Targeted Sequencing of 179 Genes Associated with Hereditary Retinal Dystrophies and 10 Candidate Genes Identifies Novel and Known Mutations in Patients with Various Retinal Diseases

Abstract

Purpose.: Hereditary retinal dystrophies (HRDs) are a group of monogenic diseases characterized by an irreversible loss of photoreceptors. HRDs exhibit significant genetic and clinical heterogeneities challenging traditional techniques for determining disease-causal mutations. This study aims to develop an efficient molecular diagnostic platform for HRDs, and to determine the genetic basis for 25 randomly collected Chinese families with a variety of HRDs.

Methods.: We designed a high throughput sequence capture microarray targeting 179 genes associated with HRDs and 10 candidate genes. We combined sequence capture with next-generation sequencing (NGS) to screen for mutations in the cohort of Chinese families. Variants detected by NGS were filtered, validated, and prioritized by pathogenicity analysis. Genotypes and phenotypes were correlated.

Results.: We identified four recurrent single mutations, two compound mutations, and eight novel putative causative mutations, including five putative pathogenic alleles (e.g., premature stop codons and frame shifts) and three novel missense variants that are very likely pathogenic. These findings provided specific genetic diagnoses in 14 of 25 families (56%). Among these, identification of a mutation in VCAN in a family with a complicated phenotype helped to finalize the clinical diagnosis as Wagner syndrome. In another five families, 11 potential novel pathogenic variants were identified.

Conclusions.: A substantial number of potential new genes and new mutations associated with HRDs remain to be discovered. Identification of the novel HRDs-causing mutations in our study not only provides a better understanding of genotype–phenotype relationships in these diseases, but also demonstrates that the approach described herein is an effective method for large scale mutation detection among diverse and complicated HRDs cases.

Introduction

Hereditary retinal dystrophies (HRDs) consist of a number of severe retinal degenerative diseases with great genetic and clinical heterogeneity. Retinitis pigmentosa (RP) is the most common form of HRDs, with a prevalence of 1 in 3500 live births in Western countries ¹ and even higher (1 in 1000) in Beijing. ² The disease predominantly affects rod photoreceptors, ³ leading to night blindness, visual restriction, and gradual loss of the peripheral visual field. Cone function can be secondarily or simultaneously affected in the process of RP, ⁴ and, thus, central vision can be eventually impaired. ³ Cone–Rod dystrophy (CRD) ⁵ and Leber congenital amaurosis (LCA), ⁶ two other types of HRDs, involve severe deficits in both cones and rods and, therefore, have overlapping phenotypes with RP. Other isolated HRDs include macular degenerative diseases (e.g., Stargardt and best vitelliform macular dystrophy [BVMD]), congenital stationary night blindness, and some complicated retinopathies with variable ocular involvements. HRDs also refer to syndromic diseases with retinopathies such as Bardet–Biedl syndrome (BBS, Online Mendelian Inheritance in Man [OMIM]: 209900) and Usher syndrome (OMIM: 276900). The phenotypic overlaps and variability in presentation of HRDs often make clinical diagnosis uncertain and complicated. 

A specific molecular diagnosis of HRDs can help with the clinical diagnosis, improve psychological well-being of patients, ⁷ direct therapeutic interventions including reproductive counseling and provide a better understanding of disease pathogenic mechanisms. However, it has been problematic to determine the specific mutations in HRDs because of their extensive genetic heterogeneity, which so far involves 179 known genes (www.RetNet.org), and potentially many new genes. Genetic background overlaps (e.g., LCA, RP, and CRD have overlapping genes and mutations ^8–10 ) and clinical ambiguities of some types of HRDs further increase the difficulties to determine causative mutations because no specific panel of HRDs genes can be predicted in these circumstances. 

Detection of specific mutations among a large number of candidate genes by traditional techniques is challenging. Linkage analysis and homozygosity mapping, two standard approaches to localize a genetic defect, require either a sufficient number of informative meiotic events in a family or a number of individual families with similar phenotypes. Sanger sequencing can be used for mutation screening, but is often cost and time prohibitive if a large number of genes need to be tested. The arrayed primer extension technique (APEX) designed for HRDs is an alternative screening method, but it can only detect a fixed set of known mutations. Compared with these methods, next-generation massively parallel sequencing (NGS) with various applications provides high sensitivity and speed of variation detection and is currently considered the most efficient approach for mutation screening. ^11–13 While several chip-based approaches for the NGS platform have been previously developed to screen various panels of HRDs genes (e.g., for autosomal dominant and recessive RP genes [adRP genes and arRP genes] ^14,15 ) exome sequencing is also considered a first tier screening method for genetically heterogeneous diseases. ¹⁶ Targeted gene approaches save time, cost less, and achieve higher enrichment of the targeted sequences, but at the expense of missing mutations in genes not previously associated with HRDs. Whole exome sequencing surveys a broad list of nearly all exons and, thus, has the potential to discover new disease-causing genes. However, compared with targeted approaches, whole exome sequencing is more expensive and requires more data analysis. 

