Toward the Mutational Landscape of Autosomal Dominant Retinitis Pigmentosa: A Comprehensive Analysis of 258 Spanish Families

Abstract

Purpose: To provide a comprehensive overview of the molecular basis of autosomal dominant retinitis pigmentosa (adRP) in Spanish families. Thus, we established the molecular characterization rate, gene prevalence, and mutational spectrum in the largest European cohort reported to date.

Methods: A total of 258 unrelated Spanish families with a clinical diagnosis of RP and suspected autosomal dominant inheritance were included. Clinical diagnosis was based on complete ophthalmologic examination and family history. Retrospective and prospective analysis of Spanish adRP families was carried out using a combined strategy consisting of classic genetic techniques and next-generation sequencing (NGS) for single-nucleotide variants and copy number variation (CNV) screening.

Results: Overall, 60% of our families were genetically solved. Interestingly, 3.1% of the cohort carried pathogenic CNVs. Disease-causing variants were found in an autosomal dominant gene in 55% of the families; however, X-linked and autosomal recessive forms were also identified in 3% and 2%, respectively. Four genes (RHO, PRPF31, RP1, and PRPH2) explained up to 62% of the solved families. Missense changes were most frequently found in adRP-associated genes; however, CNVs represented a relevant disease cause in PRPF31- and CRX-associated forms.

Conclusions: Implementation of NGS technologies in the adRP study clearly increased the diagnostic yield compared with classic approaches. Our study outcome expands the spectrum of disease-causing variants, provides accurate data on mutation gene prevalence, and highlights the implication of CNVs as important contributors to adRP etiology.

Retinitis pigmentosa (RP) is the most prevalent form of inherited retinal dystrophy (IRD), with an overall prevalence of 1:4000.¹ RP is characterized by primary degeneration and death of rod photoreceptor cells followed by secondary loss of cone photoreceptors.² Symptoms usually start with night blindness, followed by progressive restriction of the peripheral visual field, culminating in progressive visual acuity loss.¹ Ophthalmological examination shows pigment deposits typically seen as bone spicules in the peripheral retina, attenuation of blood vessels, and waxy pallor of the optic disc evidenced on slit-lamp examination of the eye fundus, and diminished or even nonrecordable signals on ERG.^1,3 However, clinical features, disease onset, and progression may vary significantly between patients, even those belonging to the same family.² 

RP is characterized by its extreme genetic heterogeneity, with all inheritance patterns described.³ Autosomal dominant RP (adRP) accounts for approximately 15% of Spanish RP families.⁴ Currently, adRP-causing variants have been described in 28 genes (in the public domain, https://sph.uth.edu/retnet/sum-dis.htm#B-diseases), with incomplete penetrance reported in some specific genes.⁵ Given the underlying complexity of adRP, identifying its genetic basis is a challenging task, although one that may not only aid in clinical diagnosis and counseling, but also contribute to the development of future gene-based therapies. 

Different molecular approaches, such as indirect screening methods and genotyping microarrays, have been developed to improve the genetic testing of adRP.^6,7 However, these costly and laborious techniques cannot address the full spectrum of disease-causing variants, thus leading to low diagnostic rates. Nowadays, next-generation sequencing (NGS), mainly gene panel–based strategies, represents the primary first-tier approach for all forms of IRD, yielding higher diagnostic rates.⁸ The application of NGS in patients with IRD allows for a hypothesis-free analysis of the genes involved, regardless of the phenotype or inheritance pattern.^9–11 Although NGS is expected to increase diagnostic rates in adRP,⁸ only the exhaustive analysis of a large cohort of American patients has been reported⁵; therefore, the actual performance of NGS on adRP forms has not been yet fully explored. 

Copy number variants (CNVs) also play a substantial role in adRP¹² and could explain, at least partially, some of the missing heritability associated with adRP. Recently, it has been demonstrated the capability to detect CNVs by read-depth strategies using NGS data sets.¹³ However, the exact contribution of CNVs in adRP remains largely unknown due to lack of comprehensive screening in most of the NGS studies. 

In this work, we provide a comprehensive overview of the molecular basis and mutational spectrum of adRP in the largest cohort of European origin reported to date, comprising Spanish families, using different genetic approaches, including NGS and CNV screening. 

