Homozygosity Mapping and Genetic Analysis of Autosomal Recessive Retinal Dystrophies in 144 Consanguineous Pakistani Families.

Purpose: The Pakistan Punjab population has been a rich source for identifying genes causing or contributing to autosomal recessive retinal degenerations (arRD). This study was carried out to delineate the genetic architecture of arRD in the Pakistani population.
Methods: The genetic origin of arRD in a total of 144 families selected only for having consanguineous marriages and multiple members affected with arRD was examined. Of these, causative mutations had been identified in 62 families while only the locus had been identified for an additional 15. The remaining 67 families were subjected to homozygosity exclusion mapping by screening of closely flanking microsatellite markers at 180 known candidate genes/loci followed by sequencing of the candidate gene for pathogenic changes.
Results: Of these 67 families subjected to homozygosity mapping, 38 showed homozygosity for at least one of the 180 regions, and sequencing of the corresponding genes showed homozygous cosegregating mutations in 27 families. Overall, mutations were detected in approximately 61.8 % (89/144) of arRD families tested, with another 10.4% (15/144) being mapped to a locus but without a gene identified. Conclusions: These results suggest the involvement of unmapped novel genes in the remaining 27.8% (40/144) of families. In addition, this study demonstrates that homozygosity mapping remains a powerful tool for identifying the genetic defect underlying genetically heterogeneous arRD disorders in consanguineous marriages for both research and clinical applications.

Hereditary retinal dystrophies (RD) constitute a group of inherited retinal diseases characterized by chronic and progressive visual impairment, genetic heterogeneity, and significant clinical overlap among the different disorders. To date, mutations in more than 200 genes are known to cause the different forms of RD[1] as summarized in RetNet (https://sph.uth.edu/Retnet/; in the public domain). Retinal dystrophies can be inherited as autosomal recessive (ar), autosomal dominant (ad), and X-linked (xl), as well as rare mitochondrial and digenic traits.[2] Rod or cone photoreceptor degenerations are two main groups of RDs.

Among the rod RDs, retinitis pigmentosa (RP, MIM no. 268000), a genetically and clinically heterogeneous retinal degeneration, is the most common worldwide,[3] having a worldwide prevalence estimated to be approximately 1 in 4000 individuals.[4-14] Currently, mutations associated with RP have been identified in more than 82 genes, of which 58 have been shown to be relevant to arRP (RetNet). However, that these 82 genes are responsible only for around 60% of RP[15,16] suggests that the number of currently unidentified genes causing RP might be quite high. In addition, the cone RDs, including cone- or cone–rod dystrophies (CORD) and the macular dystrophies, which mainly affect the central vision, have been associated with more than 30 genes (RetNet).

Pakistan has the highest prevalence of consanguineous marriages in the world, presumably because this practice provides a number of social and economic advantages.[17] In a review of all published retinal degeneration cases in Pakistan, only 4 families with compound heterozygous mutations were identified in 146 (2.7%) genetically resolved arRD families,[18] further supporting the utility of homozygosity mapping in this population. Therefore to identify causative mutations in an ongoing study of large Pakistani arRD families with multiple affected individuals, we carried out homozygosity mapping of known RD loci followed by mutation screening of the genes in homozygous loci in 67 consanguineous families with arRD from Pakistan as a part of an ongoing international collaboration between the National Eye Institute (NEI), National Institutes of Health (NIH), United States, and the National Centre of Excellence in Molecular Biology (NCEMB), Allama Iqbal Medical College, and the National Centre for Genetic Diseases, Shaheed Zulfiqar Ali Bhutto Medical University in Pakistan.

