Autozygome‐guided exome‐first study in a consanguineous cohort with early‐onset retinal disease uncovers an isolated RIMS2 phenotype and a retina‐enriched RIMS2 isoform

Leber congenital amaurosis (LCA) and early‐onset retinal degeneration (EORD) are inherited retinal diseases (IRD) characterized by early‐onset vision impairment. Herein, we studied 15 Saudi families by whole exome sequencing (WES) and run‐of‐homozygosity (ROH) detection via AutoMap in 12/15 consanguineous families. This revealed (likely) pathogenic variants in 11/15 families (73%). A potential founder variant was found in RPGRIP1. Homozygous pathogenic variants were identified in known IRD genes (ATF6, CRB1, CABP4, RDH12, RIMS2, RPGRIP1, SPATA7). We established genotype‐driven clinical reclassifications for ATF6, CABP4, and RIMS2. Specifically, we observed isolated IRD in the individual with the novel RIMS2 variant, and we found a retina‐enriched RIMS2 isoform conserved but not annotated in mouse. The latter illustrates potential different phenotypic consequences of pathogenic variants depending on the particular tissue/cell‐type specific isoforms they affect. Lastly, a compound heterozygous genotype in GUCY2D in one non‐consanguineous family was demonstrated, and homozygous variants in novel candidate genes ATG2B and RUFY3 were found in the two remaining consanguineous families. Reporting these genes will allow to validate them in other IRD cohorts. Finally, the missing heritability of the two unsolved IRD cases may be attributed to variants in non‐coding regions or structural variants that remained undetected, warranting future WGS studies.

is not explained after routine clinical testing of all known genes. 1,2e use of whole exome sequencing (WES) beyond clinical IRD gene panels allows to reveal new gene-disease associations. 3The latter, however, is hampered by sporadic cases with only one affected individual or by lack of replication of the findings in a second unrelated family.For these reasons, genomic studies in consanguineous families have taken advantage of runs of homozygosity (ROH) in the search of pathogenic variants causing disease, 4,5 which has accelerated the discovery of novel candidate genes. 6In the field of IRD, several studies have identified novel disease genes in unsolved families using this approach, for instance from Saudi Arabia where a high degree of consanguinity is observed. 3,7,8In these studies, autozygosity mapping has been performed using high-density SNP arrays and different algorithms developed for WES data. 6,9A more recent high-performance homozygosity mapping tool is AutoMap, which uses WES or WGS data as input to generate ROH data of high specificity and sensitivity. 9other advantage of the use of WES in the study of heterogeneous diseases such as IRD is the simultaneous analysis of genes associated with non-syndromic and syndromic IRD, including overlapping groups of disease varying from photoreceptor degeneration to involvement of extra-ocular systems or organs. 10Examples are the most severe forms of IRD, namely Leber Congenital Amaurosis (LCA) and early-onset retinal degeneration (EORD), characterized by a profound visual impairment or blindness from birth or before the age of 5 years, respectively, with an estimated prevalence of 1 in 80 000 children worldwide. 11At least 25 genes have been associated with LCA including RPE65, being a target for FDA-approved gene therapy. 12To add to this complexity, also syndromic conditions such as Senior-Løken or Joubert syndrome can display LCA as part of the syndrome. 13re we present an exome-first study of 15 LCA/EORD families from a Saudi cohort, 12 of which have reported consanguinity.Apart from causative variants in known LCA genes and clinical reclassifications including non-syndromic RIMS2-IRD, we identified a retinaenriched RIMS2 isoform and propose two novel candidate IRD genes.

| Subjects and samples
Nineteen patients (8 females and 11 males) from 15 IRD families, ranging between 2 and 9 years, were recruited for this study.Fourteen families were recruited from Dhahran Eye Specialist Hospital (DESH), a tertiary care ophthalmic hospital in the eastern province of Saudi-Arabia in addition to one family that was recruited from the ophthalmic genetics clinic at Department of Ophthalmology, College of Medicine, King Saud University (KSU), Riyadh, Saudi Arabia.Twelve of these families originate from this eastern province, while three of them come from the central region of Riyadh.A detailed report including family history and pedigree assessments was obtained from all the families.The study received approval from KSU Institutional Review Board, IRB Project NO.E-23-7817.Written informed consent was obtained from the parents of all participating children before inclusion in this study.

