Genetic spectrum of retinal dystrophies in Tunisia

We report the molecular basis of the largest Tunisian cohort with inherited retinal dystrophies (IRD) reported to date, identify disease-causing pathogenic variants and describe genotype–phenotype correlations. A subset of 26 families from a cohort of 73 families with clinical diagnosis of autosomal recessive IRD (AR-IRD) excluding Usher syndrome was analyzed by whole exome sequencing and autozygosity mapping. Causative pathogenic variants were identified in 50 families (68.4%), 42% of which were novel. The most prevalent pathogenic variants were observed in ABCA4 (14%) and RPE65, CRB1 and CERKL (8% each). 26 variants (8 novel and 18 known) in 19 genes were identified in 26 families (14 missense substitutions, 5 deletions, 4 nonsense pathogenic variants and 3 splice site variants), with further allelic heterogeneity arising from different pathogenic variants in the same gene. The most common phenotype in our cohort is retinitis pigmentosa (23%) and cone rod dystrophy (23%) followed by Leber congenital amaurosis (19.2%). We report the association of new disease phenotypes. This research was carried out in Tunisian patients with IRD in order to delineate the genetic population architecture.

Scientific RepoRtS | (2020) 10:11199 | https://doi.org/10.1038/s41598-020-67792-y www.nature.com/scientificreports/ genic variant in PRPH2 in family 17 (F17) (3.8%). Segregation of the mutant allele was confirmed in the majority of the families. For missense variant the substituted amino acid residues are highly conserved across species, and in silico pathogenicity prediction tools PolyPhen2 and SIFT predicted these changes to be deleterious. After molecular testing, all patients were re-evaluated to monitor whether their retinal phenotype was similar to previously described retinopathies caused by pathogenic variants in the same gene. In case of discrepancy, the respective phenotypes were considered as potential novel genotype-phenotype correlations.
Information for each patient is presented in Table 2. Below we present the families with novel P/LP variants.
LCA (Fig. 1, 2). Clinical data of patients from families F1, F2, F3, F4 and F5 revealed an age of onset of disease from birth with nystagmus and photophobia. BCVA was limited to light perception. Patient in F3 was monophthalmic of the right eye (RE) (Fig. 1).
F1-RPGRIP1 (Fig. 1A). One novel homozygous deletion (NM_020366: c.3113-3114delCT, p.T1038Rfs*8) in RPGRIP1 was identified in F1 with 2 affected members. The deletion of the CT in exon 10 results in a frameshift with a premature stop codon at position amino acid 1046.
F2-GUCY2D (Fig. 1B). The second novel homozygous likely pathogenic variant (NM_000180: c.2660 T > G, p.V887G) in GUCY2D was found in F2 in one affected individual. This likely pathogenic variant has previously been published in the validation of a targeted array but no phenotype was presented 6 .
F4-IQCB1. The homozygous likely pathogenic variant (NM_001023570: c.994C > T, p.R332*) in IQCB1 was found in F4 in two affected children ( Fig. 1F and G). Fundus appearance in both probands revealed normal structure. Renal function and ultrasound were normal.
F16-NMNAT1 (Fig. 4F).. The homozygous pathogenic variant (NM_001297778.1: c.37G > A, p.A13T) in NMNAT1 segregated with the disease in F16, with two affected members (F16-V.1, F16-V.4) (Fig. 2). This pathogenic variant has previously been reported as causing LCA 8 . In contrary to this report, our affected patients had relatively preserved visual acuity until the third decade (Table 2). On fundus examination, we found normal macular aspect with few bone spicule shaped pigment deposits and white dot deposits, large areas of atrophy in the mid periphery (Fig. 4F).
Bardet-Biedl (Fig. 2,4). Clinical reassessment of extraocular symptoms was also performed in 4 patients from two families (F22, F23) who were shown to have Bardet-Biedl syndrome with retinal dystrophy, obesity and     Fig. 4M and N).