Materials and Methods

Enrollment and Clinical Assessment of arRD Families

This study was approved by the Institutional Review Boards of the National Centre of Excellence in Molecular Biology and the Combined NeuroScience Institutional Review Board at the National Institutes of Health. Written informed consent consistent with the tenets of the Declaration of Helsinki was obtained from participating individuals or their guardians before the study. Families segregating arRD with three or more affected individuals were identified by visiting eye hospitals in Pakistan, mostly in the Punjab. Blood samples were drawn from potentially informative family members, and genomic DNA was extracted from leukocytes according to standard protocols.[19] All participants underwent a detailed family, ophthalmic, and medical history, and selected individuals were evaluated by visual acuity, best-corrected visual acuity, slit-lamp biomicroscopy, ophthalmoscopy, fundus photography, and electroretinography (ERG). Previously, as part of this project, 77 families had been mapped by linkage analysis to specific chromosomal locations; the causative gene and mutations had been identified in 62 while for the remaining 15 only the locus had been identified. For the current homozygosity exclusion mapping study, 67 families were selected from the remaining unlinked families based on the availability of DNA samples and consanguineous marriages of the parents of affected individuals. Eight of these families have undergone whole genome linkage analysis without identified causative genes while the remaining 59 families were not screened. Families with possible dominant or X-linked inheritance were excluded, and some families with only two affected offspring of consanguineous matings were included in the early parts of the study.

Homozygosity Mapping and Linkage Analysis

One hundred eighty genes or loci associated with inherited retinal diseases were selected from RetNet (https://sph.uth.edu/Retnet/) and screened by homozygosity exclusion mapping (Table 1). Homozygosity genotyping of 188 microsatellite markers was done in one affected individual of each family. Loci homozygous in the first individual were genotyped in a second affected family member and, if also homozygous, in a third affected offspring of consanguineous parents. When when no single marker with a heterozygosity of 75% or greater was available, two markers were tested. We tested two markers within 1 to 2 cM of the candidate gene with heterozygosities over 50%, but approximately 50% of families would be expected to show discordant results for these markers. Thus our assessment was that most discordant homozygosity would be the result of low information content rather than recombination between the two markers. Families uninformative for these markers were genotyped using additional surrounding markers. Families in which homozygosity was shared only by all affected siblings were further investigated by genotyping additional individuals for confirmation of cosegregation. A variant of the multiplexing short tandem repeat with tailed primers (MSTP) approach described by Oetting et. al.,[20] using fluorescently labeled tagged primers homologous to extensions on initial primers in a two-PCR approach, was used to genotype these microsatellite markers. The PCR products were multiplex electrophoresed on an ABI 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA), and fragment sizes were determined by GeneMapper version 4.0 (Applied Biosystems). Primer sequences and PCR conditions are shown in Supplementary Table S1.

Table 1

Summary of Mic rosatellite Markers for Homozygous Mapping

Two-point linkage analyses were performed using the FASTLINK modification of the MLINK program in the LINKAGE program package.[21,22] Maximum logarithm of the odds (LOD) scores were calculated using ILINK, and LINKMAP was used for multipoint analysis. Autosomal recessive RD was analyzed as a fully penetrant trait with an affected allele frequency of 0.00001. The criteria for establishing linkage have been described previously.[23] The length of the homozygous regions was 3 Mb on average. Haplotypes were generated using the Cyrillic 2.1 program (Cyrillic Software, Wallingford, Oxfordshire, UK) and confirmed by inspection.

Screening Candidate Genes

Based upon cosegregation of the risk haplotypes in a family, mutations in the exons and 100 bp of flanking intronic regions of the included known candidate gene associated with inherited retinal diseases were analyzed by Sanger sequencing using ABI PRISM 3130 automated sequencers (Applied Biosystems) and assembled and analyzed with Seqman software (DNAStar Lasergene 8; Madison, WI, USA) and Mutation Surveyor (SoftGenetics, State College, PA, USA). Mutations were submitted to the LOVD (http://databases.lovd.nl/shared/variants; in the public domain).

Assessing Pathogenicity of Identified Variants

A mutation was considered novel if it was not present in the Human Mutation Database Professional Version on Biobase (https://portal.biobase-international.com/cgi-bin/portal/login.cgi; in the public domain) or the National Center for Biotechnology Information dbSNP database (http://www.ncbi.nlm.nih.gov/snp/; in the public domain), and sequence changes were considered pathogenic when they segregated with the disease in the family as well as their absence in 192 ethnically matched control chromosomes or at a frequency > 1% in the ExAC database (http://exac.broadinstitute.org/; in the public domain); and for missense changes were judged pathogenic in a computational test for mutations, including sorting intolerant from tolerant (SIFT and PROVEAN, http://sift.jcvi.org/; in the public domain) analysis, polymorphism phenotyping (PolyPhen2, http://genetics.bwh.harvard.edu/pph2/; in the public domain), and Condel (http://bg.upf.edu/fannsdb/; in the public domain). A SIFT score below the cutoff of 0.05 for a given substitution is classified as damaging while those with scores higher than this value are considered tolerated. In addition, we used Condel[24] (CONsensus DELeteriousness score of missense SNVs), which computes a weighted average of the scores (WAS) of five tools: SIFT, PolyPhen2, MAPP (Multivariate Analysis of Protein Polymorphism), LogR Pfam E-value, and Mutation Assessor. Splicing changes were predicted using Automated Splice Site Analyses (http://www.fruitfly.org/seq_tools/splice.html; in the public domain).

