Splice-site mutations identified in PDE6A responsible for retinitis pigmentosa in consanguineous Pakistani families

• Introduction
• Methods
• Results
• Discussion

Abstract

Purpose: This study was conducted to localize and identify causal mutations associated with autosomal recessive retinitis pigmentosa (RP) in consanguineous familial cases of Pakistani origin.

Methods: Ophthalmic examinations that included funduscopy and electroretinography (ERG) were performed to confirm the affectation status. Blood samples were collected from all participating individuals, and genomic DNA was extracted. A genome-wide scan was performed, and two-point logarithm of odds (LOD) scores were calculated. Sanger sequencing was performed to identify the causative variants. Subsequently, we performed whole exome sequencing to rule out the possibility of a second causal variant within the linkage interval. Sequence conservation was performed with alignment analyses of PDE6A orthologs, and in silico splicing analysis was completed with Human Splicing Finder version 2.4.1.

Results: A large multigenerational consanguineous family diagnosed with early-onset RP was ascertained. An ophthalmic clinical examination consisting of fundus photography and electroretinography confirmed the diagnosis of RP. A genome-wide scan was performed, and suggestive two-point LOD scores were observed with markers on chromosome 5q. Haplotype analyses identified the region; however, the region did not segregate with the disease phenotype in the family. Subsequently, we performed a second genome-wide scan that excluded the entire genome except the chromosome 5q region harboring PDE6A. Next-generation whole exome sequencing identified a splice acceptor site mutation in intron 16: c.2028–1G>A, which was completely conserved in PDE6A orthologs and was absent in ethnically matched 350 control chromosomes, the 1000 Genomes database, and the NHLBI Exome Sequencing Project. Subsequently, we investigated our entire cohort of RP familial cases and identified a second family who harbored a splice acceptor site mutation in intron 10: c.1408–2A>G. In silico analysis suggested that these mutations will result in the elimination of wild-type splice acceptor sites that would result in either skipping of the respective exon or the creation of a new cryptic splice acceptor site; both possibilities would result in retinal photoreceptor cells that lack PDE6A wild-type protein.

Conclusions: we report two splice acceptor site variations in PDE6A in consanguineous Pakistani families who manifested cardinal symptoms of RP. Taken together with our previously published work, our data suggest that mutations in PDE6A account for about 2% of the total genetic load of RP in our cohort and possibly in the Pakistani population as well.

The term retinitis pigmentosa (RP) refers to bone spicule-like pigmentation in the mid-peripheral fundus that stimulates inflammation and was first used by the German physician Donders in 1857 [1]. RP is a progressive rod-cone dystrophy that primarily affects the rod photoreceptors whereas the function of the cone receptors is compromised as the disease progresses [2]. Ocular findings comprise atrophic changes in the photoreceptors and in the retinal pigment epithelium (RPE) followed by the appearance of melanin-containing structures in the retinal vascular layer [2]. The typical fundus appearance includes attenuated arterioles, bone spicule-like pigmentation, and waxy pallor of the optic disc. Affected individuals often have severely abnormal or non-detectable rod responses in electroretinogram (ERG) recordings even in the early stage of the disease [2].

RP is the most common inherited retinal dystrophy, affecting approximately 1 in 5,000 individuals worldwide [3]. RP is clinically and genetically heterogeneous, which is inherited as a dominant, recessive, as well as an X-linked trait [4]. RP presents intra- and interfamilial variations in expressivity, penetrance, progression, and age of onset [5]. Among the several causes of variable phenotypes are alternative alleles, environmental factors, modifier genes, or a combination of these factors [6,7]. Autosomal recessive RP (arRP) is the most common form of RP worldwide, and to date, 36 genes/loci have been associated with autosomal recessive RP (RetNet).

