Systematic Validation of RNF213 Coding Variants in Japanese Patients With Moyamoya Disease

Abstract

Background: A founder variant of RNF213, p.R4810K (c.14429G>A, rs112735431), was recently identified as a major genetic risk factor for moyamoya disease (MMD) in Japan. Although the association of p.R4810K was reported to be highly significant and reproducible, the disease susceptibility of other RNF213 variants remains largely unknown. In the present study, we systematically evaluated the coding variants detected in Japanese patients and controls for associations with MMD.

Methods and results: To detect variants of RNF213, all coding exons were sequenced in 27 Japanese MMD patients without p.R4810K. We also validated all previously reported variants in our case-control samples and tested for associations in combination with previous Japanese study cohorts, including the 1000 Genomes Project data set, as population-based controls. Forty-six missense variants other than p.R4810K were identified among 370 combined patients and 279 combined controls in Japan. Sixteen of 46 variants were polymorphisms with minor allele frequency >1%, and, after conditioning on the p.R4810K genotype, were not associated with MMD. We conducted a variable threshold test using Combined Annotation-Dependent Depletion on the remaining 30 rare variants (minor allele frequency <1%), and the results showed that the frequency of potentially functional variants was significantly higher in patients than in controls (permutation, minimum P=0.045).

Conclusions: Not only p.4810K but also other functional missense variants of RNF213 conferred susceptibility to MMD. Our analysis also revealed that ≈20% of Japanese MMD patients did not harbor susceptibility variants of RNF213, indicating the presence of other susceptibility genes for MMD.

Keywords: cerebrovascular disorders; epidemiology; genetics; risk factors.

Figure 1

Association of the p.R4810K variant (c.14429G>A, rs112735431) with moyamoya disease (MMD) across 4 Japanese studies. A meta-analysis was performed assuming a fixed-effects model. Squares and horizontal lines represent odds ratios and 95% CIs for individual studies. A diamond represents the summary odds ratio estimate and 95% CIs for the meta-analyses of 4 cohorts.

Figure 2

Disease-specific LD pattern within the RNF213 locus. Thirteen common SNPs (MAF >10%) and p.R4810K were selected from the present resequencing data. The pairwise LD of |D′| is displayed by the GOLD heat map. The gene structure was shown as an equal interval scale of SNPs in the LD map. LD indicates linkage disequilibrium; MAF, minor allele frequency; MMD, moyamoya disease; SNPs, single-nucleotide polymorphisms.

Figure 3

Variable threshold test using combined annotation dependent depletion for rare missense variants in RNF213. A, −log₁₀ P values for burden tests were plotted against C-score thresholds (diamonds). The empirical threshold was given if the P value was minimized (P_min; shaded circle). Corrections for multiple comparisons were achieved by 1000 permutations on phenotypes. B, Fractions of diseased mutants per total mutants were plotted against C-score thresholds (open circles). The black line represents the least-squares regression line.

Figure 4

Boxplots of C-scores for missense SNPs in the GWAS catalog and 15 candidate variants for MMD. OR <1.15 represents binary trait loci with ORs in the bottom quartile; OR>1.3 indicates binary trait loci with ORs above median. The dotted line connects each average value. The black line over bars indicates no significant differences from the MMD candidate variants (Steel’s test). CCVD indicates cardiocerebrovascular disease loci; GWAS, genomewide association study; LDL, quantitative trait loci associated with serum low-density lipoprotein cholesterol level; MMD, moyamoya disease; OR, odds ratio; SNP, single-nucleotide polymorphism.

Figure 5

Clinical characteristics according to RNF213 genotypes. A, Box plots of age at onset. The blue line connects each average value. B, Differences in clinical characteristics reflecting the disease severity. Compound hetero indicates compound heterozygotes for p.R4810K and the other rare missense variants; HA, presenting headaches only; Homo, homozygotes for p.R4810K; PCA, posterior cerebral artery involvement.