In an attempt to devise an appropriate technical approach, we reasoned that targeted sequencing of a full set of HRDs-related genes as well as some important candidate genes might be an efficient and feasible method for primary mutation screening. Here, we report an in house designed targeted gene capture microarray on the Nimblegen platform, which was demonstrated to be capable of capturing consensus coding DNA sequences (CCDS) and necessary intronic sequences of 179 genes, including all genes known to be associated with HRDs at the time of the study. In an effort to search novel genes for HRDs, 10 additional candidate genes for HRDs were included in the capture array (the array is hereafter referred as RDs189-array). By this approach, we accomplished targeted sequencing of the 189 genes in 25 distinct Chinese families with diverse categories of HRDs including adRP and arRP, Stargardt disease (STGD), Usher syndrome, BBS syndrome, and Wagner syndrome. We were able to identify 14 mutations that are most likely the causative mutations for the disease phenotypes in the corresponding families. Identification of these mutations generates specific molecular diagnoses for particular patients, improves our understanding of the genetic architecture of HRDs, and demonstrates that the targeted sequencing approach described herein is an efficient method for HRDs mutation detection with potential significance in clinical practice. 

Methods

Patients with HRDs and Controls

In this study, a cohort of 25 Chinese families with a variety of HRDs was clinically collected and examined in multiple hospitals within China (see Supplementary Material and Supplementary Table S1). In order to appropriately evaluate the efficiency of the targeted gene sequencing approach, all pedigrees were recruited within a period of 1 year with broad selection criteria, so that the pedigrees may represent the occurrence of HRDs in different parts of China. All patients underwent routine ophthalmic examinations as described previously, ¹⁷ with visual field and electroretinography (ERG) tests. Optical coherence tomography (OCT) was performed on patients with macular involvement. Systemic examinations were performed whenever necessary. The clinical diagnosis and, when possible, Mendelian inheritance mode for each family is summarized in Supplemental Table S1 (see Supplementary Material and Supplementary Table S1). 

Blood samples were collected from all participants (48 affected members and 34 unaffected members) and 300 unrelated controls with neither eye diseases as demonstrated by routine ophthalmic examinations, nor family history of any type of HRDs. Informed consent was obtained from all cases for sample collection and molecular analysis, and human studies were prospectively reviewed and approved by local institutional ethical review boards according to the Declaration of Helsinki. Genomic DNA was isolated from peripheral leukocytes. 

Design of RDs189-Array

The RDs189-array was designed and constructed on a 2.1 Megabase NimbleGen sequence capture microarray platform (Roche NimbleGen, Inc., Madison, WI). HRDs related genes were defined as known causes and/or risk alleles for isolated and/or syndomic HRDs. By these criteria, 179 genes associated with HRDs were selected from the Retnet database for inherited retinal dystrophies (http://www.RetNet.org), OMIM database (http://www.ncbi.nlm.nih.gov/omim), and publications (Pubmed search queries: hereditary retinal dystrophy). In addition, 10 genes encoding pre-mRNA splicing factors that are within U4/U6-U5 tri-snRNP complex¹⁸ were selected as potential disease-causing genes; six other splicing factors within or associated with the same complex have been implicated in adRP^19–24 emphasizing the role of pre-mRNA deficiency in the etiology of RP. Information on all 189 genes is detailed in Supplemental Table S2 (see Supplementary Material and Supplementary Table S2). CCDS of the 189 genes were downloaded from the genome database (hg19 version) of University of California at Santa Cruz (UCSC, http://genome.ucsc.edu/). In total, 4641 exons with at least 100 bp of flanking intron sequences, and 5′ and 3′ untranslated regions (UTR) for the 189 genes were defined as targeted regions and 250 to 300 base probes were designed accordingly for sequence capture. This design would allow the array to capture as much as 200 bp of flanking intron sequences for a given exon. Repetitive elements leading to cross-hybridization were identified by RepeatMasker shareware (http://www.repeatmasker.org) and discarded (2.5% of total bases), and the exons of genes that were removed is summarized in Supplemental Table S3 (see Supplementary Material and Supplementary Table S3). All probes were tiled on the RDs189-array by Roche NimbleGen, Inc., and their sequences were merged into a BED file that can be analyzed by UCSC online browser. 

