IGF2 ICR (4, 6). This ICR is normally unmethylated on the maternal
allele and methylated one the paternal allele. This figure shows bisulfite sequencing information of six paternal and five
maternal alleles. DNA sequence polymorphisms between B6 and Casteneus strains of mice are shown by arrowheads.
Stars denote CpGs. A CG on the sequence means that that cytosine was methylated while a TG means that CG was
unmethylated.
3.4. Bisulfute Bisulfite conversion is the selective deamination of unmethylated

Mutagenesis cytosines to uracils by treatment with sodium bisulfite. This proce-
dure is performed using the commercially available Imprint DNA
Modification Kit (Sigma; Cat. No. MOD50). See Note 16.
1. To the freshly isolated DNA add
(a) 1 μl of a 7 ng carrier RNA solution.
(b) 1 μl balance solution from the kit.
2. Refer to the two-step method for bisulfite mutagenesis (from
the Sigma Imprint kit manual) and follow the protocol as per
manufacturer’s directions (see Notes 17–19).
Figure 2 shows an example of bisulfite sequencing data obtained
from one blastocyst.
4. Notes
1. Bisulfite mutagenesis is a lengthy procedure and it may not be

possible to mutagenize the DNA on the same day the RNA
isolation is performed. We have obtained good quality bisulfite
converted DNA months after RNA collection by simply pre-
serving the DNA in solution in a 70% EtOH (refer to step 1 of
Subheading 3.3) at −80 °C for several months.
2. Care should be taken to prevent somatic cell contamination of
the sample.
3. Care should be taken during the entire procedure to pipet
from the front of the tube (i.e., away from the beads which will
be pulled to the back of the tube by the magnet) so as not to
accidentally aspirate the beads.
Fig. 3. Plastic rack used to facilitate the handling of samples during the cDNA synthesis
step. The rack with the tubes is placed sideways in a hybridization oven and fastened to
the rotating arm with tape.
4. A foam adapter for the vortexer that holds microcentrifuge

tubes makes this task very simple.
5. Avoid disturbing/pipetting the RNA-bound beads to prevent
RNA loss
6. The supernatant contains the sample’s DNA.
7. Sometimes the supernatant does not become fully clear during
the Buffer B washes (some of the Dynabeads/mRNA com-
plexes will not readily be attracted by the magnet). When this
occurs, gently aspirate the supernatant with a pipette and very
slowly add back to the tube. This usually helps clear the fluid.
8. Process no more than three samples at a time. We have noticed
that if the beads dry up while waiting to be resuspended in
cDNAMM they may “jump” out of the tube and are lost.
9. In order to make this task simple we use a plastic rack (Fig. 3).
10. Process no more than six samples at a time to prevent
reannealing.
11. We have successfully used single embryo cDNA libraries 10–20
times. The TIT may evaporate over long periods of storage.
Add more TIT as necessary.
12. May be stored at −80 °C for several months with good results.
13. It should be noted that the pellet can be hardly seen at the bot-
tom of the tube.
14. Use the smallest pipet tip possible
15. Care should be taken not to have any ethanol left on the tube,
nor overdry the pellet.
16. Use a PCR machine for all incubations.
17. Please note, once the DNA Modification Mixture is added, all
samples are treated as being light sensitive.
18. We have found that the following modifications to the proto-
col improve our success
(a) Half-way through the 90 min incubation at 65.0 °C,
remove the samples from the PCR machine, briefly mix by
inverting the tubes, centrifuge, and return to the PCR
machine for the remainder of the incubation.
(b) We perform all manual-described centrifugations for 1 min
instead of 20 s.
(c) Use 23 μl of sterile water to elute the bisulfite-converted
DNA from the column.
(d) Since bisulfite-converted DNA is single-stranded, storage
at −80.0 °C prevents further degradation.
(e) In addition, aliquoting the samples before storing prevents
freeze–thaw damage.
19. Other information: We have noticed that restricting the genomic
DNA with an enzyme that does not have a recognition site within
the region of interest prior to bisulfite mutagenesis improves the
success rate of PCR amplification after conversion.
Acknowledgments
The original procedures for RNA isolation from single mouse

embryos were first published by Dr. Melissa Mann (6) while in the
laboratory of Dr. Marisa Bartolomei at the University of Pennsylvania.
Dr. John Huntriss, University of Leeds, provided information of
commercial kit for isolation of DNA from single embryos. The data
used in Fig. 1 was generated by Mr. Franklin Echevarria.
References
1. Nagy A, Gertsenstein M, Vintersten K, Manipulations of mouse embryos prior to

Behringer R (2003) Manipulating the mouse implantation result in aberrant expression of
embryo: a laboratory manual, 3rd edn. Cold imprinted genes on day 9.5 of development.
Spring Harbor Laboratory Press, Cold Spring Hum Mol Genet 17:1–14
Harbor, NY 5. Doherty AS, Bartolomei MS, Schultz RM
2. Morison IM, Ramsay JP, Spencer HG (2005) (2002) Regulation of stage-specific nuclear
A census of mammalian imprinting. Trends translocation of Dnmt1o during preimplanta-
Genet 21:457–465 tion mouse development. Dev Biol 242:
3. Verona RI, Mann MR, Bartolomei MS (2003) 255–266
Genomic imprinting: intricacies of epigenetic 6. Mann MR, Lee SS, Doherty AS, Verona RI,
regulation in clusters. Annu Rev Cell Dev Biol Nolen LD, Schultz RM, Bartolomei MS (2004)
19:237–259 Selective loss of imprinting in the placenta fol-
4. Rivera RM, Stein P, Weaver JR, Mager J, lowing preimplantation development in cul-
Schultz RM, Bartolomei MS (2008) ture. Development 131:3727–3735
Chapter 14
Generation of cDNA Libraries from RNP-Derived

Regulatory Noncoding RNAs
Mathieu Rederstorff
Abstract
Next-generation sequencing of noncoding RNA (ncRNA) libraries has become an essential tool for the
profiling of ncRNAs and the identification of novel ncRNA species. Here, we describe the generation of a
ncRNA-derived complementary DNA (cDNA) library by 3¢-tailing of ncRNAs by CTP and poly(A) poly-
merase, followed by 5¢-adapter ligation by T4 RNA ligase and reverse transcription of ncRNAs with an
oligo-d(G) anchor primer. Preliminary selection of ncRNAs from ribonucleoprotein particles (RNPs)
enables a strong enrichment of the generated libraries with functional regulatory ncRNAs compared to
classical approaches.
Key words: Noncoding RNAs, Ribonucleoprotein particle, cDNA library, High-throughput sequencing
(or next-generation sequencing NGS)
1. Introduction
More than 90% of the human genome appears to be actively

transcribed, although only a minor portion of transcripts code for
proteins (1). Therefore, it has been proposed that most of the tran-
scripts correspond to regulatory noncoding RNAs (ncRNAs)
(2–8). Over the last decade, the importance of ncRNAs has
increased constantly (9–13). NcRNAs play key regulatory func-
tions in multiple biological pathways, from DNA replication and
chromosome maintenance, to RNA modification or regulation of
translation (14–24). NcRNAs are also involved in regulating the
transcriptional machinery, controlling transcription factors directly
or mediating epigenetic changes through the interaction with pro-
teins that modify DNA or histones (25). NcRNAs involved in epi-
genetic regulation are diverse. They comprise the shortest ncRNAs
yet described, namely, short interfering RNAs (siRNAs, 20–22
211
212 M. Rederstorff
Fig. 1. Overview of the library generation approach.
nucleotides), acting within the RNA-induced transcriptional silencing

(RITS) complex (26, 27), as well as some of the longest ncRNAs
known to date, such as Xist (19 kb) (28) or Air (108 kb) (29),
which recruit Polycomb group (PcG) complexes, like the Polycomb
repressive complex 2 (PRC2), to mediate silencing (30).
To identify novel ncRNAs, a method of choice consists in the
generation and sequencing of cDNA libraries, which has been con-
siderably improved owing to the development of next-generation
sequencing techniques. Previously, cDNA libraries were generated
using RNA that was size-separated on denaturing gels (12, 31, 32)
leading mostly to the cloning and sequencing of RNA species cor-
responding to ribosomal RNAs (rRNAs) or other known ncRNA
species (33). We developed an alternative procedure, based on the
size-selection of ribonucleoprotein (RNP) particles on glycerol
gradients prior to RNA extraction, which enables a strong enrichment
of the libraries with functional ncRNAs species (Fig. 1) (34, 35).
NcRNAs are next 3¢-tailed with CTP and poly(A) polymerase,
followed by 5¢-adapter ligation by T4 RNA ligase and reverse
transcription with an oligo-d(G) anchor primer. Libraries obtained
are compatible with most high-throughput sequencing techniques
available (36).
14 Detection of Small Non-coding RNAs 213
2. Materials
Diethylpyrocarbonate (DEPC)-treated RNase-free water should be

used to prepare all solutions, as well as to dissolve RNA pellets after
ethanol precipitation. To prepare DEPC-treated RNase-free water,
mix double-distilled water with 0.01% (vol/vol) DEPC in RNase-
free bottles. Let stand overnight and autoclave. RNase-free water
can be stored at room temperature (20 °C) for several months.
2.1. RNP 1. Prepare nuclear or cytoplasmic extracts according to Dignam

Sedimentation et al. (35, 37).
2. L7-65 ultracentrifuge (Beckman).
3. SW41 rotor (Beckman).
4. Polyallomer centrifuge tubes: thin wall, 13.2 ml, 14 × 89 mm
(Beckman 331372).
2.2. RNA Preparation 1. Phenol–Chloroform–Isoamyl alcohol: 25/24/1 solution

(see Note 1).
2. Chloroform.
3. Sodium acetate 3 M (pH 5.2).
4. Ethanol.
5. Glycoblue™: 15 mg/ml glycogen (Ambion AM 9516)
(see Note 2).
2.3. cDNA Library 1. RiboLock™ RNAse Inhibitor (Fermentas Life Sciences

Generation EO0381) (see Note 3).
2. A-Plus™ poly(A) Polymerase kit, including 1× Tailing buffer:
50 mM Tris–HCl (pH 8.0), 250 mM NaCl, 10 mM MgCl2
(Epicentre Biotechnologies PAP5104H).
3. Cytidine triphosphate, CTP.
4. TAP: Tobacco Acid Pyrophosphatase, including 1× TAP
buffer: 50 mM sodium acetate (pH 6.0), 1 mM EDTA, 1%
b-mercaptoethanol, 0.01% Triton® X-100 (Epicentre
Biotechnologies T19250).
5. T4 RNA Ligase, including 1× RNA Ligation buffer: 50 mM
Tris–HCl (pH 7.8), 10 mM MgCl2, 10 mM DTT and 1 mg/
ml BSA, 10 mM ATP (Fermentas Life Sciences EL0021).
6. 5¢-Adapter (Ribonucleotides underlined): 5¢-GTC AGC AAT
CCC TAA C GAG-3¢ (Pharmacon) (see Note 4).
7. SuperScript™ II Reverse Transcriptase, including 5× First strand
buffer: 250 mM Tris–HCl (pH 8.3), 375 mM KCl, 15 mM
MgCl2 (Life technologies Invitrogen 100004925).
8. Deoxynucleotide triphosphate, dNTP.
214 M. Rederstorff
9. Anchor primer (for reverse transcription): 5¢-AGG AGC CAT

CGT ATG TCG GGG GGG GH-3¢ (Microsynth) (see Note 4).
10. Taq DNA Polymerase I, including 1× Taq buffer: 10 mM Tris–
HCl (pH 9.0), 50 mM KCl, 0.1% Triton® X-100 (Life
Technologies Invitrogen 10342020) (see Note 5).
11. 5¢ PCR primer: 5¢-(sequencing tag*)-GTC AGC AAT CCC
TAA CGA G-3¢ (Microsynth).
12. 3¢ PCR primer: 5¢-(sequencing tag*)-AGG AGC CAT CGT
ATG TCG-3¢ (Microsynth).
*The sequencing tags depend on the high-throughput sequencing
technology chosen (see Note 4).
3. Methods
3.1. RNP 1. Layer the nuclear, cytoplasmic, or total extract onto a 10–30%
Sedimentation glycerol gradient (see Note 6).
2. Spin the gradient at 198,000 × g (40,000 rpm) for 13 h at 4 °C
in a Beckman SW41 rotor.
3. Collect the fractions of interest (see Note 7).
3.2. RNA Preparation 1. Extract RNA from each fraction with phenol–chloroform (1:1)
(see Note 1).
2. Thoroughly vortex the tubes for 5 min at 20 °C to completely
dissociate RNPs and extract RNA.
3. Centrifuge the tubes for 5 min at 20 °C at 10,000 × g to sepa-
rate the organic (lower) and water (upper) phases.
4. Collect the RNA containing upper phase (see Note 8).
5. Mix the sample with chloroform (1/1).
6. Vortex the tubes for 5 min at 20 °C to extract remaining traces
of phenol from the sample.
7. Centrifuge the tubes for 5 min at 20 °C at 10,000 × g.
8. Collect the RNA-containing upper phase.
9. Mix the sample with 3 M sodium acetate (pH 5.2) (1/10) and
100% ethanol (3/1) to precipitate RNA (see Note 2).
10. Gently vortex the tubes and precipitate RNA for 30 min at
−80 °C (see Note 9).
11. Centrifuge the tubes for 15 min at 10,000 × g at 4 °C.
12. Remove the supernatant.
13. Gently wash the RNA pellet with 70% DEPC-ethanol (see
Note 10).
14. Centrifuge the tubes for 5 min at 10,000 × g at 4 °C.
15. Discard the supernatant and air dry the RNA pellet (see Note
11).
16. Dissolve the RNA pellet in 20 ml DEPC-treated water.
17. Estimate RNA concentration using a spectrophotometer.
3.3. cDNA Library C-Tailing of ncRNAs (see Note 12)

Generation
1. Tail 150 ng to 1 mg of RNA in a final reaction volume of 20 ml
containing 1× Tailing buffer, 4 U of A-Plus™ poly(A) poly-
merase, 20 U of RiboLock™ RNAse Inhibitor and 100 mM
CTP (see Note 13).
2. Incubate the tube for 90 min at 37 °C.
3. Fill up the volume to 200 ml with DEPC-treated water and
stop the reaction by phenol–chloroform extraction (steps 1–8
Subheading 3.2).
4. Precipitate RNA (steps 9–15 Subheading 3.2).
5. Dissolve the RNA pellet in 7 ml DEPC-treated water (see
Note 14).
Decapping of ncRNAs (see Note 15)
6. Add 1× TAP buffer, 10 U of TAP and 40 U of RiboLock™
RNAse Inhibitor in a final reaction volume of 10 ml.
8. Fill up the volume to 200 ml with DEPC-treated water.
9. Phenol–chloroform extract and precipitate RNA (steps 1–15
Subheading 3.2).
5¢-Adapter Ligation
11. Add 2 ml 5¢-adapter (100 pmol/ml).
13. Add 1× RNA Ligation buffer, 1 mM ATP, 0.1 mg/ml BSA,
40 U of RiboLock™ RNAse Inhibitor and 10–20 U of T4
RNA Ligase in a final reaction volume of 20 ml.
14. Incubate the tube overnight at 4 °C (see Notes 16 and 17).
15. Phenol–chloroform extract and precipitate RNA (steps 1–15
Subheading 3.2).
Reverse-Transcription
17. Add 1 ml Anchor primer (100 pmol/ml) and 1 ml dNTP
(10 mM each).
18. Denaturate the mixture for 5 min at 65 °C and immediately
cool on ice.
216 M. Rederstorff
19. Add 4 ml 5× First strand buffer, 2 ml DTT (100 mM), 40 U of

RiboLock™ RNAse Inhibitor.
21. Add 200 U of SuperScript™ II Reverse Transcriptase.
cDNA Amplification
23. To 1/5 (4 ml) of the cDNA obtained, add 1× Taq buffer,
1 pmol/ml of both PCR primers, 0.25 mM MgCl2, 0.1 mM
dNTP and 1 U of Taq DNA Polymerase I in a final reaction
volume of 50 ml (see Note 5).
24. Denature the sample for 2 min at 94 °C in a thermal cycler.
25. Amplify the cDNA with 20–25 cycles of PCR as follows: 1 min
at 94 °C, 1 min at 54 °C, 1 min at 72 °C.
26. Finalize amplification for 5 min at 72 °C (see Notes 18 and 19).
27. The library can be sequenced with the chosen high-through-
put sequencing platform (see Note 4).
4. Notes
1. The pH of the phenol–chloroform mixture used for the first

extraction can be acidic (below 7, ideally 4.5 or 5) in order to
denature DNA that may possibly be present as a contamina-
tion. At pH 7, DNA and RNA will appear in the water phase.
At pH 5, DNA is partly denatured and will be directed to the
organic phase, while RNA remains in the water phase.
2. For the first precipitation step, a carrier, e.g., glycogen or
Glycoblue™, can be added since the amounts of material
retrieved after the gradient may be rather low.
3. Employing RNAse inhibitors throughout the procedure in
addition to using DEPC-treated water significantly reduces
RNA degradation, especially during long incubation periods.
4. Additional specific sequences may have to be introduced in the
5¢-adapter and anchor primer as well. These tags enable the
cDNAs to bind on the sequencing slides. Regarding the 3¢
extremity of ncRNAs, alternative protocols describe ligation of
a 3¢ adapter; however, we found it always more efficient and
productive to proceed with C-tailing of the 3¢ extremity.
5. The PCR step can also be carried out with a homemade Taq
polymerase if available. However, high-fidelity enzymes are
recommended to avoid too many errors.
6. The glycerol content of the sample should not exceed 10% to

stay on top of the gradient when loaded. Dilute sample in an
adequate buffer to reach a 10% final glycerol concentration.
7. We recommend collecting fractions starting from the bottom
of the tube to avoid perturbation of the gradient, using a peri-
staltic pump or a syringe pump if available. Very bottom and
top fractions (1 ml each) should be discarded. These fractions
contain rRNA contaminations and various RNA degradation
products, respectively. Remaining fractions can be pooled if no
particular RNP size (10–30S) is required. One can decide to
work only with specific subfractions (e.g., 15S).
8. Steps 1–4 can be repeated once.
9. Alternatively, RNA can be precipitated overnight at −20 °C.
10. To avoid washing out of small RNAs (e.g., miRNAs), pellets
should be extremely gently washed throughout the protocol.
11. A pellet that is too dry may be harder to dissolve.
12. Since ncRNAs do not contain a poly(A) tail (as do mRNAs),
ncRNAs are, in a first step, tailed with CTP and poly(A) poly-
merase (which also uses CTP as a substrate).
13. Poly(A) polymerase has a higher affinity for ATP than for CTP,
therefore, increasing CTP concentration in the reaction buffer
to 100 mM (10 mM for ATP), strongly improves the yield of
the C-tailing reaction.
14. It is not necessary to quantify RNA here and after the forth-
coming precipitation steps. Pellets are now dissolved in the
adequate volume of DEPC-treated water and all the material is
used for the next step.
15. This step enables all RNAs to feature a 5¢ phosphorylated ter-
minus, which is necessary for 5¢-adapter ligation.
16. For the best efficiency of overnight ligation, samples should be
incubated in melting ice (water + ice), rather than in crushed
ice only. To avoid having the ice melt too fast, the incubation
should take place in a cold room.
17. After overnight ligation at 4 °C, adding 10 more units of T4
RNA ligase for 30 min at 37 °C will increase the yield of the
reaction.
18. PCR product can be purified on a native polyacrylamide gel.
The user thus has the possibility of performing a size selection,
e.g., products of sizes comprised between 20 and 500 nucle-
otides (nt), prior to high-throughput sequencing.
19. Quality control (agarose gel electrophoresis, cloning, and
Sanger sequencing to check the presence of the deep-sequencing
adaptors) can be performed but is generally done by the com-
pany/platform in charge of the deep-sequencing.
218 M. Rederstorff
References
1. Birney E et al (2007) Identification and analysis 20. Rhoades MW et al (2002) Prediction of plant
of functional elements in 1% of the human microRNA targets. Cell 110:513–520
genome by the ENCODE pilot project. Nature 21. Storz G (2002) An expanding universe of non-
447:799–816 coding RNAs. Science 296:1260–1263
2. Brosius J (2005) Waste not, want not—tran- 22. Tuschl T (2002) Expanding small RNA inter-
script excess in multicellular eukaryotes. Trends ference. Nat Biotechnol 20:446–448
Genet 21:287–288 23. Tuschl T (2003) Functional genomics: RNA
3. Carninci P et al (2005) The transcriptional sets the standard. Nature 421:220–221
landscape of the mammalian genome. Science 24. Volpe TA et al (2002) Regulation of hetero-
309:1559–1563 chromatic silencing and histone H3 lysine-9
4. Cheng J et al (2005) Transcriptional maps of methylation by RNAi. Science 297:
10 human chromosomes at 5-nucleotide reso- 1833–1837
lution. Science 308:1149–1154 25. Barrandon C, Spiluttini B, Bensaude O (2008)
5. Kampa D et al (2004) Novel RNAs identified Non-coding RNAs regulating the transcrip-
from an in-depth analysis of the transcriptome tional machinery. Biol Cell 100:83–95
of human chromosomes 21 and 22. Genome 26. Morris KV et al (2004) Small interfering RNA-
Res 14:331–342 induced transcriptional gene silencing in human
6. Mattick JS, Makunin IV (2005) Small regula- cells. Science 305:1289–1292
tory RNAs in mammals. Hum Mol Genet 27. Verdel A et al (2004) RNAi-mediated targeting
14(1):R121–R132 of heterochromatin by the RITS complex.
7. Mattick JS, Makunin IV (2006) Non-coding Science 303:672–676
RNA. Hum Mol Genet 15(1):R17–R29 28. Maenner S et al (2010) 2-D structure of the A
8. Willingham AT, Gingeras TR (2006) TUF love region of Xist RNA and its implication for
for “junk” DNA. Cell 125:1215–1220 PRC2 association. PLoS Biol 8:e1000276
9. Couzin J (2002) Breakthrough of the year. Small 29. Braidotti G et al (2004) The Air noncoding
RNAs make big splash. Science 298:2296–2297 RNA: an imprinted cis-silencing transcript.
10. Dennis C (2002) Small RNAs: the genome’s Cold Spring Harb Symp Quant Biol 69:55–66
guiding hand? Nature 420:732 30. Margueron R, Reinberg D (2011) The
11. Dennis C (2002) The brave new world of RNA. Polycomb complex PRC2 and its mark in life.
Nature 418:122–124 Nature 469:343–349
12. Huttenhofer A, Brosius J, Bachellerie JP (2002) 31. Huttenhofer A et al (2001) RNomics: an
RNomics: identification and function of small, experimental approach that identifies 201 can-
non-messenger RNAs. Curr Opin Chem Biol didates for novel, small, non-messenger RNAs
6:835–843 in mouse. EMBO J 20:2943–2953
13. Rederstorff M, Huttenhofer A (2010) Small 32. Huttenhofer A, Vogel J (2006) Experimental
non-coding RNAs in disease development and approaches to identify non-coding RNAs.
host–pathogen interactions. Curr Opin Mol Nucleic Acids Res 34:635–646
Ther 12:684–694 33. Jochl C et al (2008) Small ncRNA transcriptome
14. Ambros V (2001) microRNAs: tiny regulators analysis from Aspergillus fumigatus suggests a
with great potential. Cell 107:823–826 novel mechanism for regulation of protein syn-
15. Bachellerie JP, Cavaille J, Huttenhofer A thesis. Nucleic Acids Res 36:2677–2689
(2002) The expanding snoRNA world. 34. Rederstorff M et al (2010) RNPomics: defining
Biochimie 84:775–790 the ncRNA transcriptome by cDNA library
16. Gottesman S (2002) Stealth regulation: bio- generation from ribonucleo-protein particles.
logical circuits with small RNA switches. Genes Nucleic Acids Res 38:e113
Dev 16:2829–2842 35. Rederstorff M, Huttenhofer A (2011) cDNA
17. Kiss T (2002) Small nucleolar RNAs: an abun- library generation from ribonucleoprotein par-
dant group of noncoding RNAs with diverse ticles. Nat Protoc 6:166–174
cellular functions. Cell 109:145–148 36. Metzker ML (2010) Sequencing technologies -
18. Lau NC et al (2001) An abundant class of tiny the next generation. Nat Rev Genet 11:31–46
RNAs with probable regulatory roles in 37. Dignam JD, Lebovitz RM, Roeder RG (1983)
Caenorhabditis elegans. Science 294:858–862 Accurate transcription initiation by RNA poly-
19. Lee RC, Ambros V (2001) An extensive class merase II in a soluble extract from isolated
of small RNAs in Caenorhabditis elegans. mammalian nuclei. Nucleic Acids Res 11:
Science 294:862–864 1475–1489
Chapter 15
Co-Immunoprecipitation of Long Noncoding RNAs

Victoria A. Moran, Courtney N. Niland, and Ahmad M. Khalil
Abstract
It is now estimated that the human genome encodes thousands of long noncoding (lnc)RNAs. These novel
molecules are causing a paradigm shift in the field of molecular biology as a number of lncRNAs have been
shown to be involved in a wide range of biological functions including regulation of gene expression. Also,
misregulation of lncRNAs has been observed in human diseases such as cancer and neurological disorders.
These findings have spurred a huge interest in elucidating the functions and mechanisms of lncRNAs; and
therefore, the need for new methods to do so. In this chapter, we discuss RIP-Seq, a method that is utilized
to discover the lncRNA partners of a specific protein. This procedure involves immunoprecipitation of a
protein from cross-linked cell lysate followed by reverse-cross-linking, isolation, and deep sequencing of
RNAs, leading to the identification of all lncRNAs that are associated with a specific protein complex.
Key words: Long noncoding RNA, Large noncoding RNA, lncRNA, lincRNA, Ribonucleic protein
complex, RNP, RNA coimmunoprecipitation, RIP, RNA–protein interactions
1. Introduction
One of the most surprising findings of the International Human