Discussion
The data presented here showed that a number of genes can cause IRD in this Tunisian cohort. Taking together with our previous report, the analysis of 73 Tunisian families highlights the mutational load in IRD by identifying likely disease-causing genes in more than 25 genes in 50 families associated with different forms of IRD. A total of 50 likely disease causing alleles were identified, including 8 nonsense pathogenic variants, 10 deletions, 1 duplication, 1 complex rearrangement, 6 splice-site alleles and 24 missense alleles considered potentially pathogenic, 42% of which were novel. In addition, a novel finding from this study was the evidence of high frequency of ABCA4, RPE65, CRB1 and CERKL pathogenic variants in Tunisian families with IRD 9-12 (Fig. 5). Homozygosity mapping combined with systematic screening of known genes resulted in a positive molecular diagnosis in 68.4% families. This is in accordance with several reports 13,14 and is similar to the 75% frequency reported in Saudi Arabia 15 . However, in Spanish cohort studying large sporadic IRD groups (877 patients), the diagnostic yield was 44% 16 . On one hand, this difference might be explained by the selection of the cohort analyzed, as we chose families with either two affected individuals or sporadic cases with particular phenotype and we excluded patients with Usher syndrome. On the other hand, this may be due to the high frequency of consanguineous marriages in our cohort. This work provides an overview of the mutational spectrum of IRD in Tunisian cohort (Fig. 5) which gives an the most frequent genes in our cohort of patients with retinal disorders were: 14% ABCA4 (p.E1087K, p.W782*, p.Y639C, p.P1380L, p.N380K and dup32-40; del45-47), 8% RPE65 (p.R91W, p.H182Y, p.R234* and c.1129-2A > G), 8% CRB1 (p.R764H, p.P836T, p.Y702C and c.3542 + 1G > A) with variable phenotypes of severe IRD, ranging from LCA to RP as previously reported 17 and 8% CERKL (c.1133 + 3_1133 + 6delAAGT). As expected, our results look more similar to the Spanish cohort where CRB1 (7%), ABCA4 (7%), CERKL (4%) and EYS (4%) were the most frequent mutated genes 16 . Compared to other ethnic groups, however, the most prevalent pathogenic genes in Saudi Arabia were KCNV2, RP1, TULP1, RPGRIP1, CRB1 and RPE65 respectively 15   www.nature.com/scientificreports/ In this study, we also highlighted the importance of combining molecular and clinical data to correctly diagnose IRDs. We would like to point out that in the subset of families analyzed in this study potential diseasecausing variants were detected in 19 genes, out of which 8 have not yet been described in association with the observed IRD phenotype: RPGRIP1, GUCY2D, CRB1, CDHR1, C8ORF37, ABCA4, CNGB1 and TRPM1. Ophthalmic and genetic findings are presented in Table 3.
Ophthalmic investigation identified characteristic signs and symptoms of LCA in five families (F1, F2, F3, F4 and F5). This group of the most severe and the earliest occurring IRD resulting in congenital blindness 18 typically becomes evident in the first year of life like in our five families represent 19.2% of our cohort. To date, genetic heterogeneity of LCA is well known, with 24 genes currently implicated in its pathogenesis 19 . Molecular analysis in our families identified three new pathogenic variants: novel homozygous deletion c.3113_3114delCT identified in RPGRIP1. The second missense pathogenic variant p.V887G is localized in GUCY2D, gene, estimated to account for 20% of LCA cases 20 and constitute the most common cause of the disease. The third is a novel splice variant c.3542 + 1G > A in CRB1, the most commonly mutated gene in our cohort (8%) with variable phenotypes of severe IRD, ranging from LCA to RP as previously reported 21 .
NGS allows for the screening of a large number of genes implicated in the pathophysiology of IRD. In F4, despite having a previously reported homozygous pathogenic variant p.R332* in IQCB1, the phenotypes of our index patient and their affected sister never showed dysplasia in any organ; rather, they only had LCA. Usually, defects in this gene result in Senior-Loken syndrome type 5 (SLSN5), where degenerative phenotypes involving kidneys and eyes are common 22 , but sometimes the phenotype only shows LCA, as presented in this family 23,24 .
We report ophthalmic and genetic findings of patients with RP, composed of 10 patients, with AR-RP presented in 8 patients and 2 patients with AD-RP phenotype, which represent 23% of our cohort. patients with  Most LCA-associated pathogenic variants are located in a segment that encodes two C2 domains 45 . Some RPand LCA-causing pathogenic variants in either RHD or RID were shown to impair the interaction between the two 46 . These data may explain this phenotypic variation in our patient  25 . We also describe a novel homozygous splicing pathogenic variant c.470 + 1G > T in C8orf37, observed in F9. The localization of genetic abnormalities has previously been described by Rahner et al., where 56% of the pathogenic variants are located in exon 6 in the C-terminal region of C8orf37 and the majority of reported variants are splicing variants 26 . We identified new homozygous pathogenic variant p.Y639C in family F11. ABCA4 pathogenic variants were responsible for 14% of cases in our cohort for a wide variety of IRD phenotypes from AR Stargardt disease to CRD and, in some advanced cases RP 27 . Therefore, no clear genotype-phenotype could be established. Three different pathogenic variants were identified in CNGB1, which represent 6% of our cohort. One of these variants is a novel homozygous pathogenic variant p.R765C in F19 where the index patient presented typical symptoms of RP. Unexpectedly, two probands in family F16 with RP had damaging missense pathogenic variant, p.A13T in nicotinamide adenine dinucleotide (NAD) synthase gene NMNAT1. This pathogenic variant has been previously identified in patients with LCA 8 . Although the proband's phenotype is consistent with RP, and the pathogenic variants are predicted to be deleterious, our patients showed well preserved visual acuity. Fundus examination revealed bone spicule-like pigment deposits and white spot deposits at the mid-periphery.