Patients and Positive Controls for RDs189 Array Analysis

DNA samples from 32 individuals (29 patients and three unaffected members) from 25 families were selected for RDs189-array capture followed by NGS (See Supplementary Material and Supplementary Table S4). To evaluate the targeting efficiency of the array, samples from three RP patients with previously identified suspected mutations in three of the 189 genes were included as positive controls (see Supplementary Material and Supplementary Table S5). 

Targeted Sequence Capture and NGS

Template preparation, hybridization with the RDs189-array, and NGS were carried out in collaboration with Beijing Genome Institute (BGI, Shenzhen, China) as described previously. ²⁵ For details, please see Supplemental Method (see Supplementary Material and Supplementary Method). 

Bioinformatics Analysis

See Supplemental Method (see Supplementary Material and Supplementary Method). 

Sanger Sequencing and Public Database

See Supplemental Method (see Supplementary Material and Supplementary Method). 

In Vitro Analysis of RDH5 Variant

See Supplemental Method (see Supplementary Material and Supplementary Method). 

Results

Evaluation of RDs189-Array

NGS of the captured libraries generated an average of 1.88 million reads per sample, 82.2% of which were mapped to the targeted regions. The average call rates for targeted bases was greater than 99.9%. The average of mean depth for the targeted regions among all samples was 76.47 ± 15.42-fold, and an average of 99.6% and 97.6% of targeted bases were covered at 4- and 20-fold, respectively. Coverage analysis for each sample was detailed in Figure 1. 

Figure 1

 

Coverage analysis of RDs189-array capture sequencing for the 32 individuals. (A) The numbers of total reads (blue) and reads mapped to the targeted regions (purple) for each individual. (B) Mean depth of the targeted regions for each individual. (C) Fraction of targeted bases covered at 4× (orange) and 20× (green) for each individual.

Short oligonucleotide analysis package (SOAPsnp) was used to assemble the consensus sequence and call genotypes in targeted regions. Genotypes different from reference were extracted as candidate Single nucleotide variants (SNVs), and approximately 2252 to 2558 SNVs/sample were initially picked up. Approximately 85 to 100 indels/sample were also initially identified. Preliminary bioinformatics filtering of all variations was performed in collaboration with BGI. Variations with poor sequence quality (see Materials and Methods) and intronic variations, with the exception of those localized within 10 bp from exon flanking boundaries, were discarded. This filtering step led to a significant reduction in candidate variations; only 5907 unique variations including single nucleotide polymorphisms (SNPs) and indels remained in all 32 individuals and three positive controls after application of the preliminary filter. All three of the previously identified SNVs carried by the positive controls were detected at this point (see Supplementary Material and Supplementary Table S5), demonstrating that this approach is efficient and reliable for detecting SNVs. 

Customized Filtering Significantly Reduces the Number of Candidate Variations

To efficiently screen the 5907 variations to find pathogenic changes, we set up a customized bioinformatic analysis pipeline comprising several filters for variations selection (Fig. 2). First, we removed synonymous variations and noncoding variations and only focused on nonsynonymous (NS) variants, intronic variants localized within 10 bp from exon flanking boundaries that may affect splice sites (SS), and short frame shift coding indels ranging from 1 to 50 bp, all of which were more likely to be pathogenic mutations. Furthermore, we compared our NS/SS/indel variants with several SNPs databases including dbSNP132, exome-sequenced HapMap samples (pilot1, 2, 3), and the SNP release of the 1000 Genomes Project (20100804 release), Exome Variant Server and YH databases. We considered that any variant present in a homozygous state in the SNPs databases, including in unaffected subjects, would be unlikely pathogenic for the diseases for all types of inheritance modes. We, therefore, removed such SNPs from all families, whether detected as homozygous or heterozygous in our study. Furthermore, for the heterozygous variants detected in all families, we removed those found in SNPs databases including unaffected subjects. Any variants with previously known clinical relevance were retained during SNPs filtering for further analysis. In the next step, we discarded the variants with obvious conflict with the Mendelian inheritance mode of the corresponding family; we removed heterozygous variants from families with clear recessive HRDs, and removed homozygous variants from families with dominant disease. We also analyzed the prevalence of a given variation among all 32 samples sequenced, and removed those shared by unaffected family members or shared by more than four unrelated individuals, assuming that a pathogenic mutation would be unlikely to have such a high prevalence (∼15%) in a variety of HRDs patients. All remaining variations were further subjected to validation by Sanger sequencing, and 101 variations in total were confirmed. 

Figure 2

 

Flow chart of variant analyses. Customized filtering (upper panel) and pathogenicity analysis of variants were included in the process. *SNPs database analysis was carried out in two steps: (1) variants that present in the homozygous state in SNP databases including unaffected subjects were removed, whether detected as homozygous or heterozygous in our study, and (2) for heterozygous variants detected in all families, we removed those found in SNP databases including unaffected subjects. *Overall analysis among all samples refers to the evaluation of the presence of each individual variant among all 32 samples subjected to NGS.