| Genomic testing
Whole exome sequencing was performed in the 15 index cases of each family.
WES data analysis was performed using our in-house developed analysis platform Seqplorer, a graphical web interface that executes SQLite Gemini queries on an underlying database through straightforward dropdown menus and presents the results in a clear manner (https://github.ugent.be/cmgg/seqplorer),and evaluating gnomAD v4.0 population frequencies and in silico missense predictions (REVEL, 14 CADD, 15 MetaRNN 16 ) and splicing predictions (SpliceAI 17 ) in 14 and 1 index cases, respectively.A customized panel comprising 289 IRD genes (version 5 of the in-house RetNet panel, available at https://www.cmgg.be/assets/bestanden/GENPANEL-RETNET.pdf) was used for variant filtering.Only variants with a minor allele frequency (MAF) <0.005 were selected for the analysis.Splicing variants were further analyzed using Alamut Visual Plus Software (v.1.4)(SOPHiA GENETICS, Lausanne, Switzerland).Copy number variant (CNV) detection was performed using the ExomeDepth algorithm. 18l candidate (likely) pathogenic variants were confirmed by Sanger sequencing using the BigDye Terminator v.3.1 kit (Life Technologies) and underwent segregation analysis in all available family members.
Variants were classified according to the criteria of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP). 19

| Run of homozygosity (ROH) detection and variant prioritization
All 15 index cases underwent homozygosity mapping using AutoMap version 1.2 9 via the command-line package and using default parameters and hg38 as reference genome.The files generated included the predicted ROHs which were used afterwards for variant prioritization in consanguineous cases (n = 12).

| Haplotype analysis
For two cases sharing the same RPGRIP1 variant, haplotype analysis using the WES data was performed.Homozygous alternate and heterozygous variants located in the ROH involving RPGRIP1 were retained for further analysis and compared between the two index cases.The maximal shared region between both cases was calculated using the genomic coordinates of the two closest variants to the mutation in which the genotype of both cases was different.

| In silico analyses of novel candidate IRD genes
For the two unsolved consanguineous cases additional analysis was performed to identify potential candidate IRD genes.Variants located in the largest ROHs and with a MAF < 0.005 were selected and the associated genes were evaluated thoroughly through a literature search and various in silico tools for missense and splicing predictions described above.Tissue and single-cell expression of candidate genes was investigated using the Human Protein Atlas (HPA) (https://www.proteinatlas.org/) and retinal single-cell datasets such as Spectacle 20 and the IOB Retina Atlas. 21nctional studies in animal models were assessed in the Mouse Genome Database (MGD; http://www.informatics.jax.org). 22Only variants found in genes for which the retina was found to have a top ranked expression level in retina, according to the consensus dataset of HPA, consisting of normalized expression levels (TPM) for 55 tissues, with reported functional evidence and/or segregating with the disease, were proposed as potential causative variants.

| Transcriptional study of retina-enriched RIMS2 isoform
Paired-end FASTQ files (GSE115828) with transcriptome data derived from postmortem retina samples (Ratnapriya et al.) were retrieved. 23ly samples derived from donor retinas showing no features of agerelated macular degeneration were evaluated (n = 102).Transcripts were quantified through pseudoalignment by Kallisto (v.0.46.1) 24 using default parameters.Maximum likelihood expression estimates for all annotated (Ensembl human release 106) RIMS2 isoforms were retrieved in transcripts per million (TPM).To further assess the retinaspecificity of these isoforms, we inspected the whole locus using an integration of multiple publicly available multi-omics datasets derived from human retina (Table S1).Potential functional conservation in mouse was evaluated based on PhyloP conservation scores and mining of transcriptomic and epigenomic datasets generated from murine photoreceptors.