Subjects and Methods

Cohort Description

All subjects were selected by reviewing the database of IRD patients at Fundación Jiménez Díaz University Hospital (FJD, Madrid, Spain), spanning 27 years of collected data. The adRP cohort includes 258 unrelated Spanish families with clinical diagnosis of RP reached after a complete ophthalmologic examination^4,14 and at least two consecutive affected generations. Informed consent was signed by all the participating subjects. DNA samples were collected from the FJD Biobank. This study was reviewed and approved by the Ethics Committee of our hospital, according to the tenets of the Declaration of Helsinki and subsequent revisions. 

Mutational Screening of Single-Nucleotide Variants (SNVs)

Different genetic tools were used over time as first-tier diagnostic approach (Table 1): (1) a “classical” algorithm based on screening of the most prevalent adRP-associated genes by single-strand conformation polymorphism (SSCP), CG-clamped denaturing gradient gel electrophoresis (DGGE),⁶ Sanger sequencing, and/or commercial Arrayed Primer Extension (APEX)-based genotyping microarrays (Asper Biotech, Tartu, Estonia) to detect previously known mutations in adRP genes⁷; and (2) targeted NGS of 74 IRD-associated genes (Supplementary Table S1) using the TruSightOne Sequencing Panel (Illumina, San Diego, CA, USA). This targeted clinical exome captures 12 Mb of genomic sequence enriched for 62,000 exons, spanning 4813 genes associated with known inherited diseases in Online Mendelian Inheritance in Man (in the public domain, https://www.omim.org/) and Human Gene Mutation Database (HGMD; in the public domain, http://www.hgmd.cf.ac.uk). Only coding exons and 10 bp of flanking 5′ and 3′ intronic sequence were covered with this approach. Libraries were prepared according to the manufacturer's instructions and sequenced in groups of 24 samples on the NextSeq500 platform (Illumina) using the NextSeq HighOutput kit. Bioinformatic analysis was performed using the online platform BaseSpace (Illumina) and, subsequently, variant annotation and prioritization was made with the VariantStudio software (Illumina). During this analysis, potentially pathogenic variants were prioritized in a subpanel of 74 IRD-associated genes. 

Table 1

 

Molecular Strategy Used in Our Cohort of Spanish Patients With adRP

 

A total of 105 index cases uncharacterized with the classical algorithm were further analyzed by NGS using custom gene panels or whole-exome sequencing (WES), as previously reported.^8,15,16 

Pathogenicity of variants was established according to their allele frequency in gnomAD (in the public domain, http://gnomad.broadinstitute.org/), in silico prediction tools for novel splicing and missense variants as implemented in the commercial Alamut software (Interactive Biosoftware, Rouen, France), and segregation in the family when other relatives were available.⁸ Besides, all the variants were also manually checked in HGMD and ClinVar databases (in the public domain, https://www.ncbi.nlm.nih.gov/clinvar/), and in the literature. All identified variants classified as pathogenic or likely to be pathogenic were confirmed by Sanger sequencing. 

CNV Analysis

CNV analysis of adRP-associated genes was performed using two different approaches. First, Multiplex Ligation-dependent Probe Amplification (MLPA) using the P235 kit (MRC-Holland, Amsterdam, The Netherlands) was conducted in 103 uncharacterized families after classical approach or custom NGS gene panel.¹² 

Additionally, in the cases with NGS data available, we applied the open source CoNVading software (Copy Number Variation Detection In Next-Generation Sequencing Gene Panels, in the public domain, https://molgenis.gitbooks.io/molgenis-pipelines/content/convading/CoNVaDING.html), which uses a read-depth strategy that is capable of identifying CNVs at the exon level. Default parameters were used. Final results were obtained for each sample after manual and bioinformatics data curation. Further validation of positive CNV calls was carried out by MLPA using the P235 or P221 kit. 

CNVs were further delimited by array-based comparative genomic hybridization (aCGH).¹² In cases of CRX deletions, a commercial or a customized 60K aCGH was performed (Agilent Technologies Inc, Santa Clara, CA, USA). 

