Molecular genetic analysis of retinitis pigmentosa in Indonesia using genome-wide homozygosity mapping

Subjects

Fourteen Indonesian families segregating RP—13 living on Java and one on Sulawesi islands—were enrolled in the study (total number of 56 individuals, 33 affected, 23 unaffected). The affected individuals were 20 males and 13 females of 10–71 years old, with the onset of the disease ranging from the age of 5 to 25. All families appear to be autosomal recessive, except for W09–0049, which is more suggestive of an X-linked recessive pattern. This type of inheritance is also possible in families W09–0048 and W09–0050 (Figure 1). Affected individuals were diagnosed at the Kariadi Hospital, the William Booth Hospital, or the Panti Wilasa Citarum in Semarang Central Java, Indonesia, and examined by direct ophthalmoscopy and visual acuity tests using a Snellen chart. The presumed mode of inheritance was determined by anamnesis and pedigree analysis. Informed consent was obtained from all participating individuals or in the case of under-age participants, from their parents. This study adhered to the tenets of the Declaration of Helsinki. In addition, 149 ethnically matched and unrelated control individuals participated in this study.

Clinical characterization

Clinical characterization that included funduscopy and visual acuity measurements revealed that the majority of affected individuals had poor vision and showed typical hallmarks of RP on funduscopy, e.g., bone-spicule pigmentation, attenuated retinal vessels, and/or a pale appearance of the optic disc (Table 1). Only in family W09–0036 is the visual acuity not dramatically decreased.

Linkage analysis and homozygosity mapping

Genomic DNA of participating individuals was isolated from peripheral blood using a standard salting out procedure [9]. Affected individuals were selected for genome-wide single nucleotide polymorphism (SNP) analysis. For three consanguineous families (W09–0041, W09–0042, and W09–0046), the low-resolution Illumina 6k SNP array (Illumina, San Diego, CA) that contains 6,020 SNPs was used, whereas the patients from the remaining 11 families were genotyped on a high-resolution Affymetrix 5.0 array (Affymetrix, Santa Clara, CA) that contains ~500,000 SNPs. For the three consanguineous families that were analyzed on the Illumina array, multipoint parametric linkage analysis was performed using the GeneHunter 2.1r5 program in the easyLinkage v5.052beta software package [10]. For the other families, homozygous regions were calculated using PLINK software [11], with a cut-off of 3 Mb and allowing two heterozygous SNP calls per window of 50 SNPs.

Mutation analysis

Exons and intron–exon boundaries of all known arRP genes residing in homozygous regions and X-linked RP genes retinitis pigmentosa GTPase regulator (RPGR) and retinitis pigmentosa 2 (X-linked recessive) (RP2) were sequenced in the corresponding probands using the Big Dye Terminator Cycle Sequencing Kit v3.1 on a 3130 automated sequencer (Applied Biosystems, Foster City, CA). Upon identification of an unknown variant, available family members and/or ethnically matched control individuals were analyzed using direct sequencing, the amplification refractory mutation system or restriction fragment length polymorphism analysis. Primer sequences and PCR conditions are available upon request. The analysis of mutations was conducted with AlaMut 1.5 software (Interactive Biosoftware, Rouen, France). Splice site scores were calculated using an internal AlaMut module that combines the scores of GeneSplicer, MaxEntScan, SpliceSiteFinder-like, and NNSPLICE. For the missense changes that were identified, the evolutionary conservation of the substituted amino acids was assessed by aligning the corresponding protein sequences of several species, using the Align program in VectorNTI Advance 11.0 software (Invitrogen, Carlsbad, CA). The pathogenicity of these mutations was evaluated using SIFT [12] and PolyPhen-2 [13] prediction programs. In addition, Grantham scores, which define the physicochemical difference between amino acids [14], and PhyloP scores, which are a measure of evolutionary nucleotide conservation (PhyloP44wayAll), were determined.

