Sex-specific genetic architecture of blood pressure

Abstract

The genetic and genomic basis of sex differences in blood pressure (BP) traits remain unstudied at scale. Here, we conducted sex-stratified and combined-sex genome-wide association studies of BP traits using the UK Biobank resource, identifying 1,346 previously reported and 29 new BP trait-associated loci. Among associated loci, 412 were female-specific (P_female ≤ 5 × 10^-8; P_male > 5 × 10^-8) and 142 were male-specific (P_male ≤ 5 × 10^-8; P_female > 5 × 10^-8); these sex-specific loci were enriched for hormone-related transcription factors, in particular, estrogen receptor 1. Analyses of gene-by-sex interactions and sexually dimorphic effects identified four genomic regions, showing female-specific associations with diastolic BP or pulse pressure, including the chromosome 13q34-COL4A1/COL4A2 locus. Notably, female-specific pulse pressure-associated loci exhibited enriched acetylated histone H3 Lys27 modifications in arterial tissues and a female-specific association with fibromuscular dysplasia, a female-biased vascular disease; colocalization signals included Chr13q34: COL4A1/COL4A2, Chr9p21: CDKN2B-AS1 and Chr4q32.1: MAP9 regions. Sex-specific and sex-biased polygenic associations of BP traits were associated with multiple cardiovascular traits. These findings suggest potentially clinically significant and BP sex-specific pleiotropic effects on cardiovascular diseases.

Extended-Data-Fig.1.. BP gene-by-sex effects.

Volcano plots of LD-pruned (a) SBP, (b) DBP, (c) and PP loci are shown for: (i) all sex-specific trait-associated loci, (ii) unique sex-specific trait-associated loci identified by sex-stratified BP GWAS only, and (iii) loci from sex-dimorphic effects analysis (SDE P < 1x10⁻⁵). Each line represents a locus, connecting males-only (blue dots) and females-only (purple dots) GWAS results, based on linear mixed models adjusted with BP-associated covariates. The two-sided P values reported are unadjusted. Line length indicates differences in effect size or association strength, with horizontal differences showing directional discrepancies and vertical differences indicating significance magnitude. (d) LocusZoom plots for regional association results from females-only, males-only, and SDE DBP GWAS are shown, for chromosomes 13q34-COL4A1/COL4A2 and 19q13.2 ACTN4/CAPN12, with female-specific associations (P_female≤5x10⁻⁸, P_male>5x10⁻⁸) and strong SDE effects (SDE P_GC < 5x10⁻⁵). 1000G EUR LD reference was used.

Extended-Data-Fig.2.. DBP gene-sex interaction TWAS analysis identified COL4A1.

Manhattan plots show gene-based transcriptome-wide association studies (TWAS) for DBP using genotype-sex summary statistics using the GEM program (Fig. 3c) based on a linear mixed model adjusted with BP-associated covariates. MAGMA software was used to map all SNPs to 19,184 protein coding genes based on genomic locations. P-values presented are unadjusted and from two-tailed tests. The red dashed line indicates the genome-wide significance threshold at P=2.61×10⁻⁶, defined as P=0.05/19184. Results are shown for (a) gene-sex interaction analysis with matched sample size (N_male=145,375 and N_female= 143,102) (b) gene-sex interaction analysis included all available UKB samples (N_male=166,379 and N_female=193,523).

Extended-Data-Fig.3.. Comparison of colocalization findings at sex-specific BP loci.

(a) Arterial eQTL-associated genes (eGenes) colocalized (PP.H4>0.5) with female-specific, male-specific, or non-sex-specific BP-associated loci are summarized with Venn diagrams. (b) Frequencies of arterial eGenes colocalized with sex-specific BP-associated regions were compared to non-sex-specific BP-associated regions using a Fisher’s exact test; the table presents a summary of all BP loci, as well as specific BP trait loci. (c) Arterial eGenes for sex-specific BP loci with sex-biased colocalization are shown; These loci specifically colocalize with the combined-sex eQTL data in only one sex of the GWAS data, indicated by pp.abf.H4>50% in one sex and pp.abf.H4<50% in the other sex. (d) eGenes for GWAS loci exhibiting sex-specific associations and sex-biased expression in arterial tissues, based on the GTEx v8 sex-biased genes database; heatmap values represent the largest effect size of differential expression by sex (F-M) in aorta, coronary artery, and tibial artery, indicating the magnitude of female-biased (red) or male-biased (blue) regulation for each gene. Male-specific associated loci and female-specific associated loci are denoted as 1 and 2, respectively. Strong functional candidate genes are labeled through colocalization of GWAS and eQTL in (e). O: pp.abf.H4>0.9. o: pp.abf.H4>0.75. c: pp.abf.H4>0.5.

