Next-generation genetic testing for retinitis pigmentosa

Abstract

Molecular diagnostics for patients with retinitis pigmentosa (RP) has been hampered by extreme genetic and clinical heterogeneity, with 52 causative genes known to date. Here, we developed a comprehensive next-generation sequencing (NGS) approach for the clinical molecular diagnostics of RP. All known inherited retinal disease genes (n = 111) were captured and simultaneously analyzed using NGS in 100 RP patients without a molecular diagnosis. A systematic data analysis pipeline was developed and validated to prioritize and predict the pathogenicity of all genetic variants identified in each patient, which enabled us to reduce the number of potential pathogenic variants from approximately 1,200 to zero to nine per patient. Subsequent segregation analysis and in silico predictions of pathogenicity resulted in a molecular diagnosis in 36 RP patients, comprising 27 recessive, six dominant, and three X-linked cases. Intriguingly, De novo mutations were present in at least three out of 28 isolated cases with causative mutations. This study demonstrates the enormous potential and clinical utility of NGS in molecular diagnosis of genetically heterogeneous diseases such as RP. De novo dominant mutations appear to play a significant role in patients with isolated RP, having major implications for genetic counselling.

Figure 1

Sequencing statistics. A: Schematic drawing of a 12-plex 2.1M NimbleGen sequence-capture array. Samples were barcoded and hybridized individually. After hybridization, all 12 samples were eluted simultaneously, allowing to proceed with one pooled sample consisting of 12 different DNAs. B: Histogram of median target coverage. Only 15 targets were covered less than five times. Most targets show a coverage of 20–30×. C: The minimum coverage of a percentage of targets. Solid line indicates the average across 100 samples, whereas dotted and dash lines indicate first and second standard deviation respectively. D: Evenness of coverage for all samples. The average evenness score over all samples is 97.3%. E: Example of a poorly covered target. This screenshot shows the coverage of some exons of GRM6. The box highlights exon 1, which is poorly covered in comparison with the rest of the gene. F: The 15 targets that are poorly covered (median coverage of less than five times) have either a high GC content (often an UTR or exon 1) or are containing repeat-rich regions.

Figure 2

Systematic variant prioritization tool to efficiently reduce the number of identified variants. A: Filtering of identified variants in the 12 samples, each carrying two variants in recessive RD genes. 14,144 genetic variants were automatically detected, including 21 of the 24 known mutations. A further systematic prioritization (described in Material and Methods) reduced the number of variants to 97, including 18 of the 24 known pathogenic mutations (step 2). Of these 97 variants, all variants were selected that were consistent with the known inheritance pattern of the respective gene, resulting in 44 remaining variants, including 16 of the known variants (step 3). A manual search for a variant on the second allele was performed for recessive genes where only one variant was found (step 4), resulting in a total number of 48 variants, now including 20 of the 24 known mutations (step 5). B: In the group of 100 RP samples with unknown cause of disease, in total 128,557 variants (including six larger deletions) were automatically annotated (step 1). After applying the same filtering steps as for the 12 known RD samples, 359 variants remained after prioritization (step 5). RD, retinal dystrophy.

Figure 3

De novo mutations in isolated RP patients. A: De novo mutation in PRPF31. In patient 27,790, a heterozygous nonsense mutation was detected in the autosomal dominant RP gene PRPF31 that was not present in both the unaffected parents. To confirm that individuals DNA11-01005 and DNA11-00996 are the biological parents, 16 highly polymorphic markers distributed across the genome were analyzed, showing a perfect Mendelian inheritance for all markers, thereby confirming the de novo event (Supp. Fig. S4). B: De novo mutation in RHO. In patient 22,315, a heterozygous missense mutation was detected that was predicted to be pathogenic (indicated in red). The mutation was not present in three unaffected siblings and the unaffected mother. Because the unaffected father was deceased, seven microsatellite markers surrounding and in close proximity of RHO were analyzed to determine haplotypes and the likelihood of a de novo event. The proband 22,315 and his three siblings all appeared to have inherited the same RHO allele from the deceased father, strongly suggesting that the RHO mutation has occurred by a de novo event. The genomic position of the microsatellite markers and RHO is indicated between parentheses. C: De novo mutation in USH2A. In patient 28,557, compound heterozygous mutations were detected in the autosomal recessive RP gene USH2A. One mutation (M1) was inherited from the mother, but the second mutation (M2) was not present in both parents, and as such, had occurred by a de novo event. Again, parental testing confirmed that DNA11-02253 and DNA11-02247 were the biological parents of the patient (Supp. Fig. S4).