Large-Scale European GWAS Metaanalyses

The first large-scale metaanalyses of GWAS results for SBP, DBP, and case/control hypertension were published in 2009 . Two consortia analyzed approximately 2.5 M SNPs in large numbers of individuals (~30,000) of European ancestry and followed up their top ten independent signals from each scan by performing a simultaneous reciprocal exchange of association results. Overall, a total of 13 novel loci were discovered with genome-wide significant associations. Subsequently, an analysis of a single large cohort of 23,019 individuals from the Women’s Genome Health Study led to the discovery of an association near the BLK-GATA4 genes . The majority of the genome-wide significant SNPs were associated with both SBP and DBP and risk of hypertension, with the same direction of effect. The associated SNPs were mostly common, with a minor allele frequency (MAF) ≥5%, and modest effect sizes of ≤1.0 mmHg for SBP and ≤0.5 mmHg for DBP.

The International Consortium for Blood Pressure (ICBP) reported two of the largest metaanalyses of GWAS in 2011 using 2.5 M SNPs, one for associations with SBP and DBP, the other for mean arterial pressure (MAP) and pulse pressure (PP) phenotypes. The metaanalysis of GWAS for SBP and DBP was performed in 69,395 individuals of European ancestry and was followed by a three-stage validation study using 133,661 additional individuals of European ancestry . Similar sample sizes were deployed in the MAP and PP study. These GWAS metaanalyses identified 16 novel genome-wide significant associations with SBP and BP, and 8 for MAP and PP .

GWAS of Non-European Ancestries

From 2009 to 2011, many of the BP-SNP associations discovered in individuals of European ancestry were replicated in samples of different ancestries (East Asian, South Asian, and African) . The ICBP consortium also showed that some of the 29 SNPs associated with SBP and DBP had significant associations in populations of East- or South-Asian ancestry (or both) after correction for multiple testing .

Large GWAS have also been performed in cohorts of non-European ancestry. A GWAS metaanalysis discovered a SNP located near CASZ1 in samples of East Asian (Japanese) ancestry in 2010 . In 2011, the Asian Genetic Epidemiology Network Blood Pressure consortium identified six significant associations with BP from GWAS metaanalyses of 19,608 individuals of East Asian ancestry, some of which overlapped the ICBP GWAS associations .

The largest BP discovery metaanalysis in African Americans was published in 2013 by Franceschini et al. , with 29,378 individuals from the Continental Origins and Genetic Epidemiology Network, follow-up in the ICBP dataset, and further replication in additional African Americans and East Asian samples (Total ~29,000 individuals). Overall, three new BP loci ( RSPO3 , PLEKHG1 , and EVX1-HOXA ) were identified from the combined metaanalyses.

Kato et al. in 2015 reported a large transethnic metaanalyses of GWAS for SBP, DBP, MAP, and hypertension with 99,994 individuals (31,516 East Asians, 35,352 Europeans, and 33,126 South Asians) from the International Genomics of BP Consortium , which conducted a metaanalysis with the ICBP dataset and further replication in 133,052 additional samples (48,268 East Asian, 68,456 European, and 16,328 South Asian). Overall, 19 novel SNPs were identified with little heterogeneity observed between the different ethnic groups, although two variants identified in ethnic specific analyses did not validate. In 2012, it was observed that a quarter of the BP loci (8/34) were common across ethnic groups (excluding African Americans), with the remaining 26 loci showing BP-trait associations in only two ethnic groups or a single ethnic group . With larger metaanalyses of GWAS now being performed across different ethnic groups, our knowledge of shared BP loci will become clearer.

Candidate Gene Studies

Loci have also been discovered via candidate gene studies, which only test a selection of genetic variants. For example, an analysis of SNPs at 30 genes that were known targets for antihypertensive drugs provided genome-wide significant associations at two loci, AGT and ADRB1 .