Extended-Data-Fig.4.. TFBS enrichment at sex-specific BP-associated loci.

(a) Beeswarm plots show the top 10 ranked TFs based on TFBS enrichment tests of autosomal sex-specific BP loci using LOLA's reference ChIP-Seq datasets (268 TFs out of 4166 datasets). LD-pruned GWAS loci associated with BP in females only (SBP: 86 loci, DBP: 183 loci, PP: 140 loci), males only (SBP: 37 loci, DBP: 44 loci, PP: 60 loci), and their combinations were tested using Unibind in a +/−500 Kb window, comparing genomic locations of arterial expressed genes as background. Each plot displays the distribution of enrichment test results for each TF. P values are unadjusted. The heatmap presents each of top 10 enriched TFs from these 12 separate tests of TFBS enrichment. (b) Beeswarm plots show the top 10 ranked TFs based on differential TFBS enrichment analyses within a 500Kb window, comparing sex-specific and non-sex-specific BP loci. P values are unadjusted.

Extended-Data-Fig.5.. Sex-specific epigenomic tissue enrichments.

Results were in comparison to results from the combined-sex BP GWAS. ENCODE database's human H3K27ac ChIP-seq experiments narrow peak data (N=294), consisting of 158 females from 40 tissues, 131 males from 34 tissues, and 3 individuals of unknown sex, were utilized to examine the epigenomic features of GWAS loci associated with sex-specific BP traits. (a) GWAS female-specific loci (SBP: 86 loci, DBP: 183 loci, PP: 140 loci), (b) male-specific loci (SBP: 37 loci, DBP: 44 loci, PP: 60 loci), and (c) all BP loci identified in the combined-sex samples (288 SBP loci, 505 DBP loci, 473 PP loci), were tested for their enrichment of open chromatin regions in tissues of the same sex. The GREGOR program (using EUR as the LD reference) was employed which was based on a non-parametric statistical method to assess the probability that the overlap with control SNPs surpasses that observed with GWAS index SNPs. P values are unadjusted. The best P value for each tissue was reported. Blue filled bars indicate enrichments that passed the Bonferroni correction P value threshold. F: female. M: male. O: unknown sex.

Extended-Data-Fig.6.. Predicting hypertension in MGI using sex-stratified polygenic scores.

(a) The AUC of ROC and Brier score were utilized to assess the PGS for SBP, DBP, and PP for hypertension (HTN) prediction. PGS's were estimated for each MGI sample by combining risk scores of the top loci identified in the female-only (F), male-only (M), or combined-sex (C) UKB BP GWAS. Logistic regression models were trained on 80% of the data and tested on the remaining 20% using a total of 45,500 individuals from MGI (N_female=12,427 controls and 10,502 cases; N_male=8,691 controls and 11,880 cases). The analyses included PGS_BP-only model, or models with PGS_BP and relevant covariates (age, age², BMI, sex (if applicable), and PC1-PC5) as variables to predict HTN in female or male individuals. (b) Forest plots illustrate odds ratios (OR) and minus log10(P) demonstrating the associations between PGS's and HTN status (N_female=12,427 controls and 10,502 cases; N_male=8,691 controls and 11,880 cases). This analysis employed logistic regression with identical covariates and the same MGI samples as in (a). Sensitivity analyses were conducted by using only the top 50 loci from each GWAS to derive PGS's. The middle points represent the mean odds ratios for each study, and the lines indicate the 95% confidence intervals (mean+/−1.96 SD) for these ratios. P values are unadjusted (two-tailed). (c) Proportion of variance in phenotype explained (PVE) by a given SNP. PVE was calculated based on all top LD-pruned loci identified in UKB female, male, or combined-sex BP GWAS association or only the top 50 loci from each study.

Extended-Data-Fig.7.. Cardiovascular risk in high and low SBP polygenic score groups.