| Variants in known IRD genes underlie the clinical diagnosis in 73% of the cohort
WES in the index cases of 15 IRD families, followed by AutoMap analysis in all patients of consanguineous origin (12/15 families), revealed likely causative variants in 11 families (Table 1).This corresponds to a global detection rate of 73% (11/15) in this cohort.Most of the variants met ACMG criteria for (likely) pathogenicity except two variants classified as variants of uncertain significance (VUS) (Table 1).
After investigating IRD genes located in the ROHs in the 12 consanguineous families, 8 homozygous variants were found in 10/12 families (83%).The total size of identified ROHs per patient ranged from 79.73 to 504.53 Mb (Table 1).Segregation analysis allowed us to confirm the homozygous state of the identified variants in 4 additional affected members from whom a DNA sample was available.
Three different variants in CRB1 were found in 3 families, one of which c.2969T>C; p.(Leu990Ser) being novel.In 2 families we identified the same RPGRIP1 variant c.1107del; p.(Glu370Asnfs*5), with a common origin as confirmed via haplotype analysis.A maximal shared region of 23 kb between the SNPs rs746359185 and rs3748357 was observed in the WES data.The other homozygous causal variants were identified in the ATF6, CABP4, RDH12, SPATA7, and RIMS2 genes (Table 1).
In the remaining 3 non-consanguineous families we prioritized rare variants located in known IRD genes.In one family, we identified two variants in the GUCY2D gene in compound heterozygous state, for which c.3053T>C; p.(Ile1018Thr) variant is reported here for the first time.For one of the two unsolved non-consanguineous families, the total size of ROH was 158.2 Mb, suggestive of relatedness.However, no potential candidate variants were found within those regions.upstream (Figure 2E).This first exon, despite not being annotated, is syntenically conserved in mouse and exhibits retinal, particularly photoreceptor expression, thus suggesting functional conservation (Figure S1).We compared this retina-enriched isoform with the MANE isoform (ENST00000696799.1,RefSeq NM_001348484) and the isoform displayed in HGMD (ENST00000507740.6,Refseq NM_014677.5)(Figure 3 and Table S2).As shown in Figure 3, the exons expressed in retina correspond to the ones of the retinaenriched isoform.The evidence of functionality of this retina-specific isoform together with the isolated retinal phenotype of this patient point towards a specific effect of the variant in the retina.We analyzed the potential splicing defect of this variant due to its location in the first nucleotide of a donor splice site in the MANE isoform of Ensembl (ENST00000696799.).This variant is absent in reference population databases and in silico predictors classified it as pathogenic (Table 3).Segregation analysis confirmed the heterozygous state of this variant in both parents (Figure 4A).According to the HPA, ATG2B has its highest expression in retina (11.8 TPM), after cerebellum (12.1 TPM), and single-cell expression data revealed a predominance in horizontal, rod and cone photoreceptors and bipolar cells (Figure 4B and 4C).Available phenotypic data of a null/knockout mice in MGD showed an abnormal retinal morphology.

| Clinical reevaluation and genotype-driven reclassification
In the case of F13, a candidate pathogenic variant was located within the second largest ROH with a size of 49.59 Mb (99.55% homozygosity), within a total ROH content of 332.87 Mb.A homozygous missense variant was identified in RUFY3 (chr4:70773568G>C; c.754G>C; p.(Glu252Gln)), a gene with highest expression in retina (116.7 TPM), and expression in most retinal cell types, according to single-cell RNA-seq data (Figure 4B and 4C).The variant is only present once in heterozygous state in the Turkish Variome 25 database and once in gnomAD v4.0.In silico predictions for this variant pointing towards a benign consequence, CADD excepted (Table 3).Segregation analysis in the parents and the unaffected sister revealed that both parents are heterozygous carriers while the variant is absent in the latter (Figure 4A).