Intragenic Haplotype Analysis for Families Sharing the Same Variation

If the same variation was detected in more than one family, haplotypes of single nucleotide polymorphisms (SNPs) intragenic or within 1 Mb of the mutated gene were genotyped. The frequency of the risk haplotype in the general population was calculated from 96 unrelated Pakistani controls via the CHM algorithm as implemented in the Golden Helix SVS package (Golden Helix, Bozeman, MT, USA).

Results

Patient Cohort and Homozygosity Screening

This cohort included 67 consanguineous families with more than two affected siblings. A total of 67 probands and all affected siblings or ancestors who received a clinical diagnosis of arRP were included in this study. Up to three affected individuals in each family were subjected to homozygosity screening in a stepwise manner. After screening the first affected patient with all 188 microsatellite markers, the average number of excluded markers for each family was 126 (126/188 = 67% efficiency), with exclusion rates of 85% after screening a second affected individual and 93.0% a third, so that the average number of markers to be analyzed was reduced from 188 to 13 in each family (Fig. 1). After fine mapping using microsatellite markers, 29 families were excluded from linkage to any known RP gene, making the exclusion rate 43.3% for families undergoing homozygosity screening. The remaining 38 families showed homozygosity and cosegregation in at least 1 of the 180 regions containing known arRD genes or loci.

Figure 1

Workflow of this study.

After sequencing the included candidate genes for which the remaining 38 families showed homozygosity and cosegregation, the underlying pathogenic mutations were revealed in 27 families (Table 2).[25-28] In addition, variations considered benign by in silico prediction were detected in another four families (Table 3). The two-point LOD scores for markers in the homozygous chromosomal segments in these families are shown in Table 4. In seven families, no variations were identified by sequencing the known candidate genes within identified homozygous chromosomal regions, and their significance remains unclear.

Table 2

Mutations Detected in 67 arRD Families Subjected to Homozygosity Mapping

Table 3

Variations of Unknown Significance Detected in Three Genes of 67 arRD Families

Table 4

Two-Point LOD Scores of arRD Gene Markers in 31 Families

Identification of Disease-Causing Variants in Homozygous Regions

Overall, 24 causative mutations of 20 genes were identified in 27 families, including 12 missense, 6 nonsense, 4 indel-induced frameshift mutations, and 2 splice-site mutations (the splicing site change in CNGB1 was induced by an indel mutation). In all, 11 mutations were novel, and were not found in mutation databases or in 192 ethnically matched control chromosomes, while the remaining 13 mutations had been reported previously. Each of the mutations was located within a homozygous region and cosegregated with the disease. For missense mutations, the substituted amino acid residues are highly conserved across species (Figs. 2–7), and in silico pathogenicity evaluation by PolyPhen2, SIFT, and Condel of the 12 missense mutations predicted these changes to be deleterious. Some variations were detected in two families, including c.536A>G (p.(K179R), family 61198 and 61199) and c.758T>G (p.(M253R), family 61035 and 61126) in RDH5, c.1466A>G (p.(K489P), family 61301 and 61309) in TULP1, and the probably nonpathogenic c.1946C>T (p.(S649L), family 61237 and 61267) in PROM1. Affected families who shared the same variations also shared a common haplotype of alleles at nearby intragenic SNPs, suggesting that the mutant allele was probably derived from a common ancestor (Supplementary Table S2). Although genes or loci previously associated only with autosomal dominant inherited retinal disease were also screened in the study, no mutations were identified in these genes or loci.