Samples in the top 10% PGS_SBP groups were compared to the bottom 10% PGS_SBP group for their hypertension (HTN), coronary artery disease (CAD), and stroke frequencies in UKB, using a Fisher’s exact test. Forest plots depict odds ratios (OR) and minus log10(P).(N_female=174,664, N_female=174,664 males; including sample sizes for HTN cases: N_female=43,171, N_male=53,612; CAD cases: N_female=2323, N_male=8471; stroke cases: N_female=1831, N_male=3116) and MGI (N_female=23,921, N_male=21,077; including sample sizes for HTN cases: N_female=11,171, N_male=12,265; CAD cases, N_female=2089, N_male=3819; stroke cases: N_female=980, N_male=1089). The middle points represent the mean odds ratios for each study, and the lines indicate the 95% confidence intervals (mean +/− 1.96 SD) for these ratios. P values (wo-tailed) are unadjusted.

Extended-Data-Fig.8.. Cross-trait sex-specific colocalization of BP-associated loci with CVDs.

For loci identified in the analyses of SBP (a), DBP (b) and PP (c) with colocalization posterior probability >0.50 in only one sex and the differences in colocalization posterior probability between females and males >0. 5, are shown in a 250 Kb window. The closest gene to the index SNP is shown for each region. Heatmap values reflect the differences (ranging from −1 to 1) in the posterior probabilities from the colocalization of each CVD trait and female-only SBP associations, and the colocalization of each CVD trait and male-only SBP associations. Blue and red colors represent regions with male-specific and female-specific colocalization, respectively. AS: any stroke. AIS: any ischemic stroke. LAS: large artery stroke. CES: cardioembolic stroke. SVS: small vessel stroke. ICA: intracranial aneurysms. SAH: subarachnoid hemorrhage. uIA: unruptured intracranial aneurysms; HTN: hypertension. CAD: Coronary artery disease. MI: myocardial infarction. FMD: fibromuscular dysplasia. PE: preeclampsia. HF: heart failure; AF: atrial fibrillation; AA: aortic aneurysm; AA.TAA: thoracic aortic aneurysm; AortaD.aa: thoracic aorta dimensions, ascending aortic diameter; AortaD.da: thoracic aorta dimensions, descending aortic diameter; SCAD: spontaneous coronary artery dissection.

Extended-Data-Fig.9.. Regional association plots of loci with female-specific colocalization signals.

LocusZoom plots are shown for the BP sex-stratified GWAS and FMD GWAS, SCAD GWAS, and PE GWAS, using 1000G EUR as LD reference (r²>0.2). P value presented are two-tailed and un-adjusted. (a) The chromosome 9p21 CDKN2B-AS1 region, with a lead SNP rs1333047 in PP females and a lead SNP rs2383206 in FMD (LD r²=0.9). (b) The 4q32.1 MAP9/GUCY1A3 region, with the same lead SNP, rs17033041, in PP in females and in FMD (c) The 21q22.11 LINC00310 region lead SNP rs9305545 in DBP in females s and a lead SNP of rs28451064 in SCAD or FMD (LD r²=0.81). (d) The 4q21.21 FGF5 region lead SNP rs11099097 in PP in females and lead SNP rs13125101 in PE (LD r²=0.95 for these two SNPs).

Extended-Data-Fig.10.. Tissue gene expression of COL4A1 and COL4A2.

Gene expression of (a) COL4A1 and (b) COL4A2 was queried in GTEx v8 portal (dbGaP Accession phs000424.v8. p2) for all available human tissues. Expression values shown in transcripts per million (TPM) were calculated from a gene model with isoforms collapses to a single gene, for which box plots are shown as median, 25%, and 75%. Outliers are defined as pointsoutside the 1.5 x interquartile range. N_{Adipose - Subcutaneous}=663, N_{Adipose - Visceral}=541, N_{Adrenal Gland}=258, N_{Artery - Aorta}=432, N_{Artery - Coronary}=240, N_{Artery - Tibial}=663, N_Bladder=21, N_Brain - _Amygdala=152, N_{Brain - Anterior cingulate cortex}=176, N_{Brain - Caudate}=246, N_{Brain - Cerebellar Hemisphere}=215, N_{Brain - Cerebellum}=241, N_{Brain - Cortex}=255, N_{Brain - Frontal Cortex}=209, N_{Brain - Hippocampus}=197, N_Brain- _Hypothalamus - =202, N_{Brain - Nucleus accumbens (basal ganglia)}=246, N_{Brain - Putamen (basal ganglia)}=205, N_{Brain - Spinal cord (cervical c-1)}=159, N_{Brain - Substantia nigra}=139, N_{Breast - Mammary Tissue}=459, N_{Cells - Cultured fibroblasts}=504, N_{Cells - EBV-transformed lymphocytes}=174, N_{Cervix - Ectocervix}=9, N_{Cervix - Endocervix}=10, N_{Colon - Sigmoid}=373, N_{Colon - Transverse}=406, N_{Esophagus - Gastroesophageal Junction}=375, N_{Esophagus - Mucosa}=555, N_{Esophagus - Muscularis=}515, N_{Fallopian Tube}=9, N_{Heart - Atrial Appendage}=429, N_{Heart - Left Ventricle}=432, N_{Kidney - Cortex}=85, N_{Kidney - Medulla}=4, N_Liver=226, N_Lung=578, N_{Minor Salivary Gland}=162, N_{Muscle - Skeletal}=803, N_{Nerve - Tibial}=619, N_Ovary=180, N_Pancreas=328, N_Pituitary=283, N_Prostate=245, N_{Skin - Not Sun Exposed (Suprapubic)}=604, N_{Skin - Sun Exposed (Lower leg)}=701, N_{Small Intestine - Terminal Ileum}=187, N_Spleen=241, N_Stomach=359, N_Testis=361, N_Thyroid=653, N_Uterus=142, N_Vagina=156, N_{Whole Blood}=755.