Pathogenicity Analysis Identifies Specific Mutations in 14 Families

The 101 post filter variants were subjected to several analyses to determine whether they were likely pathogenic (Fig. 2). First, Sanger sequencing was used to analyze whether a variation segregated with the disease phenotype in the corresponding family, if additional family members were available. Variants that did not segregate with the disease phenotype were discarded and human gene mutation databases and publications were searched for the remaining variants to determine whether they had been previously identified. By this comparison, we found six families carrying altogether eight recurrent mutations that were previously determined to be pathogenic (Fig. 2, Table). The eight recurrent mutations include four single and two compound mutations (Table). RHO p.Pro53Arg in family HD09 was previously reported to cause RP. ²⁶ CRB1 p.Arg526* in family HD12 and CYP4V2 p.Gly95Arg in family HD37 were known mutations deposited in the Human Gene Mutation Data Base (accession number: CM082582 and CM056576). Homozygous mutations of CYP4V2 have been previously implicated in Bietti crystalline corneoretinal dystrophy, an autosomal recessive retinal dystrophy characterized by intraretinal crystals scattered over the fundus, sclerosis of the choroidal vessels, and phenotypes similar to that of RP including retinal degeneration, night blindness, and constricted visual field (OMIM: 608614). The fact that we found a heterozygous p.Gly95Arg of CYP4V2 in an adRP family would presumably indicate the clinical heterogeneity of retinal degenerations associated with CYP4V2 mutations. A splice acceptor site mutation of VCAN c. 4004-1G>T identified in family HD28 has been implicated in Wagner syndrome. ²⁷ The two compound mutations of ABCA4 gene, p.(Phe2188Ser; Asn965Ser) (refers to two changes in one allele) and p.(Ala1598Asp);(Glu328Val) (refers to one change in each allele) identified in family HD32 and HD33, respectively, are novel. However, each of the four SNVs was previously identified as single or concurrent causative mutations for STGD. Consistent with our results, previous studies implicate homozygous or compound mutations of ABCA4 gene as common causes of STGD. ²⁸  

Table

 

Specific Mutations Identified in the 14 Families

Family ID	Final Diagnosis	Gene	Exon N	Nucleotide Change	Amino Acid Change	Status	Novel or HGMD N	PolyPhen-2	Sift	Conservation Analysis
HD01	arRP	CNGA1	5	c.472delC	p.Leu158Phefs*4	Hom	Novel	N/A	N/A	N/A
HD09	adRP	RHO	1	c.158C>G	p.Pro53Arg	Het.	CM920608	Prob/0.999	Damaging/0.00	Conserved
HD12	adRP	CRB1	6	c.1576C>T	p.Arg526*	Het.	CM082582	N/A	N/A	N/A
HD13	BBS	BBS7	10	c.1002delT	p.Ile334Ilefs*15	Hom	Novel	N/A	N/A	N/A
HD16	arRP	PROM1	intron 23	c.2373+5G>T	Unknown	Hom	Novel	N/A	N/A	Conserved‡
HD21	adRP	VCAN	8	c.7870G>A	p.Glu2624Lys	Het	Novel	Poss/0.890	Tolerated/0.08	Conserved
HD23	xlRP	RPGR	15	c.2038_2041delGACA	p.Asp680Argfs*16	Hem	Novel	N/A	N/A	N/A
HD24	US	GPR98	70	c.14366G>A	p.Arg4789Gln	Hom	Novel	Prob/0.998	Damaging/0.00	Conserved
HD28	WS	VCAN	intron 7	c.4004-1G>T	p.Gly1335Asp1348del†	Het	CS063286	N/A	N/A	N/A
HD30	FA	RDH5	4	c.625C>T	p.Arg209*	Hom	Novel	N/A	N/A	N/A
HD32	STGD	ABCA4	48	c.[6563T > C;	p.[Phe2188Ser;	Het	CM970005	Poss/0.765	Damaging/0.05	Conserved
			19	2894A>G]	Asn965Ser]	Het	CM023878	Prob/0.99	Damaging/0.00	Conserved
HD33	STGD	ABCA4	34	c.[4793C>A];	p.[Ala1598Asp];	Het	CM003386	Poss/0.669	Damaging/0.01	Conserved
			8	[983A>T]	[Glu328Val]	Het	CM003363	Benign/0.227	Damaging/0.03	Conserved
HD37	adRP	CYP4V2	2	c.283G>A	p.Gly95Arg	Het	CM056576	Prob/1.000	Damaging/0.01	Conserved
HD40	adRP	TOPORS	3	c.2570G>T	p.Arg857Met	Het	Novel	Poss/0.641	Damaging/0.05	Conserved

 Conservation analysis is performed among Homo sapiens, Gorilla gorilla, Equus caballus, Canis lupus familiaris, Bos taurus, and Mus musculus. The two compound mutations are bolded. HGMD, human genome mutation database; Prob, probably damaging; Poss, possibly damaging; N/A, not applicable; US, Usher syndrome; WS, Wagner syndrome; FA, fundus albipunctatus.