Longitudinal Data, Gene–Lifestyle Interactions, and Multitrait Analyses

GWAS utilizing different methodological approaches have also been used for discovery of BP loci. These include using long-term averages of repeated BP measures within longitudinal cohorts and performing an analysis of gene–age interactions or stratified analyses within different age subgroups . Furthermore, a study has performed a multitrait analysis by analyzing all BP traits simultaneously . These analyses have identified seven new loci ( KCNK3 , CRIP3 ; MIR1263 , CDC25A , EHBP1L1 ; and IGFBP1-IGFBP3 and CDH17 ), validated previously reported loci, and identified additional variants within known regions.

Bespoke Genotyping Arrays

Alongside genome-wide arrays and imputation methods for the discovery of trait-associated loci, the scientific community has developed genotyping arrays with bespoke content. The general purpose of these arrays is to provide cost-effective genotyping of prespecified SNPs. Two arrays have been used for analyses of BP and other cardiovascular traits; these are the gene-centric HumanCVD BeadChip and the Cardio-MetaboChip array.

The HumanCVD BeadChip contains approximately 50,000 SNPs that provide dense coverage of ~2000 genes considered to be more likely to have functional effects on cardiovascular traits . An analysis of 25,118 individuals of European ancestry for all BP traits with follow-up of the most associated signals in a further 59,349 individuals led to the discovery of eight BP loci in 2011 . Four of the associated loci were simultaneously reported from the ICBP consortium . This study analyzed fewer SNPs in smaller sample numbers ( N ~25,000) than the larger metaanalyses of GWAS and indicated that discovery analyses in independent samples will discover new loci because the BP genetic architecture involves possibly hundreds of variants, and each study will detect more or less random subsets of BP variants. Other BP loci have also been identified from the HumanCVD BeadChip: two further loci ( MDM4 and HRH1ATG7 ) were reported in 2013 , and in 2014, 11 novel loci were reported in a discovery sample of 87,736 Europeans and independent replication in 68,368 Europeans . An analysis of HumanCVD BeadChip genotypes in up to 8600 African-American individuals (a relatively small discovery sample) did not yield any new BP loci , although three previously known loci were validated.

The Cardio-MetaboChip array comprises 196,725 variants, including ~5000 SNPs with nominal ( P <0.016) evidence of BP association from previous GWAS metaanalyses . It also includes several genomic regions for fine mapping of selected loci . A metaanalyses of 201,529 individuals of European ancestry was performed, including 109,096 individuals genotyped on Cardio-MetaboChip and 92,433 individuals with imputed genotype data at SNPs overlapping the variants on Cardio-MetaboChip . Sixty-seven loci attained genome-wide significance, of which 18 were novel. To validate the novel SNPs, a further metaanalyses combining the association summary statistics with an additional 140,886 individuals was performed, which validated 17 of the 18 loci. As the Cardio-MetaboChip had high-density coverage across 21 published BP loci and 3 newly identified loci, a fine-mapping analysis was performed to refine the localization of likely functional variants. Using a Bayesian approach, a credible set of variants was defined at each locus, with 99% probability for the set containing or tagging the causal variant. The 99% credible sets included only the index variants at three BP loci ( SLC39A8 , ZC3HC1 , and PLCE1 ).

Rare Genetic Variants and BP

The contribution of low frequency and rare SNPs to BP traits has been largely unexplored by GWAS. However, these SNPs may explain some of the missing phenotypic variance, by having a larger phenotypic effect than common SNPs . The Exome chip is a bespoke genotyping array which consists of ~250,000 mostly rare (MAF≤0.01) and low-frequency (0.01<MAF<0.05) nonsynonymous coding variants. These were selected after sequencing the genomes or exomes of ~18,000 genes in 12,000 individuals, primarily of European ancestry. The array also includes other content (e.g., tagging variants from GWAS and ancestry informative markers).