Genome Sequencing Consortium was that only about 2% of the
entire human genome is protein-coding (1), and is accounted for
by the 20,000 or so protein-coding genes (1). It is has also been
suggested that as much as 90% of the genome can be transcribed (2).
While some believe that a majority of this disparity can be attrib-
uted to so-called “transcriptional noise” (3), other evidence points
to a large number of genes encoding functional noncoding RNAs.
RNA has long been known to play a myriad of different functions
distinct from its classical protein-coding role. In this way, the
“Central Dogma of Molecular Biology” has been at best an
oversimplification and at worst, a scientific falsity for decades. The
functional roles of ribosomal (r)RNAs, transfer (t)RNAs, snRNAs,
and snoRNAs have long been elucidated. Recent advances in
219
220 V.A. Moran et al.
technology such as tilling arrays and deep sequencing of RNAs have

led to the discovery of additional classes of noncoding RNAs
including small noncoding RNAs (e.g., microRNAs and piRNAs)
and long (or large) noncoding RNAs (e.g., NAT and lincRNAs)
(2, 4–13). Large noncoding RNAs (lncRNAs) are especially abun-
dant in mammalian genomes, with some groups estimating the
number of lncRNA encoding genes at approximately 23,000 (13).
The recent stir of this novel class of RNAs prompted a univer-
sal definition. A lncRNA is defined as any endogenous RNA that is
over 200 nucleotides in length, but lacks a sensible open reading
frame capable of producing a protein product, usually defined as an
ORF of more than 100 amino acids (14, 15). Many new lncRNA
molecules are being discovered and characterized at a rapid pace,
and new functions and mechanisms are being elucidated as well.
However, at this time, a relatively small number of lncRNAs have
been studied in an extensive manner. Of these, arguably the best
known and studied is Xist (X inactive specific transcript) which has
an essential role in the initiation of X inactivation in mammalian
female somatic cells as a means of dosage compensation (16, 17).
In addition to Xist function in dosage compensation, other bio-
chemical roles for lncRNAs have also been identified. So far, lncR-
NAs have been shown to have a prominent role in epigenomic
control, genomic imprinting, transcriptional control, nuclear com-
partmentalization, as well as stem cell proliferation and differentia-
tion (4). In addition, their misregulation has been observed in a
host of human disorders, including cancer metastasis, neurological
disease, cardiac disease, and others (18, 19). While lncRNAs func-
tion in obviously different cellular pathways, many lncRNAs function
as a part of ribonucleoprotein complexes (RNPs).
RNPs are defined as any complex in which RNAs associate
with protein partners to carry out a function cooperatively. These
can be further broken down into classes, based on the biochemical
role of the RNA itself. While outside the scope of this chapter, it is
important to note a few classical examples of noncoding (nc)RNPs.
Ribosomes are the most extensively studied, where the ribosomal
RNA serves as the catalytic core of the ncRNP complex. Other
examples of classical ncRNPs include eukaryotic telomerase and
the spliceosomal snRNPs. This chapter focuses on the newly
emerging field of lncRNAs and their various protein partner com-
plexes. Although lncRNAs function in different manners, patterns
are emerging based on their relationship to their protein partners.
These patterns can be separated into discrete functional classes,
which are highlighted below.
1.1. Ribonucleic Acids One of the biggest gaps in our collective knowledge of chromatin-
Act to Target Protein modifying complexes is how they are recruited to specific regions
Partners to Specific of the genome. Many chromatin-modifying complexes are capable
Genomic Loci of adding or removing histone modifications (20, 21), but many of
15 RIP-Seq of Long Non-Coding RNAs 221
these complexes lack DNA binding domains. Multiple recent studies

have shown that some lncRNAs, which bind to chromatin-modifying
complexes, are capable of recruiting them to specific genomic tar-
gets. For example, during the process of mammalian X inactiva-
tion, Xist recruits polycomb repressive complex 2 (PRC2) to initiate
transcriptional repression (22). Also, the lncRNA HOTAIR, which
is transcribed from the HOXC locus, was shown to target PRC2
initially to HOXD genes (23), and more recently, to many regions
throughout the genome (24, 25). This trend of lncRNAs acting as
“traffic controllers” of repressive chromatin complexes is likely to
be more widespread. It has recently been shown that as many as
40% of intergenic lncRNAs (lincRNAs) are capable of directly asso-
ciating with several chromatin-modifying complexes including
PRC2, CoREST, and SMCX (10).
LncRNAs have also been shown to recruit chromatin-modifying
complexes to imprinted genes in an allele-specific manner. Somatic
cells have two copies of each autosome, one is inherited paternally
and the other maternally. At most gene loci, both alleles are
expressed. However, some loci are under imprinted regulation,
that is, those specific loci are expressed from either the maternal or
paternal allele, depending on the specific locus. Recently, it has
been shown that lncRNAs may have a role in the recruitment of
chromatin-modifying complexes to imprinted loci. One such
example is the lncRNA Air. Air is expressed from a promoter in
the antisense direction to the Igf2r gene locus, and it is only
expressed from the paternal copy (26). Air associates with the
chromatin-modifying complex G9a, a repressive histone methyl-
transferase, and it targets G9a in cis to the Igf2r locus which results
in the silencing of several imprinted genes (27). Another example
of a RNP acting to facilitate imprinting has been observed in the
Kcnq1 gene locus. Another lncRNA, Kcnq1ot1, is expressed only
from the paternally inherited chromosome and antisense to the
Kcnq1 gene. Kcnq1ot1 silences adjacent genes via the recruitment
of chromatin repressor complexes. Similar to Air, Kcnq1ot1 has
been shown to interact with G9a, and additionally, components of
the PRC2 complex (28). Interestingly, studies have also indicated
that Kcnq1ot1 associates with DNMT1, a DNA methyltransferase
which is responsible for the maintenance of DNA methylation pat-
terns (29). In this way, a single lncRNA may act through multiple
pathways to regulate gene expression.
1.2. LncRNAs Perform Another function of lncRNAs in RNPs is to serve as scaffolds,

Scaffolding Functions bringing proteins with different catalytic domains into close prox-
to Bridge Distinct imity (19). For example, HOTAIR binds to both PRC2 and the
Protein Partners histone demethylase LSD1, thus serving as a scaffold between
these two distinct complexes (24). Tsai and colleagues found
evidence that different regions of HOTAIR bind to PRC2 and
LSD1, potentially coupling their catalytic functions together (24).
This mechanism of RNA serving a structural scaffolding role is by

no means restricted to mammals. A notable example of lncRNAs
portraying such a function can also be seen in the dosage compen-
sation pathway in Drosophila melanogaster. In fruit flies, dosage
compensation takes an opposite approach to that of mammals in
order to equalize sex-linked gene products. In mammals, dosage
compensation is achieved by silencing one X-chromosome in XX
females. In flies, dosage compensation is enacted by upregulation
of the single X in XY males. This upregulation requires the Male Sex
Lethal Complex (MSL), which coats the single male X, promoting
active chromatin state. This MSL complex is held together by
the lncRNAs, roX1 and roX2 (30). In addition, the roX RNAs
facilitate the spreading of the complex over the entire X-chromosome
(31, 32).
1.3. LncRNAs Function In 1999, the first lncRNA was characterized that displayed a bio-
as Coregulators of chemical role outside dosage compensation and imprinting. The
Transcription lncRNA SRA (steroid receptor hormone activator) was discovered
Machinery in screens for proteins acting in the regulation pathways of hor-
monal nuclear receptors (33). Nuclear receptors are a unique class
of transcription factors, which only actively regulate transcription
in the presence of a ligand. Specifically, ligand binding induces a
conformational change in the nuclear receptors, which are then
activated to bind to genomic cognate response elements, as well as
to potentially recruit other coregulatory factors. It is believed that
SRA forms a RNP with a common coactivator complex, named
SRC-1, contributing to its binding potential, and thus regulation,
of nuclear receptors (34).
Another example of coregulation of transcription can be seen
in the corepressive activity of another lncRNA, Gas5. Gas (growth
arrest specific) 5 is highly expressed in cells that have arrested
growth function, due to the underavailability of necessary nutri-
ents. It has been shown to be a negative regulator of glucocorticoid
receptors, a specific class of nuclear receptors (35). While operating
on receptor molecules as does SRA, the mechanism behind Gas5
interaction is radically different. Gas5 interacts directly with the
receptor’s DNA binding domain, preventing it from binding to its
DNA response element and effectively acting as a molecular decoy
(35). In the absence of binding of the glucocorticoid receptor
(GR), expression of GR target genes are severely downregulated.
This mechanism clearly shows the possibility of a direct role in tran-
scriptional regulation by lncRNA in a protein complex context.
1.4. RIP-Seq The examples of RNPs discussed above only scratch the surface of
the numerous potential lncRNA–protein interactions. As noted
above, as many as 40% of lincRNAs are capable of directly associat-
ing with several chromatin-modifying complexes (i.e. PRC2,
CoREST, and SMCX) (10). According to this observation, vast
numbers of interactions of lncRNA to chromatin-modifying
complexes, transcriptional regulators, and other protein complexes

are yet to be characterized (19). RIP-Seq (RNA coimmunoprecipi-
tation followed by deep sequencing) is a method that can be uti-
lized to discover all lncRNAs that interact with a specific protein of
interest genome-wide. Briefly, RNA is first cross-linked to proteins
by formaldehyde, protein of interest is immunoprecipitated by a
specific antibody, then the cross-linking is reversed, and RNA is
isolated and prepared for deep sequencing. The RIP method
described here is an adaptation from previous protocols with
modifications that we worked out in our laboratory to optimize the
technology (24, 36). The method described here utilizes low per-
centage formaldehyde (~0.3%) for cross-linking RNA to proteins
prior to cell lysis, thus avoiding potential interactions that may occur
after cell lysis (37). However, recent studies suggest that native (no
cross-linking) RIP can also be used to identify lncRNA–protein
interactions with high specificity and reproducibility (10, 38).
1.5. RIP-Seq Protocol This technology depends highly on the quality of the antibody
used; therefore a first step is to obtain a high quality antibody and
test it in your laboratory. Also, all reagents and materials must be
RNase-free to prevent RNA degradation. It is critical to point out
that many factors play a role in the success of this procedure includ-
ing antibody quality, cell type used, and successful preparation of
the cell lysate; therefore, this protocol is a good starting point, and
notes at each step in the following protocol can help optimize your
results on a case-by-case basis.
2. Materials
1. Protein A/G magnetic beads.

2. Antibody against protein of interest, and IgG Antibody.
3. RIPA Buffer: 150 mM NaCl, 1.0% NP-40, 0.5% Sodium
deoxycholate, 0.1% SDS, 50 mM Tris–HCl (pH 7.4), 1.0 mM
EDTA.
4. Cells grown to 80–90% confluency.
5. Dulbecco’s Modified Eagle’s Media, supplemented with 10%
Fetal Bovine Serum.
6. Cold 1× PBS.
7. 37% Formaldehyde.
8. 1.25 M Glycine in PBS.
9. Protease Inhibitors; 10× stock solution.
10. RNAse Inhibitor.
11. High Salt RIPA Buffer: 1.0 M NaCl, 1.0% NP-40, 0.5% Sodium
deoxycholate, 0.1% SDS, 50 mM Tris–HCl (pH 7.4), 1.0 mM
EDTA.
12. Buffer C: 150 mM NaCl, 50 mM Tris–HCl (pH 7.4), 5 mM
EDTA, 10 mM DTT, 1.0% SDS.
13. Proteinase K.
3. Methods
3.1. Prepare Beads for 1. Take 2 aliquots of 50 μL each of protein A/G magnetic beads,
Immunoprecipitation and wash two times with 300 μL RIPA buffer (see Note 1).
2. Remove wash supernatant with a magnet and resuspend beads
in 500 μL of RIPA buffer.
3. Add 8 μg of antibody against your protein of interest to the first
aliquot. To the second, add 8 μg of nonspecific IgG antibody
(see Note 2).
4. Incubate beads with antibodies for 2 h at 4 °C with gentle
rotation.
3.2. Lysate Preparation 1. Grow cells to ~90% confluency. Six 15 cm plates are usually
and Cross-linking sufficient for most cell types; larger cells may require more
plates to start with (see Note 3).
2. Harvest cells by trypsinization, adding an equivalent amount
of media to quench the reaction. Collect cells in a 15 mL coni-
cal tube and pellet by centrifuging at 500 × g for 10 min.
3. Wash cells twice with PBS and collect cells after each wash by
pelleting at 500 × g for 5 min.
4. Resuspend cellular pellet in 10 mL of PBS. Using a hemocy-
tometer, or other quantification method, calculate the con-
centration of cells. Dilute suspension to two million cells
per mL.
5. To perform cross-linking, add 37% formaldehyde to a final
concentration of 0.3%. Incubate for 10 min, with gentle rota-
tion at room temperature (see Note 4).
6. Quench the cross-linking reaction by adding 1.25 M glycine to
a final concentration of 0.125 M. Incubate at room tempera-
ture for 5 min (see Note 5).
7. Pellet cells by centrifuging at 500 × g for 5 min.
8. Wash cells twice, each time with 10 mL of 1× PBS. Spin and
pellet as before.
9. Resuspend pellet in 2.2 mL of RIPA buffer, supplemented
with protease and RNAse inhibitors.
10. Incubate at 37 °C for 30 min. Vortex every 5 min for 30 s

intervals for the duration of the incubation.
11. Homogenize the sample using a dounce homogenizer. Make
certain to use enough force to disrupt the cellular membranes
(see Note 6).
12. Centrifuge the lysate at maximum speed (≥10,000 × g) using a
microcentrifuge for 10 min. Collect the supernatant, discarding
the pellet.
13. Optional: Perform pre-clearing of the lysate. Wash 30 μL of
protein A/G beads two times in RIPA buffer. Remove wash
buffer and add the supernatant from step 16. Incubate for 1 h
with gentle rotation at 4 °C. Recover supernatant and proceed
to step 18 (see Note 7).
14. Wash beads from step 4 twice with 500 μL of RIPA buffer.
Discard supernatant.
15. Set aside a 100 μL aliquot of lysate to be used as an Input
sample. Divide the rest of the lysate between the two antibody-
mounted beads (1 mL each).
16. Incubate for 4 h or overnight at 4 °C with gentle rotation
(see Note 8).
17. Remove supernatant by pelleting beads using a magnet and
wash four times with high salt RIPA buffer (see Note 9).
18. Wash one time with 1× PBS.
3.3. For Protein At this point it is necessary to ensure that the antibody used works
Analysis in IP experiments, therefore, western blot analysis of the protein is
critical to ensure that your antibody immunoprecipitates your pro-
tein of interest efficiently.
1. Resuspend beads in 100 μL of Laemmeli Lysis Buffer. Also
add 100 μL of Laemmeli buffer to Input.
2. Incubate at 95 °C for 5 min to denature proteins and to reverse
the cross-linking.
3. Run on a denaturing SDS-PAGE Gel, and perform western
blot analysis on input, the IP using an antibody against your
protein of interest, and IP with IgG antibody as a negative
control. You should see a correct size band for your protein of
interest in your input and specific antibody IP but not in your
IgG IP. You should also include a positive control on your gel,
such as a total cell extract with no treatment.
3.4. For RNA Analysis 1. Resuspend beads in 100 μL of Buffer C. Additionally, add
10 μg of proteinase K. Also, add 100 μL of Buffer C and pro-
teinase K to the input sample (input volume is now 200 μL).
2. Incubate for 30 min at 42 °C, for proteinase K digestion.
3. Incubate for 4 h at 65 °C to reverse the formaldehyde cross-links.
4. Perform RNA Isolation and synthesize cDNA using random

hexamers. Proceed to qPCR to analyze retrieval efficiency of
known lncRNA that bind to your protein of interest. If none
are known, then proceed to deep sequencing of RNA on an
Illumina platform according to manufacturer’s protocol (see
Note 10).
4. Notes
1. Sepharose or Agarose beads can easily be used in place of mag-

netic beads. Since these beads tend to have higher background
than magnetic beads, the preclearing step 17 should not be
skipped. Additionally, protein A or protein G beads may be
more optimal depending on your particular antibody.
2. Depending on the affinity of your antibody to your RNP com-
ponent of interest, the mass of antibody added can be altered.
While 8 μg is a good starting point, a range from 6 to 12 μg
can be used.
3. Be sure to use enough starting material to produce the extract.
Many lncRNAs are expressed at relatively low levels in cells;
thus, it is important to certify that sufficient extract is pro-
duced. A good starting point is about ten million cells, then
increase or decrease from that starting point.
4. Depending on your RNA of interest, more or less formalde-
hyde may be added for optimal co-IP results. While 0.3% has
worked best in most cases, a concentration from 0.1 to 1% may
be used. Incubation time may be altered to vary from 5 to
60 min as well.
5. If higher concentrations of formaldehyde are used, such as 1%,
increase final glycine concentration to 0.25 M to ensure com-
plete quenching.
6. Keep in mind that the more formaldehyde used to cross-link,
the more rigid the cells will be. More force may be necessary to
disrupt cellular membranes. You can increase the incubation
and vortexing time in step 14 to 1 h to remedy this or use soni-
cation. It is critical to point out that sonication will result in
some RNA degradation which may affect the results of your
experiment.
7. This step may not be necessary, depending on the specificity of
your antibodies and the type of beads that are used. It is a good
idea to do a preclear if using sepharose or agarose beads, due
to the high potential of nonspecific interactions with these
beads. If using magnetic beads, this may not be necessary.
8. It is a good idea to incubate the input along with the IPs. This
way, if degradation should somehow occur, it will be equalized
among all the samples.
9. If performing protein pull-down analysis, an aliquot of super-
natant may be saved to determine the IP efficiency.
10. In such a procedure where so many steps can be optimized
specific to each cell type and protein of interest, it is important
to verify the efficiency of the pulldown for a protein by western
blotting analysis.
References
1. International Human Genome Sequencing Bernstein BE, van Oudenaarden A et al (2009)
Consortium (2004) Finishing the euchromatic Many human large intergenic noncoding RNAs
sequence of the human genome. Nature 431: associate with chromatin-modifying complexes
931–945 and affect gene expression. Proc Natl Acad Sci
2. Kapranov P, Cheng J, Dike S, Nix DA, U S A 106:11667–11672
Duttagupta R, Willingham AT, Stadler PF, 11. Guttman M, Amit I, Garber M, French C,
Hertel J, Hackermuller J, Hofacker IL et al Lin MF, Feldser D, Huarte M, Zuk O,
(2007) RNA maps reveal new RNA classes and Carey BW, Cassady JP et al (2009) Chromatin
a possible function for pervasive transcription. signature reveals over a thousand highly con-
Science 316:1484–1488 served large non-coding RNAs in mammals.
3. Wang J, Zhang J, Zheng H, Li J, Liu D, Li H, Nature 458:223–227
Samudrala R, Yu J, Wong GK (2004) Mouse 12. Amaral PP, Dinger ME, Mercer TR, Mattick JS
transcriptome: neutral evolution of ‘non-cod- (2008) The eukaryotic genome as an RNA
ing’ complementary DNAs. Nature 431: 1 p machine. Science 319:1787–1789
following 757; discussion following 757 13. Carninci P, Hayashizaki Y (2007) Noncoding
4. Nagano T, Fraser P (2011) No-nonsense func- RNA transcription beyond annotated genes.
tions for long noncoding RNAs. Cell 145: Curr Opin Genet Dev 17:139–144
178–181 14. Ponting CP, Oliver PL, Reik W (2009)
5. Guttman M, Donaghey J, Carey BW, Garber Evolution and functions of long noncoding
M, Grenier JK, Munson G, Young G, Lucas RNAs. Cell 136:629–641
AB, Ach R, Bruhn L et al (2011) lincRNAs act 15. Lipovich L, Johnson R, Lin CY (2010)
in the circuitry controlling pluripotency and MacroRNA underdogs in a microRNA world:
differentiation. Nature 477:295–300 evolutionary, regulatory, and biomedical
6. Clark MB, Mattick JS (2011) Long noncoding significance of mammalian long non-protein-
RNAs in cell biology. Semin Cell Dev Biol coding RNA. Biochim Biophy 1799:597–615
22:366–376 16. Pontier DB, Gribnau J (2011) Xist regulation
7. Cabili MN, Trapnell C, Goff L, Koziol M, and function explored. Hum Genet 130:
Tazon-Vega B, Regev A, Rinn JL (2011) 223–236
Integrative annotation of human large intergenic 17. Augui S, Nora EP, Heard E (2011) Regulation
noncoding RNAs reveals global properties and of X-chromosome inactivation by the
specific subclasses. Genes Dev 25:1915–1927 X-inactivation centre. Nat Rev Genet 12:
8. Orom UA, Derrien T, Beringer M, Gumireddy 429–442
K, Gardini A, Bussotti G, Lai F, Zytnicki M, 18. Wapinski O, Chang HY (2011) Long noncod-
Notredame C, Huang Q et al (2010) Long ing RNAs and human disease. Trends Cell Biol
noncoding RNAs with enhancer-like function 21:354–361
in human cells. Cell 143:46–58 19. Khalil AM, Rinn JL (2011) RNA-protein inter-
9. Khalil A, Huarte M, Rinn J (2010) The emerg- actions in human health and disease. Semin
ing non-coding RNA World. In: Slack F (ed) Cell Dev Biol 22(4):359–365
MicroRNAs in development and cancer, vol 1. 20. Kouzarides T (2007) Chromatin modifications
Imperial College Press, London, pp 17–44 and their function. Cell 128:693–705
10. Khalil AM, Guttman M, Huarte M, Garber M, 21. Khalil AM, Boyar FZ, Driscoll DJ (2004)
Raj A, Rivea Morales D, Thomas K, Presser A, Dynamic histone modifications mark sex
chromosome inactivation and reactivation Ordered assembly of roX RNAs into MSL
during mammalian spermatogenesis. Proc Natl complexes on the dosage-compensated X chro-
Acad Sci U S A 101:16583–16587 mosome in Drosophila. Curr Biol 10:136–143
22. Zhao J, Sun BK, Erwin JA, Song JJ, Lee JT 31. Kelley RL, Meller VH, Gordadze PR, Roman G,
(2008) Polycomb proteins targeted by a short Davis RL, Kuroda MI (1999) Epigenetic
repeat RNA to the mouse X chromosome. spreading of the Drosophila dosage compensa-
Science 322:750–756 tion complex from roX RNA genes into flanking
23. Rinn JL, Kertesz M, Wang JK, Squazzo SL, Xu chromatin. Cell 98:513–522
X, Brugmann SA, Goodnough LH, Helms JA, 32. Park Y, Kelley RL, Oh H, Kuroda MI, Meller
Farnham PJ, Segal E et al (2007) Functional VH (2002) Extent of chromatin spreading
demarcation of active and silent chromatin determined by roX RNA recruitment of MSL
domains in human HOX loci by noncoding proteins. Science 298:1620–1623
RNAs. Cell 129:1311–1323 33. Lanz RB, McKenna NJ, Onate SA, Albrecht U,
24. Tsai MC, Manor O, Wan Y, Mosammaparast N, Wong J, Tsai SY, Tsai MJ, O’Malley BW (1999)
Wang JK, Lan F, Shi Y, Segal E, Chang HY A steroid receptor coactivator, SRA, functions
(2010) Long noncoding RNA as modular scaf- as an RNA and is present in an SRC-1 complex.
fold of histone modification complexes. Science Cell 97:17–27
329:689–693 34. Watanabe M, Yanagisawa J, Kitagawa H,
25. Gupta RA, Shah N, Wang KC, Kim J, Horlings Takeyama K, Ogawa S, Arao Y, Suzawa M,
HM, Wong DJ, Tsai MC, Hung T, Argani P, Kobayashi Y, Yano T, Yoshikawa H et al (2001)
Rinn JL et al (2010) Long non-coding RNA A subfamily of RNA-binding DEAD-box pro-
HOTAIR reprograms chromatin state to proteins acts as an estrogen receptor alpha coacti-
mote cancer metastasis. Nature 464:1071–1076 vator through the N-terminal activation domain
26. Sleutels F, Zwart R, Barlow DP (2002) The (AF-1) with an RNA coactivator, SRA. EMBO
non-coding Air RNA is required for silencing J 20:1341–1352
autosomal imprinted genes. Nature 35. Kino T, Hurt DE, Ichijo T, Nader N,
415:810–813 Chrousos GP (2010) Noncoding RNA gas5 is
27. Nagano T, Mitchell JA, Sanz LA, Pauler FM, a growth arrest- and starvation-associated
Ferguson-Smith AC, Feil R, Fraser P (2008) repressor of the glucocorticoid receptor. Sci
The air noncoding RNA epigenetically silences Signal 3:ra8
transcription by targeting G9a to chromatin. 36. Niranjanakumari S, Lasda E, Brazas R, Garcia-
Science 322:1717–1720 Blanco MA (2002) Reversible cross-linking
28. Pandey RR, Mondal T, Mohammad F, Enroth S, combined with immunoprecipitation to study
Redrup L, Komorowski J, Nagano T, Mancini- RNA-protein interactions in vivo. Methods
Dinardo D, Kanduri C (2008) Kcnq1ot1 anti- 26:182–190
sense noncoding RNA mediates lineage-specific 37. Mili S, Steitz JA (2004) Evidence for reasso-
transcriptional silencing through chromatin- ciation of RNA-binding proteins after cell
level regulation. Mol Cell 32:232–246 lysis: implications for the interpretation of
29. Mohammad F, Mondal T, Guseva N, Pandey immunoprecipitation analyses. RNA 10:
GK, Kanduri C (2010) Kcnq1ot1 noncoding 1692–1694
RNA mediates transcriptional gene silencing by 38. Zhao J, Ohsumi TK, Kung JT, Ogawa Y, Grau
interacting with Dnmt1. Development 137: DJ, Sarma K, Song JJ, Kingston RE, Borowsky
2493–2499 M, Lee JT (2010) Genome-wide identification
30. Meller VH, Gordadze PR, Park Y, Chu X, of polycomb-associated RNAs by RIP-seq. Mol
Stuckenholz C, Kelley RL, Kuroda MI (2000) Cell 40:939–953
Part VI
Imprinting in Plants
Chapter 16
Specialized Technologies for Epigenetics in Plants

Wenyan Xiao
Abstract
Plants are excellent systems for discovering and studying epigenetic phenomena, such as transposon
silencing, RNAi, imprinting, and DNA methylation. Imprinting, referring to preferential expression of
maternal or paternal alleles, plays an important role in reproduction development of both mammals and
plants. DNA methylation is critical for determining whether the maternal or paternal alleles of an imprinted
gene is expressed or silenced. In flowering plants, there is a double fertilization event in reproduction: one
sperm fertilizes the egg cell to form embryo and a second sperm fuses with the central cell to give rise to
endosperm. Endosperm is the tissue where imprinting occurs in plants. MEDEA (MEA), a SET domain
Polycomb group gene, was the first plant gene shown to be imprinted in endosperm, and its maternal
expression is controlled by DNA methylation and demethylation. Recently there has been significant prog-
ress in identifying imprinted genes as well as understanding molecular mechanisms of imprinting in plants.
Up to date, approximately 350 genes were found to have differential parent-of-origin expression in plant
endosperm (Arabidopsis, corn, and rice). In Arabidopsis, many imprinted genes are regulated by the DNA
METHYLTRANSFERASE1 (MET1) and the DNA-demethylating glycosylase DEMETER (DME), and/
or their chromatin states regulated by Polycomb group proteins (PRC2). There are also maternally
expressed genes regulated by unknown mechanisms in endosperm. In this protocol, we describe in detail
how to perform a genetic cross, isolate the endosperm tissue from seed, determine the imprinting status
of a gene, and analyze DNA methylation of imprinted genes by bisulfite sequencing in Arabidopsis.
Key words: Imprinting, Epigenetics, DNA methylation, Bisulfite sequencing, Parent-of-origin,

Endosperm, Arabidopsis, Plant
1. Introduction
Plants have been used as model organisms for studying genetics

and epigenetics since Gregor Mendel did his series of seminal
genetic experiments using the garden pea (Pisum sativum) in
1856–1863 and Barbara McClintock discovered transposable ele-
ments in maize (Zea mays) half a century ago (1, 2). Plant research
also provided the earliest evidence for RNAi. For example, silenc-
ing of a homologous endogenous gene by the transgene in petunia
231
232 W. Xiao
is now viewed as the earliest evidence of gene silencing by RNAi

(3). Studies using potato spindle tuber viroid were the earliest evi-
dence that RNA is an intermediate in transcriptional gene silencing
(4). Arabidopsis thaliana is another excellent model organism
for studying epigenetic mechanisms, such as imprinting and DNA
methylation. The loss-of-function null mutant of DNA methyl-
transferase is viable in plants, thus providing a system to study how
loss of DNA methylation in a genome affects imprinting, growth
and development.
Imprinting is an epigenetic phenomenon that is involved in
growth and development of plants and mammals. Although genomic
imprinting occurs in flowering plants and mammals, it evolved
independently (5). One of the widely accepted theories of genomic
imprinting is the parental conflict theory (5–7), which hypothe-
sizes that female and male parents have different interests in fitness
of their progeny when one female can mate with multiple males.
The female wants all her offspring to survive and thus prefers
expression of genes that allocate resources equally to all her off-
spring, whereas a male is only interested in his progeny, therefore
favoring expression of genes that maximize allocation of resources
to his offspring. Thus, different interests of female and male par-
ents result in parent-of-origin effects on expression of parental
alleles of imprinted genes.
In mammals, fertilization is resulted from fusion of two haploid
cells that are direct products of the preceding meiosis, whereas in
plants there is a haploid gametophytic growth stage, i.e., postmei-
otic cell divisions before fertilization. In Arabidopsis, a haploid
megaspore at the end of meiosis undergoes three mitotic divisions
to form an eight-nucleus, seven-cell female gametophyte containing
the egg, central, synergid, and antipodal cells; the fusion of two
haploid nuclei makes the nucleus of the diploid central cell. In the
male gametophyte, a haploid microspore at the end of meiosis
undergoes a mitotic cell division to give rise to one vegetative and
one generative nucleus; then the generative cell undergoes a mitotic
cell division to result in two generative nuclei (two haploid sperm
cells). There is also a double fertilization event that underlies gene
imprinting in flowering plants. Fertilization of the egg cell by a
sperm cell gives rise to a diploid embryo that ultimately generates
the organs, tissues, and meristems of the plant. Fertilization of the
central cell by a second sperm generates the triploid endosperm that
supports embryo or seedling growth by producing storage proteins,
lipids, and starch, and by mediating the transfer of maternal-derived
nutrients to be absorbed by the embryo. Imprinting mainly occurs
in the endosperm tissue in plants, whereas imprinting can occur in
embryos and in different adult tissues in mammals (8, 9).
Through classical experimental approaches, 11 imprinted
genes (eight maternally expressed and three paternally expressed)
were identified in Arabidopsis endosperm (10), two maternally
16 Imprinting in Plants 233
expressed imprinted genes (fie1, mee1) in maize (11–13) and one

maternal expressed imprinted gene (OsFIE1) in rice (14). By using
extensive sequencing of cDNA libraries as well as examination of
allele-specific expression of candidate genes by RT-PCR (10),
Hsieh and colleagues discovered an additional 112 maternally
expressed and 9 paternally expressed genes in Arabidopsis endosperm.
These imprinted genes have various functions in plants and include
transcription factors, components of hormone signaling, proteins
involved in the ubiquitin-mediated protein degradation, and genes
regulating DNA methylation, histone modifications, and small
RNA pathways. By a similar approach of deep sequencing RNA
profiles of F1 seeds of Arabidopsis Col-0 and Bur-0 ecotypes, Wolff
and colleagues found that more than 60 genes (six maternally
expressed genes overlapping those discovered by Hsieh and col-
leagues) that have potential parent-of-origin specific expression
(15). In rice seeds, Luo and colleagues uncovered 262 putative
imprinted loci in endosperm and 3 in the embryo (168 genic and
97 nongenic) via deep sequencing of cDNA libraries (16).
Mechanistically, genomic imprinting is complex and is involved in
many aspects of growth and development in plants. MEA and
PHE1, representing maternal and paternal expressed genes, respec-
tively, are the best characterized at the molecular level. For exam-
ple, MET1 is responsible for maintaining CG methylation at the
MEA locus (17), whereas DME excises 5-methyl cytosine of the
maternal MEA allele, thus activating its expression (18).
Interestingly, silencing of the paternal MEA allele is not controlled
by DNA methylation. Rather, the maternally expressed Polycomb
group proteins, including MEA, maintain the paternal MEA allele
silencing (18). PHE1 is maternally silenced and paternally expressed.
Silencing of the maternal PHE1 allele requires the FERTILIZATION
INDEPENDENT SEED (FIS) Polycomb group complex as well
as direct tandem repeats located downstream of the PHE1 locus.
It has been hypothesized that the differential methylation of the
tandem repeats regulates the PHE1 imprinting (19, 20).
No imprinted gene has been detected in Arabidopsis embryo
(10). However, mee1 and Os10g05750 are preferentially maternally
expressed in both embryo and endosperm in maize and rice,
respectively (11, 16). By deep sequencing of RNA transcripts and
genetic analyses of early embryogenesis in Arabidopsis, it has been
shown that there is genome-wide dominance of maternal tran-
scripts at the 2–4 cell stage of embryo, and the relative paternal
transcripts increase at the globular stage due to a gradual activation
of the paternal genome (21). Furthermore, the authors identify
two antagonistic maternal epigenetic pathways that regulate paren-
tal contributions in plant embryos. It seems that imprinting has
evolved independently in Arabidopsis and rice since imprinted loci
identified through genomic approaches do not have extensive
sequence conservation (10, 16).
234 W. Xiao
The principle of bisulfite sequencing is that unmethylated

cytosine will be converted to uracil due to hydrolytic deamination
by high concentration of sodium bisulfite at pH 5.0 which will be
amplified as thymine in PCR product, while 5-methyl cytosine will
not be modified by sodium bisulfite and remain to be cytosine after
PCR amplification (22, 23).
In this protocol, the procedures for determining if a candidate
gene is imprinted in Arabidopsis endosperm and determining the DNA
methylation status of the imprinted genes in endosperm are described.
The major procedures involved in the protocol are: to perform
genetic crossing (emasculating the female parent and pollinating the
pistil with the male parent); to isolate endosperm tissues; to carry
out RNA isolation and RT-PCR; and to perform bisulfite sequenc-
ing of the genomic DNA (bisulfite treatment, designing strand-
specific primers, PCR amplification, and sequence analysis).
2. Materials
Prepare all solutions using sterile deionized distilled water and ana-
lytical grade reagents unless indicated. Prepare all reagents at room
temperature unless indicated. All DNA and RNA manipulations
follow standard procedures unless indicated otherwise.
2.1. Reagents 1. 70 % Ethanol: Add 30 mL water to 70 mL anhydrous ethanol.