†  The effect of c.4004-1G>T mutation on protein level is previously characterized. ²⁷

‡  Conservation analysis is performed on the genomic sequence of PROM1.

Variants not present in mutation databases were further evaluated for potential pathogenicity. Sanger sequencing was first used to determine whether a given variation was present in 300 unrelated controls and, if present, the variation was considered to be benign (Fig. 2). We next assessed the impact of a given variant at the protein level by considering its nature, and we, thus, identified five novel most likely pathogenic alleles, each of which was present in a different family (Fig. 3). These novel pathogenic alleles include three frame shift variants (CNGA1 c.472delC, BBS7 c.1002delT, and RPGR c.2038_2041delGACA), one nonsense variant (RDH5 c.625C>T) that results in a premature stop codon (p.Arg209*), and one splice site variant (PROM1 c.2373+5G>T) that affects a conserved nucleotide by genomic conservation analysis (Fig. 3E). The potential splicing effect of PROM1 c.2373+5G>T was further predicted using the NNSplice program. ²⁹ This variant significantly changes splice-site scores compared with the reference sequence (0.00 vs. 0.98). Each of the five pathogenic variants is most likely the causative mutation for the disease phenotype in the corresponding family (Table, Fig. 3) because: (1) none of the variants was present in 300 unrelated controls, (2) none of the five families has other potential pathogenic variants detected by this study, and (3) the well-established correlations between the genes mutated and the disease phenotypes for each family. 

Figure 3

 

Identification of 5 novel pathogenic alleles in the corresponding families. (A) Homozygous CNGA1 c.472delC (p.Leu158Phefs*4) mutation in pedigree HD01. (B) Homozygous BBS7 c.1002delT (p.Ile334Ilefs*15) mutation in pedigree HD13. (C) Hemizygous RPGR c.2038_2041delGACA (p.Asp680Argfs*16) in pedigree HD23. (D) Homozygous RDH5 c.625C>T (p.Arg209*) mutation in pedigree HD30. (E) Homozygous PROM1 c.2373+5G>T mutation in pedigree HD16 and genomic conservation analysis of the mutated nucleotide. Cosegregation analyses were performed on all collected family members as denoted by asterisk (upper left) and genotype data (underneath). Affected members are indicated by filled symbols and unaffected members are represented by open symbols. Black arrow denotes the proband. Sequence of BBS7 and RPGR are reversed. Hom, homozygous; Het, heterozygous; Hem, hemizygous.

Subsequent functional predictions were performed on the novel missense candidate variants including conservation analysis and in silico analyses by the PolyPhen-2 and Sorting Intolerant From Tolerant (SIFT) programs. These analyses revealed three putative pathogenic novel missense mutations, VCAN p.Glu2624Lys, GPR98 p.Arg4789Gln, TOPORS p.Arg857Met, each of which is the specific mutation that consegregates with the phenotype in the corresponding family, is absent in 300 unrelated controls, and is predicted to be damaging by PolyPhen-2 and/or SIFT (Fig. 4, Table). Moreover, each of the three novel missense mutations affects a rather conserved amino acid residue (Fig. 4). The genes in which the novel missense variants were identified have an existing correlation with the disease phenotypes of the families and, taking into account their likely functional impact, these novel variants are very likely pathogenic mutations. 

Figure 4

 

Identification of novel missense mutations in three families with conservation analyses. (A) VCAN c.7870G>A (p.Glu2624Lys) mutation in pedigree HD21. (B) GPR98 c.14366G>A (Arg4789Gln) mutation in pedigree HD24. (C) TOPORS c.2570G>T (Arg857Met) in pedigree HD40. Cosegregation analyses were performed on all collected family members as denoted by asterisk (upper left) and genotype data (underneath). Affected members are indicated by filled symbols and unaffected members are represented by open symbols. Black arrow denotes the proband.

