Genome-wide association studyfor resistant hypertension
To clarify the genetic architecture of resistant hypertension, we conducted a GWAS in a Japanese population consisting of 2,705 resistant hypertension cases and 21,296 mild hypertension controls. We evaluated the possibility of population substructure for our sample population by comparison to HapMap samples using principal component analysis (PCA). Although all cases and controls were clustered in the Asian population, a very small portion of the samples was clustered in the Chinese population (Supplementary Fig. S1a and S1b). We then selected only samples from the major Japanese cluster for further analysis. After whole-genome imputation using the 1000 Genomes Projects as a reference, we examined the association of 6,012,291 SNPs with minor allele frequency (MAF) of more than 5% and an estimated imputation accuracy of greater than 0.8. The quantile-quantile (Q-Q) plot shows the distribution of observed versus expected P values, while the corresponding genomic inflation factor (λGC) of 1.057 suggests a low possibility of false-positive associations resulting from population stratification or cryptic relatedness (Supplementary Fig. S2). The Manhattan plot, plotting -log10(P value) from the GWAS and imputation analysis against the chromosome position, is shown in Fig. 1. Our GWAS identified one genetic locus achieving genome-wide significance (P < 5× 10−8) and 17 loci showing suggestive association (P < 1 × 10−5) with resistant hypertension in the Japanese population (Table 1). We examined each locus for whether it included any variants that were previously reported to be significantly associated with blood pressure phenotypes, hypertension, or resistant hypertension in the 1Mb-flanking region of each lead variant. We detected one novel locus with significant association and three novel loci with suggestive association. The lead variant of the novel significant locus was rs1442386 (odds ratio (95% CI) = 0.85 (0.81–0.90), P = 3.75 × 10−8), which is located in the intron region of DLG associated protein 1 (DLGAP1) on chromosome 18p11.3 (Fig. 2). The three novel suggestive loci were PQLC3 (2p25), LOC105369874 (12q14), and MED4 (13q14) locus. The other 14 suggestive loci included variants previously reported to be associated with hypertension or blood pressure phenotypes. For example, variants of the CYP11B2 locus have been frequently validated to be associated with hypertension in multiple populations29,30.
Functional annotation and expression quantitative trait loci (eQTL) analysis
We used HaploReg to perform functional analysis of a total of 18 lead variants showing association with resistant hypertension in the GWAS results. All of these variants were located in non-coding regions (nine intronic and nine intergenic) (Supplementary Table S2). Seven variants were located in gene expression regulatory motifs, such as enhancers, promoters, open chromatins and protein-binding sites in various tissue types. We found that several variants have been identified as eQTLs of their nearest genes in various tissue types (Supplementary Table S3). Among them, two variants including rs2075571 of 1q22 and rs9271382 of 6p21 had associations (P < 0.05) with the expression levels of some genes, such as rs2075571 at the THBS3 locus showing an association with THBS3, GBA, RP11-263K19.6, GBAP1 and MUC1 expression, and rs9271382 at the HLA-DQA1 locus showing an association with HLA-DQA1, HLA-DQA2, HLA-DQB1, HLA-DQB1-AS1, HLA-DQB2, HLA-DRB1, HLA-DRB5, and HLA-DRB6 expression in various tissue types. Functional analysis of the rs1442386 variant which reached genome-wide significance showed one altered regulatory motif (GCM; glia cells missing) and a significant association with DLGAP1 expression in whole blood at P = 0.0018.