2. 95 % Ethanol: Add 5 mL water to 70 mL anhydrous ethanol.
3. 100 % Ethanol: 100 mL anhydrous ethanol.
4. 0.3 M Sorbitol and 5 mM MES—pH 5.7: Dissolve 5.47 g
Sorbitol (Aldrich #: 240850) and 98 mg MES (Sigma #:
M5287, anhydrous) in 90 mL water. Adjust pH to 5.7 and
bring the final volume to 100 mL.
5. Cetyltrimethyl Ammonium Bromide (CTAB) buffer: 2 %
CTAB (Sigma), 100 mM Tris–HCl (pH 8.0), 20 mM EDTA
(pH 8.0), 1.4 M NaCl, 1 % PVP (polyvinyl-polypyrrolidone,
Mr 40,000). After making the buffer, autoclave and store it.
6. Chloroform: Sigma, >99.8 %.
7. Restriction Enzymes: XhoI, NdeI, and PstI or HindIII for the
MEA promoter.
8. 3 M NaOH: Dissolve 1.2 g NaOH in 10 mL water.
9. 6.3 M NaOH: Dissolve 2.52 g NaOH in 10 mL water.
10. 6.24 M Urea/4 M Sodium Bisulfite: Dissolve 7.5 g of urea in
10 mL of sterile distilled water; slowly add 7.6 g of sodium
metabisulfite (Sigma) over 1–2 h with heating (see Note 15);
adjust the pH to 5.0 with freshly made 10 M NaOH; add ster-
ile distilled water to a final volume to 20 mL (see Note 16).
11. Sterile distilled H2O.

12. 10 mM Hydroquinone: Dissolve 0.011 g 99 % Hydroquinone
in 10 mL water.
13. Wizard DNA Clean-Up System: Promega.
14. 10 M NH4OAc: Dissolve 7.7 g NH4OAc in 10 mL water.
15. 20 mg/ml tRNA: Dissolve 2 mg tRNA in 100 mL water.
16. TE buffer: 10 mM Tris–Cl, pH 7.5. 1 mM EDTA, pH 8.0. To
make 1 L TE buffer: 10 mL of 1 M stock of Tris–Cl (pH 7.5),
2 mL of 500 mM stock of EDTA (pH 8.0), and 988 mL water.
17. The TOPO TA Cloning Kit: Invitrogen’s TOPO® TA cloning
Kit with the pCR® 2.1 vector.
2.2. Supplies 1. Dissecting Microscope.

and Equipment 2. Scissors.
3. Fine Tip Forceps.
4. Jewelry Tag.
5. Plant Stakes.
6. String or Twist-Ties.
7. 4” × 2” × 8” Polyethylene Bags.
8. 3” × 1” × 1.0 mm Microscope Slides.
9. Liquid Nitrogen.
10. Liquid Nitrogen Containers.
11. Heat Block.
12. PCR Tubes.
13. Thermocycler.
14. Microcentrifuge Tubes.
15. Microcentrifuge.
16. Gel Electrophoresis facility.
17. Different Arabidopsis thaliana ecotypes (Columbia-0,
Landsberg erecta, RLD, Ws, or Cvi).
18. Total RNA isolation kits.
19. NanoDrop spectrometer.
3. Methods
1. Emasculating the female parent. In order to distinguish the

3.1. Genetic Crossing
maternal and paternal alleles of the imprinted MEA gene, two
different ecotypes, e.g., Columbia-0 (Col-0) and RLD, will be
chosen as female and male parents (see Note 1). One can emas-
culate the female parent by using a dissecting microscope,
236 W. Xiao
magnifying visor, or naked eye (see Note 2). Locate stage-12

flowers and remove any flowers or siliques above and below
them by clipping the base of the pedicel with the scissors.
Sterilize forceps by dipping the base of the tip gently into a
beaker of 95 % ethanol, which will remove any pollen grains on
the forceps and kill the pollen as well. Gently pry apart the
flower buds using forceps, and gently remove four sepals, four
petals, and six stamens, leaving the pistil bare and intact (see
Note 3).
2. Picking the pollen donor and carrying out pollination. Choose an
open flower from a different ecotype plant as a male (see Note
4). Grab the flower at the base and just above the pedicel with
forceps, which will cause the flower to spread open. Dust the
stigma of the prepared female pistil with the anther (see Note
5). After the pollination, the stigma will be covered with yellow
pollen, which can be easily observed under a microscope.
3. Labeling the cross. After pollinating all the emasculated pistils
on a plant, we label the cross with a jewelry tag and write down
information of female and male parents and date on the tag.
Place a stake in the soil close to the plant, use a string to tie the
stem of the inflorescence to the stake, and cover the pollinated
pistils with a plastic bag (see Note 6).
3.2. Isolation of
1. Preparing materials. Get necessary materials ready before har-
Endosperm Tissues
vesting the endosperm and embryo tissues: a liquid nitrogen
in Arabidopsis
tank, liquid nitrogen, a dissecting microscope, two new pairs
of fine-tip forceps (5 INOX. FST by Dumont biology,
Switzerland), microscope glass slides (3” × 1” × 1.0 mm), and a
pH 5.7 solution of 0.3 M sorbitol and 5 mM MES.
2. Collecting siliques. At 7 or 8 day-after-pollination (DAP), the
seeds are ready to be harvested at the mid- to late torpedo
stage of embryogenesis and dissected for endosperm and
embryo (see Note 7). Locate the crossed silique and cut the
silique pedicel with small scissors.
3. Isolating endosperm and embryo. Put an 8-DAP silique on a
glass slide under a microscope, use a pair of forceps to hold the
silique pedicel and use the tip of another pair of forceps to slide
open the silique in the margin where two carpels fuse (see ref.
24). Use a pair of forceps to pick up one seed onto pH 5.7
solutions of 0.3 M sorbitol and 5 mM MES (see Note 8), make
a small cut at the micropyle end (see Fig. 1a) to slide out
embryo (see Fig. 1b), and then carefully separate endosperm
(see Fig. 1c) from seed coat (see Fig. 1d). Put the embryo and
endosperm into separate microtubes in liquid nitrogen.
Continue this until accumulating enough embryo or endosperm
from 15 to 20 siliques and then store the tube in a −80 °C
freezer (see Note 9).
Fig. 1. An Arabidopsis seed and dissected embryo, endosperm, and seed coat. (a) A seed at the late heart stage is ready to
dissect out embryo, endosperm, and seed coat. A dissected embryo (b), endosperm (c), and seed coat (d) from an
Arabidopsis seed at the torpedo stage.
Fig. 2. Determining the imprinting status of MEA. (a) Shows a sequence polymorphism in the exon 17 of MEA between
Col-0 and RLD that can be used to convert to a dCAPS marker. After PCR amplification of the fragment, digestion of the
PCR fragment with BamHI will give a 239-bp band in Col-0 and two bands (207 and 32 bp) in RLD. (b) The maternal MEA
allele is specifically expressed in the reciprocal crosses between Col-0 and RLD (C × R and R × C) and the paternal MEA
allele is silenced in Arabidopsis endosperm. Biallelic MEA expression is detected in Arabidopsis embryo.
3.3. RNA Isolation 1. Isolate total RNA from the endosperm tissue collected above
and RT-PCR (see Note 10).
2. Adjust RNA concentration to 100 ng/mL and use total 1 mg
total RNA (10 mL of 100 ng/mL) for reverse transcription
(RT) reaction (see Note 11).
3. Identify polymorphisms between two ecotypes within the cod-
ing region of your candidate gene. Design a dCAPS marker.
For MEA, there is a polymorphism between Col-0 (or Ler)
and RLD at the end of the coding region (exon 17) that can
be used to design a dCAPS marker to distinguish Col-0 and
RLD alleles (see Fig. 2a and Note 12).
4. Use 1 mL of RT reaction in a total volume of 20 mL PCR
reaction.
5. Perform restriction enzyme digestion after the PCR reaction.
In the PCR reaction for amplifying MEA fragments, perform
BamHI reaction for 6–12 h.
6. Run a 3 % agarose gel for 1 h to distinguish the Col-0 and
RLD allele (see Fig. 2b and Note 13). The maternal MEA
238 W. Xiao
allele is expressed and the paternal allele is silenced in Arabidopsis

endosperm, whereas biallelic MEA expression is detected in
embryo (see Fig. 2b).
3.4. Bisulfite 1. Preparing required reagents. Cetyltrimethyl ammonium bro-

Sequencing mide (CTAB) for isolating genomic DNA, restriction enzymes,
3 M NaOH (Freshly made), 6.42 M urea/4 M sodium bisulfite
(2 M sodium metabisulfite, Sigma-Aldrich, S9000, Na2S2O5,
Molecular Weight: 190), 10 mM hydroquinone, a DNA
purification kit (Promega Wizard DNA Clean-up System, Cat.
# A7280), TE buffer, 6.3 M NaOH (freshly made), 10 M
NH4OAc, 20 mg/mL tRNA, and 100 % ethanol.
2. Treating genomic DNA with the sodium bisulfite.
(a) Perform genetic cross (see Subheading 3.1) and collect
endosperm tissue (see Subheading 3.2) as described.
(b) Isolate genomic endosperm DNA using a CTAB proce-
dure (see Note 14).
(c) Digest 100 ng to 2 mg of endosperm genomic DNA in
20–100 mL total volume with restriction enzymes that cut
outside the region to be analyzed. For the MEA promoter,
we use XhoI, NdeI, and PstI or HindIII.
(d) Denature the restriction enzymes by boiling the DNA for
5 min after the restriction enzyme digestion and then
quench on ice.
(e) Add 1/9 volume (2.2 mL for 20 mL digested DNA) of
3 M NaOH and incubate at 37 °C for 15 min.
(f) Transfer the solution to a 250 mL PCR tube.
(g) Dissolve 7.5 g of urea in 10 mL of sterile distilled water;
slowly add 7.6 g of sodium metabisulfite (Sigma Cat. #
S900-1KG) over 1–2 h and heating usually helps dissolv-
ing (see Note 15); adjust the pH to 5.0 with freshly made
10 M NaOH; add sterile distilled water to a final volume
to 20 mL. This is 6.42 M urea/4 M sodium bisulfite solu-
tion (see Note 16).
(h) Add 6.42 M urea/4 M sodium bisulfite solution to a final
concentration of 5.36 M and 3.44 M, respectively (see ref.
25). For example, add 208 mL of 6.42 M urea/4 M sodium
bisulfite solution to the above 20 mL of starting genomic
DNA reaction.
(i) Add 10 mM hydroquinone to the DNA at a final concen-
tration of 0.5 mM (12 mL for 20 mL digestion).
(j) Conduct bisulfite treatment in a PCR machine: 30 cycles
of 55 °C for 15 min and 95 °C for 30 s (see Note 17).
(k) Desalt the bisulfite treated DNA using the Wizard DNA
Clean-Up System from Promega and follow up the proto-
col (see Note 18).
Table 1
Amounts of solutions being added depends on the volume
of TE recovered
TE recovered (mL) 44 46 48 50 52 54
6.3 M NaOH (mL) 2.2 2.3 2.4 2.5 2.6 2.7
10 M NH4Ac (mL) 21 21.95 22.9 23.86 24.82 25.77
100 % EtOH (mL) 206 215.9 224.7 234 243 252.4
(l) Measure the exact volume of TE recovered from the

column after desalting. Add 6.3 M NaOH to a final con-
centration of 0.3 M. Incubate at 37 °C for 15 min (see
Table 1 and Note 19).
(m) Add 10 M NH4OAc (pH 7.0) to a final concentration of
3 M, 2 mL of 20 mg/mL tRNA, and 3 volumes of 100 %
ethanol and then mix (see Table 1 and Note 19). Centrifuge
for 15 min at 16,800 ´ g.
(n) Wash the pellet once with 70 % ethanol, do a short centri-
fuge, and remove extra ethanol.
(o) Dry pellet in a speedvac for 5–10 min and resuspend in
25–100 mL TE buffer depending on starting amount of
DNA. The sodium bisulfite-treated DNA is now ready for
PCR analysis.
3.5. PCR Amplification 1. To sequence the 4-kb MEA promoter, we designed many sets
of primers and amplified 14 overlapping fragments to cover the
entire region (see Note 20).
2. To sequence the top-strand (see Fig. 3a), in designing a for-
ward primer, (1) choose a G (guanine)-rich region in order to
have a higher annealing temperature without extra long nucle-
otides in the primers; (2) change C (cytosine ) to Y (pyrimi-
dine) at CG and CNG contexts and change the remaining C to
T (thymine). In designing a reverse primer, (1) choose a C-rich
region (5¢–3¢ strand direction); (2) change G to R (purine) at
CG and CNG contexts and change the remaining G to A
(adenine).
3. To sequence the bottom-strand (see Fig. 3b), in designing a
forward primer, (1) choose a C-rich region (5¢–3¢ strand direc-
tion); (2) change G to R at CG and CNG contexts and change
the remaining G to A. In designing a reverse primer, (1) choose
a G-rich region (5¢–3¢ strand direction); (2) change C to Y
(pyrimidine) at CG and CNG contexts and change the remain-
ing C to T.
240 W. Xiao
Fig. 3. General rule in designing primers for bisulfite sequencing. (a) Strategy for designing forward and reverse primers to
sequence the top-strand in bisulfite sequencing. In designing a forward primer, choose a G (guanine)-rich region in order
to have a higher annealing temperature without extra long nucleotides in the primers; change C (cytosine) to Y (pyrimidine)
at CG and CNG contexts and change the remaining C to T (thymine). In designing a reverse primer, choose a C-rich region;
change G to R (purine) at CG and CNG contexts and change the remaining G to A (adenine). (b) Strategy for designing for-
ward and reverse primers to sequence the bottom-strand in bisulfite sequencing. In designing a forward primer, choose a
C-rich region; change G to R at CG and CNG contexts and change the remaining G to A. In designing a reverse primer,
choose a G-rich region; change C to Y (pyrimidine) at CG and CNG contexts and change the remaining C to T.
4. We usually use 1–2 mL of the sodium bisulfite-treated DNA as

a template for each PCR amplification (see Note 21). The PCR
product needs to be analyzed by gel electrophoresis to confirm
the correct size of the fragment, then to be gel purified and
cloned into the TOPO TA cloning vector pCR2.1 (Invitrogen)
as an insert. A single colony is picked up and cultured; plasmid
DNA extracted and sent for sequencing.
5. Sequence analysis. After obtaining the sequencing result, we
compare it with the strand-specific template that is used for PCR
amplification (see Fig. 4a). If a cytosine residue in the template
reads as a thymine in the sequencing result, it indicates that the
cytosine is not methylated (see Fig. 4b). If a cytosine residue in
the template remains a cytosine in the sequencing, it means that
the cytosine is methylated (see Fig. 4b). After sequencing many
colonies, one can compile the bisulfite sequencing result together
as shown in Fig. 5 (see Note 22).
4. Notes
1. For choosing a female parent for genetic crossing, it is important

to choose a healthy plant to emasculate. Young plants with large
un-open flower buds are easier to emasculate for beginners.
2. In emasculation, different people can use different approaches.
For beginners, the emasculation procedure should be done
Fig. 4. Strategy for bisulfite sequencing. (a) Shows an outline of the major bisulfite sequencing procedures. (b) Shows the
genomic sequence of the MEA promoter between −587 and −519 bp. There are five methylated cytosine at five CG sites
in the promoter region. Fig. 4b shows how one can compare the bisulfite sequencing result with the original genomic
sequence and deduce whether a cytosine residue is methylated or not.
under a dissecting microscope to make sure that all six stamens

are removed from the flower. A very experienced person can
do emasculation with naked eyes. One common problem for
beginners is the damage made to the plant during emascula-
tion. Thus, it is critical to gently grasp or hold the plant stems
without damaging them.
3. Try to avoid damage to the carpels during this process. Another
option is that you wait 24–48 h to allow ovules from emascu-
lated plants to reach maturity and synchronize to the same
growth stage before performing pollination.
242 W. Xiao
Fig. 5. The methylation status of the maternal MEA allele in −500 bp region of the MEA promoter. (a) and (b) show the
methylation status of five methylated CpG sites in the sequenced clones of (Col gl X RLD) endosperm and embryo, respec-
tively. Black filled circles and white unfilled circles indicate methylated and unmethylated cytosines, respectively. Number
of sequences is relative to the translation start site of MEA (Xiao et al., 2003; Gehring et al., 2006).
4. The best pollen donors are anthesed open flowers at the stage
14 with the petals extending at a 90° angle to the pistil (26), in
which a lot of pollen is shedding. For examining MEA imprinting
status, we have used reciprocal crosses between Col-0 and
RLD, both crosses work.
5. Pollination should be done under a microscope because it is
almost impossible to see whether enough pollen has been put
onto the stigma with naked eyes.
6. Covering the pollinated pistils with a plastic bag has two pur-
poses: one is to avoid cross-pollination contamination and the
other is to avoid water evaporation from the emasculated pistil.
The emasculated pistil can easily wilted if not covered with a
plastic bag.
7. Sometimes, it might take nine DAP if the emasculated flowers
are too young. Actual time depends on the plant growth con-
dition. If it is too early, the endosperm is not cellularized at all
but fluid. If it is too late, the endosperm is completely cellular-
ized, the endosperm tissue might not be optimal for imprinting
analysis.
8. When embryo and endosperm are located in the buffer solu-
tion, sometimes it is not easy to pick them up and put into the
eppendorf tube in the liquid nitrogen tank. One can accumu-
late endosperm tissues from several seeds before putting them
into the eppendorf tube in the liquid nitrogen tank. Another
optional is not using the buffer solution and isolating the
embryo and endosperm directly on the top of a slide. After dis-
secting embryo and endosperm from one seed, immediately put
into the eppendorf tube in the liquid nitrogen tank. This isola-
tion method needs to reduce exposure of dissected endosperm
or embryo in the air to the minimum before freezing them in
liquid nitrogen in order to avoid degradation of RNA.
9. It is relatively easy to separate embryo from endosperm and
seed coat, but it is tedious to separate endosperm from seed
coat, especially for seed at the heart stage of embryogenesis.
Since seed coat only contributes a very small amount of tissue,
for some genes, e.g., MEA and FWA, we do not have to sepa-
rate endosperm from seed coat. That means that we can isolate
RNAs from a mixture of endosperm and seed coat tissues,
check expression of the maternal and paternal alleles of a gene,
compare with expression in embryo, and determine the
imprinting status of the gene.
10. Total RNA can be isolated using your own method. We have
used kits from Ambion and Qiagen and they both work fine.
11. Total RNA amount used for RT reaction is variable. We have
used any amount between 500 ng and 2 mg depending expres-
sion levels of your candidate gene and total amount RNA
244 W. Xiao
available. We have used RT kits from Ambion and New England

Biolab and both are good.
12. For MEA, we use the following two primers to amplify the
PCR fragments to distinguish Col-0 and RLD alleles: MEA-
L1: 5¢-GACCTAACTGCTACGCCAAG-3¢; MEA-R2d: 5¢-AA
GGACTGCTTGAATTGCTGCTTCTCCTCGGATC-3¢ (27).
13. A 3 % agarose gel is very easy to solidify after dissolving. Get
gel box and comb ready and immediately pour the gel after
dissolving. If you only need to determine imprinting status of
your candidate gene, you are done after you run a gel as shown
in Fig. 2b.
14. Isolate genomic DNA using CTAB procedure is fine (28).
Make sure to thoroughly clean the bench to avoid contamina-
tion before starting DNA isolation.
15. It is not easy to dissolve sodium bisulfite. Start with a little bit
sodium bisulfite until it is dissolved and gradually add the rest
of it over a 1–2 h period.
16. Another way to make sodium bisulfite solution is: dissolve
40.5 g of sodium bisulfite (Fisher Cat. #S654-500) in 80 mL
distilled water with slow stirring to avoid aeration. Adjust pH
to 5.1 with freshly made 10 M NaOH. Add 3.3 mL of 20 mM
Hydroquinone. Adjust the final volume to 100 mL (29).
17. Different labs have used different protocol to conduct bisulfite
treatment. We have tried different bisulfite treatments in a
PCR machine in order to make the PCR amplification work-
ing. Another bisulfite treatment condition is: 55 °C for 16 h in
the dark with a jolt to 95 °C for 5 min every 3 h (29). Recently,
several plant biology labs have leaded the field of bisulfite
sequencing and sequenced the whole methylome at single-base
resolution in Arabidopsis and human by combining the bisulfite
treatment with deep sequencing technology (30–35). Several
commercial kits have been used successfully, which provide
easier alternatives to the classical bisulfite treatment (Dr.
Hsieh’s personal communication). Another important devel-
opment is that a higher concentration of ammonium bisulfite
solution (10 M) has been shown to be more effective in con-
verting cytosine to uracil than traditional sodium bisulfite solu-
tion (4–5 M) (36). These kits are: The Imprint® DNA
Modification Kit (Sigma-Aldrich), the EpiTect Bisulfite Kits
(Qiagen), the MethylEasy Xceed kit (Human Genetic Signatures,
NSW, Australia) (34), the EZ DNA Methylation-Gold™ Kit
(Zymo Research), and the MethylCode kit (Invitrogen, Life
Technologies).
18. For desalting the bisulfite treated DNA, just follow Promega pro-
tocol of the Wizard DNA Clean-Up System (Cat. #: 288742).
19. The exact amount of 6.3 M NaOH, 10 M NH4Ac and 100 %