Haplotype Analysis

Three microsatellite markers with high heterozygosity (D19S606, D19S596, and D19S879) flanking 1.5 Mb the CRX gene and an intragenic polymorphic marker (D19S902) were analyzed in two families with apparently the same causative variant to determine a possible common ancestor. 

Results

Mutation Detection Rate

The overall diagnostic detection rate in our Spanish adRP cohort was 60% (155/258) after applying sequential SNV and CNV screening over the past 27 years. Overall, our families have been analyzed using several molecular approaches, each obtaining different molecular diagnostic yields (Table 1): a classical approach in 226 families, NGS in 32 families, or both approaches in 105 families. The diagnostic rate has clearly increased over time with the implementation of new high-throughput approaches, from approximately 38% (87/226) obtained with classical molecular screening to 66% (21/32) of current characterization yield for “naïve” patients analyzed by NGS, the first-tier approach used since 2015. Additionally, NGS analysis allowed us to characterize 47 of 105 retrospective negative adRP cases. More specifically, disease-causing variants were detected by the following: initial direct (Sanger) or indirect (SSCP/DGGE) screening of the most prevalent genes in 46 families; screening of known mutations through an adRP mutation chip in 41 families; NGS in 60 cases; and MLPA or CNV analysis in 8 families. A total of 103 (40%) families remain unsolved; however, 34 of them could not be analyzed by the most recent NGS studies. 

Mutational Spectrum

Our cohort showed high allelic and genetic heterogeneity. A total of 114 different, likely causal SNVs and CNVs were found (Tables 2, 3). Fifty-three of them have been identified in our adRP cohort for the first time. Eighteen variants showed recurrence in two families, and only seven variants in the RHO, NR2E3, C1QTNF5, and PRPF3 genes can be considered as common, being present in three or more unrelated families and accounting for 13.6% (35/258) of the families (Supplementary Table S2). The most prevalent mutations seen in our cohort were p.(Pro347Leu) in the RHO gene, followed by p.(Gly56Arg) in the NR2E3 gene. 

Table 2

 

Spectrum of Likely Pathogenic SNVs Identified in Our Cohort of 258 adRP Spanish Families

 

Table 3

 

Large Genomic Rearrangements on 19q13.42, 19q13.32, and 3q22.1, Affecting PRPF31, CRX, and RHO, Respectively

 

Mutations in 26 different genes were identified in our study (Table 4): 17 adRP-related genes in 138 families (53.5%), 1 autosomal dominant late-onset retinal degeneration gene in 4 different families (1.6%), 3 X-linked RP associated genes in 8 families (3.1%), and 5 different autosomal recessive IRD-associated genes in 5 families (1.9%). Pedigree and segregation analysis for these reclassified families are shown in Supplementary Figure S1. 

Table 4

 

Prevalence of Disease-Causing Mutation in 258 adRP Spanish Families, Including Types of Mutation in Each Gene

 

Four adRP-associated genes accounted for 62% (96/155) of the characterized families, with RHO being the most prevalent gene, followed by PRPF31, RP1, and PRPH2. In general, most of the adRP-associated causal variants were missense, except for PRPF31, RP1, CRX, and TOPORS, in which loss-of-function variants were the most prevalent (Table 4). 

Copy Number Variation

We identified eight CNVs affecting three different adRP genes in seven unrelated families (Table 3) using a combination of MLPA and bioinformatic analysis of NGS data. The prevalence of CNVs in our adRP cohort was 3.1% (8/258), and this type of mutational event accounted for 5% (8/160) of all the identified variants. 

PRPF31 had the highest prevalence of CNVs in our series, with five different variations identified in four families (1.9%: 5/258), including two contiguous gene deletions, two partial deletions involving a single exon, and one partial multi-exon duplication. In a particularly noteworthy finding, we report a novel partial deletion spanning exon 9 in family RP-0076 identified using NGS data (Fig.); all remaining deletions have been previously described.¹² 

Figure

 