Gene-based association analysis
We used VEGAS2 to obtain gene-based P values for phenotypic association from SNP-based P values, using the 1000 Genomes Projects EAS phase 3 reference set. The genes of which the gene-based P values exceeded a Bonferroni-corrected threshold of P < 2.07 × 10-6 are given in Supplementary Table S4. Gene-based tests identified 21 genes associated with resistant hypertension, including GBX1, AGAP3, ASB10, ABCF2 and TMUB1 on chromosome 7q36, ESRP1 and LOC100288748 on 8q22, CYP11B2, CYP11B, GML, LY6D, LYNX1_1, LYNX1_2 and LOC100133669 on 8q24, DLGAP1, DLGAP1-AS3, DLGAP1-AS4 and MIR6718 on 18p11, ERG on 21q22, and GAB4 and CECR7 on 22q11. These genes were located not only at one locus with significant association, but also at five loci with suggestive association, in the current GWAS. For four lead variants, rs253447 of 5q31, rs77163128 of 7p12, rs200741614 of 12q14, and rs11619475 of 13q14, no genes were identified by VEGAS2, because these variants were intergenic and were located over 50kb outside the neighboring genes, resulting in them being outside the subject for gene-based association analysis.
Pathway-based association analysis
To further investigate the biological processes involved in resistant hypertension, we performed pathway-based association analysis using the VEGAS2Pathway approach. Figure 3 shows 35 Gene Ontology (GO) terms of biological process (BP) and cellular component (CC) that reached a genome-wide, pathway-based significant P value of less than 1 × 10−5 (Supplementary Table S5). Among them, we observed three prominent sets of GO terms that were highly associated with resistant hypertension. The most numerous set consisted of synapse (GO:0045202) and excitatory synapse (GO:0060076), especially involving postsynaptic compartments (CC term) for chemical synaptic transmission (GO:0007268) (BP term). Subsequently, a set of plasma membrane region (GO:0098590), postsynaptic membrane (GO:0045211), and membrane region (GO:0098589) for regulation of transmembrane transport (GO:0034762), and a set of neuron part (GO:0097458) and neuron projection (GO:0043005) for neuron development (GO:0048666) and neurological system process (GO:0050877) were also highly associated. These results suggest important pathways of the nervous system that may be involved in resistant hypertension.
Evaluation of previously reported variants
To verify previously reported loci showing an association with resistant hypertension, we performed analysis in the current Japanese GWAS dataset (Supplementary Table S6). These 26 SNPs have been previously evaluated in a multi-ethnic GWAS dataset including Caucasian, Hispanic, and African American subjects26-28,31. These SNPs, however, did not show a significant association with resistant hypertension in the Japanese population. We further examined whether variants previously associated with blood pressure phenotypes or hypertension showed an association with resistant hypertension in the current GWAS dataset. A total of 2074 and 172 SNPs were selected as variants previously associated with blood pressure and hypertension from the NHGRI-EBI GWAS catalog, respectively (listed in Supplementary Tables S7 and S8). Among these blood pressure-associated variants, eight at three loci showed suggestive associations with resistant hypertension (P < 1 × 10-5) (Supplementary Table S7). The most significant association was rs62525059 of 8q24 at the CYP11B2 locus (P = 2.58 × 10−7). The next suggestive associated variants were rs4072037 of 1q22 near MUC1/GBAP1 (P = 5.30 × 10−6) and rs3774427 of 3p21 near CACNA1D (P = 6.51 × 10−6). Also, two loci, including rs62525059 (CYP11B2) and rs3774427 (CACNA1D), showed a suggestive association with resistant hypertension in variants previously associated with hypertension (Supplementary Table S8), the same as those previously associated with blood pressure. These results suggest the possibility that CYP11B2 (the aldosterone synthase gene) and CACNA1D (a member of the voltage-gated calcium channel gene family) may be involved in the development not only of hypertension, but also of resistant hypertension. However, the current GWAS data that were used to assess the association with resistant hypertension did not successfully replicate a large number of previous GWAS findings. Most of the previously reported variants associated with blood pressure were established from studies of quantitative traits of blood pressure phenotypes. In addition, previous GWASs of hypertension frequently adopted non-hypertensive subjects or general populations as the control. On the contrary, the present study evaluated a binary outcome using mild hypertensive controls, which may have led to a reduction in statistical power. These differences in our GWAS data may have resulted in the discrepancy in genetic correlations from previous findings in quantitative outcomes or studies using non-hypertensive controls.