Two large-scale Exome chip metaanalyses for BP traits have recently been undertaken. One was a European-led consortium that genotyped up to 193,000 individuals (165,276 Europeans and 27,487 South Asians) and assessed association of SNPs with SBP, DBP, PP, and hypertension . The consortium-selected lead variants from 80 novel loci for follow up in an independent set of samples from The Cohorts for Heart and Aging Research in Genomic Epidemiology+ (CHARGE+) Exome Chip BP Working Group ( N ~147,000), and validated 30 novel loci associations. The associations included rare nonsynonymous SNPs (nsSNPs) in 3 novel genes: COL21A1 , RBM47 , and RRAS . The other Exome chip metaanalyses were undertaken by the CHARGE+ Exome Chip BP Working Group for SBP, DBP, MAP, PP, and hypertension . The discovery stage included up to 147,000 individuals comprising European, African, and Hispanic individuals; SNPs were followed-up in metaanalyses of 180,726 samples from the European-led consortium. The analysis identified 31 novel loci and included 3 low-frequency nsSNPs in the novel genes SVEP1 and PTPMT1 , and the previously implicated BP gene NPR1 . In total, 52 distinct novel BP loci were identified using the Exome chip across both consortia.

Each consortium also performed gene-based tests using the Burden test and the sequence kernel association test (SKAT). Burden tests detect associations when all variants contribute to effects in a concordant direction , and SKAT detects effects of alleles that collectively contribute to higher or lower BP effects . These tests have increased statistical power to detect associations in genes harboring rare variants. Analysis was restricted to coding variants with MAF <5% or MAF <1%. One gene was reported by the European consortium to be significantly associated with hypertension ( A2ML1 ), containing multiple rare variant associations. The CHARGE+Consortium identified three significant genes; two of these ( DBH and NPR1 ) overlap loci identified using single SNP testing, PTPMT1 was a novel locus.

A review of the effect sizes across BP SNPs with comparative MAF (≥5%) indicates the majority to have small-effect sizes (mmHg, mean, and s.d.) on BP individually [SBP: 0.51 (0.22); DBP: 0.35 (0.23), and PP: 0.31 (0.09)]. There are four SNPs that have relatively larger effect sizes. These include the three rare nsSNPs discovered by Exome chip analyses: variant rs61760904 at the RRAS locus, associated with SBP, with an effect size of 1.5 mmHg per allele; rs200999181 at the COL21A1 locus associated with PP, with an effect size of 3.14 mmHg per allele and r235529250 at the RBM47 locus associated with SBP, with an effect size of 1.61 mmHg per allele, and rs16833934 at the MIR1263 locus . The fourth SNP, rs16833934 at the MIR1263 locus is associated with DBP. It is a common variant (MAF=26%) and the effect size is 1.63 mmHg per allele. This association was discovered using a gene×age interaction analysis for BP. The SNP was significantly associated with BP amongst 20–29 year olds, within one of the specific age–bin-stratified subgroup analyses. Hence, the effect is more extreme within a particular age group, rather than at an overall population level .

In 2011, the percentage of BP variance explained for 29 BP SNPS was <1% for SBP and DBP . In order to show the added contribution of the recent genetic discoveries to the percentage of variance explained, we have firstly recalculated these results from the 29 SNP genetic-risk score (GRS) and then extended the GRS model to include all currently known BP SNPs, using data from the 1958 Birth Cohort ( N =5639). Of the 29 SNPs from the ICBP GRS, 26 were available in our data. For the full GRS, we included 143 linkage disequilibrium (LD)-filtered ( r ² <0.2) variants that were available from the 163 known BP SNPs. The GRS was constructed according to the trait-increasing alleles and weighted by effect estimates from summary data. The 26 SNP GRS explained 0.67%, 0.72%, and 0.22% of the variance for SBP, DBP, and PP, respectively. This increases to 1.48, 1.66, and 0.47%, respectively, showing at least a twofold increase in the percentage variance explained due to the variants identified since 2011. We note that the percentage variance explained results in the 1958 Birth Cohort and those reported in the literature are variable, as results are population specific and differ according to the sample size tested. In general, we conclude, however, that the percentage variance explained overall is still very low, consistent with large numbers of common variants with weak effects.