A frameshift mutation in golden retriever dogs with progressive retinal atrophy endorses SLC4A3 as a candidate gene for human retinal degenerations

Abstract

Progressive retinal atrophy (PRA) in dogs, the canine equivalent of retinitis pigmentosa (RP) in humans, is characterised by vision loss due to degeneration of the photoreceptor cells in the retina, eventually leading to complete blindness. It affects more than 100 dog breeds, and is caused by numerous mutations. RP affects 1 in 4000 people in the Western world and 70% of causal mutations remain unknown. Canine diseases are natural models for the study of human diseases and are becoming increasingly useful for the development of therapies in humans. One variant, prcd-PRA, only accounts for a small proportion of PRA cases in the Golden Retriever (GR) breed. Using genome-wide association with 27 cases and 19 controls we identified a novel PRA locus on CFA37 (p(raw) = 1.94×10(-10), p(genome) = 1.0×10(-5)), where a 644 kb region was homozygous within cases. A frameshift mutation was identified in a solute carrier anion exchanger gene (SLC4A3) located within this region. This variant was present in 56% of PRA cases and 87% of obligate carriers, and displayed a recessive mode of inheritance with full penetrance within those lineages in which it segregated. Allele frequencies are approximately 4% in the UK, 6% in Sweden and 2% in France, but the variant has not been found in GRs from the US. A large proportion of cases (approximately 44%) remain unexplained, indicating that PRA in this breed is genetically heterogeneous and caused by at least three mutations. SLC4A3 is important for retinal function and has not previously been associated with spontaneously occurring retinal degenerations in any other species, including humans.

Figure 1. Genome-wide association mapping of PRA in Golden Retrievers.

−Log₁₀ of p-values after correction for multiple testing and population stratification with 100 000 permutations and IBS clustering, respectively. The dashed lines indicate the 5% significance level. A) −Log₁₀ plot of genome-wide association results show a strong statistical signal on CFA37 (P_raw = 1.94×10⁻¹⁰, P_genome = 1.00×10⁻⁵). CFA39 represents the X chromosome. The most significant of the raw and permuted values are indicated. B) The statistically associated SNPs on CFA37 span 1.6 Mb from 28.331 Mb to 29.847 Mb.

Figure 2. Fine mapping using haplotype analysis.

SNP genotypes for 27 PRA cases and 19 PRA controls, over the 1.6 Mb region identified during the GWA study. Fifteen of the cases share a homozygous haplotype of eight SNPS (Left Block) and 15 cases share another homozygous haplotype of nine SNPS (Right Block). These two homozygous blocks overlap by just one SNP (Shared Block), defining the critical region of 644 kb between 28.633 Mb and 29.277 Mb on CFA37, a region containing 27 genes. Four of the cases do not share any haplotypes with the other 23 cases, but do share haplotypes with the controls.

Figure 3. Graphical comparison of the exons and exon-intron boundaries of human and canine SLC4A3.

A) Human (Homo sapien) SLC4A3. B) Canine (Canis familiaris) SLC4A3 as predicted by Ensembl genebuild. Sixteen of the genebuild exons predicted are identical to the human exons (black). Exons 1, 4, 5 and 7 (grey) have differences in 5′ and/or 3′ exon-intron boundaries. Exon 6 (grey) shows no sequence similarity to its human equivalent and is probably incorrect. C) Canine SLC4A3 exons confirmed by sequencing the retinal transcript. Exons 1, C1 and the correct exon 7 have not been predicted by Ensembl genebuild. The locations of the five fully-segregating sequence variants (in two PRA cases, two obligate carriers and four normal samples) are indicated, but only two are exonic and nonsynonymous. One of these is in the first exon of the cardiac isoform only (c: A44C) while the other is found in both the cardiac (c: 1981_1982insC) and full-length (fl: 2601_2602insC) isoforms of SLC4A3.

Figure 4. Graphical representation of the SLC4A3 protein.

A) Both isoforms are made up of a transmembrane domain flanked on both sides by a cytoplasmic domain (represented by C*). The number of amino acids that make up each domain are indicated. Several amino acids at the N-terminus are unique to either the full-length (280 residues) or cardiac (73 residues) isoform, but the remainder of the protein is identical in both isoforms. B) The locations of the two nonsynonymous SNPs in the confirmed canine SLC4A3 transcript. C) The cardiac isoform alone is affected by the A44C SNP mutation (p.Q15P). Both isoforms are affected by the 2601_2602insC mutation. 867 amino acids at the N-terminus of the full-length protein and 660 of the cardiac isoform are normal. However the insertion causes a shift in the reading frame of 104 amino acids, leading to a premature termination codon. This results in a truncated protein product, lacking 266 residues of the C-terminus of both isoforms, if expressed. Therefore a large part of the transmembrane region and the entire c-terminal cytoplasmic region is absent in the mutant proteins.

Figure 5. Segregation of GR-PRA1 in a family.

GR-PRA1 appears to be recessive and fully penetrant, with the exception of one dog not yet diagnosed. Individual 29 is not clinically affected with PRA, but did show ophthalmascopic signs of the disease at the age of 6. Individuals 31 and 32 were only 4 years old when examined at the same time as a number of relatives including their mother, and were diagnosed with early-stage PRA. Individuals that formed part of the genome-wide association study are indicated with “*”.