In families HD20, 34, 36, 38, and 39, 11 potential pathogenic variants passed through our filtering process and pathogenicity analyses (see Supplementary Material and Supplementary Table S6). However, these families either carry more than one potential pathogenic variant or have no second DNA sample available to confirm the mutant allele (See Supplementary Material and Supplementary Table S6). Thus, our current results are not sufficient to generate specific molecular diagnoses in these families. In the remaining six families, no potential pathogenic variations were identified. Overall, the identification of four recurrent mutations, two new compound mutations, and eight novel single mutations, and the agreement between genes in which the mutations were identified and the clinical phenotypes provide very likely specific molecular diagnosis in 14 families (56%). 

In Vitro Pathogenicity Analysis of RDH5 p.Arg209*

The nonsense mutation, RDH5 c.625C>T, identified in Family HD30 should generate a premature stop codon (PSC), p.Arg209*, in the mature RNA. Because transcripts harboring PSCs can be eliminated before translation by nonsense mediated decay (NMD), we further examined the pathogenic effect of RDH5 p.Arg209* in transiently transfected HEK293 cells. Similar transfection efficiencies of pEGFP-RDH5^Arg209* and pEGFP-RDH5^WT were determined by semiquantitative PCR analysis (25 cycles) on the exogenous DNA templates (Fig. 5A). However, the transcript level of RDH5^p.Arg209* , in contrast to RDH5^WT , was undetectable after 40 cycles of RT-PCR analysis (Fig. 5B), suggesting the truncated transcript was degraded via NMD. Moreover, the protein level of RDH5^p.Arg209* was not detectable by immunoblot using antibodies against either RDH5 or GFP (Figure 5C). Thus, the RDH5 p.Arg209* mutation likely results in a complete loss of function via NMD. 

Figure 5

 

In vitro functional analysis of RDH5 p.Arg209*. (A) Semiquantitative PCR (25 cycles) analysis of exogenous DNA template obtained from the transfected HEK293 cells. 5 μL PCR products generated form pEGFP-RDH5^p.Arg209* and pEGFP-RDH5^WT transfected cells revealed similar patterns of bands on agrose gel. (B) RT-PCR analysis (40 cycles) of mRNA on the transfected HEK293 cells. The transcript of RDH5 (467 bp) was only detected in pEGFP-RDH5^WT transfected cells, suggesting very low levels of truncated transcript for pEGFP-RDH5^p.Arg209* . Samples with no RT treatment during the process of RT-PCR serve as negative control. (C) Immunoblot analysis using antibodies against either RDH5 (upper panel) or GFP (middle panel) only detected RDH5-GFP fusion protein (62 KDa) in pEGFP-RDH5^WT transfected cells.

Identification of a Causative Mutation in Family YB Finalized a Clinical Diagnosis of Wagner Syndrome

The family YB (Fig. 6) has retinopathy with a complicated phenotype that is likely inherited in an autosomal dominant mode (Fig. 6A). All three patients in this family developed night blindness beginning in early childhood accompanied with progressively decreased central vision and restricted visual fields. Fundus examinations reveal global retinal degeneration in patients II:2 and II:3 with RP-like pigment deposits in the peripheral retina and macular region (Fig. 6B). OCT confirmed the global retinal attenuation (Fig. 6C). Consistent with these findings, decreased rod and cone functions were demonstrated by ERG in all three patients. Notably, ectopic fovea (Fig. 6B) that results in pseudoextropia is present in all three patients. No definitive clinical diagnosis was assigned for this family before our genetic analysis. Because we identified a recurrent mutation in the VCAN gene (Fig. 6D), 4004-1G>T, ²⁷ we finalized the clinical diagnosis for the family as Wagner syndrome. 

Figure 6

 

Clinical phenotypes and mutation of VCAN identified in family HD28. (A) Pedigree of the family. Cosegregation analysis was performed on all collected family members as denoted by asterisk (upper left) and genotype data (underneath). Affected members are indicated by filled symbols and unaffected members are represented by open symbols. Black arrow denotes the proband. Fundus photographs (B) and OCT examination (C) of patient II:2 revealed severe global retinal degeneration with pigment deposits (denoted by white arrows) in macular and peripheral regions. The ectopic fovea is indicated by a red circle. (D) Sanger sequencing of VCAN for patient II:2 (lower trace) confirmed the c.4004-1G>T mutation (denoted by arrowhead), whereas the unaffected sibling II:4 (upper trace) shows reference sequence. The mutation is located at the acceptor splice site (boxed) upstream of exon 8.