Overview of novel CNVs identified in this study. (A) Schematic representation of deletions in the CRX and PRPF31 genes in three patients by read-depth NGS data analysis. Each exon is represented by a circle. Normal exons are indicated in black and deleted exons are indicated in red. (B) Refinement by aCGH analysis of CRX deletions on chromosome 19 in families RP-1192 and RP-1092. Red bar represents the genomic positions of the minimum and maximum deletion size of 8010 bp (chr19:48339345–48347355) and 11905 bp (chr19:48338052–48349957), respectively, in family RP-1192. Green bar represents the genomic positions of the minimum and maximum deletion size of 6313 bp (chr19:48339216–48345529) and 12007 bp (chr19:48336134–48348141), respectively, in family RP-1092. The intragenic marker D19S902 is represented by a black rectangle. Schematic representation of the complete intron-exon structure of CRX in blue and the repeat element (SINE) in black are shown. Exons are indicated by rectangles. (C) Validation of the deletion in the PRPF31 gene by MLPA and segregation analysis in the RP-0076 family. The segregation analysis showed incomplete penetrance, as individual III:3 was an asymptomatic carrier.

A partial deletion in the CRX gene encompassing exons 3 and 4 was identified in two adRP families (RP-1092 and RP-1192) by NGS (Fig.). Further MLPA and aCGH analysis allowed us to confirm, segregate, and refine descriptions of these deletions. Both probands had an overlapping genomic region spanning from intron 2 to the 3′ untranslated region of CRX; however, we could not delimit the exact breakpoints. No other genes were involved in the distal 3′ region of either deletion. Haplotype analysis ruled out a possible founder effect for this CNV (Supplementary Figure S2). 

Clinical and Genetic Reclassification of X-linked and Recessive Families

NGS analysis revealed that 13 families (5%) had been misclassified as autosomal dominant (Supplementary Figure S1). A closer analysis of these families showed likely disease-causing X-linked variants in RPGR (four families), RP2 (two families), and CHM (two families). Similarly, we identified biallelic pathogenic variants compatible with autosomal recessive inheritance in five families, including a homozygous splice-site variant in EYS, and compound heterozygous variants for ABCA4, CNGA1, RDH5, and USH2A. The families carrying the RDH5 and USH2A variants were only partially solved and, to date, additional variants explaining pseudo-dominant inheritance have not yet been identified. 

Ophthalmologic reevaluation was performed in those recessive cases in which the respective genes were associated with a clinical diagnosis other than RP. In the RDH5 case, the proband had night blindness, nonrecordable scotopic ERG, normal disc and retina vessels, small white dots in the midperiphery to the posterior pole, and no macular involvement on fundus examination. As with the genetic findings, these symptoms were compatible with the features of retinitis punctata albescens. The proband with variants in USH2A, did not have hearing loss, and no such symptom was reported in the family history. Therefore, the phenotype was suggestive of RP rather than Usher syndrome. In the family carrying ABCA4 mutations, the phenotype was compatible with a cone-rod dystrophy, with progressive loss of visual acuity followed by visual field constriction and total blindness at 45 years of age. 

Discussion

We present the largest mutational screening carried out in an adRP cohort with European ancestry, specifically constituted of Spanish families. Therefore, our study provides an accurate overview of the prevalence of adRP-associated genes and mutational events. This cohort has been screened for more than 25 years by means of the molecular strategies available at the time, which gives us an interesting insight into how diagnostic rates for adRP have changed over time with the development of more innovative and high-throughput techniques. Direct or indirect screening of the most frequently mutated genes in combination with genotyping microarrays achieved a significant diagnostic rate of 38%. After NGS and CNV analysis, the overall detection rate in our cohort increased to 60%. This is higher than that found in the German (41%)¹⁰ and Belgian cohorts (56%),¹⁷ but lower than the 78.5% reported by the most comprehensive study to date, performed on a large North American cohort.⁵ The presence of the founder p.Pro23His mutation in RHO, which is almost exclusively from the United States, in 13.2% of American adRP families⁵ could explain the difference in detection rates between the two cohorts. Additionally, our cohort had greater allelic diversity. Our study is also somewhat biased, as 34 (13.2%) of our 258 adRP families could not be studied by NGS; therefore, it is likely that the rate of characterized families has been slightly underestimated. Taking into account that 66% of naïve cases were characterized when NGS was applied as a first-tier approach, the actual diagnostic yield for adRP is expected to be between 60% and 70%. 

