GWAS summary statistics for the traits analyzed in this study were obtained from the GWAS Catalog (Buniello et al., 2019), with detailed information provided in Supplementary Table S1. Summary data for cardiometabolic traits—including systolic blood pressure (SBP), diastolic blood pressure (DBP), high-density lipoprotein (HDL), low-density lipoprotein (LDL), and triglycerides (TG)—were derived from a meta-analysis of East Asian cohorts (Chen et al., 2023), incorporating 92,615 participants from the Taiwan Biobank, Biobank Japan, and the East Asian subset of UK Biobank.

We assessed the genetic correlation (rg) between CVDS and cardiometabolic traits using linkage disequilibrium score regression (LDSC). LDSC estimates rg by analyzing the relationship between GWAS test statistics and LD scores, which represent the cumulative LD between a given SNP and its neighboring variants. This method operates on summary-level GWAS data, making it well-suited for large-scale meta-analyses and robust to confounding from sample overlap. In contrast to approaches requiring individual-level genotypes, LDSC is less susceptible to biases from population stratification or cryptic relatedness. As such, it offers a reliable framework for characterizing the genetic overlap between traits and for informing the selection of phenotypes in subsequent causal inference analyses (Bulik-Sullivan et al., 2015). The formula used in LDSC is as follows: E β j γ j = N 1 N 2 r g M l + N s r N 1 N 2 where β j and γ j   represent the effect sizes of S N P j on the two traits being tested, N 1 and N 2 are the sample sizes for the two traits, N s is the number of overlapping samples between the two traits, r is the phenotypic correlation in the overlapping samples, and, l j is the LD score. In this analysis, precomputed LDSC for HapMap3 SNPs, derived from individuals of European ancestry in the 1000 Genomes Project (Bulik-Sullivan et al., 2015), were used. Variants with an imputation INFO score greater than 0.9 were included for the analysis (Bulik-Sullivan et al., 2015).

Stratified LD Score Regression (s-LDSC) extends the LDSC framework by quantifying the contribution of specific genomic annotations to the heritability of complex traits (Trynka et al., 2013). By modeling heritability across predefined functional categories while adjusting for linkage disequilibrium, s-LDSC enables the identification of biologically relevant regions contributing to disease risk. Unlike conventional approaches that rely on genome-wide significant variants, s-LDSC utilizes all SNPs to provide a more comprehensive view of the polygenic architecture. This method is computationally scalable and suitable for large-scale GWAS datasets. Importantly, s-LDSC can pinpoint enrichment in cell type–specific regulatory elements and functional annotations, aiding in the prioritization of genomic features for follow-up studies. Its integration into cross-trait analyses also allows for more nuanced interpretation of genetic correlations by revealing functional categories that drive shared heritability (Finucane et al., 2015).

To dissect tissue- and cell type-specific contributions to trait heritability, we employed s-LDSC using functional annotations derived from six chromatin marks (DHS, H3K27ac, H3K36me3, H3K4me1, H3K4me3, and H3K9ac) across 88 tissues and cell types from the Roadmap Epigenomics Consortium (Finucane et al., 2015). For each histone mark, annotations were grouped into nine biological categories: adipose, cardiovascular, central nervous system, digestive, immune/blood, liver, pancreas, musculoskeletal/connective tissue, and others. Trait-specific heritability enrichment was calculated for each annotation, and the results were visualized through hierarchical clustering based on normalized enrichment scores. This approach enabled the discovery of distinct enrichment profiles across traits and tissues, revealing shared biological underpinnings among genetically correlated phenotypes. Chromatin mark-specific signals provided further granularity, highlighting regulatory elements that may play central roles in tissue-relevant pathways. Together, these analyses support the functional interpretation of GWAS findings and inform the prioritization of candidate tissues and regulatory mechanisms involved in complex trait etiology.