Discussion

HRDs have great genetic heterogeneity involving over 170 genes, many of which have a large number of known disease alleles. HRD patients often lack sufficient family history to determine a specific gene or even a definitive inheritance pattern. Therefore, it has been a great challenge for traditional methods to detect specific mutations in a large number of candidate genes. The difficulty is compounded by the fact that, in a significant fraction of individuals with HRDs (particularly with RP), the causative genes/mutations have not yet been identified. ³⁰ Thus, developing a validated high throughput platform that can simultaneously analyze numerous HRD genes is indeed needed to address the issue. Microarray-based targeted gene sequencing has offered such an opportunity with reduced cost compared with conventional sequencing. ³¹ In this study, we applied targeted gene sequencing by using our self-designed RDs189-array on 25 families representing six disease categories of HRDs. Because these families were collected in the course of normal clinical practice, most do not have enough participants to perform linkage analysis, or even enough familial history to determine their inheritance pattern. Despite these factors, we were able to rapidly determine likely pathogenic mutations in 14 families and establish specific molecular diagnoses, demonstrating that the RDs189-array with NGS approach is an effective technique for identifying causative mutations in HRD patients and has significant advantages over traditional methods. 

Several microarrays have been developed recently to capture various panels of HRD genes.^{14,15,31–33} Most of these arrays were generated on the Affymetrix resequencing chip platform (Affymetrix, Inc., Cleveland, OH).^15,31,32 The nucleotide call rates for the previously reported resequencing chips ranged from 90% to 99%.¹⁵ The RDs189-array yielded superior call rates (>99.9%), presumably due to the removal of repetitive elements (see Supplementary Material and Supplementary Table S3). The high call rates of the RDs189-array demonstrate that it is able to capture efficiently nearly all the targeted sequences of the 189 genes. Compared with the aforementioned earlier capture arrays,^14,15,31,33 the RDs189-array has a significantly expanded panel of targeted genes including, to our knowledge, all known HRD related genes, and, thus, can be applied to almost all types of HRD patients. Also, simultaneously screening all known HRD genes will theoretically increase the mutation detection rates. Indeed, while the two chips designed for 14 arRP genes and for 44 adRP genes showed mutation detection rates of 14% and 24%, respectively,^14,15 our results and a parallel study that also employed a chip targeting a broad list of HRD genes³⁴ revealed similar mutation call rates of 56% and 57%, respectively. In this study, we included 10 candidate genes associated with pre-mRNA splicing in the RDs189-array (see Supplementary Material and Supplementary Table S2), which may potentially increase the cost for first tier screening. However, the RDs189-array offers for the first time an opportunity to screen a relatively full set of pre-mRNA splicing genes within the U4/U6-U5 complex, which has an important role in adRP etiology,^23,35 and, thus, raises the possibility of identifying novel splicing genes associated with RP. Although we did not find any putative pathogenic variants in the 10 candidate genes in the present study, it's likely that mutations in some of these candidate genes will be correlated with RP in the future. Indeed, while we were designing the chip, mutations in the sixth splicing gene, PRPF6, were identified in patients with adRP.²⁴  

With the capability to screen all known HRD genes, the RDs189-array has shown a particular benefit for analyzing patients that have an unclear clinical diagnosis due to phenotypic ambiguity. This is demonstrated by our study of family HD28. HD28 has rather complicated phenotypes (see Results), such that neither we, nor our colleagues, were unable to differentiate the disease from LCA, CRD, and RP. Thus, screening genes associated with the three diseases (over 100 genes) would have been necessary, but not sufficient to identify the causative mutation in the family. The RDs189-array allowed us to rapidly detect a mutation of VCAN, c.4004-1G>T, which has been reported in one Dutch family with Wagner syndrome, ²⁷ a disease with vitreous abnormality, choroidalretinal degeneration, CRD, and other varied ocular and systemic involvements (OMIM: 143200). Wagner syndrome is known to be caused by VCAN mutations with clinical expressivity varying from unaffected genetic carrier to bilateral blindness. In a previously reported Japanese family, an adjacent mutation, c.4004-2A>G, was identified as affecting the same acceptor splice site of VCAN. ³⁶ There are dramatic similarities between the Japanese and Chinese families including rod and cone degeneration, pigment deposition, and ectopic fovea with consequent pseudoextropia. However, the vitreous abnormality was not remarkable in the Chinese family, which in turn made the primary clinical diagnosis unclear. Nevertheless, the identification of a recurrent mutation in VCAN and consistency between the genotype and phenotype established a specific genetic diagnosis for this small pedigree with rather complicated phenotypes, and subsequently finalized the clinical diagnosis as Wagner syndrome. This example strongly suggests that the targeted gene sequencing approach employing the RDs189-array is a feasible and promising tool for establishing final molecular and clinical diagnoses for HRD patients. 