ethanol solutions being added depends on the volume of TE
recovered (see Table 1).
20. Since unmethylated cytosines are converted to uracil by bisulfite-
treatment, it is difficult to amplify a large fragment using the
bisulfite-treated DNA as a template. Thus, we usually design
primers to amplify a product no longer than 500 bp (17).
21. Sodium bisulfite-treatment causes damage to DNA. It is chal-
lenging to amplify your expected fragment from a bisulfite-
treated genomic DNA for beginners and this can be the most
frustrating part in the bisulfite sequencing. After restriction
enzyme digestion, the genomic DNA needs to be completely
denatured into single-stranded since the bisulfite deamination
almost does not work on the cytosine residues in double-
stranded DNA structures. Another key is that bisulfite treat-
ment needs to be complete. If a known unmethylated cytosine
is not converted to thymine in the bisulfite sequencing, it indi-
cates that the bisulfite treatment is not conducted properly.
Alternatively, if many cytosine residues in an unknown region
are not changed to thymine in the bisulfite sequencing, it is
most likely due to amplification of unconverted DNA rather
than indication of all methylated cytosine residues in the
region. For bisulfite sequencing, the designed primers need to
be strand-specific. If it is impossible to amplify an expected
fragment from the bisulfite-treated genomic DNA due to dam-
aged DNA template during sodium bisulfite treatment, we
usually try different sets of PCR primers in different regions
and shorter fragments. When you are sequencing an unknown
region and you do not know whether a particular cytosine is
methylated or not, you can use a degenerate nucleotide (Y for
C and T; R for G and A) in a primer in order to increase the
chance of pairing between the primer and template. For exam-
ple, as shown in Fig. 3a, in designing a forward primer to
sequence the top-strand, you can change both C to Y in the
forward primer. Finally, if your bisulfite sequencing is still not
working after you have tried all of the above, you can use com-
mercially available bisulfite conversion kits (see Note 17).
These kits are easier to use, less time-consuming, and have a
higher successful rate.
22. Sodium bisulfite sequencing can precisely reveal whether a par-
ticular cytosine residue is methylated or not in a genome if the
experiment is conducted correctly (37). Combining the sodium
bisulfite sequencing with recent high-throughput sequencing
techniques, such as the Illumina Genome Analyzer high-
throughput sequencing platform, one can map the epigenome
at single-base resolution in plants and mammals (31–35).
246 W. Xiao
Acknowledgments
The author thanks colleagues in the lab for discussion and Dr.
Tzung-Fu Hsieh for critical reading of the manuscript. This work
is supported by startup fund from Saint Louis University and
National Institutes of Health grants 1R15GM086846-01 and
3R15GM086846-01S1.
References
1. McClintock B (1951) Chromosome organiza- imprinting suggest distinct functions. Plant
tion and genic expression. Cold Spring Harbor Cell 15:425–438
Symp Quant Biol 16:13–47 14. He G et al (2010) Global epigenetic and tran-
2. McClintock B (1965) The control of gene scriptional trends among two rice subspecies
action in maize. Brookhaven Symp Biol 18: and their reciprocal hybrids. Plant Cell 22:
162–184 17–33
3. Napoli C, Lemieux C, Jorgensen R (1990) 15. Wolff P et al (2011) High-resolution analysis of
Introduction of a chimeric chalcone synthase parent-of-origin allelic expression in the arabi-
gene into petunia results in reversible co-sup- dopsis endosperm. PLoS Genet 7:e1002126
pression of homologous genes in trans. Plant 16. Luo M, Taylor JM, Spriggs A, Zhang H, Wu X,
Cell 2:279–289 Russell S, Singh M, Koltunow A (2011) A
4. Wassenegger M, Heimes S, Sanger HL (1994) genome-wide survey of imprinted genes in rice
An infectious viroid RNA replicon evolved seeds reveals imprinting primarily occurs in the
from an in vitro-generated non-infectious endosperm. PLoS Genet 7:e1002125
viroid deletion mutant via a complementary 17. Xiao W et al (2003) Imprinting of the MEA
deletion in vivo. EMBO J 13:6172–6177 Polycomb gene is controlled by antagonism
5. Feil R, Berger F (2007) Convergent evolution between MET1 methyltransferase and DME
of genomic imprinting in plants and mammals. glycosylase. Dev Cell 5:891–901
Trends Genet 23:192–199 18. Gehring M et al (2006) DEMETER DNA gly-
6. Haig D, Westoby M (1989) Parent-specific cosylase establishes MEDEA polycomb gene
gene expression and the triploid endosperm. self-imprinting by allele-specific demethylation.
Am Nat 134:147–155 Cell 124:495–506
7. Moore T, Haig D (1991) Genomic imprinting 19. Makarevich G, Villar CB, Erilova A, Kohler C
in mammalian development: a parental tug- (2008) Mechanism of PHERES1 imprinting in
of-war. Trends Genet 7:45–48 Arabidopsis. J Cell Sci 121:906–912
8. Gregg C et al (2010) High-resolution analysis 20. Villar CB, Erilova A, Makarevich G, Trosch R,
of parent-of-origin allelic expression in the Kohler C (2009) Control of PHERES1
mouse brain. Science 329:643–648 imprinting in Arabidopsis by direct tandem
9. Constancia M, Kelsey G, Reik W (2004) repeats. Mol Plant 2:654–660
Resourceful imprinting. Nature 432:53–57 21. Autran D et al (2011) Maternal epigenetic
10. Hsieh TF et al (2011) Regulation of imprinted pathways control parental contributions to
gene expression in Arabidopsis endosperm. Arabidopsis early embryogenesis. Cell 145:
Proc Natl Acad Sci U S A 108:1755–1762 707–719
11. Jahnke S, Scholten S (2009) Epigenetic reset- 22. Frommer M et al (1992) A genomic sequenc-
ting of a gene imprinted in plant embryos. Curr ing protocol that yields a positive display of
Biol 19:1677–1681 5-methylcytosine residues in individual DNA
12. Hermon P, Srilunchang KO, Zou J, Dresselhaus strands. Proc Natl Acad Sci U S A 89:
T, Danilevskaya ON (2007) Activation of the 1827–1831
imprinted Polycomb Group Fie1 gene in maize 23. Clark SJ, Harrison J, Paul CL, Frommer M
endosperm requires demethylation of the (1994) High sensitivity mapping of methylated
maternal allele. Plant Mol Biol 64:387–395 cytosines. Nucleic Acids Res 22:2990–2997
13. Danilevskaya ON et al (2003) Duplicated fie 24. Rea M et al (2011) Determination of
genes in maize: expression pattern and DNA methylation of imprinted genes in
Arabidopsis endosperm. J Vis Exp 47, http:// during seed development underlies gene
www.jove.com/index/Details.stp?ID=2327, imprinting. Science 324:1447–1451
doi: 10.3791/2327 31. Cokus SJ et al (2008) Shotgun bisulphite
25. Paulin R, Grigg GW, Davey MW, Piper AA sequencing of the Arabidopsis genome reveals
(1998) Urea improves efficiency of bisulphite- DNA methylation patterning. Nature 452:
mediated sequencing of 5¢-methylcytosine in 215–219
genomic DNA. Nucleic Acids Res 26: 32. Hsieh TF et al (2009) Genome-wide demethy-
5009–5010 lation of Arabidopsis endosperm. Science 324:
26. Smyth DR, Bowman JL, Meyerowitz EM 1451–1454
(1990) Early flower development in Arabidopsis. 33. Lister R et al (2008) Highly integrated single-
Plant Cell 2:755–767 base resolution maps of the epigenome in
27. Kinoshita T, Yadegari R, Harada JJ, Goldberg RB, Arabidopsis. Cell 133:523–536
Fischer RL (1999) Imprinting of the MEDEA 34. Lister R et al (2009) Human DNA methylomes
polycomb gene in the Arabidopsis endosperm. at base resolution show widespread epigenomic
Plant Cell 11:1945–1952 differences. Nature 462:315–322
28. Rogers SO, Bendich AJ (1988) Extraction of 35. Lister R et al (2011) Hotspots of aberrant epig-
DNA from plant tissues. Plant Mol Biol Manual enomic reprogramming in human induced
A6:1–10 pluripotent stem cells. Nature 471:68–73
29. Jacobsen SE, Sakai H, Finnegan EJ, Cao X, 36. Hayatsu H, Negishi K, Wataya Y (2009)
Meyerowitz EM (2000) Ectopic hypermethy- Progress in the bisulfite modification of nucleic
lation of flower-specific genes in Arabidopsis. acids. Nucleic Acids Symp Ser 53:217
Curr Biol 10:179–186 37. Henderson IR, Chan SR, Cao X, Johnson L,
30. Gehring M, Bubb KL, Henikoff S (2009) Jacobsen SE (2010) Accurate sodium bisulfite
Extensive demethylation of repetitive elements sequencing in plants. Epigenetics 5:47–49
Part VII
Evolution of Imprinted Genes

Chapter 17
Computational Studies of Imprinted Genes

Martina Paulsen
Abstract
Computational studies on imprinted genes can have very different purposes: one major aim of these studies
is the identification of DNA elements that distinguish imprinted genes from biallelically expressed genes.
Comparative studies may help to identify imprinting regulatory elements and to understand common
mechanisms of imprinted gene regulation in mammalian species. To date, the continuously growing
number of genomic and epigenetic data sets makes detailed, genome-wide analyses on imprinted genes
feasible. However, imprinted genes are characterized by genomic features that can influence statistics and
can make such studies difficult. Hence, comparative computational studies can get very complex and
require a tight interaction between bioinformaticians and biologists. Furthermore, analyses of raw data
that are generated by micro-array hybridization and high-throughput sequencing technologies require
computational approaches that have been designed especially for the epigenetic field. This chapter gives an
overview about databases and software that is suitable for analyses of imprinted genes. Furthermore, possible
difficulties that are typical for computational and statistical analyses of imprinted genes are described.
Key words: Imprinting, Bioinformatics, Comparative genomics
1. Introduction
In mammalian genomes most genes are active on both parental

chromosomes. However, quite a number of genes are silenced on one
of the two chromosomes, i.e., they are mono-allelically expressed.
Among these genes are genes on the X chromosomes in females,
olfactory receptor genes, and imprinted genes. In all these cases,
epigenetic modifications such as DNA methylation and histone
modifications appear to be the key players in silencing one of the
gene copies. Imprinted genes are particularly interesting since the
parental origin decides which of the two gene copies remains active
and which is repressed. Silencing of one of the two gene copies is
initiated in one of the parental germ lines, and is maintained after
fertilization.
251
252 M. Paulsen
Although allele-specific DNA methylation and histone

modification patterns are prerequisites for epigenetic regulation of
imprinted genes, the genetic basis for this regulatory mechanism is
not understood, i.e., we do not know what DNA elements in
imprinted regions attract the epigenetic modification machinery to
these genes in one germ line but keep them free of the respective
modifications in the other germ line. Similarly, it is unclear how
such parental origin-specific modifications survive remodeling of
the epigenome after fertilization. Hence, during the last years many
studies aimed at identifying the DNA elements that are responsible
for allele-specific epigenetic modifications (1–6). Furthermore, it
has been questioned if the mono-allelic regulation of these genes is
associated with specific expression patterns and functions in distinct
tissues or developmental stages (7, 8). Such questions can be asked
for individual imprinted genes or can also address the features of
imprinted genes in general. In both cases biocomputational studies
can help to identify specific DNA elements, cell type-specific
expression patterns, or putative functions. In turn, the design of
experimental studies can be optimized, or hypotheses or models on
the functional or regulatory particularities of imprinted genes can
be developed.
We are currently facing a continuously growing number of data
sets comprising genomic sequences, epigenome, and transcriptome
data in a genome-wide scale. In parallel, computational hardware
capacities have been expanded and software tools have been devel-
oped that allow the fast processing and annotation of raw data and
subsequently complex statistical analyses. Hence, computational
analyses of imprinted genes will benefit from this growing amount
of available data. However, analyses of this data will require a com-
bination of bioinformaticians who provide the computational tools
for such analyses, statisticians for statistical support, and biologists
who are experienced in the field of genomic imprinting for the ini-
tial design of such studies and interpretation of the obtained results.
The success of such interdisciplinary projects depends to a high
degree on the interaction of these scientists and on a basic knowl-
edge of the particularities of imprinted genes and computational
analyses and statistics of all participants. Therefore, this review aims
to help strategies for computational analyses on imprinted genes
and tries to point out possible difficulties that are related to the
particular genetic and epigenetic properties of these genes.
2. Materials
2.1. Databases During the past two decades much effort has been invested in the
of Imprinted Genes identification of imprinted genes. Screens for imprinted genes are
based on different methodologies. The first imprinted genes were
17 Computational Studies of Imprinted Genes 253
discovered by investigations of genomic regions in mouse and

human, where mutations show a parental origin-specific inheri-
tance pattern for the phenotype. Further imprinted genes have
been identified in vicinity of already known genes. Furthermore,
computational approaches have been used for the prediction of
candidate genes that are likely to be imprinted. For some of these
candidate genes imprinting was verified by additional experiments
in the lab. Collections of imprinted genes are provided by several
databases that focus on different aspects of genomic imprinting
and are therefore shortly described in the following.
2.1.1. The Catalogue The database at the University of Otago (New Zealand) provides
of Parent of Origin Effects lists of imprinted genes in various mammalian species (http://igc.
otago.ac.nz/home.html) (9). In addition to defined genes,
genomic regions are listed for which diseases or phenotypes with
parental-specific biases in inheritance patterns have been observed.
The database comprises currently more than 450 entries for the
human and mouse genomes. The information on imprinted genes
and phenotypes is derived from screens of the available literature,
i.e., numerous different studies, and is regularly updated. For each
entry the database provides a summary on features such as the
expressing parental allele, imprinting status of the respective gene
in other species, chromosomal location, tissue-specific imprinting
effects, literature references, etc. The Internet surface allows searches
by species, chromosome location, and name of the gene.
2.1.2. MRC Harwell This database focuses on imprinted genes in mouse (http://www.
Database on Imprinted har.mrc.ac.uk/research/genomic_imprinting/index.html) (10).
Genes in the Mouse The information provided is structured similarly to the Otago
database and is complemented by graphic maps on the chromo-
somal locations of imprinted genes.
2.1.3. Geneimprint This database is a service of the Jirtle Laboratory at Duke University
Database (USA) (http://www.geneimprint.com/). In addition to experi-
mentally identified imprinted genes, the database also provides lists
of genes that have been predicted to be imprinted by computa-
tional approaches (4, 5). For each gene information on cDNA and
genomic sequences are given. In addition, the genes are linked to
the SNP, Gene, and PubMed databases at the NCBI and to the
UCSC genome browser.
2.1.4. The ncRNAimprint The database is run by the Key Laboratory of Gene Engineering of
Database Ministry of Education at Sun Yat-Sen University (China) (http://
rnaqueen.sysu.edu.cn/ncRNAimprint/index.php) (11). The data-
base focuses on imprinted noncoding RNAs and encompasses a
comprehensive collection of microRNAs, snoRNAs, antisenseR-
NAs, etc. that are located in imprinted regions.
254 M. Paulsen
2.1.5. Transcriptome Data In the past few years, systematic experimental screens that cover
Sets on Imprinted Gene the entire human or mouse genome have been applied and compu-
Expression tational approaches have helped to identify further imprinted
genes. One frequently used approach is the evaluation of mono-
allelic gene expression on the basis of SNP’s occurrence in cDNA
sequences. Subjects of these studies are usually public EST libraries
derived from human or mouse tissues. The EST populations of
individual genes are investigated for biases towards one sequence
variant at heterozygous SNP positions. Such biases can be caused
either by parental imprinting or by genetic variability. By combin-
ing the information on allelic biases from different EST libraries
genetic effects can often be distinguished from parental imprinting
effects. Another interesting approach is the comparison of gene
expression levels in androgenetic and parthenogenetic embryos.
Approximately 2,000 genes show differential expression patterns
in these embryos. However, of the genes that were identified as
possibly imprinted genes by transcriptome analyses, only very few
have been verified as being indeed imprinted. Hence, the lists of
candidate genes may contain rather high numbers of false positives.
Some studies that provide lists of imprinted candidate genes are
listed below:
● Candidates for imprinting based on SNP distributions in EST
libraries in the human (12).
● Two data sets on transcriptome sequences derived from brain
samples of F1 mice (13, 14).
● Differentially methylated regions in reciprocal human unipa-
rental disomy samples (15).
● Differentially expressed genes in parthenogenetic and andro-
genetic mouse embryos (16, 17).
2.1.6. Computationally A substantial number of imprinted genes have been predicted for
Predicted Imprinted Genes the human and mouse genomes based on the special densities and
distribution of repetitive elements in imprinted regions. The pre-
dictions were performed by statistical classifiers that had been
trained on the special distributions of repetitive elements (4, 5) and
epigenetic properties of imprinted regions (18). Similar to the
transcriptome data sets, imprinted expression has been proven for
only a few of these candidate genes.
2.2. Databases for For quite a number of mammalian genomes assembled versions
Information on exist that have been made available by the UCSC Genome Browser,
Genomes and DNA and the ENSEMBL Genome Browser. For these genomes, both
Sequences of Different browsers annotate a high number of different genome features that
Species can be downloaded. The annotated features encompass informa-
tion on genome organization such as gene organization, CpG
islands, repetitive elements, etc. Information on different aspects
of sequence conservation is available such as the annotation of
SNPs and highly conserved elements. Furthermore, information

on gene expression and epigenetic modifications is given.
The NCBI provides a broad platform (http://www.ncbi.nlm.
nih.gov/sites/genome) for the access to additional genome
resources. This includes links to information on genomes that are
not annotated in the UCSC or ENSEMBL genome browsers, for
example, for the genome of species that are not considered common
model organisms in medicine or evolutionary research. In addition,
the NCBI provides information on the current status of genome
sequencing projects that are still in progress (see Note 1).
2.3. Databases The Genevestigator database (https://www.genevestigator.com/

for Information gv/index.jsp) is a joint project of the the NEBION company and
on Gene Expression the ETH Zuerich, Switzerland (19). Currently, the database includes
and Epigenetic gene expression data for 14 species. The Genevestigator homepage
Modifications offers several options for doing some basic statistical analyses.
2.3.1. Gene Expression Epigenome data are to some extent already annotated in the UCSC
Data genome browser. More data sets and information from ongoing
epigenome projects can be accessed via the browser of the Epigenome
2.3.2. Epigenome Data
Roadmap project of the NIH Roadmap Epigenomics Mapping
Consortium (http://www.roadmapepigenomics.org/) (20).
2.4. Software for the The appearance of newly sequenced genomes may make it neces-
Annotation of Genomic sary to annotate genomic features, and even in case of genomes
Features that are available in the UCSC or ENSEMBL genome browsers, it
may sometimes be necessary to annotate genomic features anew.
Newly released genomes or sequences that have been generated by
the investigators’ personal project may require special annotations.
A selection of useful software for the annotation of genomic fea-
tures of imprinted genes is listed below:
Identification of repetitive elements:
● RepeatMasker (http://www.repeatmasker.org/): Detection of
(retro-)transposed elements.
● Tandem Repeats Finder (http://tandem.bu.edu/trf/trf.html):
Identification of tandem repeat arrays (21).
Identification of CpG-rich regions:
● EMBOSS CpG plot: http://www.ebi.ac.uk/Tools/emboss/
cpgplot/.
● CpG island searcher: http://cpgislands.usc.edu/ (22).
● CpG cluster: http://genius-index.com/cluster.aspx (23).
● More comprehensive overviews on the performance of avail-
able software for CpG island identification are given in two
recent reviews (24, 25).
256 M. Paulsen
Identification of transcription factor binding sites:

● Transfac database: http://www.gene-regulation.com/pub/
databases.html.
● An overview of the computational analyses and discovery of
transcription factor binding sites is given in a recent review
published in Methods in Molecular Biology (26).
Epigenetic features:
● Epigraph: Relationships between DNA sequence and epige-
netic modifications (http://epigraph.mpi-inf.mpg.de/
WebGRAPH/), (27).
● BIQ Analyzer HAT: Processing and statistical analyses of
bisulfite sequencing data (http://biq-analyzer-ht.bioinf.mpi-
inf.mpg.de/) (28).
● Methvisual package: Visualization and statistical analyses of
bisulfite sequencing data (http://methvisual.molgen.mpg.
de/) (29).
● The Methdb database (http://www.methdb.de/links.html)
provides a comprehensive list of computational tools that are
specifically designed for applications in epigenetics (30).
Data organization and statistical analyses:
● Bioconductor provides the framework for sequence retrieval,
annotation, and statistical analyses (http://www.bioconductor.
org/).
● R package: Open source package for statistics (http://www.r-
project.org/).
3. Methods
Genome-wide studies on imprinted genes consist usually of four

phases that will be the subject of the following paragraphs:
1. Selection of gene sets
2. Sequence retrieval—Evaluation of sequence quality
3. Annotation of genomic and epigenomic features
4. Statistical analyses of different parameters
3.1. Selection of Gene If studies on a rather broad group of imprinted genes are planned
Sets the selection of sets of imprinted genes can be hampered by various
factors that are to some extent specific for this special class of
3.1.1. Collecting Gene
genes.
Sets of Suitable Size
The major problem with selecting a set of imprinted genes is
the danger of too small a sample size with a gene number below
hundred, or even below 50. In order to ensure a sound statistical
analysis gene numbers below 30 should be avoided.
3.1.2. Manually Collected In most cases the selection of imprinted genes is driven by the need
Data Sets to get a set of genes whose imprinting status is out of question. As
a primary source, public databases on imprinted genes are usually
adequate. For consistent selection, several parameters can be used
as criteria, such as a minimal number of cited publications, conser-
vation of imprinting in at least two species, application of at least
two different methods for detection of allele-specific expression or
allele-specific modifications, successful tests for imprinted expres-
sion in a distinct number of samples or individuals, etc. Examples
of such manually collected lists can be found in (8, 31, 32).
3.1.3. Selection Another option can be the usage of one of the transcriptome data
of Imprinted Candidate sets on imprinted expression described in Subheading 2.1.5. This
Genes from Transcriptome has the advantage that the imprinted candidate genes are all dis-
Data Sets covered by the same method. From such a data set imprinted genes
can be chosen by setting a threshold for a score for parental allele-
specific expression that has to be passed. This procedure has the
advantage of being free of any biases caused by manual curation of
gene lists. However, there are indications that such a procedure
might be biased as well. For example, data sets on allele-specific
gene expression are biased towards strongly expressed genes that
achieve higher scores for allele-specific expression than weakly
expressed genes (33).
3.1.4. Imprinted Genes Can Imprinted regions in the human and mouse genomes contain many
Be Divided into Protein- genes that encode untranslated RNAs. Among these are long non-
Encoding Genes and RNA spliced transcripts that appear to be involved in epigenetic silencing
Genes and spliced noncoding RNAs such as the H19 RNA whose func-
tion in imprinted gene regulation is still an enigma (34, 35).
Especially the microRNA and snoRNA clusters of imprinted
regions are famous for their high numbers of small RNAs (36–38).
For studies that aim to identify features that distinguish imprinted
genes from non-imprinted genes one should consider analyzing
imprinted protein-coding genes and untranslated RNA genes sepa-
rately since genes of noncoding RNAs and protein-encoding genes
differ strongly in terms of their exon/intron structure.
3.2. Sequence The next step in computational analysis is usually the download of
Retrieval: Evaluation genomic DNA sequences. Usually, this will be done via the server
of Sequence Quality of the UCSC or Ensembl Genome browser or via the Genome
project homepage of the NCBI.
Due to the differences in progress of the numerous sequencing
projects, genomic sequences can show pronounced differences in
quality as is nicely described for comparison of the Gnas locus in
three species (39). Especially if several unfinished genomes are to
be compared, an eye should be kept on parameters such as read
coverage and contig length of the chosen genomes, and it might
be a good idea to select for comparative studies only genomes with
similar average contig length or sequence coverage.
258 M. Paulsen
In past studies (own unpublished data) we have seen that the

average contig length of assembled sequences can differ significantly
between the genomes of different species. We also observed differ-
ences in contig lengths between imprinted genes and other auto-
somal genes. For example repetitive elements may cause assembly
problems. Hence, sequence and assembly quality can cause biases
in statistical analyses on features such as the density of repetitive
elements, and CpG and G+C content.
3.3. Annotation of Before starting with the actual annotation of DNA sequences it is
Sequence Features always very useful to develop a plan for the consistent annotation
of the physical structure of all genes. This means that the investiga-
3.3.1. Organization of a
tors need to define the starts and ends of transcriptional units and
Systematic Annotation
the extent of intergenic regions. This sounds rather trivial but can
be difficult in case of overlapping genes, gene with more than one
promoter, etc. Another typical problem is the multiple annotation
of transcripts of the same gene. Such a phenomenon can be caused
by the presence of more than one reference sequence for one gene.
For example, if this affects mostly strongly expressed genes, such
multiple annotations may bias studies on gene expression towards
strongly expressed genes.
As already mentioned above, many features are annotated in
the UCSC or Ensembl genome browsers. Positions, scores, etc.
can be downloaded in table formats and can subsequently be used
for statistical comparisons of imprinted genes vs. control genes.
In rare cases, for example if newly sequenced genomes are part of the
analyses such sequences have to be annotated by the investigator.
Some useful software suites are listed above (see Subheading 2.3).
Especially when DNA sequences need to be converted to dif-
ferent formats or if basic statistical analyses on general sequence
features are planned, the Bioconductor package can be very useful
(http://www.bioconductor.org/). The Bioconductor package is
an open source package. It provides tools for DNA sequence analy-
ses, investigations on gene expression patterns, and also software
for epigenetic analyses. The incorporated statistics packages use R
as statistical programming language.
3.3.2. Annotation of DNA Although the information derived from some of the epigenetic
Methylation Patterns data sets has been incorporated into the Ensembl and UCSC
browsers, a lot of information is only available in the form of raw
data. One example for a software that helps to determine DNA
methylation patterns using bisulfite sequencing raw data is the
BiQAnalyzer HT package. The recently updated version is available
as a download and allows the processing of large sets of raw data as
they are typically generated by high-throughput sequencing (28).
Implemented into the package are some useful tools for the visual-
ization of methylation data and tools for statistical analyses of DNA
methylation patterns. An alternative for statistical analyses and the
visual presentation of bisulfite data is the Methvisual package

(http://methvisual.molgen.mpg.de/) (29). This software package
is implemented in an R/Bioconductor environment.
3.3.3. How to Identify A common question in epigenetics is if a specific pattern of epige-

Genomic Features that are netic modifications is associated with special features of DNA
Associated with Epigenetic sequence (see Note 2). For example investigators might be inter-
Modifications ested if differentially methylated regions in imprinted regions show
specific DNA features. For such questions the EpiGraph suite
provides useful computational tools for the identification and sta-
tistical evaluation of DNA features that are associated with a specific
type of epigenetic modifications (27). Furthermore, the tool allows
the prediction of further genomic regions that are likely to show
the same epigenetic modification patterns as the test samples.
3.4. Statistical After annotation of sequence features, the major task of genome-
Analyses wide studies is to compare imprinted vs. non-imprinted genes or to
compare the imprinted genes of different species. Usually tables of
3.4.1. Planning Statistical
annotated elements are taken as a basis for the subsequent statisti-
Tests
cal analyses. Due to the large amount of available genome data
statistics can become very complex. Therefore, it is recommended
to use from the beginning a professional software such as the R
package (http://www.r-project.org/).
Especially researchers with little experiences with statistical
analyses should keep in mind that not every statistical test can be
applied to every question or every sample. Hence, it is usually helpful
to check first if the chosen statistical test can indeed be applied to
the selected samples. A frequent mistake is for example the applica-
tion of t-tests to samples that do not show a normal distribution.
Often a study encompasses not just one but several similarly
structured comparisons of different data sets. In such cases the
consistency of the entire study can be improved if a statistical test
is chosen that can be used for all planned comparisons.
3.4.2. Statistical Problems Rarely do comparative studies on imprinted genes encompass

with Sample Sizes more than 50 genes. In special cases such as differentially methy-
lated regions or retrotansposed imprinted genes numbers drop to
less than 20 or even 10. In such cases, most statistical methods are
not suitable.
Another statistical problem can be the comparison of a moder-
ately sized group of imprinted genes (approx. 50) to a large control
group, for example all other autosomal genes (approx. 20,000).
Due to the extremely large number of control genes, such com-
parisons produce easily significant p-values. One option to show
that an observed difference is unlikely to occur by chance is back-
ground testing, i.e., multiple comparison of a randomly selected,
similarly sized group of genes to all other genes (8).
260 M. Paulsen
4. Notes
Mammalian genomes consist of segments with different G+C

content, the so-called isochores. Different isochores are character-
ized by differences in gene densities and gene activities. Isochores
are typical examples of how the global structure of a genomic
region can influence gene activity. Thus, general features such as
G+C content, repetitive elements, etc. have a major input on the
organization of imprinted regions. Such differences in genomic
features can influence comparative studies on imprinted genes, and
it is therefore sensible to keep a close eye on possible interdepen-
dencies of genomic features and species-specific effects.
1. Species-specific differences
Studies on the identification of imprinted genes focus on human
and mouse, and information on imprinted gene expression in
other species is scarce. However, imprinted gene expression is
apparently well conserved among different mammals: a survey
on orthologous protein-encoding genes that have been investi-
gated in somatic human and mouse tissues revealed that 87 %
were imprinted in both species (8). Despite the conservation of
imprinted gene expression in human tissues and the murine
embryo proper, imprinted expression in the placenta might
differ considerably between the two species. For the mouse,
quite a number of genes have been identified that show a bias
towards maternal expression in the placenta but not in the
embryo proper. Although this phenomenon has been known
for quite a while, for many of these genes it is still unclear if the
maternal bias is indeed caused by a true imprinting effect, or
rather by contamination with maternal tissues (40).
Furthermore, not only in imprinted genes, but also in the
entire genome, features such as G+C and CpG contents and
repetitive elements can differ considerably between different
species. For example, the mouse possesses fewer and shorter
CpG islands than the human (24). In addition, different species
possess different types of repetitive elements, for example Alu
elements are primate-specific repetitive elements and are absent
in other mammals.
2. Correlations between different genomic features
As already mentioned above, genome-wide comparative studies
on imprinted genes can be hampered by technical biases in
association with sequence quality. Genome-wide studies should
take into account that quite a number of interdependencies
exist between different genomic parameters. For example, dif-
ferences in intron length between imprinted and control genes
may have some impact on the intragenic repetitive element or
CpG island content (32). Hence, statistical analyses should be

normalized against the lengths of intragenic and intergenic
regions. A second parameter that can have a pronounced
influence on the analyses of genomic features is the G+C
content. Imprinted genes show highly variable G+C contents.
On average, the G+C content of imprinted genes is slightly but
not significantly elevated in comparison to other autosomal
genes. This slight increase, nevertheless, may be associated
with an elevated CpG content, and may therefore have some
impact on CpG island identification. Hence, an elevated CpG
island content might simply be the result of an elevated G+C
content as is might be also seen in other (non-imprinted)
regions of the genome.
In the human, many Alu elements contain CpG islands.
Hence, in human genomes there are many CpG islands within
repetitive elements. As a consequence, the Alu depletion that is
typical for human imprinted genes influences the CpG island
content of these genes in a species-specific manner (2). One
rather simple option to avoid such complications is to apply
repeat masking to sequences and to analyze single copy sequences
and repetitive elements separately.
Nevertheless, the described interdependencies between
different parameters call for a more detailed analysis using dif-
ferent approaches such as multivariate statistics.
Acknowledgments
Computational studies in the Paulsen group are supported by the