Linkage analysis and homozygosity mapping in arRP families

To determine the location of the genetic defect in the 14 presumed arRP families, affected individuals were genotyped using whole-genome SNP arrays (Figure 1). In three consanguineous families with multiple affected individuals (W09–0041, W09–0042, and W09–0046), the Illumina 6k array was used. Multipoint linkage analysis was performed to detect genomic regions that might contain the genetic defect. In family W09–0041, only a single peak with a maximum multipoint LOD score of 3.00 on chromosome 15 was identified (Figure 2A). In families W09–0042 and W09–0046, multiple peaks potentially harboring the genetic defect were identified, with multipoint LOD scores of 1.80 (Figure 2B,C). As expected in patients born from consanguineous marriages, all regions with the highest LOD scores in the three families represented continuous stretches of homozygous SNP calls (Table 2), with identical haplotypes shared between affected siblings. In the remaining 11 arRP families, high-resolution SNP genotyping was performed using Affymetrix 5.0 arrays. Homozygous regions were calculated and considered to be significant if they spanned more than 3 Mb of genomic DNA. For the families with multiple affected siblings, only those regions that were homozygous in all affected individuals were considered to potentially harbor the causative gene. In four families, no regions that were homozygous in all affected siblings were identified, whereas in a few other families, many of those regions were present. In the non-consanguineous families, the number of homozygous segments per family varied from 0 to 5, with a maximum length of 20 Mb.

Mutation analysis of autosomal recessive RP genes in homozygous regions

In the majority of the families, one or more homozygous regions contained one of the 31 known arRP genes. Sequence analysis of these genes in the corresponding families revealed seven novel mutations likely to be causative for RP in the respective families (Table 2). The pathogenicity of the missense and splice site mutations were evaluated using a detailed in silico analysis (Table 3).

In family W09–0038, the largest homozygous region contained the c-mer proto-oncogene tyrosine kinase (MERTK) gene. Exon 15 failed to amplify in the affected individuals but not in the unaffected family members, suggesting a genomic deletion (Figure 3A). Detailed analysis revealed the occurrence of a complex rearrangement that included a 1732-bp deletion containing exon 15 of MERTK (Figure 3B). The absence of exon 15 from MERTK mRNA is predicted to result in a frameshift and premature stop codon, which may lead to truncation of the MERTK protein (p.G654AfsX41) or trigger nonsense-mediated mRNA decay. The affected individuals of family W09–0045 displayed three shared homozygous regions, of which the second largest contained MERTK. Sequence analysis revealed a homozygous mutation (c.2487–2A>G) that affects the consensus splice site of exon 19 and most likely results in aberrant splicing and premature termination of the MERTK protein, including the highly conserved tyrosine kinase domain (Table 3).

In family W09–0041, four affected siblings shared only one large homozygous region, on chromosome 15, harboring the nuclear receptor subfamily 2, group E, member 3 (NR2E3) gene. Mutation analysis of NR2E3 identified a homozygous mutation substituting a glycine for a valine residue (c.1025T>G; p.V342G). The PolyPhen-2 program predicts this variant to be probably damaging, and SIFT indicates that in this position only a valine or isoleucine is tolerated (Table 3). Furthermore, the mutation was homozygously present in five affected family members and was absent or heterozygously present in three unaffected siblings. Finally, this variant was not detected in 298 ethnically matched control alleles. NR2E3 encodes a transcription factor that belongs to the family of nuclear hormone receptors, with a DNA-binding domain (DBD) at the N-terminus and a ligand-binding domain (LBD) located more C-terminally [15]. Most NR2E3 mutations that cause retinal dystrophy affect amino acid residues located in these two domains [16]. The LBD of NR2E3 is structurally similar to that of other nuclear hormone receptor family members and consists of 12 structurally conserved α-helices [17,18]. The valine residue at position 342 is highly conserved, both in the NR2E3 proteins of other species and in the LBD of other nuclear hormone receptors, being either a valine or a similar isoleucine residue in all other protein sequences (Figure 4B). This high degree of conservation suggests that a relatively large nonpolar amino acid must be present at this position to preserve the structure and/or function of the LBD of these proteins. The glycine residue that is present in the mutant NR2E3 protein is small and might interrupt the structure and render the protein inactive or partially active. The pathogenic character of this variant is supported by the fact that this alteration has not been found in alleles of healthy ethnically matched individuals.