Our data provide an overview of the mutational spectrum of IRD in a Tunisian cohort, which gives an idea about genes spectrum in North Africa. We demonstrate a high degree of genetic complexity in both, the causative disease genes and their associated phenotypes, highlighting uncommon genotype-phenotype correlations and contributing to the current knowledge about disease-causing variants. We realise that this study presents some limitations, such as relatively small number of patients and the lack of complete ophthalmic and other examinations. Ideally, the efficacy of genotype-phenotype correlation could be improved with a complete ophthalmic examination, including ERG in all patients.

Methods clinical data and sample collection. 150 families were evaluated at the Department B of Hedi Rais
Institute of Ophthalmology, Tunis; Tunisia, over the course of 15 years. We selected a subset of families which accepted to be part of the study, with onset of the disease in the first or second decade of life, with clinical diagnosis of AR-IRD excluding patients with unclear or unlabeled diagnosis of retinal dystrophy or for whom multimodal imaging exploration could not be carried out. 73 families fulfilled these criteria.
In this study, we draw up a report on 26 families. Consanguinity involving first-cousin marriages was observed, and in the non-consanguineous families, most marriages were between individuals from the same geographic origin and the highest number of cases was recorded in the region of Nabeul containing 15 families. A questionnaire was used to collect information which included socio-demographic data (age, gender, geographical origin, educational level, occupation, socio-economic level) family medical, surgical and ophthalmological history, age of onset and duration of symptomatology (onset date has been defined by the age of onset of the first visual symptoms), the disease course (defined as either stationary or progressive) and any additional nonocular findings, such as deafness, mental retardation, polydactyly, obesity, heart disease or other malformations. Each patient had complete ophthalmological examination including best corrected visual acuity (BCVA) using Snellen chart, slit-lamp biomicroscopy, dilated fundus examination, retino-photos, optical coherence tomography (Swept source DRI OCT-A Triton®,Topcon, Tokyo, Japan), fundus autofluorescence (Heidelberg Spectralis; Heidelberg-Engineering, Heidelberg, Germany) and some patients also received full-field electroretinogram (ERG) (Métrovision, France).
This study was approved by the Local Ethics Committee of the Hedi Rais Institute. Peripheral blood samples were obtained from the index patient and from some of the family members, including parents and affected siblings. Informed consent was obtained from all participants and a parent and/or legal guardian for participants under the age of 18 years old. Analyses were done in accordance with local guidelines. DNA was extracted from leukocytes according to the salting-out method 28 .
Whole exome sequencing (WeS). Exome capture was performed using the Roche Nimble-Gen version 2 (44.1-megabase pair) and paired-end multiplexed sequencing was carried out on an Illumina HiSeq 2000 system (Illumina, San Diego, CA, USA) by Otogenetics Corporation (Norcross, Georgia, USA) using DNA samples from all index cases. Homozygosity was evaluated from SNPs obtained by WES.
Sequence data alignment, variant calling and identification. The Illumina paired-end DNA sequence data were mapped and aligned to the reference human genome NCBI Build 37 (hg19) using the Nextgene software package v.2.3.5. (Soft-genetics, State College, PA). Median coverage of the target region was 95X with 96% of target region covered by at least 10 reads.
The resulting list of homozygous gene variants was compared to the IRD genes found in RetNet (https ://sph. uth.edu/retne t/disea se.htm).
Sanger sequencing. Identified variants were validated by Sanger sequencing and segregation analysis. DNA was amplified by PCR using FastStart PCR Master Kit (Roche, Basel, S) and sequenced as previously described 11 . Primers and condition of each PCR are provided in Table 4. Fragments were sequenced on an ABI 3100XL DNA automated sequencer (Applied Biosystems, Foster City, CA).