| DISCUSSION
0][31][32][33] In this study, 15 families of Saudi Arabia were clinically diagnosed with LCA/EORD and 73% (11/15) received a genetic diagnosis after a WES-based analysis.Of these solved families, 90% (10/11) display consanguinity and harbor homozygous pathogenic variants.Consanguineous pedigrees have been widely studied for the discovery of disease-causing variants in rare diseases 6 and in this study, pathogenic variants in IRD genes were found in 83% (10/12) of the pedigrees with reported consanguinity using autozygositydirected WES.Of the three non-consanguineous families however, only one has received a genetic diagnosis based on compound heterozygous variants in the LCA-associated gene GUCY2D.Other genomic studies of Saudi IRD cohorts have reported a diagnostic yield of 63% but when solely considering autosomal recessive cases this percentage steeply increases to 93%, with almost exclusively homozygous pathogenic variants being identified. 8st genes implicated in the current cohort are known LCA/EORD genes.Although the most prevalent LCA gene, CEP290, has not been found to be mutated in this cohort, other variants in genes with an approximate frequency of 3%-10% such as CRB1, RPGRIP1, RDH12 and SPATA7 34 were identified in 7/15 (46%) of our cases.First, three different homozygous CRB1 variants explained the disease in three families, one of which, p.(Leu990Ser), is novel.Second, a previously reported Saudi founder variant in RPGRIP1, p.
(Glu370Asnfs*5) 35 was identified in two a priori unrelated families that shared a common ROH region of 23 kb surrounding this variant.
Apart from variants in known LCA/EORD genes, variants were found in IRD genes not previously or rarely associated with LCA/EORD.First, in the patient harboring a homozygous ATF6 variant, a clinical reevaluation directed the clinical diagnosis towards achromatopsia.Second, the CABP4 gene has generally been associated with congenital stationary night blindness (CSNB) 36 and congenital cone-rod synaptic disorder (CRSD) 37 although the same variant found here has previously been reported in four siblings with an LCAlike phenotype. 38Although the oculodigital sign that has been observed in F10 has so far not been reported in patients with CABP4associated CRSD, it has been reported by us in RIMS2-associated CRSD, also known as incomplete congenital stationary night blindness (iCSNB). 28So it cannot be excluded that the LCA-like features observed here align with CABP4-CRSD.A new clinical reassessment of F10 was not possible, however.
Finally, a novel homozygous variant in the RIMS2 gene, recently associated with CRSD, was identified in a 9-years old child with a non-syndromic form of IRD.Up to date, only five RIMS2 variants have been reported in seven CRSD patients presenting neurodevelopmental disease or abnormal glucose homeostasis apart from their IRD.
However, intrafamilial phenotypic variability has been observed in patients with the same genotype. 28Only the eldest individual from a reported family with 3 clinically evaluated affected relatives presented insulin resistance, while the two others, a mother and son displayed neurological features.Given that RIMS2 has a role in the regulation of synaptic membrane exocytosis in pancreas, it is possible that patients with biallelic RIMS2 variants could develop pancreatic symptoms later in life, as previously suggested. 28Regarding the neurological phenotype, while the mother displayed only autistic behavior, her son was severely affected with marked neurodevelopmental delay at 1 year old. 28Our patient F1 was diagnosed at 3 months of age with LCA due to the presence of nystagmus and photophobia.Ophthalmological pancreas, this variant should always result in a truncated protein, similar to the other five pathogenic RIMS2 variants described before. 28 observed that exons present in the MANE isoform are not expressed in retina while exons expressed in retina are not present in the HGMD isoform, which highlights the importance of isoform-aware variant interpretation.