In our cohort, mutations in RHO and PRPF31 were the most prevalent, explaining 21% and 8% of our families, respectively, followed by RP1 and PRPH2. Supplementary Table S3 shows the characterization rate and the percentage of mutations in these genes in different adRP populations. The frequency of involvement of the adRP-associated genes is quite similar to that reported for adRP in the US population,⁵ with the exception of RP1, which is the fifth most prevalent mutated gene after PRPH2 and RPGR.⁵ RHO is known to be the most prevalent adRP-associated gene worldwide; however, the prevalence and mutational spectrum of this gene seem to vary according to the population studied, ranging from 7% in South Africa¹⁸ to 31% in American cohorts.⁵ 

Most of the variants (89 variants) identified in our population were private mutations for their respective family, 18 mutations were seen twice, and 7 different mutations in the RHO, NR2E3, C1QTNF5, and PRPF3 genes were found in three or more families, accounting for 13.6% of the solved families. Among them, the most prevalent mutations seen in our cohort were the p.(Pro347Leu) in the RHO gene, described as a hotspot,¹¹ followed by the founder mutation p.(Gly56Arg) in the NR2E3 gene in our population.¹⁹ For the other five variants p.(Arg135Trp), p.(Ala164Glu), and p.(Gly182Ser) in RHO, p.(Ser163Arg) in C1QTNF5, and p.(Thr494Met) in the PRPF3 gene, it would be necessary to perform haplotype analyses in these families to determinate if these variants are founder mutations in our population. 

Recently, pathogenic CNVs have been found to be an important and underestimated cause of human retinal dystrophies.¹³ Multi-exon deletions have been described in several IRD-associated genes, accounting for an increasing percentage of uncharacterized patients.¹³ Our study was pioneering in the implementation of comprehensive CNV analysis for molecular characterization of adRP, using both MLPA and/or read-depth NGS analysis, highlighting the important role of structural variants in the etiology of adRP. Remarkably, up to 3.1% of our patients carried this type of underestimated variant, mainly in PRPF31 and CRX. The high frequency of CNVs for both genes could be related to their genomic location on chromosome 19, which is a well-known enriched region in Alu repeats.²⁰ Similar to that described for genomic PRPF31 rearrangements,²¹ the most plausible mechanism underlying gross CRX deletions seems to be a nonallelic homologous recombination between nearly identical Alu repeats, which are present at both 5′ and 3′ boundaries of the deleted regions (Fig.). Interestingly, the proportion of structural variants in both genes was similar to that obtained for truncated variants. However, with the exception of the CRX, PRPF31, RP1, and TOPORS genes, missense variants were the most prevalent cause of adRP in our cohort, presumably due to a mechanism different from haploinsufficiency. Thus, the adRP-associated disease mechanism seems to be gene-specific, but also it may also be related to the type of defect and gene localization. 

We carried out a hypothesis-free analysis using different targeted NGS approaches that include genes not only associated with RP but also with other rare forms of IRD, which allowed us to perform clinical and genetic reclassification of some families. In four families, we found the pathogenic variant p.(Ser163Arg) in the C1QTNF5 gene that has been previously associated with a characteristic phenotype of late-onset retinal degeneration (LORD).^16,22 This retinal form is an autosomal dominant ocular disease characterized mainly by night blindness in the fifth and sixth decades of life, in which patients showed mottled yellow-white deposits, central and peripheral degeneration, and choroidal neovascularization on eye fundus.²² Index cases from these four families noticed night blindness as the first symptom at 50 to 60 years and were referred to our center with an initial diagnosis of late-onset RP, some of them presented with macular alterations and drusoid material on fundus examination. LORD can be easily misdiagnosed as AMD or classical RP in early or later stages of disease, respectively.¹⁶ In view of the relative high frequency of this variant in our cohort and the inherent difficulty of LORD clinical diagnosis, we believe that this gene should be taken into consideration during NGS screening of cases with suspected clinical diagnosis of late-onset RP. 