Figure 2

Family 61220, 61166, 61219, 61086, 61042 and 61036 structure and haplotype of flanking markers of loci identified by homozygosity mapping. DNA sequence tracings confirming the mutations are shown adjacent to the pedigrees, and cross-species conservation of amino acids showing missense mutations is shown on the right.

Figure 7

Family 61301, 61309 and 61191 structure and haplotype of flanking markers of loci identified by homozygosity mapping. DNA sequence tracings confirming the mutations are shown adjacent to the pedigrees, and cross-species conservation of amino acids showing missense mutations is shown on the right.

Family 61217, 61065, 61198, 61199,61035 and 61126 structure and haplotype of flanking markers of loci identified by homozygosity mapping. DNA sequence tracings confirming the mutations are shown adjacent to the pedigrees, and cross-species conservation of amino acids showing missense mutations is shown on the right.

Family 61113, 61262, 61231, 61312 and 61206 structure and haplotype of flanking markers of loci identified by homozygosity mapping. DNA sequence tracings confirming the mutations are shown adjacent to the pedigrees, and cross-species conservation of amino acids showing missense mutations is shown on the right.

Details of Sequence Variations Identified by Homozygosity Mapping

Information regarding the 24 sequence variations felt likely to be causative identified by homozygosity screening, and phenotypes of the 27 families in which they occurred, is provided in Table 2, and similar information for variations judged likely to be benign based on in silico predictions or presence in unaffected control individuals is provided in Table 3. The domain structure of the corresponding proteins and locations of the mutations with respect to these are shown in Figures 8 and 9, while the pedigree structure and corresponding haplotypes as well as sequence tracings and cross-species conservation for missense changes are shown in Figures 2 through 7, with the corresponding LOD scores summarized in Table 4.

Figure 8

Domain structure and mutations of proteins in which mutations were identified by homozygosity mapping: CDHR1, CEP290, CERKL, CNGA3, CNGB1, CNGB3, EYS, GRK1, GRM6, LCA5.

Figure 9

Domain structure and mutations of proteins in which mutations were identified by homozygosity mapping: LRAT, LRP5, NR2E3, PROM1, RBP3, RDH5, RDH12, RP1, RPE65, RPGRIP1, TULP1, USH2A.

Details of Sequence Variations Considered Likely to Be Deleterious

The c.847C>T, p.(R283*) mutation in CERKL seen in family 61219 is predicted to lead to truncation of the protein through a premature stop codon, and might be expected to lead to nonsense-mediated decay. However, this mutation has been shown to cause accumulation of truncated protein in the nucleus.[29] The c.55C>T, p.(R19*) mutation in GRK1 seen in family 61015 was predicted to lead to truncation of GRK1 protein and result in nonsense-mediated decay, and seems likely to result in a variant form of Oguchi disease, as described by Zhang et al.,[30] also predicted to result in nonsense-mediated decay; however, detailed clinical data sufficient for this diagnosis could not be obtained, so it is listed in Table 2 as having a stationary RD. Family 61058 (Table 2) contains a c.652C>G, p.(R218G) mutation in LCA5 segregating in a pseudo-dominant inheritance pattern, but the recessive nature of the mutation is confirmed by the sequencing results (Fig. 4). The c.227G>A p.(R76Q) mutation in the DNA-binding domain (DBD) of NR2E3 (Fig. 9) had been shown to increase dimerization significantly but to abolish DNA binding.[31]

Figure 4

Family 61058, 61076, 61169, 61150, 61237 and 61267 structure and haplotype of flanking markers of loci identified by homozygosity mapping. DNA sequence tracings confirming the mutations are shown adjacent to the pedigrees, and cross-species conservation of amino acids showing missense mutations is shown on the right.

Details of Sequence Variations Considered Likely to Be Benign

The previously described missense variation c.1208G>A, p.(R403Q) in CNGB3[32] seen in family 61221 (Fig. 3) was identified in a homozygous state in 1 of 96 control individuals. It was predicted to be probably damaging by PolyPhen2 but tolerated by SIFT and neutral by Condel, so it seems most likely only to be a rare variation (Table 3), although a modifying gene affecting the phenotype cannot be excluded. The novel c.4268C>T, p.(P1423L) variation in LRP5 was in an amino acid evolutionarily conserved from humans to zebrafish (Figs. 4, 9), had an allele frequency of 2.529 × 10−5 in the ExAC browser with no homozygotes listed in any population, and was not seen in 96 control individuals of Pakistani ethnic extraction. However, in silico analysis with PolyPhen2 predicted that it would be benign, with SIFT showing that it was tolerated, and Condel that it was neutral (Table 3). Hence, we treated this change as a rare variation even though the possibility that it might be responsible for disease in this family remains, especially since the phenotype in this family was progressive, consistent with a less severe mutation.