The RDs189-array has several limitations that are common for most sequence screening techniques. Although as much as 200 bp on both sides of a given exon can be assessed by the array, most intronic sequences are still undetectable given their absence on the array. Detection of large deletions, insertions, and copy number variations is still problematic for the array. Also, the filtering process, in which we discarded noncoding variants, may lead to missing out some pathogenic variants. Not finding any potential pathogenic variants in six families does not exclude the 189 genes as disease-causing. The significant reduction in variants by our filtering process requires follow up analysis in these cases. Similarly, identification of mutations in 14 families does not completely rule out the possibility that variants other than those we identified may contribute partially or fully to the disease etiology. Indeed, identification of a few real causative mutations among numerous variants detected by NGS remains a major challenge. ³⁷ To determine the pathogenicity of variants in this study, we applied several approaches that have been widely used ³⁷ including segregation analysis, prevalence testing in controls, and in silico analyses for amino acid conservation predicted effects on protein structure/function. Very recently, a software tool was developed to analyze exome sequencing variants in small pedigrees with Mendelian inheritance. ^38,39 Several filters and parameters integrated into this software are very similar to our approaches. However, it is currently still very difficult to conclusively determine whether a given variant is a disease-causing mutation because DNA sequence information alone is often insufficient. Nevertheless, the mutations identified in this study are most likely the causative mutations for the corresponding diseases. 

Some of the novel mutations identified herein may provide insight into the genetic structure of HRDs. Several missense mutations in RDH5 have been identified ^39–41 and proposed to be loss of function alleles on the basis of cellular and murine models. ^40,42 The RDH5 homozygous mutation p.Arg209* found in family HD30 is, to our knowledge, the only null allele reported to date. We provided the first direct evidence that a RDH5 homozygous nonsense mutation could result in complete loss of function via NMD, as one of the pathogenic mechanisms of fundus albipunctatus. Homozygous mutations of CNGA1 have been linked with arRP (OMIM: 123825). The two novel heterozygous alleles of CNGA1 indentified in family HD36 (see Supplementary Material and Supplementary Table S6), p.Arg629*, and p.Gly133Valfs*29, are very likely pathogenic, and, thus, may implicate for the first time a causative role of compound mutations of CNGA1 in RP. Unfortunately, no additional DNA samples are available in this family to confirm that the two mutations are different alleles. Among the 11 candidate pathogenic variants, three variants were identified in the genes that have no existing connections with the phenotypes in the corresponding families (see Supplementary Material and Supplementary Table S6), and may be less likely pathogenic. 

Whole exome sequencing is considered by some laboratories as the most efficient tool for mutation screening in heterogeneous diseases ¹⁶ because it can survey a broad list of genes with continuously increased exome coverage and decreased cost. Indeed, the targeted sequencing approach is limited to the currently known genes and requires updating the capture array and retesting patients. However, in our hands, the targeted gene capture approach achieves better enrichment for the sequences of interest and, therefore, decreases the false positive/negative rates and increases the simplicity of the subsequent analysis. It is presently difficult to determine which approach, whole exome sequencing or targeted sequencing, would serve as the first tier screening method in clinical practice. Nonetheless, proper selection of samples from patients who have been precisely phenotyped, using a well designed questionnaire to determine the most likely inheritance pattern, in combination with suitable in silico analysis and functional assays, are still critical to find causative mutations. 

Supplementary Materials

pdf S1 

xls S2 

xls S3 

xlsx S4 

xls S5 

xls S6 

xls S7 

Acknowledgments

The authors thank the patients, families, and physicians for making this work possible. They also thank Jingjing Jiang from Beijing Genome Institute for technical support, and Cong Hu, Jie Cen, Pancy Tam, and Sylvia Chiang for their special help with collecting the clinical data. 

Footnotes

 Supported by grants from the National Key Basic Research Program of China (973 Program No. 2013CB967500); National Natural Science Foundation of China (Grant Nos. 81170856, 81222009, and 81170867); Thousand Youth Talents Program of China (CZ); Jiangsu Outstanding Young Investigator Program (Grant No. BK2012046); Jiangsu Province's Key Provincial Talents Program (Grant No. RC201149); Applied Research Program of Science and Technology Commission Foundation of Tianjin (Grant No. 013111411); Foundation Fighting Blindness (DV); Ningxia Scientific and Technological Projects (Grant No. 2011ZYS175); and Shanghai Science and Technology Committee (114119b3000).

Footnotes

⁹  These authors contributed equally to the work presented here and should therefore be regarded as equivalent authors.

Footnotes

 Disclosure: X. Chen, None; K. Zhao, None; X. Sheng, None; Y. Li, None; X. Gao, None; X. Zhang, None; X. Kang, None; X. Pan, None; Y. Liu, None; C. Jiang, None; H. Shi, None; X. Chen, None; W. Rong, None; L.J. Chen, None; T.Y.Y. Lai, None; Y. Liu, None; X. Wang, None; S. Yuan, None; Q. Liu, None; D. Vollrath, None; C.P. Pang, None; C. Zhao, None