In family W09–0042, the largest homozygous region contains the ATP-binding cassette, sub-family A, member 4 (ABCA4) gene, which has been associated with Stargardt's disease [19], cone-rod dystrophy, and to a lesser extent, RP [20]. Sequence analysis revealed a nucleotide variant in the splice donor site of exon 3 (c.302+4A>C) homozygously present in the two affected siblings and absent or heterozygously present in their nonaffected relatives. In addition, the variant was not detected in 298 ethnically matched control chromosomes. In silico prediction of the strength of the splice donor site showed a decrease due to the alteration (e.g., 40% decrease in GeneSplicer, 24% in MaxEntScan, 12.3% SpliceSiteFinder-like, and 0.04% decrease in NNSPLICE), indicating that this variant might alter ABCA4 splicing (Table 3). However, due to the unavailability of patient RNA or lymphoblastoid cell lines, altered ABCA4 splicing could not be confirmed in vivo. Hence, the pathogenicity of this variant remains uncertain.

In family W09–0046, three regions that are homozygously shared by two affected siblings were detected. The second largest region contained two known arRP genes, namely tubby-like protein 1 (TULP1) and eyes shut homolog (Drosophila) (EYS). Sequence analysis of TULP1 did not reveal any pathogenic variants, whereas a missense variant was detected in EYS that segregated with arRP in the family (c.9082G>T; p.D3028Y). This alteration affects an aspartic acid residue localized in the fifth laminin A G-like domain of the human EYS protein. The aspartic acid residue at this position is also present in homologous positions of other EYS LG domains and is evolutionarily conserved in EYS proteins of other species (Figure 4C). Both SIFT and PolyPhen-2 predict this change to be damaging, which is further supported by high Grantham and PhyloP scores (Table 3). Intriguingly, a similar mutation (p.D2767Y), located at the homologous position in the fourth laminin A G-like domain of EYS, was recently found in a Pakistani family segregating arRP [21]. This suggests that the aspartic acid residue located at that position within the laminin A G-like domain is crucial for proper functioning of EYS and that the p.D3028Y mutation is indeed pathogenic in family W09–0046.

In two affected siblings from family W09–0047, only a single homozygous region was detected that harbored the phosphodiesterase 6A, cGMP-specific, rod, alpha (PDE6A) gene. A nonsense mutation (c.1675C>A; p.Y558X) was found to segregate with RP in this family, and this is predicted to either result in premature termination of the α subunit of the phosphodiesterase enzyme or to cause nonsense-mediated decay of PDE6A mRNA.

The largest homozygous region detected in the affected individual 50068 from family W09–0048 contained the crumbs homolog 1 (CRB1) gene, and sequence analysis revealed a transition of cytosine to thymine, resulting in the substitution of a leucine for a proline residue (c.3914C>T; p.P1305L). This alteration was not detected in 149 ethnically matched control individuals and is predicted to be probably damaging by PolyPhen, although the SIFT software predicts it to be tolerated (Table 3). The proline residue that is substituted resides in one of the EGF-like calcium-binding domains of CRB1 and neighbors one of the six core cysteine residues that are essential for the formation of disulphide bridges and proper structure of these domains [22]. The fact that the proline residue is highly conserved throughout vertebrate evolution and even in fruit fly (Figure 4A), suggests that this mutation is causative for arRP in this family.

In two of the families (W09–0049 and W09–0050), genes residing in the largest homozygous regions (family with sequence similarity 161, member A [FAM161A], ABCA4, retinal pigment epithelium-specific protein 65 kDa [RPE65], Usher syndrome 2A (autosomal recessive, mild) [USH2A], and progressive rod-cone degeneration [PRCD]) were excluded. Four families (W09–0036, W09–0039, W09–0040, and W09–0044) had no significant homozygous regions.

For the families with plausible X-linked inheritance (W09–0048, W09–0049 and W09–0050), we performed screening of the two known X-linked RP genes. RP2 and RPGR were sequenced, but no mutations were identified.