Deutsche Forschungsgemeinschaft (DFG grant PA750/3-1).
References
1. Greally JM (2002) Short interspersed transpos- 5. Luedi PP et al (2005) Genome-wide prediction

able elements (SINEs) are excluded from of imprinted murine genes. Genome Res 15:
imprinted regions in the human genome. Proc 875–884
Natl Acad Sci USA 99:327–332 6. Ke X et al (2002) The distinguishing sequence
2. Hutter B et al (2006) Tandem repeats in the characteristics of mouse imprinted genes.
CpG islands of imprinted genes. Genomics 88: Mamm Genome 13:639–645
323–332 7. Varrault A et al (2006) Zac1 regulates an
3. Kobayashi H et al (2006) Bisulfite sequencing imprinted gene network critically involved in
and dinucleotide content analysis of 15 the control of embryonic growth. Dev Cell
imprinted mouse differentially methylated 11:711–722
regions (DMRs): paternally methylated DMRs 8. Steinhoff C et al (2009) Expression profile and
contain less CpGs than maternally methylated transcription factor binding site exploration of
DMRs. Cytogenet Genome Res 113:130–137 imprinted genes in human and mouse. BMC
4. Luedi PP et al (2007) Computational and Genomics 10:15
experimental identification of novel human 9. Morison IM et al (2005) A census of mamma-
imprinted genes. Genome Res 17:1723–1730 lian imprinting. Trends Genet 21:457–465
262 M. Paulsen
10. Williamson CM et al. (2011), MRC Harwell, 26. Ladunga I (2010) An overview of the compu-
Oxfordshire. World Wide Web Site - Mouse tational analyses and discovery of transcription
Imprinting Data and References - http://www. factor binding sites. Methods Mol Biol 674:
har.mrc.ac.uk/research/genomic_imprinting/ 1–22
11. Zhang Y et al (2010) ncRNAimprint: a com- 27. Bock C et al (2010) Web-based analysis of
prehensive database of mammalian imprinted (Epi-) genome data using EpiGRAPH and
noncoding RNAs. RNA 16:1889–1901 Galaxy. Methods Mol Biol 628:275–296
12. Seoighe C et al (2006) Maximum likelihood 28. Lutsik P et al (2011) BiQ Analyzer HT: locus-
inference of imprinting and allele-specific specific analysis of DNA methylation by high-
expression from EST data. Bioinformatics throughput bisulfite sequencing. Nucleic Acids
22:3032–3039 Res 39:W551–W556
13. Wang X et al (2008) Transcriptome-wide 29. Zackay A, Steinhoff C (2010) MethVisual -
identification of novel imprinted genes in neo- visualization and exploratory statistical analysis
natal mouse brain. PLoS One 3:e3839 of DNA methylation profiles from bisulfite
14. Gregg C et al (2010) High-resolution analysis sequencing. BMC Res Notes 3:337.20
of parent-of-origin allelic expression in the 30. Negre V, Grunau C (2006) The MethDB DAS
mouse brain. Science 329:643–648 server: adding an epigenetic information layer
15. Nakabayashi K et al (2011) Methylation screen- to the human genome. Epigenetics 1:101–105
ing of reciprocal genome-wide UPDs identifies 31. Hutter B et al (2010) Divergence of imprinted
novel human-specific imprinted genes. Hum genes during mammalian evolution. BMC Evol
Mol Genet 20(16):3188–3197 Biol 10:10
16. Mizuno Y et al (2002) Asb4, Ata3, and Dcn are 32. Hutter B et al (2010) Imprinted genes show
novel imprinted genes identified by high- unique patterns of sequence conservation.
throughput screening using RIKEN cDNA BMC Genomics 11:649
microarray. Biochem Biophys Res Commun 33. Nothnagel M et al (2011) Statistical inference
290:1499–1505 of allelic imbalance from transcriptome data.
17. Nikaido I et al (2003) Discovery of imprinted Hum Mutat 32:98–106
transcripts in the mouse transcriptome using 34. Sleutels F, Zwart R, Barlow DP (2002) The non-
large-scale expression profiling. Genome Res coding Air RNA is required for silencing auto-
13:1402–1409 somal imprinted genes. Nature 415:810–813
18. Brideau CM et al (2010) Successful computa- 35. Gabory A et al (2009) H19 acts as a trans regu-
tional prediction of novel imprinted genes lator of the imprinted gene network control-
from epigenomic features. Mol Cell Biol ling growth in mice. Development 136:
30:3357–3370 3413–3421
19. Hruz T et al (2008) Genevestigator v3: a refer- 36. Seitz H et al (2003) Imprinted microRNA
ence expression database for the meta-analysis genes transcribed antisense to a reciprocally
of transcriptomes. Adv Bioinformatics imprinted retrotransposon-like gene. Nat
2008:420747 Genet 34:261–262
20. Bernstein BE et al (2010) The NIH roadmap 37. Kircher M, Bock C, Paulsen M (2008)
epigenomics mapping consortium. Nat Structural conservation versus functional diver-
Biotechnol 28:1045–1048 gence of maternally expressed microRNAs in
21. Benson G (1999) Tandem repeats finder: a the Dlk1/Gtl2 imprinting region. BMC
program to analyze DNA sequences. Nucleic Genomics 9:346
Acids Res 27:573–580 38. Cavaillé J et al (2000) Identification of brain-
22. Takai D, Jones PA (2002) Comprehensive analysis specific and imprinted small nucleolar RNA
of CpG islands in human chromosomes 21 and genes exhibiting an unusual genomic organi-
22. Proc Natl Acad Sci USA 99:3740–3745.2 zation. Proc Natl Acad Sci USA 97:
23. Hackenberg M et al (2006) CpGcluster: a dis- 14311–14316
tance-based algorithm for CpG-island detec- 39. Hart EA et al (2007) Lessons learned from the
tion. BMC Bioinformatics 7:446 initial sequencing of the pig genome: compara-
24. Hutter B et al (2009) Identifying CpG islands tive analysis of an 8 Mb region of pig chromo-
by different computational techniques. Omics-a some 17. Genome Biol 8:R168
Journal of Integrative Biology 13:153–164 40. Proudhon C, Bourc‘his D (2010) Identification
25. Zhao Z, Han L (2009) CpG islands: algorithms and resolution of artifacts in the interpretation
and applications in methylation studies. Biochem of imprinted gene expression. Brief Funct
Biophys Res Commun 382:643–645.14 Genomics 9:374–384
Chapter 18
Insights on Imprinting from Beyond Mice and Men

Andrew Pask
Abstract
Genomic imprinting is an epigenetic phenomenon that results in the silencing of alleles, dependent on
their parent of origin. Within vertebrates, this phenomenon is restricted only to the mammals and has been
identified in eutherians and marsupials but not in the egg-laying monotremes. Many hypotheses have been
put forward to explain why genomic imprinting evolved, most of which are centered on the regulation of
nutrient provisioning from parent to offspring. The three different mammalian lineages have adopted very
different modes of reproduction and, as a result, vary widely in the amount of nutrient provisioning to the
conceptus. Examining imprinting across the three mammal groups enables us to test hypotheses on the
origin of this phenomenon in mammals and also to investigate changes in the genome coincident with its
evolution.
Key words: Genomic imprinting, Eutherian, Marsupial, Monotreme, Genome evolution
1. Introduction
The class Mammalia is divided up into three extant lineages, the

eutherians, marsupials, and monotremes. Eutherian mammals last
shared a common ancestor with marsupials around 160 million
years ago and monotremes around 180 million years ago (Fig. 1)
(1). Each lineage is characterized by a unique mode of reproduc-
tion. Eutherian mammals typically give birth to well-developed
young after an extended gestation period. Fetal growth is sup-
ported by a large and invasive placenta, and pregnancies place
large physiological demands on the mother (Fig. 2). Marsupials,
like eutherian mammals, have a fully functional placenta to deliver
nutrition to their developing fetus, but it is less invasive and short-
lived. Most marsupial species give birth to very small young that
place a minimal demand on the mother’s resources during
preg-nancy. Maternal contribution to pregnancy is further reduced
in the egg-laying monotremes. Even so, the early stages of
263
264 A. Pask
Short gestation Prolonged gestation

Minimal placental Increased placental
dependence dependence
Few imprinted loci Many imprinted loci
Repeat expansion
160 Genomic imprinting
Vivipartiy
Few repeats
180
No genomic imprinting
Fig. 1. Evolutionary tree of the three extant mammalian lineages. Numbers indicate
divergence times in millions of years. Arrows indicate the acquisition of various features
associated with genomic imprinting.
5% 0.005% 0.05%
Fig. 2. Relative maternal contributions to offspring for each mammalian lineage (Eutherians—
left, marsupials—center and monotremes—right). Numbers indicate the average weight of
the offspring as a percentage of the maternal weight.
development and nutritional supplies for the egg require maternal

resource provisioning, but this places little demand on the mother
(2). Each of these strategies results in different maternal contribu-
tions of nutrients to the offspring in utero and during postnatal
care, making them especially useful models for examining the ori-
gins of genomic imprinting.
18 Comparative Insights on Imprinting 265
Genomic imprinting is the epigenetic silencing of certain alleles

dependant on their parent of origin. As a result, the eutherian
conceptus is haploinsufficient for around 100 different genes, many
of which are essential regulators of growth and development (3).
This presents a perplexing evolutionary strategy since any mutations
in the single expressed copy of each of these genes could lead to
disease. In fact, we see many disease states in humans arising from
mutations in imprinted loci (4). Many different theories have been
put forward to explain what the evolutionary advantage might be
of adopting much a mechanism, the most supported of which is
the parental conflict/kinship hypotheses (5, 6). Briefly, these theo-
ries attempt to explain the different investments of the male and
female genomes in the fitness of the offspring. Paternally expressed
genes that favor nutrient provisioning and growth of the offspring
will result in greater genetic fitness of the father, even if this occurs
at the mother’s expense. The genetic fitness of the mother is
increased by restricting the nutrients given to any one offspring, so
that she may have many successive pregnancies. In line with these
hypotheses, the vast majority of imprinted genes identified in mice
and humans are expressed in the placenta or fetus and affect growth
and nutrition in utero (7). Thus, the parental conflict/kinship
hypotheses would have very different predictions for the preva-
lence of genomic imprinting across the three extant mammalian
lineages. One would expect that the evolutionary forces that
have favored the evolution of imprinting in eutherians would be
significantly reduced in the marsupials and monotremes.
Here I review how studies in marsupials and monotremes
have contributed to our understanding of why and how genomic
imprinting evolved. Such evolutionary studies have been funda-
mental to our understanding of the mechanisms governing genomic
imprinting and the evolutionary forces selecting for it.
2. The Absence
of Genomic
Imprinting in the
Nonmammalian Genomic imprinting is presumed not to exist in nonmammalian
Vertebrates vertebrates since all lineages are capable of producing viable
parthenogenotes (embryos created from only maternal chromo-
somal contributions). This finding suggests that the genes must
not be differently silenced between the sexes. In contrast, parthe-
nogenetic mouse embryos show early embryonic lethality, due to
the loss of critical gene function, and the few mammalian parthe-
nogenetic embryos that were able to make it to post implantation
stages of development show stunted placental growth (8).
Despite the presumed absence of genomic imprinting in the
genomes of nonmammalian vertebrates, only a few studies have
actually verified this assumption. Two highly conserved genes that
266 A. Pask
show monoallelic expression in mice in humans (IGF2 and IGF2R)

were shown to be biallelically expressed in the chicken embryo (9).
Most mammalian imprinted genes exist in clusters in the genome
and are often regulated by shared imprinting control regions
(ICRs). While no such imprinting control regions have been
identified outside of the therian (marsupial and eutherian) mam-
mals, the genes themselves are clustered in the same highly con-
served arrangements in chickens and show asynchronous replication
which may have served as a prelude to genomic imprinting (10).
3. Insights on
the Evolution of
Genomic
Imprinting from Monotremes, like all oviparous animals, provide some maternal
the Monotreme nutrients to the developing embryo before the egg is laid (2).
Genome However, this contribution is often minimal, and would not have
as major an impact on maternal resources as an extended preg-
nancy (Fig. 2). Therefore, based on the parental conflict/kinship
hypotheses, one might expect a lack of genomic imprinting associ-
ated with genes regulating fetal nutrition in monotremes, and this
is indeed the case. Imprinting has not been detected in any of the
ten eutherian imprinted genes so far investigated in the platypus
(11–21) (Table 1). However, due to the protected status of this
species, the imprint status for each of these genes has only been
examined in adult material and not in the developing young or
fetal membranes. Nevertheless, it is generally concluded that
imprinting does not occur in monotremes, making them ideal
comparative models for examining the evolution of this epigenetic
phenomenon.
Many hypotheses have been put forward that attempt to
explain what the evolutionary advantage of imprinting might be,
but it is equally important to examine how such a mechanism arose.
The host defense hypothesis suggests that genomic imprinting
evolved from endogenous mechanisms that silence transposable
elements and invading foreign DNA within the genome (44). This
hypothesis is supported by the observation that most imprinted
genes in eutherians are associated with a high density of repeat
sequences and endogenous retroviruses that could have attracted
silencing to the region (45). As such, the monotreme (platypus)
genome (46) provides a unique resource to compare with marsupials
(opossum (47) and tammar (48)) and eutherians (49, 50), to
determine what changes were coincident with the evolution of
imprinting. These analyses revealed that most imprinted genes
were highly conserved at the nucleotide level in all mammals (51).
Furthermore, most genes resided in similar clusters, suggesting
their spatial arrangement predated the evolution of imprinting
(51). The regions of the platypus genome surrounding all the
Table 1
A list of the eutherian imprinted loci that have been investigated in marsupials
and monotremes
Location (human) Gene Eutherian Marsupial Monotreme
6q25 IGF2R Yes (22) Yes (13) No (13)

Air Yes (23) No (20)
7q21.3 SGCE Yes (24) No (17) No (17)
PEG10 Yes (25) Yes (17) Absent (17)
PPP1R9A Yes (26) No (17)
ASB4 Yes (27) No (17)
7q32.2 MEST Yes (28) Yes (18)
11p15 H19 Yes (29) Yes (16)
IGF2 Yes (30) Yes (31) No (14)
INS Yes (32) Yes (33)
CDKN1C Yes (34) No (35)
PHLDA2 Yes (36) No (19)
14q32 DLK1 Yes (37) No (11) No (11)
MEG3 Yes (38) Absent (20, 21)
RTL1 Yes (39) Absent (11)
DIO3 Yes (40) No (11) No (11)
15q11–12 SNURF-SNRPN Yes (41) No (15) No (15)
UBE3A Yes (42) No (15) No (15)
20q11.23 NNAT Yes (43) Absent (12)
The genes are grouped into clusters based on their arrangement in the human genome. The imprint status for each gene
is indicated for each mammalian lineage: Yes, denotes that the gene is imprinted; No, denotes the gene is not imprinted;
Absent, denotes that the gene is not present in that lineage
imprinted gene orthologues contained significantly less long-

terminal repeats (LTRs) and DNA transposable elements than in
the therian orthologous regions (51). In addition, other classes of
transposable elements (such as SINEs and LINEs) were seen to
expand in certain therian imprinted gene clusters (51). An expan-
sion in repeat elements is also seen across the entire mammalian
genome after the divergence of monotremes (46). Thus, it appears
that imprinting evolved coincident with repeat expansion in the
mammalian genome. This is consistent with the host defense
hypothesis (44), and suggests that repeats attracted epigenetic
silencing to many regions throughout the genome. Where this
silencing caused an evolutionary advantage it was selected for, and
maintained, and the gene became imprinted. If the silencing had
no effect, or a deleterious effect, it was lost (51). While the platy-
pus genome analyses provided important data in support of the
host defense hypothesis, the first direct evidence of such a mecha-
nism came from imprinted gene analyses in marsupials.
268 A. Pask
4. The Birth
of Genomic
Imprinting in
Marsupial Marsupials are viviparous and give birth to relatively small young,
Mammals placing minimal demand on maternal resources (Fig. 2) (52).
Under the predictions of the parental conflict/kinship hypotheses
we would expect to see reduced selection for imprinting in this lin-
eage. This prediction holds true, and of the 19 eutherian imprinted
orthologues investigated in marsupials, only six are imprinted
(Table 1) (11–13, 15–21, 31, 33, 35), three of which reside in the
IGF2-H19 cluster. One of these genes, insulin (INS), is exclusively
imprinted in the eutherian placenta but not elsewhere in the
embryo (53). Similarly, INS is exclusively imprinted in the placen-
tal membranes in marsupials (33). Since INS has maintained
imprinting solely in the therian placenta, evolutionary pressures in
this tissue alone must have been sufficient to drive genomic imprint-
ing (33). This provides strong support for the parental conflict/
kinship hypotheses and for the placenta being a focal point for the
selection of genomic imprinting.
In eutherians, imprinted genes are generally regulated by a dif-
ferentially methylated region (DMR) (54). However, in marsupials,
DMRs have only been found to be associated with two imprinted
genes, PEG10 (17) and H19 (16). Interestingly, these are also the
only two genes that show complete silencing from the imprinted
alleles. The remaining four marsupial imprinted genes show paren-
tal bias, resulting in diminished expression from the imprinted
allele, but not complete silencing (2). These data suggest that
genomic imprinting does not require differential methylation, and
that such a mechanism may have evolved over time, due to increased
selective pressure, to further silence imprinted loci.
The accumulation of regulatory mechanisms appears to be a
feature of many imprinted clusters across marsupial and eutherian
mammals. For example, in eutherian mammals the CDKN1C gene
is syntenic to IGF2 and oppositely imprinted. Both genes are
expressed in the placenta and are antagonistic in function (55).
CDKN1C and IGF2 are also in a syntenic arrangement in the mar-
supial genome and CDKN1C is expressed alongside IGF2 in the
tammar wallaby placenta. In spite of this conservation, only IGF2
is imprinted and CDKN1C shows biallelic expression (35, 56).
Thus, CDKN1C imprinting is not dependant on its syntenic local-
ization with an imprinted IGF2. Imprinting of CDKN1C in mice
is regulated by KCNQ1OT1, a long noncoding antisense RNA
derived from an intron of the KCNQ1 gene under the control of a
DMR (57). Although CDKN1C is not imprinted in marsupials,
they still produce the KCNQ1OT1 transcript but there is no evi-
dence of a DMR within the region. PHLDA2 is another eutherian
imprinted gene from the same region that negatively controls
placental growth in mice (58). Similar to CDKN1C, marsupial

PHLDA2 shows a conserved syntenic arrangement and expression
in the placenta, but is not imprinted (19). Together, these data
suggest that placental expression and antisense transcription pre-
ceded the acquisition of imprinting within certain gene clusters.
In addition to the accumulation of imprinting control mecha-
nisms, marsupial analyses have also shown the evolution of new
imprinted regions. Peg10 is an imprinted gene derived from the
Sushi-ichi transposon, essential for placental development in mice
(25). Analyses of PEG10 across the three extant mammalian lin-
eages showed that it was recently inserted into the mammalian
genome in the therian ancestor after the monotreme split (17).
This insertion event was coupled with methylation and the acquisi-
tion of an imprint. This imprint has remained in marsupials and
PEG10 is one of only two imprinted genes in marsupials with a
DMR (17). Interestingly, in marsupials imprinting is restricted
only to PEG10 and not the surrounding loci. However, in euthe-
rian mammals, imprinting in the PEG10 region has spread to
encompass the neighboring genes SGCE, PPP1R9A and ASB4,
suggesting that genesis of a new DMR can lead the evolution of
an entire imprinted gene cluster (2). The only other marsupial
imprinted gene with an identified DMR is H19 (16). H19 also
appears to be a recent insertion in to the therian genome, and has
not yet been found in the monotremes.
Similar retrotransposition events also appear to have driven the
evolution of the SNRPN-UBE3A (Prader–Willi/Angelman syn-
drome) region in mice and humans. This imprinted region is not
present in marsupials, in which SNRPN and UBE3A map to differ-
ent chromosomes and are both biallelically expressed. Thus, it
appears that the region was formed and acquired imprinting only
in the eutherian lineage from a major rearrangement event linking
SNRPN and UBE3A. This rearrangement was concomitant with
the insertion of retrotransposed genes and snoRNAs that regulate
imprinted gene expression from the region in mice and humans
(59). Similarly, genomic imprinting of the DLK-DIO3 region is a
recent evolutionary event and occurs only in eutherian mammals.
While the genomic arrangement of these genes is conserved
between eutherians and marsupials, both DLK and DIO3 are bial-
lelically expressed in the wallaby (11). This is in contrast to euthe-
rian mammals where both genes are imprinted (60). Comparisons
of the entire imprinted region between the mouse, human and
wallaby genomes revealed the insertion of the retrotransposed gene
RTL1 and noncoding transcripts including microRNAs and snoR-
NAs exclusively in the eutherian lineage (11, 59).
Since the insertion of retrotransposed genes and snoRNAs are
known to attract methylation, it is likely that they attracted imprint-
ing to the regions described above (2). These findings provide direct
evidence in support of the host defense hypothesis and confirm
270 A. Pask
that retrotransposed genes can acquire methylation dependent