Phosphodiesterase 6 (PDE6) is a critical enzyme of the visual phototransduction cascade in the vertebrate retina [8,9]. The protein is a heterotrimer that consists of an alpha, a beta, and two gamma subunits. PDE6A (Gene ID: 5145; OMIM: 180071) encodes for the α-subunit of cyclic guanosine monophosphate (cGMP)-phosphodiesterase that is primarily expressed in rod photoreceptors [10,11]. PDE6A consists of 22 coding exons and encodes for 860 amino acids [12,13]. Dryja and colleagues were the first to report causal mutations in PDE6A in patients with arRP [14].

We previously reported three mutations in PDE6A including a splice acceptor site mutation in three families of Pakistani origin [15]. Here, we report two additional familial cases with multiple members manifesting symptoms of RP. The clinical phenotype was established with fundus photography and electroretinography while genome-wide linkage analyses localized the disease interval to chromosome 5q. Subsequently, Sanger sequencing identified splice acceptor site mutations in these two families that were absent in ethnically matched control chromosomes.

PKRP133 was recruited from the Punjab province of Pakistan. A total of 23 individuals including five affected individuals were enrolled in the study (Figure 1). Medical records and clinical reports of earlier examinations by the family physician suggested retinal dystrophy with early most likely congenital onset. Interviews with family elders revealed the initial presenting symptoms of night blindness and decreased visual acuity experienced by affected individuals in early childhood. We performed a detailed ophthalmic examination of selected individuals. Visual acuity in affected individuals 21 and 30 was limited, progressing to only counting fingers in affected individual 21 (Table 1). The fundus examination of affected individual 21 shows severe bone spicule-like pigmentation in the mid-periphery and periphery of the retina along with milder macular involvement, arteriolar attenuation, and disc pallor (Figure 2A). The fundus examination of affected individual 30 shows only milder macular involvement, arteriolar attenuation, and disc pallor, and no bone spicule-like pigmentation was observed in the mid-periphery and periphery of the retina (Figure 2B). The affected individuals had reduced ERG responses with delayed implicit time (Figure 3A–D). Photophobia or nystagmus was not observed in any affected or unaffected family member.

Initially, a genome-wide scan with 382 STR markers with an average spacing of approximately 10 cM was performed to localize the RP phenotype in PKRP133. Suggestive linkage with alleles of microsatellite markers on chromosome 5q, D5S422 and D5S400, was observed with LOD scores of 2.67 and 2.24 at θ=0.05, respectively. Haplotype analysis identified a region distal to D5S636 based upon recombination in affected individual 17 excluding PDE6A that lies approximately 5 Mb proximal of D5S636. The causal haplotype in the critical interval was segregated with the disease phenotype within the family with the exception of individual 33 (age: 24 years), who was homozygous for the disease haplotype (Figure 1). This individual was clinically evaluated over a period of 2 years, but no RP-like symptoms were observed. As shown in Figure 2, the fundus is normal with healthy retinal vasculature and the absence of any pigmentation in the peripheral and mid-peripheral regions of the retina (Figure 2C-D). Similarly, ERG recordings show normal ranges of the a- and b-waves with normal implicit time (Figure 3E-H). Taken together, the clinical evaluations of individual 33 over the course of the study (Figure 2C–D and Figure 3E-H) confirmed that this individual is clinically unaffected and does not exhibit any symptoms of RP.

In view of the clinical and haplotype analyses of individual 33, an additional genome-wide linkage analysis was completed to rule out the possibility of additional evidence of linkage to a causal locus other than 5q33–35. Subsequently, a set of 399 additional markers were genotyped from the Linkage mapping set v.2.5-HD5, and taken together, all these markers refined the genome-wide resolution to approximately 5 cM spacing across the entire genome. A two-point LOD score of 2.84 at θ=0 was obtained with marker D5S2066 located distal to D5S422. Two-point LOD scores greater than 0.50 were also obtained for markers D3S3715, D5S471, D5S2115, D8S1771, D8S1820, D10S189, D13S173, D15S1002, D15S117, D16S3068, D16S503, D20S906, and D22S423. Linkage to all these markers was excluded either by large negative LOD scores calculated for closely flanking markers or by haplotype analysis.