In addition, variants in novel genes-candidate IRD genes-can be causative but are missed when filtering for known IRD genes.Two F I G U R E 3 Overview of the RIMS2 isoforms analyzed in this study.RNA-seq data showed that the isoform expressed in retina corresponds to the retina-enriched isoform (yellow) identified in this study.reported, supporting AGBL5 as candidate IRD gene. 45,46Here, we aimed to identify candidate IRD genes in the two remaining unsolved consanguineous families focusing on variants located in ROH and found homozygous variants in two novel genes with remarkable expression in retina, ATG2B and RUFY3.Apart from its retinal expression, functional evidence of retinal degeneration has been reported in mice lacking Atg2b in the Mouse Genome Informatics (Atg2bem1 (IMPC)J, MGI:6275189, C57BL/6NJ-Atg2bem1(IMPC)J/Mmjax mouse strain).8][49] In the case of RUFY3 there is no reported functional evidence for its role and expression in the retina.Hence, both ATG2B and RUFY3 are proposed as potential IRD candidate genes to be assessed in other IRD cohort to further validate them as IRD genes.
Finally, in the two remaining unsolved non-consanguineous families without candidate genetic variants, non-coding variants affecting splicing or SVs might explain part of the unsolved cases, as has been reported in other IRD studies. 50,51 conclusion, we took advantage of performing autozygomedriven WES in a consanguineous IRD cohort to identify causative variants in known LCA/EORD genes (CRB1, GUCY2D, RDH12, RPGRIP1, SPATA7), and to establish genotype-driven clinical reclassifications (ATF6, CABP4, RIMS2).We identified an isolated, non-syndromic RIMS2-IRD and revealed a retina-enriched RIMS2 isoform.Although the relationship of the latter with the non-syndromic RIMS2-IRD is to be proven, it may suggest potential different phenotypic consequences of pathogenic variants depending on the particular tissue/ cell-type specific isoforms they affect.We report two candidate IRD genes, which is important to be able to assess them in other IRD cohorts.Indeed, their validation as IRD gene is needed before they can be included in IRD genetic testing panels in a clinical setting.
Finally, despite the high overall diagnostic yield of 73%, the missing heritability of the unsolved IRD cases may be attributed to variants in non-coding regions or SVs that remained undetected by the WES approach, warranting WGS approaches in future studies.
All affected patients were diagnosed through an ophthalmological examination including best-corrected visual acuity (BCVA) by projected Snellen charts, slit-lamp biomicroscopy, and fundus examination.All patients underwent retinal imaging using a VX-10 Alpha Combination Mydriatic/Non-Mydriatic retinal camera (Kowa American Corporation, California, USA).Color vision was reevaluated in one patient after the genetic diagnosis (ATF6-genotype) using the Ishihara 24-plate.Full-field Electroretinography (ffERG) (Diagnosys LLC, Lowell, Massachusetts, USA) was performed in all patients.Light-and dark-adapted responses were measured in accordance with the International Society for Clinical Electrophysiology of Vision (ISCEV) extended protocol. 14Macular spectral-domain optical-coherence tomography (SD-OCT) (Heidelberg SPECTRALIS, Heidelberg, Baden-Wurttemberg, Germany), SD-OCT of the retinal nerve fiber layer thickness (RNFL-SD-OCT) was performed in one patient.