imprinting in mammals. Furthermore, they show that DMR asso-
ciated imprinting arose before the marsupial–eutherian split.
5. What Have
Studies Beyond
Mice and Men
Taught Us About Comparative analyses of imprinted regions across the three extant
Genomic mammalian groups and in birds have enabled us to draw many
conclusions on the origins and selective pressures of genomic
Imprinting?
imprinting. First, we see that the genesis of genomic imprinting in
vertebrates is coincident with the evolution of viviparity. Second,
we see that many of the genes that have retained imprinted expres-
sion in marsupials are expressed in the placenta. Together, these
data are in strong support of the parental conflict/kinship hypoth-
eses for explaining why imprinting has been maintained in mam-
mals. Furthermore, it suggests that in marsupials, even though the
placental attachment is short lived and maternal nutrient contribu-
tion to the developing fetus is minimal, there is sufficient evolu-
tionary pressure to retain imprinting. Many of the eutherian
imprinted gene orthologues that are not imprinted in marsupials
are still expressed in the placenta. This suggests that placental
expression for many loci predated the acquisition of genomic
imprinting. Furthermore, we also see a reduced prevalence of
genomic imprinting in marsupials compared to eutherians as pre-
dicted by the parental conflict/kinship hypotheses. However, it is
possible that marsupials have their own, as yet unidentified, unique
imprinted loci not present in eutherians.
Another consistent feature of imprinted regions is that they
have expanded in eutherians as compared to marsupials, to encom-
pass neighboring loci. Furthermore, we also see an increase in
the complexity of regulatory mechanisms. In many instances this
involves the expression of long noncoding antisense RNA tran-
scripts such as KCNQ1OT1 that regulates imprinting of the
CDKN1C gene. Interestingly, KCNQ1OT1 is conserved and
expressed in marsupials, but CDKN1C is not imprinted suggesting
further regulatory features are needed to initiate silencing of this
region in eutherians.
The stringency of imprinting is also increased in eutherian
mammals as compared to marsupials. The majority of marsupial
imprinted genes lack differential methylation and show biased
expression rather than complete silencing from one allele. This is in
contrast to eutherians where DMRs are a feature of imprinted
regions. Interestingly the only two marsupial imprinted genes that
show complete silencing of one allele are also the only two associ-
ated with DMRs. These data suggest that not all genomic imprint-
ing is methylation based, and that the evolution of DMRs, in at
least some clusters, evolved after the initial imprint was attracted to
the region. Selection for complete silencing and the evolution of
DMRs was greatest in the eutherian lineage, again consistent with
the parental conflict/kinship hypotheses.
In addition to providing empirical evidence explaining why
genomic imprinting evolved, comparative studies in mammals have
also shed light on how it evolved. Examination of imprinted gene
clusters across all mammals showed that the advent of imprinting
was coincident with repeat expansion and retrotransposition events
in many regions. The retrotransposition of PEG10 in therian
mammals provided the first direct evidence that such an event can
trigger the evolution of an entire imprinted region. This was also
the first direct evidence of the host defense hypothesis to explain
the origins of genomic imprinting.
These studies highlight the importance of evolutionary per-
spectives to understanding complex genetic mechanisms. By com-
paring orthologous gene clusters across the three extant mammalian
groups we have gained a deep insight into the evolutionary pres-
sures that selected for genomic imprinting, as well as the mecha-
nisms driving it.
References
1. Luo ZX, Yuan CX, Meng QJ, Ji Q (2011) A 9. Smith SM, Mefford M, Sodora D, Klase Z,
Jurassic eutherian mammal and divergence Singh M, Alexander N, Hess D, Marx PA
of marsupials and placentals. Nature 476: (2004) Topical estrogen protects against SIV
442–445 vaginal transmission without evidence of sys-
2. Renfree MB, Papenfuss AT, Shaw G, Pask AJ temic effect. AIDS 18:1637–1643
(2009) Eggs, embryos and the evolution of 10. Dunzinger U, Nanda I, Schmid M, Haaf T,
imprinting: insights from the platypus genome. Zechner U (2005) Chicken orthologues of
Reprod Fertil Dev 21:935–942 mammalian imprinted genes are clustered on
3. Radford EJ, Ferron SR, Ferguson-Smith AC macrochromosomes and replicate asynchro-
(2011) Genomic imprinting as an adaptative nously. Trends Genet 21:488–492
model of developmental plasticity. FEBS Lett 11. Edwards CA, Mungall AJ, Matthews L, Ryder
585:2059–2066 E, Gray DJ, Pask AJ, Shaw G, Graves JA,
4. Hirasawa R, Feil R (2010) Genomic imprint- Rogers J, Dunham I, Renfree MB, Ferguson-
ing and human disease. Essays Biochem 48: Smith AC (2008) The evolution of the DLK1-
187–200 DIO3 imprinted domain in mammals. PLoS
5. Haig D (2004) Genomic imprinting and kin- Biol 6:e135
ship: how good is the evidence? Annu Rev 12. Evans HK, Weidman JR, Cowley DO, Jirtle RL
Genet 38:553–585 (2005) Comparative phylogenetic analysis of
6. Moore T, Haig D (1991) Genomic imprinting blcap/nnat reveals eutherian-specific imprinted
in mammalian development: a parental tug- gene. Mol Biol Evol 22:1740–1748
of-war. Trends Genet 7:45–49 13. Killian JK, Byrd JC, Jirtle JV, Munday BL,
7. Fowden AL, Sibley C, Reik W, Constancia M Stoskopf MK, MacDonald RG, Jirtle RL
(2006) Imprinted genes, placental develop- (2000) M6P/IGF2R imprinting evolution in
ment and fetal growth. Horm Res 65(Suppl mammals. Mol Cell 5:707–716
3):50–58 14. Killian JK, Nolan CM, Stewart N, Munday BL,
8. Surani MA, Barton SC, Norris ML (1984) Andersen NA, Nicol S, Jirtle RL (2001)
Development of reconstituted mouse eggs sug- Monotreme IGF2 expression and ancestral ori-
gests imprinting of the genome during game- gin of genomic imprinting. J Exp Zool 291:
togenesis. Nature 308:548–550 205–212
272 A. Pask
15. Rapkins RW, Hore T, Smithwick M, Ager E, Ishino F (2006) Deletion of Peg10, an
Pask AJ, Renfree MB, Kohn M, Hameister H, imprinted gene acquired from a retrotranspo-
Nicholls RD, Deakin JE, Graves JA (2006) son, causes early embryonic lethality. Nat Genet
Recent assembly of an imprinted domain 38:101–106
from non-imprinted components. PLoS Genet 26. Nakabayashi K, Makino S, Minagawa S, Smith
2:e182 AC, Bamforth JS, Stanier P, Preece M, Parker-
16. Smits G, Mungall AJ, Griffiths-Jones S, Smith Katiraee L, Paton T, Oshimura M, Mill P,
P, Beury D, Matthews L, Rogers J, Pask AJ, Yoshikawa Y, Hui CC, Monk D, Moore GE,
Shaw G, VandeBerg JL, McCarrey JR, Renfree Scherer SW (2004) Genomic imprinting of
MB, Reik W, Dunham I (2008) Conservation PPP1R9A encoding neurabin I in skeletal mus-
of the H19 noncoding RNA and H19-IGF2 cle and extra-embryonic tissues. J Med Genet
imprinting mechanism in therians. Nat Genet 41:601–608
40:971–976 27. Mizuno Y, Sotomaru Y, Katsuzawa Y, Kono T,
17. Suzuki S, Ono R, Narita T, Pask AJ, Shaw G, Meguro M, Oshimura M, Kawai J, Tomaru Y,
Wang C, Kohda T, Alsop AE, Marshall Graves Kiyosawa H, Nikaido I, Amanuma H,
JA, Kohara Y, Ishino F, Renfree MB, Kaneko- Hayashizaki Y, Okazaki Y (2002) Asb4, Ata3,
Ishino T (2007) Retrotransposon silencing by and Dcn are novel imprinted genes identified
DNA methylation can drive mammalian by high-throughput screening using RIKEN
genomic imprinting. PLoS Genet 3:e55 cDNA microarray. Biochem Biophys Res
18. Suzuki S, Renfree MB, Pask AJ, Shaw G, Commun 290:1499–1505
Kobayashi S, Kohda T, Kaneko-Ishino T, Ishino 28. Kaneko-Ishino T, Kuroiwa Y, Miyoshi N,
F (2005) Genomic imprinting of IGF2, Kohda T, Suzuki R, Yokoyama M, Viville S,
p57(KIP2) and PEG1/MEST in a marsupial, Barton SC, Ishino F, Surani MA (1995) Peg1/
the tammar wallaby. Mech Dev 122:213–222 Mest imprinted gene on chromosome 6
19. Suzuki S, Shaw G, Kaneko-Ishino T, Ishino F, identified by cDNA subtraction hybridization.
Renfree MB (2011) Characterisation of marsu- Nat Genet 11:52–59
pial PHLDA2 reveals eutherian specific acquisi- 29. Zhang Y, Shields T, Crenshaw T, Hao Y,
tion of imprinting. BMC Evol Biol 11:244 Moulton T, Tycko B (1993) Imprinting of
20. Weidman JR, Dolinoy DC, Maloney KA, human H19: allele-specific CpG methylation,
Cheng JF, Jirtle RL (2006) Imprinting of opos- loss of the active allele in Wilms tumor, and
sum Igf2r in the absence of differential methy- potential for somatic allele switching. Am J
lation and air. Epigenetics 1:49–54 Hum Genet 53:113–124
21. Weidman JR, Maloney KA, Jirtle RL (2006) 30. DeChiara TM, Robertson EJ, Efstratiadis A
Comparative phylogenetic analysis reveals mul- (1991) Parental imprinting of the mouse
tiple non-imprinted isoforms of opossum Dlk1. insulin-like growth factor II gene. Cell 64:
Mamm Genome 17:157–167 849–859
22. Barlow DP, Stoger R, Herrmann BG, Saito K, 31. O’Neill MJ, Ingram RS, Vrana PB, Tilghman
Schweifer N (1991) The mouse insulin-like SM (2000) Allelic expression of IGF2 in mar-
growth factor type-2 receptor is imprinted and supials and birds. Dev Genes Evol 210:18–20
closely linked to the Tme locus. Nature 32. Moore GE, Abu-Amero SN, Bell G, Wakeling
349:84–87 EL, Kingsnorth A, Stanier P, Jauniaux E,
23. Lyle R, Watanabe D, te Vruchte D, Lerchner Bennett ST (2001) Evidence that insulin is
W, Smrzka OW, Wutz A, Schageman J, Hahner imprinted in the human yolk sac. Diabetes
L, Davies C, Barlow DP (2000) The imprinted 50:199–203
antisense RNA at the Igf2r locus overlaps but 33. Ager E, Suzuki S, Pask A, Shaw G, Ishino F,
does not imprint Mas1. Nat Genet 25:19–21 Renfree MB (2007) Insulin is imprinted in the
24. Muller B, Hedrich K, Kock N, Dragasevic N, placenta of the marsupial, Macropus eugenii.
Svetel M, Garrels J, Landt O, Nitschke M, Dev Biol 309:317–328
Pramstaller PP, Reik W, Schwinger E, Sperner 34. Hatada I, Mukai T (1995) Genomic imprint-
J, Ozelius L, Kostic V, Klein C (2002) Evidence ing of p57KIP2, a cyclin-dependent kinase
that paternal expression of the epsilon-sarco- inhibitor, in mouse. Nat Genet 11:204–206
glycan gene accounts for reduced penetrance in 35. Ager EI, Pask AJ, Gehring HM, Shaw G,
myoclonus-dystonia. Am J Hum Genet 71: Renfree MB (2008) Evolution of the CDKN1C-
1303–1311 KCNQ1 imprinted domain. BMC Evol Biol
25. Ono R, Nakamura K, Inoue K, Naruse M, 8:163
Usami T, Wakisaka-Saito N, Hino T, Suzuki- 36. Salas M, John R, Saxena A, Barton S, Frank D,
Migishima R, Ogonuki N, Miki H, Kohda T, Fitzpatrick G, Higgins MJ, Tycko B (2004)
Ogura A, Yokoyama M, Kaneko-Ishino T, Placental growth retardation due to loss of
imprinting of Phlda2. Mech Dev 121: Wakefield MJ, Olender T, Lancet D, Huttley
1199–1210 GA, Smit AF, Pask A, Temple-Smith P, Batzer
37. Wylie AA, Murphy SK, Orton TC, Jirtle RL MA, Walker JA, Konkel MK, Harris RS,
(2000) Novel imprinted DLK1/GTL2 domain Whittington CM, Wong ES, Gemmell NJ,
on human chromosome 14 contains motifs Buschiazzo E, Vargas Jentzsch IM, Merkel A,
that mimic those implicated in IGF2/H19 Schmitz J, Zemann A, Churakov G, Kriegs JO,
regulation. Genome Res 10:1711–1718 Brosius J, Murchison EP, Sachidanandam R,
38. Miyoshi N, Wagatsuma H, Wakana S, Shiroishi Smith C, Hannon GJ, Tsend-Ayush E,
T, Nomura M, Aisaka K, Kohda T, Surani MA, McMillan D, Attenborough R, Rens W,
Kaneko-Ishino T, Ishino F (2000) Identification Ferguson-Smith M, Lefevre CM, Sharp JA,
of an imprinted gene, Meg3/Gtl2 and its Nicholas KR, Ray DA, Kube M, Reinhardt R,
human homologue MEG3, first mapped on Pringle TH, Taylor J, Jones RC, Nixon B,
mouse distal chromosome 12 and human chro- Dacheux JL, Niwa H, Sekita Y, Huang X, Stark
mosome 14q. Genes Cells 5:211–220 A, Kheradpour P, Kellis M, Flicek P, Chen Y,
Webber C, Hardison R, Nelson J, Hallsworth-
39. Seitz H, Youngson N, Lin SP, Dalbert S, Pepin K, Delehaunty K, Markovic C, Minx P,
Paulsen M, Bachellerie JP, Ferguson-Smith AC, Feng Y, Kremitzki C, Mitreva M, Glasscock J,
Cavaille J (2003) Imprinted microRNA genes Wylie T, Wohldmann P, Thiru P, Nhan MN,
transcribed antisense to a reciprocally imprinted Pohl CS, Smith SM, Hou S, Nefedov M, de
retrotransposon-like gene. Nat Genet 34: Jong PJ, Renfree MB, Mardis ER, Wilson RK
261–262 (2008) Genome analysis of the platypus reveals
40. Tsai CE, Lin SP, Ito M, Takagi N, Takada S, unique signatures of evolution. Nature 453:
Ferguson-Smith AC (2002) Genomic imprint- 175–183
ing contributes to thyroid hormone metabo- 47. Mikkelsen TS, Wakefield MJ, Aken B, Amemiya
lism in the mouse embryo. Curr Biol 12: CT, Chang JL, Duke S, Garber M, Gentles AJ,
1221–1226 Goodstadt L, Heger A, Jurka J, Kamal M,
41. Leff SE, Brannan CI, Reed ML, Ozcelik T, Mauceli E, Searle SM, Sharpe T, Baker ML,
Francke U, Copeland NG, Jenkins NA (1992) Batzer MA, Benos PV, Belov K, Clamp M,
Maternal imprinting of the mouse Snrpn gene Cook A, Cuff J, Das R, Davidow L, Deakin JE,
and conserved linkage homology with the Fazzari MJ, Glass JL, Grabherr M, Greally JM,
human Prader-Willi syndrome region. Nat Genet Gu W, Hore TA, Huttley GA, Kleber M, Jirtle
2:259–264 RL, Koina E, Lee JT, Mahony S, Marra MA,
42. Herzing LB, Cook EH Jr, Ledbetter DH Miller RD, Nicholls RD, Oda M, Papenfuss
(2002) Allele-specific expression analysis by AT, Parra ZE, Pollock DD, Ray DA, Schein JE,
RNA-FISH demonstrates preferential maternal Speed TP, Thompson K, VandeBerg JL, Wade
expression of UBE3A and imprint maintenance CM, Walker JA, Waters PD, Webber C,
within 15q11–q13 duplications. Hum Mol Weidman JR, Xie X, Zody MC, Graves JA,
Genet 11:1707–1718 Ponting CP, Breen M, Samollow PB, Lander
43. Kagitani F, Kuroiwa Y, Wakana S, Shiroishi T, ES, Lindblad-Toh K (2007) Genome of the
Miyoshi N, Kobayashi S, Nishida M, Kohda T, marsupial Monodelphis domestica reveals inno-
Kaneko-Ishino T, Ishino F (1997) Peg5/ vation in non-coding sequences. Nature 447:
Neuronatin is an imprinted gene located on 167–177
sub-distal chromosome 2 in the mouse. Nucleic 48. Renfree MB, Papenfuss AT, Deakin JE, Lindsay
Acids Res 25:3428–3432 J, Heider T, Belov K, Rens W, Waters PD,
44. Barlow DP (1993) Methylation and imprint- Pharo EA, Shaw G, Wong ES, Lefevre CM,
ing: from host defense to gene regulation? Nicholas KR, Kuroki Y, Wakefield MJ, Zenger
Science 260:309–310 KR, Wang C, Ferguson-Smith M, Nicholas
45. McDonald JF, Matzke MA, Matzke AJ (2005) FW, Hickford D, Yu H, Short KR, Siddle HV,
Host defenses to transposable elements and the Frankenberg SR, Chew KY, Menzies BR,
evolution of genomic imprinting. Cytogenet Stringer JM, Suzuki S, Hore TA, Delbridge
Genome Res 110:242–249 ML, Mohammadi A, Schneider NY, Hu Y,
O’Hara W, Al Nadaf S, Wu C, Feng ZP, Cocks
46. Warren WC, Hillier LW, Marshall Graves JA, BG, Wang J, Flicek P, Searle SM, Fairley S, Beal
Birney E, Ponting CP, Grutzner F, Belov K, K, Herrero J, Carone DM, Suzuki Y, Sagano S,
Miller W, Clarke L, Chinwalla AT, Yang SP, Toyoda A, Sakaki Y, Kondo S, Nishida Y,
Heger A, Locke DP, Miethke P, Waters PD, Tatsumoto S, Mandiou I, Hsu A, McColl KA,
Veyrunes F, Fulton L, Fulton B, Graves T, Landsell B, Weinstock G, Kuczek E, McGrath
Wallis J, Puente XS, Lopez-Otin C, Ordonez A, Wilson P, Men A, Hazar-Rethinam M,
GR, Eichler EE, Chen L, Cheng Z, Deakin JE, Hall A, Davies J, Wood D, Williams S,
Alsop A, Thompson K, Kirby P, Papenfuss AT,
274 A. Pask
Sundaravadanam Y, Muzny DM, Jhangiani SN, Singer JB, Slater G, Smit A, Smith DR, Spencer
Lewis LR, Morgan MB, Okwuonu GO, Ruiz B, Stabenau A, Stange-Thomann N, Sugnet C,
SJ, Santibanez J, Nazareth L, Cree A, Fowler Suyama M, Tesler G, Thompson J, Torrents D,
G, Kovar CL, Dinh HH, Joshi V, Jing C, Lara Trevaskis E, Tromp J, Ucla C, Ureta-Vidal A,
F, Thornton R, Chen L, Deng J, Liu Y, Shen Vinson JP, Von Niederhausern AC, Wade CM,
JY, Song XZ, Edson J, Troon C, Thomas D, Wall M, Weber RJ, Weiss RB, Wendl MC, West
Stephens A, Yapa L, Levchenko T, Gibbs RA, AP, Wetterstrand K, Wheeler R, Whelan S,
Cooper DW, Speed TP, Fujiyama A, Graves JA, Wierzbowski J, Willey D, Williams S, Wilson
O’Neill RJ, Pask AJ, Forrest SM, Worley KC RK, Winter E, Worley KC, Wyman D, Yang S,
(2011) Genome sequence of an Australian kan- Yang SP, Zdobnov EM, Zody MC, Lander ES
garoo, Macropus eugenii, provides insight into (2002) Initial sequencing and comparative
the evolution of mammalian reproduction and analysis of the mouse genome. Nature
development. Genome Biol 12:R81 420:520–562
49. Waterston RH, Lindblad-Toh K, Birney E, 50. Venter JC, Adams MD, Myers EW, Li PW,
Rogers J, Abril JF, Agarwal P, Agarwala R, Mural RJ, Sutton GG, Smith HO, Yandell M,
Ainscough R, Alexandersson M, An P, Evans CA, Holt RA, Gocayne JD, Amanatides
Antonarakis SE, Attwood J, Baertsch R, Bailey P, Ballew RM, Huson DH, Wortman JR,
J, Barlow K, Beck S, Berry E, Birren B, Bloom Zhang Q, Kodira CD, Zheng XH, Chen L,
T, Bork P, Botcherby M, Bray N, Brent MR, Skupski M, Subramanian G, Thomas PD,
Brown DG, Brown SD, Bult C, Burton J, Zhang J, Gabor Miklos GL, Nelson C, Broder
Butler J, Campbell RD, Carninci P, Cawley S, S, Clark AG, Nadeau J, McKusick VA, Zinder
Chiaromonte F, Chinwalla AT, Church DM, N, Levine AJ, Roberts RJ, Simon M, Slayman
Clamp M, Clee C, Collins FS, Cook LL, Copley C, Hunkapiller M, Bolanos R, Delcher A, Dew
RR, Coulson A, Couronne O, Cuff J, Curwen I, Fasulo D, Flanigan M, Florea L, Halpern A,
V, Cutts T, Daly M, David R, Davies J, Hannenhalli S, Kravitz S, Levy S, Mobarry C,
Delehaunty KD, Deri J, Dermitzakis ET, Reinert K, Remington K, Abu-Threideh J,
Dewey C, Dickens NJ, Diekhans M, Dodge S, Beasley E, Biddick K, Bonazzi V, Brandon R,
Dubchak I, Dunn DM, Eddy SR, Elnitski L, Cargill M, Chandramouliswaran I, Charlab R,
Emes RD, Eswara P, Eyras E, Felsenfeld A, Chaturvedi K, Deng Z, Di Francesco V, Dunn
Fewell GA, Flicek P, Foley K, Frankel WN, P, Eilbeck K, Evangelista C, Gabrielian AE,
Fulton LA, Fulton RS, Furey TS, Gage D, Gan W, Ge W, Gong F, Gu Z, Guan P, Heiman
Gibbs RA, Glusman G, Gnerre S, Goldman N, TJ, Higgins ME, Ji RR, Ke Z, Ketchum KA,
Goodstadt L, Grafham D, Graves TA, Green Lai Z, Lei Y, Li Z, Li J, Liang Y, Lin X, Lu F,
ED, Gregory S, Guigo R, Guyer M, Hardison Merkulov GV, Milshina N, Moore HM, Naik
RC, Haussler D, Hayashizaki Y, Hillier LW, AK, Narayan VA, Neelam B, Nusskern D,
Hinrichs A, Hlavina W, Holzer T, Hsu F, Hua Rusch DB, Salzberg S, Shao W, Shue B, Sun J,
A, Hubbard T, Hunt A, Jackson I, Jaffe DB, Wang Z, Wang A, Wang X, Wang J, Wei M,
Johnson LS, Jones M, Jones TA, Joy A, Kamal Wides R, Xiao C, Yan C, Yao A, Ye J, Zhan M,
M, Karlsson EK, Karolchik D, Kasprzyk A, Zhang W, Zhang H, Zhao Q, Zheng L, Zhong
Kawai J, Keibler E, Kells C, Kent WJ, Kirby A, F, Zhong W, Zhu S, Zhao S, Gilbert D,
Kolbe DL, Korf I, Kucherlapati RS, Kulbokas Baumhueter S, Spier G, Carter C, Cravchik A,
EJ, Kulp D, Landers T, Leger JP, Leonard S, Woodage T, Ali F, An H, Awe A, Baldwin D,
Letunic I, Levine R, Li J, Li M, Lloyd C, Lucas Baden H, Barnstead M, Barrow I, Beeson K,
S, Ma B, Maglott DR, Mardis ER, Matthews L, Busam D, Carver A, Center A, Cheng ML,
Mauceli E, Mayer JH, McCarthy M, McCombie Curry L, Danaher S, Davenport L, Desilets R,
WR, McLaren S, McLay K, McPherson JD, Dietz S, Dodson K, Doup L, Ferriera S, Garg
Meldrim J, Meredith B, Mesirov JP, Miller W, N, Gluecksmann A, Hart B, Haynes J, Haynes
Miner TL, Mongin E, Montgomery KT, C, Heiner C, Hladun S, Hostin D, Houck J,
Morgan M, Mott R, Mullikin JC, Muzny DM, Howland T, Ibegwam C, Johnson J, Kalush F,
Nash WE, Nelson JO, Nhan MN, Nicol R, Kline L, Koduru S, Love A, Mann F, May D,
Ning Z, Nusbaum C, O’Connor MJ, Okazaki McCawley S, McIntosh T, McMullen I, Moy
Y, Oliver K, Overton-Larty E, Pachter L, Parra M, Moy L, Murphy B, Nelson K, Pfannkoch C,
G, Pepin KH, Peterson J, Pevzner P, Plumb R, Pratts E, Puri V, Qureshi H, Reardon M,
Pohl CS, Poliakov A, Ponce TC, Ponting CP, Rodriguez R, Rogers YH, Romblad D, Ruhfel
Potter S, Quail M, Reymond A, Roe BA, B, Scott R, Sitter C, Smallwood M, Stewart E,
Roskin KM, Rubin EM, Rust AG, Santos R, Strong R, Suh E, Thomas R, Tint NN, Tse S,
Sapojnikov V, Schultz B, Schultz J, Schwartz Vech C, Wang G, Wetter J, Williams S, Williams
MS, Schwartz S, Scott C, Seaman S, Searle S, M, Windsor S, Winn-Deen E, Wolfe K, Zaveri
Sharpe T, Sheridan A, Shownkeen R, Sims S, J, Zaveri K, Abril JF, Guigo R, Campbell MJ,
Sjolander KV, Karlak B, Kejariwal A, Mi H, imprinting. Crit Rev Eukaryot Gene Expr
Lazareva B, Hatton T, Narechania A, Diemer 10:241–257
K, Muruganujan A, Guo N, Sato S, Bafna V, 55. Obata Y, Kaneko-Ishino T, Koide T, Takai Y,
Istrail S, Lippert R, Schwartz R, Walenz B, Ueda T, Domeki I, Shiroishi T, Ishino F, Kono
Yooseph S, Allen D, Basu A, Baxendale J, Blick T (1998) Disruption of primary imprinting
L, Caminha M, Carnes-Stine J, Caulk P, Chiang during oocyte growth leads to the modified
YH, Coyne M, Dahlke C, Mays A, Dombroski expression of imprinted genes during embryo-
M, Donnelly M, Ely D, Esparham S, Fosler C, genesis. Development 125:1553–1560
Gire H, Glanowski S, Glasser K, Glodek A, 56. Ager EI, Pask AJ, Shaw G, Renfree MB
Gorokhov M, Graham K, Gropman B, Harris (2008) Expression and protein localisation of
M, Heil J, Henderson S, Hoover J, Jennings D, IGF2 in the marsupial placenta. BMC Dev
Jordan C, Jordan J, Kasha J, Kagan L, Kraft C, Biol 8:17
Levitsky A, Lewis M, Liu X, Lopez J, Ma D,
Majoros W, McDaniel J, Murphy S, Newman 57. Arima T, Kamikihara T, Hayashida T, Kato K,
M, Nguyen T, Nguyen N, Nodell M, Pan S, Inoue T, Shirayoshi Y, Oshimura M, Soejima
Peck J, Peterson M, Rowe W, Sanders R, Scott H, Mukai T, Wake N (2005) ZAC, LIT1
J, Simpson M, Smith T, Sprague A, Stockwell (KCNQ1OT1) and p57KIP2 (CDKN1C) are
T, Turner R, Venter E, Wang M, Wen M, Wu in an imprinted gene network that may play a
D, Wu M, Xia A, Zandieh A, Zhu X (2001) role in Beckwith-Wiedemann syndrome.
The sequence of the human genome. Science Nucleic Acids Res 33:2650–2660
291:1304–1351 58. Frank D, Fortino W, Clark L, Musalo R, Wang
W, Saxena A, Li CM, Reik W, Ludwig T, Tycko
51. Pask AJ, Papenfuss AT, Ager EI, McColl KA,
B (2002) Placental overgrowth in mice lacking
Speed TP, Renfree MB (2009) Analysis of the
the imprinted gene Ipl. Proc Natl Acad Sci
platypus genome suggests a transposon origin
USA 99:7490–7495
for mammalian imprinting. Genome Biol 10:R1
59. Runte M, Huttenhofer A, Gross S, Kiefmann
52. Tyndale-Biscoe CH, Renfree MB (1987) M, Horsthemke B, Buiting K (2001) The
Reproductive physiology of marsupials. IC-SNURF-SNRPN transcript serves as a host
Cambridge University Press, Cambridge for multiple small nucleolar RNA species and as
53. Giddings SJ, King CD, Harman KW, Flood JF, an antisense RNA for UBE3A. Hum Mol
Carnaghi LR (1994) Allele specific inactivation Genet 10:2687–2700
of insulin 1 and 2, in the mouse yolk sac, indi- 60. da Rocha ST, Edwards CA, Ito M, Ogata T,
cates imprinting. Nat Genet 6:310–313 Ferguson-Smith AC (2008) Genomic imprint-
54. Mann JR, Szabo PE, Reed MR, Singer-Sam J ing at the mammalian Dlk1-Dio3 domain.
(2000) Methylated DNA sequences in genomic Trends Genet 24:306–316
Chapter 19
Nonmammalian Parent-of-Origin Effects

Elena de la Casa-Esperón
Abstract
Chromosomes acquire different epigenetic marks during oogenesis and spermatogenesis. After fertilization,
if retained and selected, these differences may result in imprinting effects. Rather than being an oddity,
imprinting effects have been found in many sexually reproducing organisms. Interestingly, imprinting can
result in disparate effects under different selective forces. At the same time, epigenetic mechanisms and
selective pressures shared by sexually reproducing organisms could underlie common imprinting effects.
Large-scale studies are revealing that parent-of-origin effects are more common than previously thought
and supporting the important contribution of imprinting to many traits and diseases.
Key words: Parent-of-origin effects, Parental origin effects, Imprinting, Allelic expression, Chromosome
elimination, Epigenetic reprogramming
“A chromosome which passes through the male germ line acquires an “imprint” which will
result in behavior exactly opposite to the “imprint” conferred on the same chromosome by the
female germ line. In other words, the “imprint” a chromosome bears is unrelated to the genic
constitution of the chromosome and is determined only by the sex of the germ line through which
the chromosome has been inherited”—Helen Crouse, 1960.
1. Introduction:
Imprinting, a Form
of Parent-of-Origin
Effect Parent-of-origin effects (also parental origin effects, POEs) comprise
a broad range of phenomena that result from the different influence
of each parent on the offspring; therefore, POEs are caused by
sexual differences between the parents. The best-known form of
POE is imprinting. The term “imprint” was coined by Helen
Crouse to describe the differential marking of maternal and pater-
nal chromosomes (1). It refers to a reversible mark of epigenetic
nature that is differentially established during oogenesis and sper-
matogenesis, and transmitted to the offspring. By extension, the
term “imprinting” has also been applied to those POEs derived
277
278 E. de la Casa-Esperón
from the different chromatin properties of maternally and paternally

inherited chromosomes, which includes POEs affecting gene
expression, chromosome segregation, heterochromatinization and
a variety of other chromosomal functions. These disparate POEs
are called “imprinting” in their respective fields, creating some
confusion across them. For instance, mammalian researchers often
restrict the use of the term “imprinting” to imprinted gene expres-
sion and their associated epigenetic marks. This form of imprinting,
in which expression of one of the two copies of a gene is repressed
in a parental origin dependent manner, has been the focus of most
imprinting studies in mammals and plants. Only a few studies have
explored other imprinting phenomena that occur within the same
species. In contrast, analyses of other organisms have revealed a
broad diversity of effects caused by differences between the mater-
nally and paternally inherited genomes. Interestingly, the phenom-
ena affected by imprinting in these species can also be imprinted in
mammals, and the lessons learned from imprinting studies in one
species have the potential of revealing new insights about imprint-
ing phenomena in others. Therefore, the aim of this review is to go
beyond the classical imprinting studies in mammals, in order to
provide an overview of the diversity and the distribution of imprint-
ing effects, as well as to discuss their implications.
Not all POEs are imprinting effects: POEs also comprise other
phenomena, such as maternal effects, which result from the
influence of the maternally provided environment on the pheno-
type of her offspring (2). They include effects derived from mater-
nal transcripts present in the egg, maternal nursing behavior, etc.
and, therefore, they are very common. Although it is beyond the
scope of this review to summarize the broad spectrum of maternal
(and paternal) effects identified to date, it is important to notice
that they can be confused with effects derived of imprinted gene
expression. Consequently, several strategies have been proposed to
distinguish diverse types of POEs (3–5).
For the purpose of this review, the term “imprinting” is
restricted to the POEs associated with parent-of-origin dependent
epigenetic imprints (1), while those POEs unrelated to imprinting
(such as maternal effects) will not be discussed in depth. An exten-
sive analysis reveals that a plethora of imprinting effects and variants
occurs in sexually reproducing organisms, although many of them
display recurrent themes. Moreover, diverse forms of imprinting
can occur within the same organism. Imprinting diversity will be
discussed by grouping the effects according to the affected func-
tion: gene expression, whole chromosome heterochromatinization
and/or elimination, chromosomal interactions and replication,
epigenetic marks and other phenotypes. Understanding imprinting
diversity and complexity could, in turn, help us to approach the
analysis of many complex phenotypes from a broader perspective
and may unveil novel phenomena that result from the simple fact
that we have a mom and a dad.
19 Nonmammalian Parent-of-Origin Effects 279
2. Imprinting
Effects in
Expression Can
Affect Single Although a large body of imprinting studies (particularly in
Genes, Entire mammals and plants) have been performed on the expression of
Chromosomes and individual genes, whether they are isolated or clustered in imprinted
domains (6, 7), imprinting can also affect the transcription of entire
Relocated Genes
chromosomes, as well as rearranged genes and transgenes upon
exposure to foreign chromatin. In insects, imprinting of whole
chromosome gene expression has been observed in several species:
condensation and transcriptional inactivation of the entire paternal
set occurs in males of lecanoid coccids (mealybugs) and
Hypothenemus hampei (coffee berry borer) (8, 9). In Drosophila,
paternal transmission of a Dp(1;f)LJ9 mini-X chromosome results
in transcriptional silencing of more than a hundred genes (10, 11).
Chromosome-wide imprinting is not exclusive of insects: in female
mouse extraembryonic tissues (and possibly brain) and marsupial
somatic cells, the paternal X chromosome is preferentially con-
densed and transcriptionally silenced (imprinted X-chromosome
inactivation) (12–14). In addition, POEs have been observed dur-
ing the X-chromosome choice process that precedes random
X-inactivation in mouse embryonic cell lineages (15).
Unlike mammals and plants, in which there are many reports
of endogenous imprinted genes (http://www.har.mrc.ac.uk/
research/genomic_imprinting/) (7), imprinting effects on gene
expression have not been found in karyotypically normal Drosophila.
In fruit flies, imprinting effects on expression are restricted to rear-
ranged genes and transgenes that fall under the repressive influence
of heterochromatic regions (16). Spreading of this heterochroma-
tin to nearby regions results in patchy silencing of genes located in
the vicinity, a phenomenon known as position effect variegation
(PEV). Silencing of the mini-X chromosome genes is an example
of imprinted PEV, caused by rearrangements and deletions that
relocated genes under the effect of centric heterochromatin (11).
Ultraabdominal1 (Uab1) inversion, a bithorax-complex rearrange-
ment, results in a mutant phenotype that is only observed when
paternally transmitted (17). PEVs of multiple transgene insertions
in the heterochromatic Y chromosome are also subject to POEs,
implying that imprinting is a general property of the Drosophila Y
chromosome (18) (unlike other species, the Y chromosome is
necessary for Drosophila male fertility, but it does not determine
sex; therefore, it can be transmitted through both males and
females). Y chromosome imprinting can also affect PEV in other
chromosomes: transmission of the Y chromosome through a
mod(mgd4) male (a mutant of the mod(mgd4)/E(var)3-93D
enhancer of PEV) enhances eye color variegation of the X-linked
wm4h allele (a white locus juxtaposed to centric X heterochromatin).
However, and unlike typical imprinting, this Y chromosome effect