Lack of identification of any potential locus other than on chromosome 5q during two separate genome-wide linkage scans and multiple clinical examinations of unaffected individual 33 that confirmed the absence of the clinical symptoms of RP prompted us to consider the following scenarios. First, affected individual 17 had been diagnosed incorrectly by the local ophthalmologist, and/or perhaps he manifested another form of retinal degeneration that had been misdiagnosed as RP. Second, affected individual 17 was a “phenocopy” manifesting the clinical phenotype of RP due to a genetic mutation not present on chromosome 5q. Initially, individual 17 voluntarily provided a blood sample; however, he has been reluctant to travel to an ophthalmology clinic for a thorough ocular examination. We approached affected individual 17 several times during the study and requested his cooperation for a comprehensive ocular examination without any success. Since we do not have conclusive evidence of the clinical phenotype, we redesignated his affectation status as “unknown.”

Subsequently, we genotyped additional STR markers proximal to D5S636, the previously defined proximal boundary of the critical interval. Lack of homozygosity in affected individuals 21, 27, and 29 at marker D5S2090 suggested the causal mutation lies distal to marker D5S2090 extending the critical interval with significant two-point LOD scores (Table 2). The new critical interval included PDE6A, a gene previously associated with RP. We sequenced all the coding exons, exon–intron boundaries along with the 5′ and 3′ untranslated regions (UTRs) of PDE6A that identified a variation in intron 16 at the splice acceptor site substituting guanine for adenosine at position −1: c.2028–1G>A. The mutation segregated with disease phenotype in the family (Figure 1 and Figure 4A-C) except individual 17 (heterozygous carrier of the c.2028–1G>A variation) and was absent in 350 ethnically matched control chromosomes. This variation was not present in the 1000 Genomes and NHLBI Exome Sequencing Project databases.

We were confident that c.2028–1G>A is responsible for the RP phenotype in PKRP133 provided that (a) two genome scans localized the critical interval chromosome 5q, (b) PDE6A has been previously implicated in the pathogenesis of retinal dystrophies including RP, and (c) the splice acceptor site mutations at positions −1 and −2 are considered deleterious. Nonetheless, we sought additional evidence to exclude the possibility that a variant other than c.2028–1G>A present within the critical interval is responsible for the disease phenotype. We captured the entire exome of individual 29 using the SureSelect Human All Exon Kit and sequenced on the Illumina HiSeq 2000 Genome Analyzer (Illumina). We identified 37 additional variants present within the critical interval excluding the splice-acceptor site variant (c.2028–1G>A) identified with Sanger sequencing; however, none of them satisfied the inclusion criterion of causality, which includes (a) absence or a low minor allele frequency in the 1000 Genomes and NHLBI Exome Sequencing Project databases; (b) absence from ethnically matched control chromosomes; and (c) segregation with the disease phenotype in the respective family, further confirming that c.2028–1G>A is responsible for the RP phenotype in PKRP133.

Next, we interrogated our entire cohort of RP familial cases by genotyping STR markers that span the PDE6A followed by bidirectional sequencing of the coding exons and exon–intron boundaries. We identified an additional familial case, PKRP140 (Figure 4D), who harbored a splice acceptor site mutation in intron 10: c.1408–2A>G (Figure 4E-G). We recruited a total of 15 individuals of PKRP140; ten of these participating individuals manifested symptoms of RP. The splice acceptor mutation segregated with the disease phenotype (RP) in the family (Figure 4D-G) and was not found in 350 ethnically matched control chromosomes and the 1000 Genomes and NHLBI Exome Sequencing Project databases. We further evaluated the conservation of c.1408–2A and c.2028–1G in other PDE6A orthologs. As shown in Figure 5, c.1408–2A and c.2028–1G are fully conserved in primates, placental mammals, and vertebrate PDE6A orthologs.

