Genome-wide association study meta-analysis of blood pressure traits and hypertension in sub-Saharan African populations: an AWI-Gen study

Abstract

Most hypertension-related genome-wide association studies (GWASs) focus on non-African populations, despite hypertension (a major risk factor for cardiovascular disease) being highly prevalent in Africa. The AWI-Gen study GWAS meta-analysis for blood pressure (BP)-related traits (systolic and diastolic BP, pulse pressure, mean-arterial pressure and hypertension) from three sub-Saharan African geographic regions (N = 10,775), identifies two novel genome-wide significant signals (p < 5E-08): systolic BP near P2RY1 (rs77846204; intergenic variant, p = 4.95E-08) and pulse pressure near LINC01256 (rs80141533; intergenic variant, p = 1.76E-08). No genome-wide signals are detected for the AWI-Gen GWAS meta-analysis with previous African-ancestry GWASs (UK Biobank (African), Uganda Genome Resource). Suggestive signals (p < 5E-06) are observed for all traits, with 29 SNPs associating with more than one trait and several replicating known associations. Polygenic risk scores (PRSs) developed from studies on different ancestries have limited transferability, with multi-ancestry PRS providing better prediction. This study provides insights into the genetics of BP variation in African populations.

Fig. 1. Workflow summary for datasets and analyses.

Stage 1 GWAS was conducted for all five BP-related traits, and a meta-analysis was performed using the summary statistics of the GWAS for each AWI-Gen region (East, South, West) – sample sizes are indicated in brackets. Stage 2 GWAS was conducted for SBP and DBP only, and was a meta-analysis of the Stage 1 results with other African-ancestry summary statistics i.e. UGR and UKBB. Replication of associations was assessed using the GWAS Catalog, PhenoScanner and summary statistics from other studies^,. Transferability across studies was conducted via PRS based on African-ancestry (UGR, UKBBa), Multi-ancestry i.e. PAGE and European-ancestry i.e. UKBB & ICBP cohorts (discovery) used to assess the distribution of SBP and DBP risk according to PRS quintiles in the AWI-Gen cohort (target).

Fig. 2. Discovery GWAS genetic associations in AWI-Gen (Stage 1) and the meta-analysis (Stage 2).

Miami plots with results from Stage 1 and Stage 2 are shown for (a) SBP and (b) DBP and Manhattan plots for Stage 1 for (c) HTN, (d) PP and (e) MAP. GWAS included adjusting for age, age2, sex and the first 10 PCs as covariates. With GW significance = p < 5E−08. Miami and Manhattan plot shows –log10-transformed two-tailed p-value for each BP trait (y-axis) and base pair positions along the chromosomes (x-axis); red line = GW significance (p < 5E−08); purple line = threshold for suggestive association (p < 1E−06).

Fig. 3. Regional plots of novel GW significant (p < 5E−08) associations.

LocusZoom plots showing GW significant associations for a SBP and b PP in the AWI-Gen study. Regional visualization of associated SNP regions was performed using LD from AWI-Gen in LocusZoom V0.4.8, using a 1MB flanking region, which was compared to data in the GWAS Catalog 17. Lead SNPs are indicated by the purple diamond and GWAS Catalog trait labels and genes are shown below the plots. Plots shown for a SBP around the P2RY1 region (rs77846204, p = 4.95E−08 and b PP around the LINC01256 region (rs115808348, p = 1.76E−08, intergenic ELL2P2 – also consisting of rs62317311 (p = 8.92E−07), for the AWI-Gen Stage 1 GWAS (N = 10,775).

Fig. 4. Transferability of Polygenic Risk Score (PRS).

PRSs derived from four GWASs (discovery studies) and applied to the AWI-Gen cohort (target), using PRSice-2 V2.3.5: (1) African-ancestry cohort, UKBBa (n = 3058) (2) African cohort, UGR (n = 6400) (3) European cohort, UK biobank and ICBP (N = 757,601)^, and (4) Multi-ancestry cohort, PAGE (N = 49,839 with 17,152 African-ancestry)^,. a PRS stratification of SBP and DBP in the AWI-Gen target population: Point range-plots comparing the difference in BP-trait mean (mmHg) of the upper PRS quintiles from the lowest, stratified by the discovery datasets (error bars = mean ± 95% confidence intervals) are shown. b Plots showing adjusted variance explained (% R2), between phenotype and risk score estimated, by each PRS for SBP and DBP: P-value (above bar, two-tailed) and number of SNPs (within bar) are stated for the P-value threshold (PT).