is not reset in the germ line and persists over several generations, even
when the Y chromosome is transmitted through wild type flies (19).
Imprinted transgene expression has also been observed in
other organisms, such as mice and worms (20, 21). In Caenorhabditis
elegans, paternal transmission of transgenes results in greater
expression than maternal transmission. This imprint is acquired
through both male (X0) and hermaphrodite (XX) spermatogene-
sis, resulting in similar transgene expression levels, and is reset by
passage through oogenesis (21). But unlike other species in which
imprinting is considered to be fully reset during gametogenesis in
every generation, germ-line imprint erasure does not seem to be
complete in C. elegans. After transmission of the transgene through
the same germ line for several generations, expression resetting
requires multiple passages through the opposite germ line (21).
In summary, both genome-wide and individual gene studies
have shown that imprinting can affect gene expression in many dif-
ferent ways and species, being more common than previously
thought. Moreover, expression analyses of transgenes and rear-
rangements in C. elegans and Drosophila have revealed that, even in
the absence of endogenous imprinted genes, many chromosomal
regions are capable of generating imprinted expression upon inser-
tion of transgenes. This implies that such regions have parent-of-
origin dependent chromatin properties, which may be involved in
imprinting effects other than gene expression.
3. Imprinting
Effects in
Chromosome
Segregation and Paternal and maternal chromosomes may behave differently during
Elimination mitotic and meiotic segregation. These imprinting effects can be
dramatic and lead to chromosome elimination, or more subtle and
result in preferential segregation (transmission ratio distortion) of
specific chromosomes. Proper chromosome segregation depends
on centromeric integrity and telomere organization, as well as on
chromatin condensation, meiotic pairing and recombination
(which will be discussed in later sections), and, as we will see,
imprinting effects have been observed in all of them.
Although the term “imprinting” is often restricted to the
parent-of-origin dependent monoallelic expression of genes, it was
first coined by Helen Crouse (1) to explain a different phenome-
non that was also influenced by the sex of the transmitting parent:
in sciarid flies, she observed that paternal X chromosomes are selec-
tively eliminated during early embryonic stages (one chromosome
in females and two in males) and in germ cells (one chromosome
in both sexes). In addition, all paternal chromosomes are discarded
during meiosis in males (1, 22). Chromosome elimination of the
entire paternal set is also observed in other invertebrates and, in

several cases, such elimination is part of the sex determination
process. In diaspidid coccids, sex determination depends on the
elimination of the entire paternal set while, in lecanoid coccids
(including mealybugs), the paternal chromosomes are heterochro-
matized and silenced during the cleavage stage of embryogenesis.
Later on, during spermatogenesis, these paternal chromosomes are
also eliminated in lecanoids (9). Similarly, paternal chromosome
condensation in male somatic cells and elimination during sper-
matogenesis is observed in the beetle Hypothenemus hampei, the
coffee berry borer (8). In the wasp Nasonia vitripennis, males are
haploid and develop parthenogenetically from unfertilized eggs,
while fertilized eggs render diploid females. However, males can
develop from diploid eggs after abnormal paternal genome con-
densation and elimination caused by Wolbachia-induced cytoplas-
mic incompatibility or by the effect of an extranumerary paternal
sex ratio (PSR) chromosome. Wolbachia is an intracellular bacte-
rium that causes cytoplasmic incompatibility in insects; in Nasonia
embryos of uninfected females mated with infected males, it
triggers paternal chromosomal loss (23). POEs can also be caused
by the presence of a PSR chromosome in Nasonia sperm, which
results in condensation and subsequent loss of all paternal chromo-
somes except itself; PSR chromosomes are also present in other
hymenoptera with similar POEs (24–26).
Imprinting in chromosome exclusion also occurs in C. elegans:
in XX larvae grown in specific bacterial metabolites, a single X chro-
mosome (preferentially the paternal one) occasionally suffers non-
disjunction and is lost, resulting in X0 males (27); X chromosomal
loss has been observed in larvae from X0 fathers, but not from self-
fertilized hermaphrodites grown in the same conditions. However,
discarding paternally inherited chromosomes is not the universal
rule: several species of Corbicula, a mollusc, can undergo spontane-
ous androgenesis by female pronucleus extrusion in the zygote,
thus eliminating the entire maternal chromosomal set (28, 29).
In Drosophila, several mutations also show POEs on chromo-
some elimination in zygotes and early embryonic stages: maternal
chromosomes are lost in ncd (nonclaret disjunctional) mutants,
whereas pal (paternal loss) and HorkaD mutants eliminate paternal
ones (30–33). In addition, paternal chromosome loss due to
telomere fusion has been observed in Drosophila embryos fathered
by k81 mutants. K81 is a telomeric protein that protects Drosophila
telomeres during spermatogenesis and constitutes a paternal mark
that is required for the functional reestablishment of the paternal
chromosomes after fertilization (34). Therefore, the underlying
cause of these mutation-associated POEs appears to reside in the
differential contribution of the mutated gene products to the pro-
cessing of the chromosomes during spermatogenesis respect to
that of oogenesis.
Imprinting can have more subtle effects on chromosome

segregation than chromosome loss. Transmission ratio distortion
(TRD) in favor of paternal or maternal alleles has been reported in
several mouse and human loci, including imprinted regions (35–
37). TRD can be the result of preferential meiotic segregation or
postfertilization processes, and imprinting effects may act in both
(38). While chromosome loss in insects is often associated to large
chromatin differences (such as chromosome heterochromatiniza-
tion and condensation) between the parental sets (8, 22), it is
unclear which epigenetic marks mediate imprinted TRDs. Proper
chromosome segregation depends on the centromeres and the het-
erochromatic domains that surround them, and thus, parent-of-
origin dependent marks in these regions are candidates for
mediating imprinting effects in chromosome transmission (39).
4. Imprinting
Effects in
Replication and
Chromosomal Imprinting effects also modulate other chromosomal functions,
Interactions such as replication, chromosomal interactions during interphase
and meiosis (pairing and recombination), and nuclear compart-
mentalization. Imprinting has very interesting effects in replica-
tion: in mammalian somatic cells, all imprinted domains tested to
date replicate asynchronously, with the paternal allele replicating
early and the maternal one late (40). But in Drosophila, imprinting
of the Dp(1;f)LJ9 mini-X chromosome affects a different aspect of
replication: the extent of polytenization in somatic cells. During
polytenization, the DNA is replicated multiple times without divi-
sion, resulting in polyploid (polytene) chromosomes. Paternally
transmitted mini-X chromosomes undergo less endoreplication
than maternally transmitted ones. Underreplication, in turn,
reduces copy number and, consequently, the RNA levels of many
adjacent genes. In this way, Drosophila displays a unique form of
imprinting that affects both replication and expression levels (10).
During meiosis, homologous chromosomes pair and recom-
bine. A link between imprinting and meiotic pairing has been
reported in C. elegans: in XX embryos, the paternal X chromosome
is refractory to accumulating activating histone marks present in
the rest of the chromosomes. This POE is never observed past
20-cell stage in cross-progeny from X0 males, while it lasts shorter
(not past 10-cell stage) in self-progeny of XX hermaphrodites. This
observation, as well as the analyses of several mutants bearing one
or two X chromosomes, suggest that C. elegans unpaired X chro-
mosomes acquire more repressive and stable imprints than paired
X chromosomes (41). In Drosophila, imprinting effects in gene
expression are also observed within two genomic contexts in which
meiotic pairing can be deficient or absent: rearrangements and
Y-chromosomes. Meiotic unpaired DNA can sometimes acquire

repressive marks -e.g., Neurospora meiotic silencing by unpaired
DNA (MSUD) that represses the expression of homologous
unpaired genes (42). These observations have led Menon and
Meller (16) to propose that pairing disruption might be necessary
to establish Drosophila germ line-specific imprints. The reciprocal,
imprinting effects on meiotic processes have also been reported:
analyses of meiotic crossovers in the offspring of F1 hybrid mice
have shown that recombination rates can be affected by the paren-
tal origin of the chromosomes (43–45).
Aside from meiosis, imprinted regions also participate in other
chromosome interactions, sometimes with a clear preference for the
paternal or the maternal allele. These interactions have been postu-
lated to contribute to the parent-of-origin dependent activity of
imprinted genes (46–48). For instance, high throughput analyses have
revealed that imprinted loci are overrepresented among the regions
that interact with the Igf2/H19 imprinting controlling region (ICR);
some of these regions preferentially associate with the maternally
inherited ICR (49). These observations have lead to the concept of
“imprinting interactomes,” which might facilitate the regulation of
the epigenetic properties (gene expression, asynchronous replication,
etc.) of multiple imprinted regions in trans (48, 49).
Additional imprinting effects on nuclear organization and
localization of certain genomic sequences have been observed in
insects, as well as in mammals (50, 51). In Sciarid male germ line,
paternal and maternal chromosome sets occupy distinct compart-
ments in prophase I nuclei; this separation may facilitate their
meiotic segregation to opposite poles and the subsequent paternal
chromosome set elimination (22, 52). A transitory separation of
maternal and paternal chromosomes is also observed in preimplan-
tation mouse embryos (53).
In conclusion, studies in several invertebrate species have
shown links between imprinting and chromosome replication,
meiotic chromosomal interactions and nuclear compartmentaliza-
tion. These processes are also affected by imprinting in mammals
and, although the effects are very different, taken together they
indicate that parent-of-origin epigenetic differences present
through nuclear space and at critical times (replication and meio-
sis) have allowed their selection for very diverse functions.
5. Imprinting
Affects Many
Phenotypes
I have presented several examples of imprinting effects that illus-
trate the diversity of phenomena and species affected. Imprinting
can also influence other processes (Fig. 1). A few additional exam-
ples have been selected to illustrate that: (1) different imprinting
ORGANISMS: IMPRINTING EFFECTS ON: REF.
GENE EXPRESSION
Arabidopsis, maize, imprinted gene expression 6, 7, 87,
marsupial & placental 91, 96
mammals
C. elegans, Drosophila, transgenes expression 18, 20,
mouse 21
Lecanoid coccids and paternal chromosomes silencing by 8, 9
Hypothenemus hampei heterochromatinization, contributing to sex
(bettle) determination
Marsupials, mouse paternal X-chromosome silencing by 12-14
heterochromatinization (X-inactivation)
Drosophila PEV in mini-X chromosome 10, 11
Drosophila homeotic transformations by paternal Uab1 17
inversion
Plants parental ratio effect on endosperm development 90
Mammals androgenote and gynogenote death 97-99
CHR. SEGREGATION & ELIMINATION
Marsilea (fern) non-random distribution of paternal chromatids 89
Sciarid flies, coccids, H. paternal chromosomes elimination in mitosis (sex 1, 8, 9,
hampei (bettle) determination) and/or meiosis 22
C. elegans induced preferential paternal X chromosome 27
elimination in XX larvae
Corbibula (mollusc) female pronucleus extrusion in the zygote 28, 29
Nasonia (wasp) paternal chromosomes elimination induced by 23-26
Wolbachia or PSR
Isoodon obesulus paternal X chromosome elimination 100
(marsupial)
Human and mouse transmission ratio distortion 36, 37
Mouse level of aneuploidy in sperm 35
REPLICATION
Drosophila somatic endoreplication (polytenization) of mini-X 10
chromosome
Mammals asynchronous replication of imprinted regions 40
CHROMOSOME INTERACTIONS
Sciara compartimentalization of parental chromosomes 22, 52
in male germ line
Mouse nuclear compartimentalization of imprinted 50
domains
Mouse preimplantation spatial separation of parental 53
genomes
Mouse meiotic recombination levels 43-45
Human and mouse chromosome associations at imprinted regions 46-49
EPIGENETIC DIFFERENCES
Plants, worms, insects DNA methylation, histone modification and/or 6-8, 11,
and mammals condensation differences between maternal and 16, 41,
paternal chromosomes 69-71, 75
Mouse transgenerational imprinting effects 81
Fig. 1. Imprinting effects in sexually reproducing organisms. Although other POEs have been reported in additional species
and phenotypes, it remains to be determined if they are actually caused by imprinting (see text) and, therefore, have not been
included in this list.
mechanisms can affect the same phenomenon, (2) a chromosomal

imprint can have multiple effects and (3) large-scale studies have
the potential to unravel novel imprinted loci and imprinting effects
in additional species.
Sex determination exemplifies how different imprinting effects
can lead to similar outcomes. As previously discussed, sex determi-
nation is mediated by diverse imprinting effects in coccids and
sciarid flies and, under some conditions, in Nasonia and C. elegans.
The diverse mechanisms include heterochromatinization or elimi-
nation of one chromosome or an entire parental set (9, 22, 27,
54). Moreover, in Nasonia vitripensis, several lines of evidence
have demonstrated that sex determination depends on the
imprinted expression of the transformer gene (a member of the
cascade of genes that control insect sex determination) (55). In this
wasp, diploid female development depends on the transcription of
the zygotic transformer gene (Nvtra). This gene appears to be
maternally repressed in unfertilized eggs (which develop into hap-
loid males); in contrast, the presence of a paternal genome triggers
a peak of Nvtra zygotic transcription in fertilized eggs, rendering
female embryos (55, 56).
Drosophila provides an example of how the imprinting of one
chromosome (the Y) can have diverse effects: (1) on the expression
of individual Y-linked transgenes and (2) on the entire gene
expression of other chromosome. As previously discussed, studies
of chromosome Y-inserted transgenes have shown differential
expression according to their parental origin (18). In addition, a
maternally inherited Y chromosome can rescue mutations that
disrupt X chromosome dosage compensation (92). Normally, this
is a mechanism that compensates for unequal X chromosome dos-
age between males (with one X) and females (with two X) by
upregulating the transcription of the genes of the single male X
chromosome. This process is mediated by the chromatin-modifying
male-specific lethal (MSL) complex; mutations of some MSL com-
ponents (roX1 and roX2) disrupt expression upregulation and
cause male-specific lethality. However, a maternally inherited Y
chromosome can suppress this mutant male lethality by elevating
the expression of X-linked genes (92). How imprinting of the
Y-chromosome can rescue a mutant phenotype in other chromo-
some is very intriguing, because no endogenous imprinted genes
that could mediate this effect have been reported in Drosophila.
Recent years have seen tremendous advances in large-scale
genotyping, phenotyping and mapping studies, changing our view
of POEs. Until recently, imprinted expression appeared to be lim-
ited to a few genes (http://www.har.mrc.ac.uk/research/genomic_
imprinting/) (7). However, recent large-scale studies have shown
that imprinted gene expression and imprinted epigenetic marks are
more frequent and widespread than previously thought (57–62).
This conclusion is further supported by observations of imprinting
effects in methylation and expression of transgenes inserted in

multiple locations in diverse species (4, 18, 20, 21, 63). Moreover,
quantitative trait loci (QTL) analyses have revealed that POEs are
common and can affect many traits and diseases (http://igc.otago.
ac.nz/home.html; Richard Mott, personal communication) (59);
these studies have also found complex patterns of POEs (64, 65).
Therefore, large-scale expression studies are very promising for
uncovering novel POEs; however, as previously discussed, in many
cases the nature of these POEs and whether they are originated by
imprinting remains to be determined. An example is found in
chicken: while no conclusive evidence of imprinted gene expres-
sion has been reported in birds (66), QTL mapping has identified
several POEs (67). Their nature is still unknown, but some POE
candidate regions map in or close to chicken orthologues of mam-
malian imprinted genes (67). Moreover, these avian genes are clus-
tered and replicate asynchronously, as observed in their mammalian
orthologues (68).
In summary, while numerous imprinting effects have been
known for a long time, the list is still growing. Figure 1 does not
intend to be a comprehensive catalog, but illustrates the extreme
diversity of imprinting and imprinting-candidate POEs and their
widespread distribution in sexually reproducing organisms.
6. The Common
Theme: Paternal
and Maternal
Chromosomes Although not all imprinting effects have been studied with the
Bear Different same depth, most have been associated with parent-of-origin
dependent epigenetic marks, such as DNA methylation and histone
Epigenetic Marks
modifications, chromatin condensation (euchromatin vs. hetero-
chromatin), and chromatin remodeling factors (such as noncoding
RNAs and CTCF). Many of the chromatin modifications involved
in mammalian or plant imprinting (7, 8) are also associated with a
variety of imprinting effects in other organisms. For instance,
imprinted heterochromatinization of chromosomes is observed
during sex determination of lecanoid coccids and Hypothenemus
hampei, as well as in mammalian imprinted X-inactivation (39, 40,
44, 45). In coccids, the paternal DNA is hypomethylated in both
males and females. Additionally, in males, the paternal chromo-
somes are heterochromatized and condensed, accumulating
nuclear-resistant chromatin, Heterochromatin Protein 1 (HP1)
protein and repressive histone modifications (9, 69, 70). In Sciara
ocellaris, chromosome elimination in early germ nuclei and male
meiosis is accompanied by differential condensation and differen-
tial histone H3 and H4 methylation and acetylation between
maternal and paternal chromosomes (71, 72). In Drosophila, het-
erochromatin imprints affect gene expression (16). Mutant analyses
have shown that this imprinting depends on the products of

Suppressor of variegation (Su(var)) and trithorax (trx-G) chromatin
modifier genes (73). CTCF is a common player in both fruit fly
and mammalian imprinting, although the role of DNA methyla-
tion in Drosophila (74) has not been clarified yet. These data indi-
cate that histone methylation and acetylation, as well as other
well-known mammalian epigenetic marks, contribute to the imprint
of parental chromosomes also in insects.
The origin of parent-of-origin dependent epigenetic differ-
ences (and the subsequent imprinting effects) resides on a simple
fact: maternal and paternal chromosomes can retain part of the
chromatin configuration assembled during the formation of two
extremely specialized cell types: the gametes. The epigenetic repro-
gramming that occurs during oogenesis and spermatogenesis has
been extensively studied in mammals and plants (75, 76) and also
occurs in other organisms. For instance, different patterns of
histone marks (such as histone H3 lysine 9 methylation) are incor-
porated during spermatogenesis vs. oogenesis in coccids (mealybugs)
and C. elegans (41, 77). In many organisms, tight chromosome
condensation and replacement of histones for protamines occurs
only during spermatogenesis. Still, some histones can be retained
in the mature sperm and transmit imprinted epigenetic marks to
the embryo, constituting a potential source of POEs (78). After
fertilization, two pronuclei with very different chromatin organiza-
tion encounter in the zygote. Therefore and in order to generate a
totipotent embryo, many epigenetic differences between the paren-
tal genomes are removed and new ones established (79, 80). For
instance, recent reports of imprinted gene expression not only in
endosperm, but also in plant embryos, suggest that imprint erasure
and resetting mechanisms are present in plants (7). Through these
cyclic reprogramming events, the imprints are reset and established
in every generation (6, 75, 76). Nevertheless, transgenerational
POEs have been observed, suggesting incomplete gametic erasure
of certain epigenetic marks (81).
The links between gametic epigenetic differences, imprinting
and embryonic reprogramming can be illustrated with an example
found in C. elegans: after fertilization, the paternal pronucleus lacks
all tested histone modifications, but rapidly accumulates them as it
decondenses (41). As a result, maternal and paternal chromosomes
end up displaying similar histone marks in the early embryos. All
but the X chromosome transmitted by the sperm, which, as previ-
ously discussed, remains devoid of activating histone modifications
(methylated H3/Lys4 and acetylated H3) (41). This epigenetic
imprint might be a direct consequence of the transcriptional and
epigenetic history of the C. elegans X: this chromosome is devoid
of spermatogenesis-specific genes and, therefore, it remains silent
during sperm development, while it is actively transcribed during
oogenesis (41, 82). Interestingly, the paternal epigenetic imprint
of the X-chromosome is transitory: it is maintained in the embryo

through several cell divisions, but histone modifications accumu-
late at later stages and the imprint is no longer observed past the
20-cell stage (41).
This marked, although transient, epigenetic asymmetry
between the two parental genomes is also found in other organ-
isms during early embryogenesis and has two important implica-
tions: first, reprogramming involves unique mechanisms for each
of the two parental genomes (79). For instance, in paternal pronu-
clei tightly condensed and packed around protamines, these must
be replaced by histones and the chromosomes decondensed.
Preimplantation reprogramming leads to chromatin equalization
compatible with embryo development, but the differences between
the paternal and maternal chromosomes remodeling processes can
also result in the acquisition of distinct epigenetic marks. Second,
this epigenetic asymmetry manifests in a number of POEs: nuclear
separation (53), pericentric heterochromatin (39) and asynchro-
nous replication (83, 84) distinguish the two parental sets during
mammalian early embryonic stages. Although these POEs fade
after a few cell divisions, some POEs can remain at later stages; for
instance, asynchronous replication (resulting in early replication of
paternal alleles respect to maternal ones) is first observed between
paternal and maternal chromosomes after fertilization, but becomes
restricted to imprinted loci after a few cell divisions and then for
life (84). Moreover, early embryonic POEs could contribute to the
establishment of novel epigenetic imprints that may result in
imprinting effects at later stages.
7. The Lessons
of Comparative
Analyses:
Imprinting Effects In spite of the vast embryonic reprogramming, epigenetic differ-
are Widespread ences between maternal and paternal chromosomes are present in
and Diverse, yet somatic cells. The specificity of many of these parent-of-origin
dependent marks and the key roles they play in the imprinting of
They Can Be
crucial processes suggest that, besides being derived from gametic
Interrelated
epigenetic marks or early embryonic POEs, they have been selected
due to their functional relevance. Several theories about the selec-
tive forces operating at imprinting effects on expression have been
proposed (reviewed in (85)), although none of them is sufficient to
explain the evolution of all imprinted genes. Moreover, the diver-
sity of imprinting effects suggests that different selective pressures
have operated over different loci and organisms. For instance,
imprinting has no known expression effects in karyotypically nor-
mal Drosophila genes, only on PEV of rearranged and inserted
genes; therefore, Drosophila chromosomes are imprinted, but the

role and origin of this imprinting might have nothing to do with
the control gene expression as in mammals (16). Nevertheless, fruit
fly and mammalian imprinting share some common features: for
instance, the CTCF chromatin insulator contributes to both mam-
malian and Drosophila imprinting control (16, 73, 74, 86) and the
physical size of several Drosophila imprinted regions (including the
imprinted mini-X chromosome) is comparable to the size of some
mammalian imprinted clusters (10). Besides placental and marsu-
pial mammals, imprinted gene expression has been found in plants,
but not in other animals (87); therefore, imprinting is believed to
have evolved independently in those two groups (88). However,
mammals and plants share several common principles of imprinting
regulation (such as DNA methylation and other epigenetic marks)
(88) and even other imprinting effects (89–91). Therefore, imprint-
ing shows recurrent features among distantly related species and
imprinted phenomenon: as an additional example, CTCF has been
reported to contribute to asynchronous replication (92), intra- and
interchromosomal interactions (47, 93) and imprinted expression
regulation in diverse organisms (48, 74, 86). Furthermore, the
phylogenetic distribution of parent-of-origin dependent epigenetic
differences (Fig. 1) indicates that they are an ancient phenomenon
in sexually reproducing organisms. This suggests that, along with
specific selective pressures in different organisms, there might be
common selective forces operating on paternal and maternal epig-
enomes (85, 94, 95).
In summary, the development of two highly differentiated cell
types that participate in fertilization culminates with the zygotic
encounter of two chromosome sets with dramatically different
chromatin configurations. Although most epigenetic differences
must be erased and reset to generate totipotent cells, natural selec-
tion may also operate to maintain some of them for particular func-
tions (Fig. 1). Any epigenetic difference between maternally and
paternally inherited chromosomes offers the possibility of an
imprinting effect: the substrate is there. While different selective
forces have led to a large spectrum of imprinting effects (Fig. 1),
very diverse organisms, despite their different evolutionary paths,
share several imprinting effects and, maybe, common selective
pressures. Moreover, disparate imprinting effects are observed
within the same species. It is possible that intraspecific imprinting
diversity is a common occurrence and it is just our inability to
detect certain effects (or our lack of curiosity or awareness) that
keep them out of our sight. Future studies will expand our knowl-
edge of imprinting in additional species, discover novel or complex
imprinted phenomena, and unveil the links between diverse
imprinting effects.
Acknowledgments
I apologize to colleagues whose original research papers could not