Figure 3

Family 61221, 61192, 61016, 61015 and 61155 structure and haplotype of flanking markers of loci identified by homozygosity mapping. DNA sequence tracings confirming the mutations are shown adjacent to the pedigrees, and cross-species conservation of amino acids showing missense mutations is shown on the right.

A novel missense variation c.1946C>T, p.(S649L) in PROM1 was detected and cosegregated with the disease in families 61237 and 61267 (Table 3; Fig. 4). S649 is only weakly conserved from human to chicken, but the substitutions, from serine to threonine, are relatively conservative (Fig. 4). This mutation does not occur in a transmembrane domain (Fig. 9) and is predicted to be benign by three in silico analyses. Thus, we presumed that this mutation is a rare variation, even though p.(S649L) was not found in 192 control chromosomes from the Pakistani population and had an overall frequency of 5.456 × 10−5 in the ExAC database, with all six variant alleles identified in the South Asian population with no homozygotes. Affected members of the two families share a common haplotype of alleles at 10 consecutive SNPs in and around PROM1 suggesting that the variant allele is derived from a common ancestor (Supplementary Table S2).

Clinical Features

An overview of the clinical data from affected individuals in each family is shown in Table 2. The clinical symptoms, age of onset, and mode of inheritance in these families are generally consistent with the form of retinal degeneration described in the literature and on RetNet. However, the clinical data available for this study are limited to ophthalmic history and examination, fundus photographs, and ERGs, so that it is difficult to distinguish closely related forms of the retinal disease, for example, Leber congenital amaurosis (LCA) or CORD from RP. Thus, we use RD to represent those families in which the clinical findings are consistent with the previously described retinal disease, but were insufficient to distinguish unambiguously among the various diseases previously associated with that gene: RP, LCA, congenital stationary night blindness (CSNB), or fundus albipunctatus, which together account for 8 of 27 (30%) families. Four of these had an early onset, but the differentiation between early-onset arRP and LCA, difficult with the best documentation, could not be made reliably. Thirteen families showed clear signs of RP. One family was consistent with LCA with typical fundoscopy, extinguished ERGs, and a history of a congenital onset accompanied by nystagmus. Four families that carried a homozygous RDH5 mutation received a diagnosis of fundus albipunctatus by fundus photographs. Of the remaining two families, one showed the typical signs of CSNB without progression, and the second was consistent with Usher syndrome (RP with deafness). Although there was no indication of vestibular dysfunction in any affected individual in this family, suggesting Usher type 2 and consistent with the causative gene, this was simply listed as Usher syndrome because formal vestibular testing was not performed.

Discussion

Consanguineous matings have long been known to increase the risk of recessive diseases by increasing the fraction of the genome that is homozygous and identical by descent, and thus the number of potentially deleterious alleles descended from a common ancestor[33]; and homozygosity mapping provides an efficient means of localizing causative genes for recessive traits in these populations,[34] particularly in populations with high consanguinity rates. Pakistan, with consanguinity rates ranging from 17% to 38%,[35] is an optimal country in which to implement this approach.

In addition, the clinical phenotypes for some families are of particular interest. One such was family 61015, shown in Table 2 as having a stationary form of RD. The phenotype in this family, as far as it could be ascertained, was consistent with mutations in GRK1 causing Oguchi disease and was similar to that in a previous Pakistani family with a variant of Oguchi disease due to deletion of exon 3,[30] although the degree of recovery from dark adaptation could not be ascertained in this family. Thus, it is listed simply as a stationary RD, even though it is very likely to have Oguchi type CSNB. Another was family 61150, with a mutation in NR2E3. This gene primarily has been associated with enhanced S-cone syndrome and Goldmann-Favre syndrome. However, the phenotype in family 61150 is most consistent with progressive arRP, similar to that observed in a family of Portuguese “crypto-Jews” with a mutation in this gene.[36] In addition, Sharon et al.[37] have described NR2E3 mutations in a series of patients with clumped pigmentary retinal degeneration, who also might be consistent with the phenotype observed in this family.

