Simultaneous Sequencing of 24 Genes Associated with Steroid-Resistant Nephrotic Syndrome

Introduction

Approximately 20% of children with nephrotic syndrome will be resistant to steroids and other immunosuppression either at first presentation or subsequently. In the United Kingdom, between 350 and 400 children are affected, with 25% of these children currently on the end stage renal failure program (1). Steroid-resistant nephrotic syndrome (SRNS) is considered to represent glomerular disease resulting from a number of different etiologies leading to similar patterns of glomerular damage. The main renal histology detected is FSGS.

Positional cloning of causal genes for high-penetrance monogenic nephrotic syndromes has vastly increased our knowledge of the molecular genetic background to SRNS over the last decade. SRNS (including congenital nephrotic syndrome) displays considerable genetic heterogeneity, with 24 genes currently linked to this disorder (Table 1). In familial SRNS and SRNS presenting in the first 2 years of life, identifiable mutations are present in >95% of cases (2). This percentage decreases to around 40%–60% in children under 5 years, and in older children with sporadic disease, the incidence of genetic mutations is estimated to be 20% (2). SRNS associated with genetic mutations may progress more rapidly compared with idiopathic disease, and response to steroids or other immunosuppression drops from 40%–50% to 8% (3).

Although study of rare hereditary nephrotic disorders has helped our understanding of the molecular biology, the big challenges facing clinicians are how to accurately assign disease-causing roles to genetic variants detected in a specific patient, enable more accurate diagnosis/prognosis, and translate findings into concrete clinical advances that slow or abrogate disease progression.

The purpose of this study was to test the effectiveness of next generation sequencing (NGS; also known as second generation sequencing) in genes of interest in SRNS and correlate with detailed phenotypic information. The identification of causal genes has leapt forward with the advent of NGS technologies, because vastly greater amounts of DNA sequence data can now be obtained more rapidly and at less cost. However, with accumulation of very large datasets, it is also becoming increasingly apparent that correct evaluation and interpretation of any variants detected depend crucially on shared information held in open-access collections of accurate data on specific genetic disorders. The comprehensive and discriminatory phenotypic information contained within these databases can be used to curate and annotate sequence data, and at least, they indicate the degree of support for causality for a particular variant.

Our results show the obvious clinical potential for this technique, and we propose that NGS could be of equivalent quality to conventional Sanger methodology, particularly in diseases such as SRNS, where genetic heterogeneity is a defining feature.

Materials and Methods

Sequencing and Data Analyses

DNA was extracted using a QIAamp DNA Blood Mini Kit (Qiagen). Sample libraries were constructed using Illumina multiplexing sample preparation protocol (PE-930–1002; Illumina) with personal modifications (Supplemental Methods). A Roche NimbleGen sequence capture array, with probe capacity 12×135 K, was used to capture the exome of 446 genes inclusive of donor and acceptor sites. The genes were chosen on the basis of their known or potential involvement in nephrotic syndrome and glomerular function. The sequence library was hybridized for 70 hours, and then, the array was washed. Captured DNA was eluted and amplified. The final product (containing all 12 samples) was sequenced with 110-bp paired-end reads on Illumina GAIIx. Sequencing as well as base calling and demultiplexing were performed in the Transcriptomics Unit (University of Bristol).

Data analysis was performed using CLC Genomics Workbench v 4.9–5.0 (CLC Bio). Duplicate reads were removed, and reads were quality-trimmed (default parameters). Unique reads were then mapped against the human genome (hg19/GRCh37) with standard mapping parameters (Supplemental Table 1). Detection of single-nucleotide polymorphisms (SNPs) and short insertion/deletion polymorphisms (indels) was performed using listed parameters (Supplemental Table 2). Figure 1 is a flow diagram that shows variant analysis pathway. Any rare variants were subsequently validated by resequencing using Sanger sequencing (MWG, Germany). Splice site effect prediction was performed using Genomics Gateway (version 2.0 β5) identifying any SNPs that fell within 2 bp of an intron–exon boundary.