2. 3 |
Genetic and genomic studies 2.3.1 | DNA extraction EDTA blood was drawn from the 19 patients and available family members for DNA extraction and downstream molecular studies using the ReliaPrep Large Volume HT gDNA Isolation System (Promega, Leiden, The Netherlands) according to the manufacturer's protocols.

A
Abbreviations: ACMG/AMP, American College of Medical Genetics and Genomics and the Association for Molecular Pathology; het, heterozygous state; hom, homozygous state; LP, likely pathogenic; Mb, Megabase; P, pathogenic; ROHs, runs of homozygosity; VUS, variant of uncertain significance.

3. 3 |
Novel phenotypic association of RIMS2-IRD and retina-enriched RIMS2 isoform After assessing known IRD genes in the proband of consanguineous pedigree F1 we identified a novel homozygous variant in a canonical splice donor site of RIMS2.As this gene has been implicated in congenital cone-rod synaptic disorder with neurodevelopmental disease and occasional anomalies of glucose homeostasis, 28 a further systemic evaluation was performed in this patient.The pediatric neurological examination revealed no signs of neurodevelopmental problems and normal glucose homeostasis.Ophthalmological examination including color wide-field fundus photographs showed bilateral optic disc pallor with mild vessels attenuation and general RPE mottling without peripheral pigmentary migration in both eyes at age 9 (Figure2).BCVA was 20/100 in both eyes and apart from horizontal nystagmus no other eye abnormalities were observed.Full field ERG revealed absent response in light and dark conditions in both eyes.Macular SD-OCT showed retinal thinning at the level of inner retina, while SD-OCT of the retinal RNFL-SD-OCT showed bilateral temporal RNFL atrophy in both eyes (Figure2).Due to the exceptional non-syndromic and retina-specific phenotype observed in F1 with the identified novel RIMS2 variant, we performed data mining of RIMS2 expression in human.Isoform ENST00000436393.6,not annotated in RefSeq, was found to be retina-specific as supported by exclusive retinal expression of the first exon and binding of retinal specific transcription factors (CRX, OTX2) PhotophobiaNystagmusPupillary reflexStrabismus ODS Fundus appearance examination revealed a phenotype similar to the one reported by Mechaussier et al., including absent pigmentary migration and thinning of the inner retina, although in F1 the ERG was flat instead of electronegative.Given the isolated retinal phenotype of F1, we performed further analyses to assess a potential effect on splicing of the identified variant.Interestingly, we observed a unique RIMS2 transcript with a retina-specific first exon, among all the different RIMS2 transcripts.Although this retina-specific exon is also conserved in mouse, the corresponding retina-enriched mouse isoform seems unannotated in public databases.The novel RIMS2 splice variant identified in this study, located in the first donor site of exon 9, 10, or 14 of the HGMD, retina-enriched or MANE isoforms, respectively, is predicted to induce skipping of those exons, resulting in truncated proteins for all isoforms.Unless tissue-specific or cryptic splice sites would be used in these and other RIMS2 isoforms expressed in the other tissues associated with RIMS2-pathogenicity, such as brain and

F I G U R E 1
Pedigree and fundus image of the patient homozygous for the novel ATF6 c.493_494del; p.(Leu165Aspfs*64) variant identified in this study.(A) Pedigree of F9. (B) Fundus (left eye) of 8-year-old girl (F9), displaying retinal vessels attenuation and welldefined, oval-shaped retinal pigment epithelial defects at the foveal area.V1, variant allele; wt, wild-type allele.studies of Saudi cohorts, Abu-Safieh et al. and Patel et al., proposed several candidate IRD genes. 3,8In the first study, 6 genes were presented as candidate IRD genes, and since their publication in 2013, only additional IRD cases with pathogenic variants in C21orf2, KIAA1549, and ACBD5 have been reported. 9-44This is similar to the findings of Patel et al., after which only AGBL5 variants have been F I G U R E 2 Novel RIMS2 variant in a non-syndromic patient and a retina-enriched RIMS2 isoform.(A) Pedigree of family F1.(B) Color widefield fundus photograph.(C) Macular spectral-domain optical-coherence tomography (SD-OCT).(D) SD-OCT of the retinal nerve fiber layer thickness (RNFL-SD-OCT).Only images of the right eye are shown.(E) We identified a retina-enriched RIMS2 isoform: the first blue highlight corresponds to the transcription start site (TSS) of the MANE RIMS2 isoform ENST00000696799.1.We observed that the first exon of RIMS2 isoform ENST00000436393.6 is highly conserved (second blue highlight) and expressed in retina, but not in other relevant tissues such as brain.Active transcription is supported by bulk RNA-seq, Nuc-seq and CAGE-seq derived from human retina.The retinal specificity of this isoform is further confirmed by signatures of open chromatin, H3K4me2 and retinal transcription factor binding sites (CRX and OTX2).V, variant allele; wt, wild-type allele.