be cited due to space limitations. I would like to thank Elena Becker
Barroso and Jose Javier García Ramírez for their helpful comments
to this manuscript. I am also grateful to Francisco R. Jiménez Díaz
for technical support and to the Consejería de Educación y Ciencia
(ref. PPII10-0259-4347) for financial support.
References
1. Crouse HV (1960) The controlling element 12. Deakin JE, Chaumeil J, Hore TA, Marshall
in Sex chromosome behavior in sciara. Graves JA (2009) Unravelling the evolution-
Genetics 45:1429–1443 ary origins of X chromosome inactivation in
2. Wolf JB, Wade MJ (2009) What are maternal mammals: insights from marsupials and mono-
effects (and what are they not)? Philos Trans tremes. Chromosome Res 17:671–685
R Soc Lond B Biol Sci 364:1107–1115 13. Takagi N, Sasaki M (1975) Preferential inacti-
3. Hager R, Cheverud JM, Wolf JB (2008) vation of the paternally derived X chromo-
Maternal effects as the cause of parent-of-ori- some in the extraembryonic membranes of
gin effects that mimic genomic imprinting. the mouse. Nature 256:640–642
Genetics 178:1755–1762 14. Wang X, Soloway PD, Clark AG (2010)
4. Wang AD, Sharp NP, Spencer CC, Tedman- Paternally biased X inactivation in mouse neo-
Aucoin K, Agrawal AF (2009) Selection, epista- natal brain. Genome Biol 11:R79
sis, and parent-of-origin effects on deleterious 15. Chadwick LH, Willard HF (2005) Genetic
mutations across environments in Drosophila and parent-of-origin influences on X chromo-
melanogaster. Am Nat 174:863–874 some choice in Xce heterozygous mice.
5. Wittkopp PJ, Haerum BK, Clark AG (2006) Mamm Genome 16:691–699
Parent-of-origin effects on mRNA expression in 16. Menon DU, Meller VH (2010) Germ line
Drosophila melanogaster not caused by genomic imprinting in Drosophila: Epigenetics in
imprinting. Genetics 173:1817–1821 search of function. Fly (Austin) 4:48–52
6. Bartolomei MS, Ferguson-Smith AC (2011) 17. Kuhn DT, Packert G (1988) Paternal imprint-
Mammalian genomic imprinting. Cold Spring ing of inversion Uab1 causes homeotic trans-
Harb Perspect Biol 3(7). pii: a002592 formations in Drosophila. Genetics 118:
7. Raissig MT, Baroux C, Grossniklaus U (2011) 103–107
Regulation and flexibility of genomic imprinting 18. Maggert KA, Golic KG (2002) The Y chro-
during seed development. Plant Cell 23:16–26 mosome of Drosophila melanogaster exhibits
8. Brun LO, Stuart J, Gaudichon V, Aronstein chromosome-wide imprinting. Genetics
K, French-Constant RH (1995) Functional 162:1245–1258
haplodiploidy: a mechanism for the spread of 19. Dorn R, Krauss V, Reuter G, Saumweber H
insecticide resistance in an important interna- (1993) The enhancer of position-effect varie-
tional insect pest. Proc Natl Acad Sci USA 92: gation of Drosophila, E(var)3-93D, codes for
9861–9865 a chromatin protein containing a conserved
9. Khosla S, Mendiratta G, Brahmachari V domain common to several transcriptional
(2006) Genomic imprinting in the mealybugs. regulators. Proc Natl Acad Sci USA
Cytogenet Genome Res 113:41–52 90:11376–11380
10. Anaka M, Lynn A, McGinn P, Lloyd VK 20. Preis JI, Downes M, Oates NA, Rasko JE,
(2009) Genomic imprinting in Drosophila Whitelaw E (2003) Sensitive flow cytometric
has properties of both mammalian and insect analysis reveals a novel type of parent-of-ori-
imprinting. Dev Genes Evol 219:59–66 gin effect in the mouse genome. Curr Biol
11. Lloyd VK, Sinclair DA, Grigliatti TA (1999) 13:955–959
Genomic imprinting and position-effect varie- 21. Sha K, Fire A (2005) Imprinting capacity of
gation in Drosophila melanogaster. Genetics gamete lineages in Caenorhabditis elegans.
151:1503–1516 Genetics 170:1633–1652
22. Goday C, Esteban MR (2001) Chromosome fluorescence in situ hybridization. Mutat Res
elimination in sciarid flies. Bioessays 23: 372:269–278
242–250 36. Naumova AK, Leppert M, Barker DF,
23. Breeuwer JA, Werren JH (1990) Morgan K, Sapienza C (1998) Parental origin-
Microorganisms associated with chromosome dependent, male offspring-specific transmis-
destruction and reproductive isolation between sion-ratio distortion at loci on the human X
two insect species. Nature 346:558–560 chromosome. Am J Hum Genet 62:
24. Nur U, Werren JH, Eickbush DG, Burke WD, 1493–1499
Eickbush TH (1988) A “selfish” B chromo- 37. Croteau S, Andrade MF, Huang F, Greenwood
some that enhances its transmission by elimi- CM, Morgan K, Naumova AK (2002)
nating the paternal genome. Science Inheritance patterns of maternal alleles in
240:512–514 imprinted regions of the mouse genome at
25. Dobson SL, Tanouye MA (1998) Evidence for different stages of development. Mamm
a genomic imprinting sex determination mech- Genome 13:24–29
anism in Nasonia vitripennis (Hymenoptera; 38. Pardo-Manuel de Villena F, de la Casa-Esperon
Chalcidoidea). Genetics 149:233–242 E, Briscoe TL, Sapienza C (2000) A genetic
26. Werren JH, Stouthamer R (2003) PSR (pater- test to determine the origin of maternal trans-
nal sex ratio) chromosomes: the ultimate selfish mission ratio distortion. Meiotic drive at the
genetic elements. Genetica 117:85–101 mouse Om locus. Genetics 154:333–342
27. Prahlad V, Pilgrim D, Goodwin EB (2003) 39. Puschendorf M, Terranova R, Boutsma E,
Roles for mating and environment in C. Mao X, Isono K, Brykczynska U, Kolb C,
elegans sex determination. Science 302: Otte AP, Koseki H, Orkin SH, van Lohuizen
1046–1049 M, Peters AH (2008) PRC1 and Suv39h
28. Komaru A, Kawagishi T, Konishi K (1998) specify parental asymmetry at constitutive het-
Cytological evidence of spontaneous andro- erochromatin in early mouse embryos. Nat
genesis in the freshwater clam Corbicula leana Genet 40:411–420
Prime. Dev Genes Evol 208:46–50 40. Kitsberg D, Selig S, Brandeis M, Simon I,
29. Ishibashi R, Ookubo K, Aoki M, Utaki M, Keshet I, Driscoll DJ, Nicholls RD, Cedar H
Komaru A, Kawamura K (2003) Androgenetic (1993) Allele-specific replication timing of
reproduction in a freshwater diploid clam imprinted gene regions. Nature 364:459–463
Corbicula fluminea (Bivalvia: Corbiculidae). 41. Bean CJ, Schaner CE, Kelly WG (2004)
Zoolog Sci 20:727–732 Meiotic pairing and imprinted X chromatin
30. Baker BS (1975) Paternal loss (pal): a meiotic assembly in Caenorhabditis elegans. Nat
mutant in Drosophila melanogaster causing Genet 36:100–105
loss of paternal chromosomes. Genetics 42. Shiu PK, Raju NB, Zickler D, Metzenberg RL
80:267–296 (2001) Meiotic silencing by unpaired DNA.
31. Szabad J, Mathe E, Puro J (1995) Horka, a Cell 107:905–916
dominant mutation of Drosophila, induces 43. Paigen K, Szatkiewicz JP, Sawyer K, Leahy N,
nondisjunction and, through paternal effect, Parvanov ED, Ng SH, Graber JH, Broman
chromosome loss and genetic mosaics. KW, Petkov PM (2008) The recombinational
Genetics 139:1585–1599 anatomy of a mouse chromosome. PLoS
32. Szalontai T, Gaspar I, Belecz I, Kerekes I, Genet 4:e1000119
Erdelyi M, Boros I, Szabad J (2009) HorkaD, 44. Ng SH, Madeira R, Parvanov ED, Petros LM,
a chromosome instability-causing mutation in Petkov PM, Paigen K (2009) Parental origin
Drosophila, is a dominant-negative allele of of chromosomes influences crossover activity
Lodestar. Genetics 181:367–377 within the Kcnq1 transcriptionally imprinted
33. Lewis EB, Gencarella W (1952) Claret and domain of Mus musculus. BMC Mol Biol
non-disjunction in Drosophila melanogaster. 10:43
Genetics 37:600–601 45. Billings T, Sargent EE, Szatkiewicz JP, Leahy
34. Gao G, Cheng Y, Wesolowska N, Rong YS N, Kwak IY, Bektassova N, Walker M, Hassold
(2011) Paternal imprint essential for the T, Graber JH, Broman KW, Petkov PM
inheritance of telomere identity in Drosophila. (2010) Patterns of recombination activity on
Proc Natl Acad Sci USA 108:4932–4937 mouse chromosome 11 revealed by high reso-
35. Baulch JE, Lowe XR, Bishop JB, Wyrobek AJ lution mapping. PLoS One 5:e15340
(1996) Evidence for a parent-of-origin effect 46. LaSalle JM, Lalande M (1996) Homologous
on sperm aneuploidy in mice carrying association of oppositely imprinted chromo-
Robertsonian translocations as analyzed by somal domains. Science 272:725–728
47. Ling JQ, Li T, Hu JF, Vu TH, Chen HL, Qiu imprinting by transcriptome sequencing. Curr
XW, Cherry AM, Hoffman AR (2006) CTCF Biol 18:1735–1741
mediates interchromosomal colocalization 59. Gregg C, Zhang J, Weissbourd B, Luo S,
between Igf2/H19 and Wsb1/Nf1. Science Schroth GP, Haig D, Dulac C (2010) High-
312:269–272 resolution analysis of parent-of-origin allelic
48. Yang J, Corces VG (2011) Chromatin insula- expression in the mouse brain. Science
tors: a role in nuclear organization and gene 329:643–648
expression. Adv Cancer Res 110:43–76 60. Wolff P, Weinhofer I, Seguin J, Roszak P,
49. Sandhu KS, Shi C, Sjolinder M, Zhao Z, Beisel C, Donoghue MT, Spillane C,
Gondor A, Liu L, Tiwari VK, Guibert S, Nordborg M, Rehmsmeier M, Kohler C
Emilsson L, Imreh MP, Ohlsson R (2009) (2011) High-Resolution Analysis of Parent-
Nonallelic transvection of multiple imprinted of-Origin Allelic Expression in the Arabidopsis
loci is organized by the H19 imprinting con- Endosperm. PLoS Genet 7:e1002126
trol region during germline development. 61. Gehring M, Bubb KL, Henikoff S (2009)
Genes Dev 23:2598–2603 Extensive demethylation of repetitive ele-
50. Terranova R, Yokobayashi S, Stadler MB, Otte ments during seed development underlies
AP, van Lohuizen M, Orkin SH, Peters AH gene imprinting. Science 324:1447–1451
(2008) Polycomb group proteins Ezh2 and 62. Choufani S, Shapiro JS, Susiarjo M, Butcher
Rnf2 direct genomic contraction and DT, Grafodatskaya D, Lou Y, Ferreira JC,
imprinted repression in early mouse embryos. Pinto D, Scherer SW, Shaffer LG, Coullin P,
Dev Cell 15:668–679 Caniggia I, Beyene J, Slim R, Bartolomei MS,
51. Gribnau J, Hochedlinger K, Hata K, Li E, Weksberg R (2011) A novel approach
Jaenisch R (2003) Asynchronous replication identifies new differentially methylated regions
timing of imprinted loci is independent of (DMRs) associated with imprinted genes.
DNA methylation, but consistent with differ- Genome Res 21:465–476
ential subnuclear localization. Genes Dev 17: 63. Sapienza C, Peterson AC, Rossant J, Balling R
759–773 (1987) Degree of methylation of transgenes is
52. Kubai DF (1982) Meiosis in Sciara copro- dependent on gamete of origin. Nature
phila: structure of the spindle and chromo- 328:251–254
some behavior during the first meiotic division. 64. Wolf JB, Cheverud JM, Roseman C, Hager R
J Cell Biol 93:655–669 (2008) Genome-wide analysis reveals a com-
53. Mayer, W., Smith, A., Fundele, R., and Haaf, plex pattern of genomic imprinting in mice.
T (2000) Spatial separation of parental PLoS Genet 4:e1000091
genomes in preimplantation mouse embryos. 65. Cheverud JM, Lawson HA, Fawcett GL,
J Cell Biol 148:629–634 Wang B, Pletscher LS, R Fox A, Maxwell TJ,
54. Reed KM, Werren JH (1995) Induction of Ehrich TH, Kenney-Hunt JP, Wolf JB,
paternal genome loss by the paternal-sex-ratio Semenkovich CF (2011) Diet-dependent
chromosome and cytoplasmic incompatibility genetic and genomic imprinting effects on
bacteria (Wolbachia): a comparative study of obesity in mice. Obesity (Silver Spring) 19:
early embryonic events. Mol Reprod Dev 160–170
40:408–418 66. Nolan CM, Killian JK, Petitte JN, Jirtle RL
55. Verhulst EC, Beukeboom LW, van de Zande (2001) Imprint status of M6P/IGF2R and
L (2010) Maternal control of haplodiploid IGF2 in chickens. Dev Genes Evol 211:
sex determination in the wasp Nasonia. 179–183
Science 328:620–623 67. Tuiskula-Haavisto M, Vilkki J (2007) Parent-
56. Beukeboom LW, van de Zande L (2010) of-origin specific QTL–a possibility towards
Genetics of sex determination in the haplodip- understanding reciprocal effects in chicken
loid wasp Nasonia vitripennis (Hymenoptera: and the origin of imprinting. Cytogenet
Chalcidoidea). J Genet 89:333–339 Genome Res 117:305–312
57. Luedi PP, Dietrich FS, Weidman JR, Bosko 68. Dunzinger U, Haaf T, Zechner U (2007)
JM, Jirtle RL, Hartemink AJ (2007) Conserved synteny of mammalian imprinted
Computational and experimental identification genes in chicken, frog, and fish genomes.
of novel human imprinted genes. Genome Cytogenet Genome Res 117:78–85
Res 17:1723–1730 69. Bongiorni S, Cintio O, Prantera G (1999)
58. Babak T, Deveale B, Armour C, Raymond C, The relationship between DNA methylation
Cleary MA, van der Kooy D, Johnson JM, and chromosome imprinting in the coccid
Lim LP (2008) Global survey of genomic Planococcus citri. Genetics 151:1471–1478
70. Bongiorni S, Prantera G (2003) Imprinted Early Caenorhabditis elegans Embryos Can
facultative heterochromatization in mealy- Be Established by Gene Activity in the Parental
bugs. Genetica 117:271–279 Germ Cells. PLoS Genet 7:e1001391
71. Goday C, Ruiz MF (2002) Differential acety- 83. Ferreira J, Carmo-Fonseca M (1997) Genome
lation of histones H3 and H4 in paternal and replication in early mouse embryos follows a
maternal germline chromosomes during defined temporal and spatial order. J Cell Sci
development of sciarid flies. J Cell Sci 115: 110(Pt 7):889–897
4765–4775 84. May A, Reifenberg K, Zechner U, Haaf T
72. Greciano PG, Goday C (2006) Methylation of (2008) Asynchronous replication dynamics of
histone H3 at Lys4 differs between paternal and imprinted and non-imprinted chromosome
maternal chromosomes in Sciara ocellaris ger- regions in early mouse embryos. Exp Cell Res
mline development. J Cell Sci 119:4667–4677 314:2788–2795
73. Joanis V, Lloyd VK (2002) Genomic imprint- 85. de la Casa-Esperon E, Sapienza C (2003)
ing in Drosophila is maintained by the prod- Natural selection and the evolution of genome
ucts of Suppressor of variegation and trithorax imprinting. Annu Rev Genet 37:349–370
group, but not Polycomb group, genes. Mol 86. Engel N, Thorvaldsen JL, Bartolomei MS
Genet Genomics 268:103–112 (2006) CTCF binding sites promote transcrip-
74. MacDonald WA, Menon D, Bartlett NJ, tion initiation and prevent DNA methylation
Sperry GE, Rasheva V, Meller V, Lloyd VK on the maternal allele at the imprinted H19/
(2010) The Drosophila homolog of the mam- Igf2 locus. Hum Mol Genet 15:2945–2954
malian imprint regulator. CTCF, maintains 87. Renfree MB, Hore TA, Shaw G, Graves JA,
the maternal genomic imprint in Drosophila Pask AJ (2009) Evolution of genomic imprint-
melanogaster. BMC Biol 8:105 ing: insights from marsupials and monotremes.
75. Feng S, Jacobsen SE, Reik W (2010) Annu Rev Genomics Hum Genet 10:241–262
Epigenetic reprogramming in plant and ani- 88. Scott RJ, Spielman M (2006) Genomic
mal development. Science 330:622–627 imprinting in plants and mammals: how life
76. Kota SK, Feil R (2010) Epigenetic transitions history constrains convergence. Cytogenet
in germ cell development and meiosis. Dev Genome Res 113:53–67
Cell 19:675–686 89. Tourte Y, Kuligowski-Andres J, Barbier-
77. Bongiorni S, Pugnali M, Volpi S, Bizzaro D, Ramond C (1980) Different behaviour of
Singh PB, Prantera G (2009) Epigenetic paternal and maternal genomes during
marks for chromosome imprinting during embryogenesis in the fern, Marsilea (author’s
spermatogenesis in coccids. Chromosoma transl). Eur J Cell Biol 21:28–36
118:501–512 90. Kermicle JL, Alleman M (1990) Gametic
78. Hammoud SS, Nix DA, Zhang H, Purwar J, imprinting in maize in relation to the angio-
Carrell DT, Cairns BR (2009) Distinctive sperm life cycle. Dev Suppl, 9–14.
chromatin in human sperm packages genes for 91. Vielle-Calzada JP, Baskar R, Grossniklaus U
embryo development. Nature 460:473–478 (2000) Delayed activation of the paternal
79. Burton A, Torres-Padilla ME (2010) genome during seed development. Nature
Epigenetic reprogramming and development: 404:91–94
a unique heterochromatin organization in the 92. Bergstrom R, Whitehead J, Kurukuti S,
preimplantation mouse embryo. Brief Funct Ohlsson R (2007) CTCF regulates asynchro-
Genomics 9:444–454 nous replication of the imprinted H19/Igf2
80. de la Casa-Esperon E, Roy A (2009) domain. Cell Cycle 6:450–454
Mammalian gametogenesis to implantation. 93. Donohoe ME, Silva SS, Pinter SF, Xu N, Lee
In: Reproduction and Development Biology, JT (2009) The pluripotency factor Oct4 inter-
Encyclopedia of Biological, Physiological and acts with Ctcf and also controls X-chromosome
Health Sciences, Encyclopedia of Life pairing and counting. Nature 460:128–132
Support Systems(EOLSS). Eolss Publishers, 94. Pardo-Manuel de Villena F, de la Casa-
Oxford ,UK Esperon E, Sapienza C (2000) Natural selec-
81. Han Z, Mtango NR, Patel BG, Sapienza C, tion and the function of genome imprinting:
Latham KE (2008) Hybrid vigor and trans- beyond the silenced minority. Trends Genet
generational epigenetic effects on early mouse 16:573–579
embryo phenotype. Biol Reprod 79:638–648 95. Paldi A (2003) Genomic imprinting: could
82. Arico JK, Katz DJ, van der Vlag J, Kelly WG the chromatin structure be the driving force?
(2011) Epigenetic Patterns Maintained in Curr Top Dev Biol 53:115–138
96. Morison IM, Ramsay JP, Spencer HG (2005) 99. Surani MA, Barton SC, Norris ML (1984)
A census of mammalian imprinting. Trends Development of reconstituted mouse eggs
Genet 21:457–465 suggests imprinting of the genome during
97. Barton SC, Surani MA, Norris ML (1984) gametogenesis. Nature 308:548–550
Role of paternal and maternal genomes in 100. Johnston PG, Watson CM, Adams M, Paull
mouse development. Nature 311:374–376 DJ (2002) Sex chromosome elimination, X
98. McGrath J, Solter D (1984) Completion of chromosome inactivation and reactivation in
mouse embryogenesis requires both the the southern brown bandicoot Isoodon obe-
maternal and paternal genomes. Cell sulus (Marsupialia: Peramelidae). Cytogenet
37:179–183 Genome Res 99:119–124
INDEX
A DNA methylation ....................................... 5, 22, 62, 69, 90,

145, 149, 150, 159, 160, 187–197, 202, 221,
Alignment ....................................................83–86, 103, 194 232–234, 244, 251, 252, 258–259, 286, 287, 289
Alignment algorithms .......................................................85 Dosage compensation ...................................... 220, 222, 285
Allele-specific ..................................................22, 50, 84, 85,
90, 91, 159, 160, 166–171, 221, 233, 252, 257 E
Allelic imbalance (AI) ................................................. 80–87
Androgenone ................................................4, 6, 7, 9, 11–13 Ecotypes .......................................................... 233, 235–237
Annotation ............................... 195, 252, 254–256, 258–259 Electrofusion ........................................................7, 9, 11, 17
Arabidopsis...................................................... 232–238, 244 Electroporation.........................................139–142, 144, 145
Association frequency .......................174, 175, 178, 180–183 Embryonic stem cells (ESCs) ..........................21, 22, 26, 33,
34, 39, 43, 46
B Epigenome ......................................... 70, 245, 252, 255, 289
EST ................................................................................. 254
Bioconductor ..............................................74, 256, 258, 259
Eutherians ............................................................... 263–271
Bioinformatics ..........................................150, 188, 194, 252
Extraembryonic stem cells .................................................49
Biopsy .......................................................................... 40, 46
Bismark ........................................................................... 194 F
Bisulfite sequencing ........................... 90, 150, 155, 187–197,
207, 234, 238–241, 244, 245, 256, 258 Feeder cells .......................................... 24, 26–28, 32–34, 37,
Blastocyst...........................5, 22, 49, 51, 53, 54, 58, 201–209 39, 42–45, 52, 56, 57
Bone marrow ......................................................... 25, 34–36 Fibroblasts ...........................................23–25, 27–34, 39–47,
50, 52–53, 139, 157, 160, 162, 181, 183
C Freeze stocks.....................................................30, 34, 40, 44
Cauda epididymes .............................................................12 G

Chip-seq ............................................................ 92, 103, 111
Chromatin immunoprecipitation (ChIP) .......... 91, 159–171 Galaxy..................................... 93, 96, 98–101, 103–105, 130
Chromosome conformation capture (3C) ............... 173–185 Gametogenesis ............................................................ 5, 280
Computational prediction ....................89, 91, 116, 120, 124 Gametophyte ................................................................... 232
CpG islands ..................................... 100, 101, 109, 111–113, Generalized linear models (GLM) .......................... 121–125
155, 156, 187, 254, 255, 260, 261 Gene targeting ......................................................... 140–142
Cre/loxp................................................................... 137–145 Gene transduction ....................................................... 21–48
Cross-linking ...........................................160, 162, 165, 169, Genome evolution ........................................................... 263
170, 173, 174, 176–177, 181, 223–225 Genomic coordinates ......................... 84, 86, 93–96, 98–101,
CTCF.............................................. 100–103, 111, 113, 114, 103, 104, 106, 107, 114, 115, 123, 124, 129
131, 132, 160, 169, 286, 287, 289 Gynogenone .......................................................4, 5, 7, 9, 11
D H
Databases................................................. 100–102, 252–257 Hanging drop ....................................................................15
Data mining .............................................................. 89–132 Hidden Markov model (HMM) ..................................... 103
Defense hypothesis ...................................266, 267, 269, 271 High-throughput sequencing ..............................70, 74, 194,
Differentially methylated regions (DMRs) ................ 69, 75, 212, 214, 216, 217, 245, 258
76, 159, 160, 166, 167, 169–171, Histone modifications .....................................103, 104, 131,
202, 254, 259, 268–271 132, 159, 160, 166–170, 220, 233,
Differential methylation .......................74–76, 233, 268, 270 251, 252, 286–288
DOI 10.1007/978-1-62703-011-3, © Springer Science+Business Media, LLC 2012
295
GENOMIC IMPRINTING
296 Index
I Plants...............................................................231–246, 278,
279, 286, 287, 289
Immunoprecipitation .............. 69–77, 91, 149–171, 219–227 Pluripotent stem cells ............................................ 21–47, 49
Immunopurification ..........................................................61 Polar body.................................................................... 10, 14
Induced pluripotent stem cells (iPSCs) ....................... 21–47 Pollination ....................................................... 236, 241, 243
Inter-species ........................................................................4 Polycomb group proteins ................................................. 233
Inter-strain .................................................................... 4, 90 Primordial germ cells ........................................... 61–65, 162
Intracytoplasmic sperm injection (ICSI) ............. 6, 8, 11–14 Pronuclear transfer .................................................... 4, 9–11
Intramolecular ligation .............................173, 174, 177, 185
Inverted microscope .....................................7, 9, 15, 16, 139 Q
K Quantile normalization ............................................... 74, 75
Quantitative real-time PCR .................................... 166–169
Karyoplast fusion ........................................................... 9, 17
Kinship hypothesis ........................... 265, 266, 268, 270, 271 R
L Recombination ........................................137–139, 143–145,
280, 282, 283
Long non-coding RNAs ......................................... 219–227 Reduced representational bisulfite sequencing...................90
Repetitive elements ...................................92, 188, 254, 255,
M
258, 260, 261
Marmosets ..................................................23, 27, 37–39, 46 Reprogramming ................................. 3, 22, 39, 61, 287, 288
Marsupials ................................................263–270, 279, 289 Retrotransposition ................................................... 269, 271
Mef feeders ...........................................27, 39, 51–53, 56, 57 Retrovirus ...................................... 24, 26, 28, 31, 32, 35–37,
Mesenchymal stem cells (MSCs) .................... 25–26, 34–37 39–42, 45, 47, 266
Methylation profiling .................................................. 69–77 Ribonucleoprotein particles (RNPs) ................ 214, 220–222
Microarray .................................................70, 74–76, 79, 90, RNA immunoprecipitation (RIP)-seq..................... 222–223
149–151, 155, 160, 171 RNA-seq ............................................................... 80–83, 86
Microforge............................................................... 7, 15, 16 Round spermatid injection (ROSI) ............................. 13, 14
Micromanipulator ......................................................... 7, 15
Micro-RNA (miRNAs).....................................40, 100, 101, S
111, 113, 114, 131, 132, 217 Sliding window analysis............................................... 75, 76
Model training .........................106–109, 114–117, 120–122 snoRNAs ..................................................219, 253, 257, 269
Modifiers ................................................................. 5, 6, 287 SNPs................................................................80–87, 90, 91,
Monotremes .................................................... 263–267, 269 160, 168, 253–255
Sonication.................................... 71, 77, 151, 153, 155–157,
N
160, 163, 164, 169, 226
Nextgen sequencing..................................................... 80, 82 Species-specificity .............................................. 91, 260, 261
Nimblegen microarrays................................................ 70, 74 Spermatocyte nuclear transfer ......................................... 7, 8
Noncoding RNAs (ncRNAs) ......................90, 91, 211–217, Sperm injection ..............................................6, 7, 11–13, 17
219–227, 253, 257, 286 Spindle ................................................... 7, 8, 12, 15, 16, 232
Novoalign .................................................................... 83–85 Statistics ..................................... 84, 252, 256, 258, 259, 261
Stereo microscope................................................................7
P Superovulation.....................................................................8
Paired-end data .................................................................85
T
Parental conflict hypothesis .................................... 232, 265,
266, 268, 270, 271 TAMERE ....................................................... 138, 140–145
Parent-of-origin ...................................... 4, 6, 50, 69, 79, 81, Targeting vector....................................................... 140–142
89, 91, 92, 232, 233, 253, 265, 277–290 Transcriptome sequencing ............................... 79–87, 89, 92
Parthenogenone ...................................................................6 Transduction................................................................ 21–47
Perl ................................................ 83, 92, 96–100, 105–107, Transfection .....................................................24, 26–28, 31,
109, 111–117, 119, 120, 122–125, 127–132 32, 36–39, 41, 45–47
Piezo pipet driver .......................................................... 7, 10 Transgenerational inheritance .............................................6
Pipet beveler .................................................................. 7, 15 Transgenes ............................................. 5, 6, 36, 38, 46, 138,
Pipet puller .................................................................... 7, 15 142–145, 231, 279, 280, 285, 286
Placenta ................................................. 29, 49, 50, 260, 263, Transposable elements ..................................... 231, 266, 267
265, 268–270 Trophectoderm stem cells ............................................ 50–55
GENOMIC IMPRINTING
Index
297
U Y
UCSC........................................................ 83, 85, 86, 93, 96, Yamanaka .........................................................27, 37, 39, 45
98–101, 103, 107, 123, 124, 128, 129, 253–255,
257, 258
Z
Uniparental disomy ................................................... 70, 254 Zona pellucida ................................................... 9–13, 16, 17
Uniparental embryos ..................................................... 3–17 Zygote ........................................................3, 8–11, 281, 287

978 1 62703 011 3

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

978 1 62703 011 3

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

978 1 62703 011 3

Uploaded by

Copyright:

Available Formats

METHODS IN MOLECULAR BIOLOGY™

For further volumes:

Methods and Protocols

ISSN 1064-3745 ISSN 1940-6029 (electronic)

Library of Congress Control Number: 2012943828

© Springer Science+Business Media, LLC 2012

Printed on acid-free paper

Humana Press is a brand of Springer

Philadelphia, PA, USA Nora Engel

PART I PARENT-OF-ORIGIN EFFECTS

1 Uniparental Embryos in the Study of Genomic Imprinting . . . . . . . . . . . . . . . 3

PART II IDENTIFYING IMPRINTED GENES

5 Whole Genome Methylation Profiling by Immunoprecipitation

PART III IDENTIFYING THE REGULATORY FEATURES OF IMPRINTED DOMAINS

8 Engineering of Large Deletions and Duplications In Vivo . . . . . . . . . . . . . . . . 137

PART IV EPIGENETICS OF IMPRINTED REGIONS

9 Methylated DNA Immunoprecipitation (MeDIP)

10 Chromatin Immunoprecipitation to Characterize

PART V ANALYSIS OF IMPRINTED EXPRESSION

13 Isolation of RNA and DNA from Single Preimplantation

PART VI IMPRINTING IN PLANTS

16 Specialized Technologies for Epigenetics in Plants . . . . . . . . . . . . . . . . . . . . . . 231

PART VII EVOLUTION OF IMPRINTED GENES

17 Computational Studies of Imprinted Genes . . . . . . . . . . . . . . . . . . . . . . . . . . . 251

WADO AKAMATSU • Department of Physiology, School of Medicine, Keio University,

LOUIS LEFEBVRE • Department of Medical Genetics, Molecular Epigenetics Group,

Uniparental Embryos in the Study of Genomic Imprinting

Key words: Pronuclear transfer, Androgenone, Gynogenone, Parthenogenone, Uniparental embryo,

Green algae, carrots, salamanders, frogs, sea urchins, and mammals—

These early epigenetic processes are responsible for creating and

1.1. Uniparental The existence of genomic imprinting in mammals was initially

(exclusively maternal chromosomes) are unable to develop to term

PNT yielded early evidence for differential modification of the

simple to perform in comparison to microsurgery, consideration

2.1. Equipment 1. Stereo microscope (see Note 1).

2.2. Culture and 1. HEPES-buffered CZB (HCZB) or modified M2 medium as

6. Dulbecco’s PBS containing 5.6 mM glucose and 5.4 mM

2.3. Solutions 1. Cytochalasin B (Sigma, 5 mg/ml 1000 ´ stock in ethanol).

2.4. Pipets 1. Embryo transfer pipet connected to aspiration device.

of pronuclei) of appropriate morphology and granularity for

2. Place two rows of drops of HEPES-buffered manipulation

that the original opening in the zona pellucida is oriented at

3.4. Intracytoplasmic ICSI involves injecting a spermatozoon into the ooplasm of

The procedure for ROSI is almost the same as described above

and return to the incubator. Transfer the embryos to embryo

1. Stereomicroscope providing from 10.5× to 105× magnification,

used system. Sutter Instruments offers an attractive new system

CA-630 several times aspirating Igepal CA-630 (Sigma, Cat.

Derivation of Induced Pluripotent Stem Cells by Retroviral

Induced pluripotent stem cells (iPSCs) can be generated by trans-

Because of this, iPSC technology holds an enormous utility as a

(17–19). In these studies, genomic imprinting is a standard subject

2.1.2. Cell Culture 1. PBS.

6. Gelatin-coated culture dishes: To prepare 10× stock solution,

2.1.3. Preparation 1. Sterilized forceps and scissors.

neomycin-resistant gene and leukemia inhibitory factor (LIF).

2.2.2. Animals 1. NanogGFP-IRES-Puro mice (available from RIKEN BioResource