Results

Discussion

This study describes the use of targeted enrichment and NGS to screen 24 genes associated with SRNS. Patients were identified nationally using a novel rare renal disease registry (www.renalradar.org) and recruited to the United Kingdom SRNS in Childhood study. As a pilot, DNA from the first 36 patients was sequenced. These samples were the first samples received, although local clinician bias as to which patients were recruited first could not be ruled out, and it is possible that the more severe or frequently reviewed patients were recruited earlier to the cohort.

In this study, 70% of patients with familial disease and 15% of sporadic cases had a definitely or probably pathogenic variant identified, and in keeping with previous reports, age of onset correlated with risk of an identifiable pathogenic variant, with a positive finding in 86% of those patients presenting before 2 years of age (Table 5). In fact, 100% of those patients presenting before 2 years who had familial disease had an identifiable mutation, although the numbers in this pilot study were small. Interestingly, sequencing this number of genes enabled identification of mutations that would rarely be considered. This finding confirms that phenotype may not easily predict genotype and that screening all known genes is clinically relevant and may provide therapeutic options.

SRNS is known to be a disease of particular genetic heterogeneity, with the list of causative genes growing constantly (28–30). Although clinical advances have been rather limited, the benefit of this expansion in knowledge of the pathogenetic origin of disease is significant. It is accepted that accurate early identification of known pathogenic variants will modify therapeutic approach, often avoiding unnecessary immunosuppression and allowing treatment of rare conditions, such as COQ10 biosynthetic pathway disorders. In a disease where recurrence post-transplant is high but almost unknown in those patients with a genetic origin, transplantation and particularly, live related transplantation may be explored as therapeutic options earlier.

It is apparent that this condition, like others with multiple genes involved, lends itself to a clinical diagnostic service screening several genes at one time using NGS technology. The savings and efficiency of the process compared with Sanger sequencing may be obvious, but to replace or complement Sanger technique, it would need to be considered equally as accurate. This report shows that the accuracy, based on verification in 56 separate variants detected, is now approaching levels that would make it possible to develop the technology for clinical testing. It is interesting to note that, in the United Kingdom, with diagnostic laboratories already offering services using NGS (31), the Clinical Molecular Genetics Society is working to national best-practice guidelines for the use of NGS.

This report also highlights the difficulty in interpretation of results. Identification of definite mutations is clearly useful in a clinical context, but the remaining results require careful analysis. In this report, we have listed some variants as probably pathogenic (for instance, a novel homozygous variant in NPHS1); however, a previous report describes a variant at the same codon resulting in a different amino acid change in a patient with a similar phenotype or a compound heterozygous in COQ2, where one variant is previously described as pathogenic and the other is a novel indel, resulting in significant disruption of the protein. Instinctively, these variants would seem to be pathogenic, and we would advocate reporting as such with caution.

For other variants, however, where they are rare but not previously described, the impact on patient management is less tangible, except that by reporting these types of variants by our group and others with clear genotype phenotype description, a clearer picture may be developed. With the inevitable expansion of NGS data in this disease, we propose a publicly accessible combined disease-specific database of variants in nephrotic genes with phenotypic data available as the most efficient way to define genetic pathogenicity.

An additional issue to consider is variant frequency in patients of certain ethnic origins. In the open-access databases used here, certain populations are not well described (for instance, those populations from India, Pakistan, and North Africa). In fact, using data from whole-genome databases (compared with whole-exome databases) as controls does have the disadvantage that coverage is often not high; therefore, rare variants may be missed, or allele frequencies may be underestimated. In the near future, population genetics will improve the databases of different ethnic groups.

One might also argue about what constitutes rare in a disease that, if genetic in origin, is usually autosomal-recessive. Using a cutoff MAF of 1% may be too stringent; in this study, other interesting variants were found occurring at >1% but generally below 3%.

The use of HGMD as a filtering tool is difficult. Although it is very useful to have a list of all variants associated with the disease described in the literature, in-depth analysis of the individual references revealed many were nonsignificant polymorphisms.

A particular strength of this study is the use of a new comprehensively phenotyped national cohort of pediatric SRNS patients, ensuring the most appropriate inclusion criteria for testing. The results that we present are comparable with gold standard Sanger sequencing and highly informative.

We aim to propose, based on these data, that simultaneous NGS sequencing should be considered for genetic management of all patients with SRNS.

Disclosures

None.