Taken together with our previous characterization of autosomal recessive retinal degenerations (arRD) in Pakistan, we have studied a total of 144 consanguineous arRD families, in addition to 154 families with limited numbers of affected individuals or unclear inheritance patterns. Of these 144 families, we have identified putative causative variations in 40 genes and 11 loci so far, and these genes and loci collectively account for disease in 104 of the 144 families (72.2%) (Table 5; Fig. 10). The percentages of families who had variants in the 40 genes and 11 loci are shown in Figure 10, in decreasing order: RPE65 6.9% (10/144), TULP1 6.9% (10/144), RP1 4.9% (7/144), PDE6A and locus 4.9% (7/144), USH2A and new locus 3.5% (5/144), RDH5 2.8% (4/144), 11p11.2-q13.2 locus 2.1% (3/144), GRM6 2.1% (3/144), and 1p13.3 locus 2.1% (3/144). It should be noted that this tabulation counts families sharing an intragenic haplotype as separate families. Other genes or loci were identified in only one or two families in this cohort, respectively, accounting for less than 2% of the arRD population in Pakistan.

Table 5

Sequence Variations Identified in 144 Unselected Pakistani Families

Figure 10

Contributions of specific genes and loci to Pakistani families with arRP.

The most frequently mutated genes in arRD differ remarkably among the populations of different ethnic origins. RPE65 and TULP1 were the genes most frequently mutated in Pakistani patients with arRD, while the genes found to be most frequently mutated in other populations were RP1 in Saudi Arabians,[38] RDH12 in Spanish,[39] and USH2A worldwide.[15] The overall rate of variant detection was 61.8% (89/144) in this study, which is comparable to the worldwide variation detection rate of 60%.[15] A recent report that summarized 103 published Pakistani RD families[18] found AIPL1 and CRB1 to be the most frequent causative genes, probably because LCA families accounted for approximately 20% of the arRD families, while there were significantly fewer LCA families in our patient cohort. In addition, the previous summary included families screened by sequencing previously identified candidate genes, which might tend to favor those genes identified early and/or widely publicized. In our study, sharing of variations by different families is likely to be the result of the variant allele being derived from a common ancestry, since all of the families that shared the same variation also shared common intragenic SNP haplotypes for the associated gene. In addition, although we included the 79 genes/loci reportedly responsible only for autosomal dominant RD, no mutations were identified in genes previously associated only with adRD.

Taken together, we failed to uncover the pathologic variation in 27.8% of families in our arRD patient cohort. Although 38 families showed homozygosity and cosegregation in at least 1 of the 180 regions containing known arRD genes or loci, no disease-causing mutations were identified in 11 of these, so that they are excellent candidates for identification of new arRD genes residing within linked regions in which no mutations were identified in the known candidate gene. The most promising approach to discover novel genes after such a systematic analysis of many consanguineous families is next-generation sequencing, although familial locus heterogeneity or compound heterozygous mutations might explain the phenotype of a small proportion of the families, as intrafamilial locus heterogeneity was detected in 15.3% of the families studied in a recent report of Pakistani families with hearing impairment.[40] In addition, compound heterozygous mutations were identified in 2.7% of genetically resolved arRD families in a review of all published retinal degeneration cases in Pakistan.[18] Although homozygosity mapping has been proven effective, a major limitation is that this type of analysis will overlook familial locus heterogeneity or compound heterozygous mutations.

In conclusion, homozygosity mapping of known genes is relatively inexpensive while still being accurate and comprehensive. Our results provide a key bridge between bench and bedside and should make genetic diagnosis of arRD in patients more accessible and practical. This should greatly enhance the clinical genetic counseling, diagnosis, and early intervention of arRD in the Pakistani population. These results also highlight the importance of analyzing the causative genes and their exons in different ethnic groups in a systematic and population-specific fashion.

Click here for additional data file.

Click here for additional data file.