Eight families carrying pathogenic variants on X-linked genes were initially misclassified as adRP due the presence of affected females. Mutations in the X-linked RP genes, RPGR and RP2, have been previously described in up to 8.5% of families with initial diagnosis of adRP and affected carrier females.²³ It is well known that there is a large proportion of causal mutations at the 3′ end of RPGR, a highly repetitive region of exon open reading frame 15 (ORF15), that is poorly covered in the NGS assays.²² Our proportion of characterized adRP families with disease-causing mutation in RPGR and RP2 is lower (2.4%) than reported by Churchill et al.,²³ and it could be because some mutations in this region could be missed by capture-based NGS approaches. This highlights the need of ORF15 screening by alternative Sanger sequencing in all uncharacterized adRP families with no male-to-male transmission. 

Although carrier females of CHM mutations are usually asymptomatic,²⁴ there are some previously reported cases of females fully affected by choroideremia.²⁵ This complete penetrance of heterozygous CHM variants could be explained by nonrandom lyonization of chromosome X, inactivating the wild-type allele.²⁶ Choroideremia is generally a clinical diagnosis made by eye-fundus examination and family history,²⁷ although in our two families the available phenotypic and pedigree data were scarce, and therefore the diagnosis mainly relied on the results of molecular genetic testing. Therefore, it would be highly recommended to perform an ophthalmologic examination in these families. 

In five additional cases, unexpected biallelic pathogenic variants for known recessive genes were identified. In the RP-1217 family, the observed pseudo-dominant pattern was explained by the segregation of a homozygous variant in several consecutive generations in a context of high inbreeding. Additionally, pseudo-dominant inheritance in other families may also be justified by the relatively high carrier frequency of some mutations in the Spanish population for some IRD genes with high allelic heterogeneity.²⁸ This is the case of family RP-1455, with three different pathogenic ABCA4 variants segregating over several generations. 

There are several reasons that would explain the lack of causative variants in the noncharacterized studied patients. First, some families could carry disease-causing variants in regions not targeted by our NGS approaches, such as deep intronic regions of the studied genes, recently identified adRP-associated genes (ADIPOR1, ARL3, PRPF4, and SPP2), or novel genes not yet discovered. Second, despite the successful identification of genomic rearrangements in this study, we could not rule out the possibility that there may be additional CNVs involving partially covered regions by the NGS approaches used here. Therefore, further WES and/or whole-genome sequencing studies in these uncharacterized patients could help us to identify novel adRP genes. 

In conclusion, the combination of different strategies in adRP molecular testing over time has allowed us to achieve high diagnostic rates in our Spanish cohort. Our study confirms the great diagnostic value of gene panel–based NGS tools, which currently represent the most reliable, cost-effective, and efficient approaches for identifying not only disease-causing SNVs, but also CNVs in adRP families. It was also evidenced that the use of a hypothesis-free approach allows for the genetic and/or phenotypic reclassification of a relevant number of families, which could have a huge impact on family counseling, clinical follow-up, and future enrollment in gene therapy–based treatments. 

Acknowledgments

The authors thank the Genetics and Ophthalmology Departments of the Fundación Jiménez Díaz University Hospital (Madrid, Spain), Oliver Shaw for English review and editing the manuscript, and all patients and doctors who participated in the study. 

Supported by grants from the Instituto de Salud Carlos III (ISCIII) of the Spanish Ministry of Health, including CIBERER (06/07/0036), FIS–FEDER (European Regional Development Fund) (PI16/00425), and IIS-FJD Biobank PT13/0010/0012. In addition, the Spanish National Organization for the Blind (ONCE), the Spanish Fighting Blindness Foundation (FUNDALUCE), and the Ramon Areces Foundation also supported this work. IM-M is sponsored by the IIS-Fundación Jiménez Díaz-UAM Genomic Medicine Chair. MC is supported by the Miguel Servet Program (CP12/03256) from ISCIII. 

Disclosure: I. Martin-Merida, None; D. Aguilera-Garcia, None; P. Fernandez-San Jose, None; F. Blanco-Kelly, None; O. Zurita, None; B. Almoguera, None; B. Garcia-Sandoval, None; A. Avila-Fernandez, None; A. Arteche, None; P. Minguez, None; M. Carballo, None; M. Corton, None; C. Ayuso, None