Fig.1,. Study Overview.

This study aimed to define sex-specific and sexually dimorphic genetic associations with blood pressure (BP) traits and understand genetic influences on gene regulation and their relationships with hypertension (HTN) and cardiovascular diseases (CVDs). We conducted sex-stratified and combined-sex BP genetic association studies using samples from the UKB resource. Findings were further evaluated in the MGI and GERA cohorts. Gene regulatory effects of identified variants were analyzed. Pleiotropy of sex-specific BP effects were further investigated for general HTN, female-biased forms of HTN, and several CVDs. SBP, systolic BP; DBP, diastolic BP; PP, pulse pressure; TFBS, transcription factor binding site; eQTL, expression quantitative trait loci. Blue: male effects. Purple: female effects.

Fig.2.. Sex differences in the complex genetic architectures of BP traits and effects on gene regulation.

(a) Miami plots and Venn diagrams summarizing sex-stratified and sex-combined GWAS results for SBP, DBP, and PP traits in the UKB dataset. Miami plots overlay female-only (purple) and male-only (blue) GWAS results on combined-sex GWAS results (gray). Venn diagrams depict independent genetic regions identified through LD pruning of the GWAS results (r²=0.2, +/−500 Kb). (b) Volcano plots show LD-pruned loci for SBP, DBP, and PP categorized into groups of all male-specific and all female-specific trait-associated loci, and loci identified from analyses of sex-dimorphic effects (SDE, P<1x10⁻⁵). Each line connects male-only (blue dots) and female-only (purple dots) GWAS results, representing a locus. Longer lines indicate larger differences in male and female associations, in effect size (Z-scores) or association strength (P values). Horizontal differences imply directional disparities, while vertical differences indicate statistical significance magnitude. (c) Summary of colocalized arterial eQTL-associated genes for GWAS loci exhibiting female-specific associations (purple cells) or male-specific associations (blue cells) with BP traits meeting the genome-wide significance threshold of P_GC < 5x10⁻⁸, using GTEx v8 arterial tissues (tibial, coronary, aorta) eQTL data from combined-sex samples. X indicates loci uniquely identified from sex-stratified GWAS, D indicates loci with effect directions that differed by sex, S indicates loci colocalized with GTEx arterial eQTLs in only one sex, and E indicates loci arterial eQTL associated genes with gene-sex interactions in the GTEx gene expression resource. (d) Motif enrichments of known nuclear receptors in the promoter regions of sex-biased gene candidates identified from sex-stratified GWAS findings, based on a hypergeometric method. Un-adjusted P values are presented(e) The top 10 differential enriched TFBS when comparing all sex-specific BP loci to non-sex-specific loci within a 500Kb window. P values presented are un-adjusted.

Fig.3:. Statistical analysis of sex interactions by evaluating sex- dimorphic effects (SDE) and gene-by-sex interactions.