We used HSF, a freely available online bioinformatics tool, to predict the effect of the c.1408–2A and c.2028–1G variations on PDE6A mRNA splicing. The HSF analysis generated consensus values of 81.30 and 52.35 for the wild-type (c.1408–2A) and mutant (of c.1408–2G) nucleotides, respectively. The predicted consensus deviation value of −35.61% for c.1408–2A>G suggests the loss of the wild-type splice site that would result in the skipping of exon 11 of PDE6A (Figure 6A). Likewise, HSF analysis predicted the consensus values of 82.23 and 53.28 for the wild-type (c.2028–1G) and mutant (c.2028–1A) nucleotides, respectively. The predicted deviation of −35.21% for c.2028–1G>A suggests the loss of the wild-type splice acceptor site of exon 17 (Figure 6B-C). In parallel, HSF also detected the creation of a new cryptic splice acceptor site using the G nucleotide (c.2028G) in exon 17 (Figure 6D). The algorithm predicted consensus values of 45.37 and 74.31 for the wild-type (c.2028–1G) and the new cryptic splice site (c.2028G), respectively. The predicted consensus value deviation of +63.81% for the new cryptic splice acceptor site suggests the loss of the wild-type splice site that would result in a frameshift and eventually would lead to a premature stop codon (p.K677Rfs24*) in the protein.

We report in this study two splicing acceptor site mutations in PDE6A associated with autosomal recessive RP in two large consanguineous Pakistani families. Two-point LOD scores of 3.66 and 3.17 for markers D5S2026 and D5S636, respectively, the absence of any scores >1.5 observed during genome-wide linkage scan other than on 5q, identification of variations in evolutionary conserved bases, c.1408–2A and c.2028–1G, and the absence of these variations in ethnically matched control chromosomes, 1000 Genomes and NHLBI Exome Sequencing Project databases strongly support our conclusion that the two splice acceptor site mutations are responsible for the retinal phenotype of the patients reported in this study.

In silico analysis predicted that a splice-site mutation (c.1408–2A>G) would destroy the splice acceptor site, which would result in the skipping of exon 11. Exon 11 consists of 66 base pairs, and thus, skipping of exon 11 would introduce the deletion of 22 amino acids (p.K470_L491del) without disturbing the reading frame of PDE6A. The simple modular architecture research tool (SMART) identified two cGMP-specific phosphodiesterases, adenylyl cyclases and FhlA (GAF) domains (73–232 and 254–441 amino acids), two low-complexity regions (478–495 and 823–851 amino acids), and one PDEase catalytic domain (3′5′-cyclic nucleotide phosphodiesterase) ranging from 556 to 734 amino acids. Based on the PDE6A protein domain structure, this deletion will eliminate the low-complexity region (478–495 amino acids).

However, two possible scenarios are predicted for the splice site c.2028–1G>A mutation: (a) the creation of a new cryptic splice acceptor site using the first G nucleotide of exon 17 and (b) the less likely scenario of the complete skipping of exon 17. The inclusion of this cryptic acceptor site would lead to a frameshift and would ultimately result in a premature termination: p.K677Rfs24* (Figure 6D). The mutant mRNA would be degraded by the nonsense mediated decay (NMD) pathway. Even if the protein escapes the NMD pathway, the truncated protein would lack 160 C-terminal amino acids including the cGMP-binding site and one low-complexity region. However, skipping exon 17, which comprises 108 bp, would result in a deletion composed of 36 amino acids (p.K676_M712del) without altering the reading frame of PDE6A but would disrupt the PDEase catalytic domain (Figure 6D).

To date, 20 causal mutations in PDE6A have been associated with RP including ten missense, five nonsense, two splice site, two frameshift, and a deletion encompassing the entire gene [14,15,20–26]. Dryja and colleagues screened sporadic and recessive RP cases and found that mutations in PDE6A are responsible for the clinical phenotype in 3% of Caucasian patients [21]. Previously, our group identified mutations in PDE6A in three consanguineous families [15]. Taking these results together, we have identified five families in our cohort of >300 familial cases suggesting an approximate 2% contribution to the total genetic load of RP.

In conclusion, we identified two splice acceptor site mutations segregated in two large consanguineous pedigrees responsible for arRP. The identification of causal mutations will enhance diagnostic capabilities leading to the identification of carrier status, which would help in the prevention of the familial form of arRP.