(a) Scatter plots displaying beta coefficients and Z-scores (normalized beta coefficients) for male-only and female-only BP GWAS with SDE-P<0.05. Two-sided P-values were unadjusted for multiple testing. Variants located in the second and fourth quadrants indicate sex-dimorphic effect directions. The color of the dots indicates the range of SDE-P-values: gray for 10⁻⁴ to 0.05, orange for 10⁻⁵ to 10⁻⁴, blue for 5x10⁻⁸ to 10⁻⁵, and purple for 0 to 5x10⁻⁸. (b) Manhattan plot of the SDE analysis results of sex-stratified BP GWAS for DBP in the UKB data set. Orange, P<10⁻⁴; blue, P<10⁻⁵; and purple, P<5x10⁻⁸. (c) Manhattan plot of the genotype-by-sex interaction results of DBP in UKB, using linear mixed models adjusted for BP-associated covariates. Red line indicates P=5x10⁻⁸. (d) Regional genotype-sex interaction plot of the DBP association locus (index SNP: rs12429386 and rs2000660, +/−400 kbp) on chromosome 13q34-COL4A1/COL4A2. P values were derived from a two-sided test and are reported here un-adjusted. LD r² was estimated based on the 1000G EUR population. LD R² between rs12429386 and rs2000660 is 0.0016.

Fig.4.. Associations of sex-stratified BP polygenic scores with cardiovascular traits and diseases defined sex-dimorphic or sex-specific associations.

(a) Logarithmic odds ratio (OR) forest plots display the associations between polygenic scores (PGS) BP traits (SBP, DBP, and PP) derived from female-only (F, purple), male-only (M, blue), or combined-sex (C, orange) GWAS and various cardiovascular diseases (CVD). Each PGS was constructed using the top loci identified in the corresponding GWAS of UKB data. Association tests were conducted using an approximation of regression analysis based on summary statistics, without the application of multiple test correction. The middle points of the forest plots represent the mean odds ratios for each study, and the lines indicate the 95% confidence intervals (mean +/− 1.96 SD) for these ratios. Asterisks (*) denote sexually dimorphic effects. (b) Volcano plots compare the effect sizes, using Z-score statistics, of the association between PGS of BP from female-only and male-only GWAS and each CVD, as labeled. The plots depict Z-scores and corresponding p-values. CVD abbreviations: AS - any stroke, AIS - any ischemic stroke, LAS - large artery stroke, CES - cardioembolic stroke, SVS - small vessel stroke, ICA - intracranial aneurysms, SAH - subarachnoid hemorrhage, uIA - unruptured intracranial aneurysms, HTN - hypertension, CAD - coronary artery disease, MI - myocardial infarction, FMD - fibromuscular dysplasia, PE - preeclampsia, HF - heart failure, AF - atrial fibrillation, AA - aortic aneurysm, AA.TAA - thoracic aortic aneurysm, AortaD.aa - thoracic aorta dimensions, ascending aortic diameter, AortaD.da - thoracic aorta dimensions, descending aortic diameter, SCAD - spontaneous coronary artery. N_HTN=67389, N_PE=349327, N_MI=639221, N_CAD=1165690, N_AS=1308460, N_AIS=1296908, N_LAS=1241207, N_CES=1245612, N_SVS=1241619, N_ICA=79429, N_SAH=77074, N_uIA=74004, N_HF=1665481, N_AF=2339188, N_AA.TAA=19646, N_AA=653950, N_AortaD.da=38694, N_AortaD.aa=39688, N_FMD=5656, N_SCAD=11209.

Fig.5.. Fine-mapping of the chromosome 13q34-COL4A1/COL4A2 region overlapping DBP and PP signals.

(a) LocusZoom plots illustrate the genetic association of DBP_F, PP_F, SCAD, and FMD GWAS, using 1000G EUR as the LD reference (r²>0.2). Results of stepwise model selection using multi-SNP-based conditional and joint association analyses in the region of chromosome 13:110793123-111043309, with prioritized SNPs marked. rs9521636/rs7326444, a female-specific DBP-associated locus, aligns to SCAD association in this region (upper panel); rs11838776/rs55940034 a female-specific DBP- and PP-associated locus aligned with both FMD (4^th panel) and SCAD (3rd panel) GWAS results. LD R² (rs7326444, rs9521636) =0.98, R² (rs11838776, rs55940034) =0.89, R² (rs11838776, rs12867262) =0.302, and R² (rs7986871, rs4773143) =1.0. (b) LD matrix of chromosome 13:110793123-111043309, inferred from 1000G CEU (European ancestry) using LDlink, is shown, highlighting 5 LD blocks in the region. (c) LocusZoom plots for the corresponding regions shows association results in the male-only DBP or PP GWAS.