To complement the genome-wide genetic correlation estimates obtained from LDSC, we applied ρ-HESS to assess local genetic correlations between trait pairs (Shi et al., 2017). While LDSC provides a genome-wide average estimate, ρ-HESS partitions the genetic covariance across 1,703 approximately independent genomic regions, enabling locus-level interrogation of shared genetic architecture. This method uses GWAS summary statistics and accounts for linkage disequilibrium patterns and potential sample overlap, without assuming specific distributions of effect sizes. For each region, ρ-HESS estimates local SNP heritability and covariance by projecting GWAS effect size vectors onto LD-derived eigenvectors. We applied this approach to all trait pairs with significant global genetic correlations from LDSC, using a Bonferroni-corrected threshold (P < 0.05/1,703) to determine statistical significance. This analysis allowed us to identify specific genomic intervals that disproportionately contribute to the observed genome-wide correlations, revealing heterogeneity in shared genetic architecture across loci.

To enhance locus discovery and statistical power, we applied Multi-Trait Analysis of GWAS (MTAG), a method that leverages shared genetic architecture across genetically correlated traits to identify trait-specific SNP associations (Turley et al., 2018). MTAG uses GWAS summary statistics and estimates pairwise genetic correlations via LDSC (Bulik-Sullivan et al., 2015), accounting for sample overlap and trait correlation to generate unbiased SNP-level effect estimates. In our analysis, the included GWAS datasets were largely derived from non-overlapping cohorts; nonetheless, MTAG’s internal framework corrects for any residual overlap using covariance estimates from LDSC, thereby reducing inflation in test statistics. This framework is particularly suited for complex, polygenic traits with overlapping etiology. After stringent variant filtering, we retained SNPs that achieved genome-wide significance (P < 5 × 10⁻⁸) in MTAG and showed suggestive associations (P < 0.01) in the original single-trait GWAS, ensuring robustness and biological relevance. This approach enabled the identification of pleiotropic loci contributing to multiple cardiovascular phenotypes.

We performed genome-wide genetic correlation analysis using LDSC to evaluate shared genetic architecture among four major CVDs and five cardiometabolic traits (Figure 1; Supplementary Table S2). MI exhibited strong and significant positive genetic correlations with SBP (rg = 0.35, SE = 0.05, P = 1.59 × 10⁻¹⁴), DBP (rg = 0.30, SE = 0.05, P = 7.02 × 10⁻⁹), LDL (rg = 0.27, SE = 0.05, P = 1.15 × 10⁻⁷), and TG (rg = 0.21, SE = 0.03, P = 1.92 × 10⁻¹⁰). A significant inverse correlation was observed with HDL (rg = −0.20, SE = 0.05, P = 8.50 × 10⁻⁵). These results confirm the known involvement of blood pressure and lipid metabolism pathways in MI pathophysiology. HF also showed robust genetic correlations with both SBP (rg = 0.52, SE = 0.0891, P = 4.81 × 10⁻⁹) and DBP (rg = 0.54, SE = 0.0937, P = 9.84 × 10⁻⁹), highlighting a shared genetic basis linked to hemodynamic stress. Additionally, HF was modestly associated with TG (rg = 0.22, SE = 0.0699, P = 0.0021), and negatively correlated with HDL (rg = −0.19, SE = 0.0757, P = 0.0142), further suggesting convergence on metabolic dysregulation. In contrast, AF demonstrated weaker genetic correlations with cardiometabolic traits, with no significant associations observed for SBP or DBP. Modest but nominal associations were observed with LDL (rg = −0.22, SE = 0.10, P = 0.0296) and TG (rg = −0.18, SE = 0.07, P = 0.0109), suggesting a partially distinct genetic basis. Arrhythmia showed moderate genetic correlations with SBP (rg = 0.22, SE = 0.0486, P = 8.14 × 10⁻⁶) and DBP (rg = 0.18, SE = 0.0498, P = 0.0004), but weaker or no associations with lipid traits, including non-significant correlations with HDL and LDL. A small negative correlation with TG was noted (rg = −0.10, SE = 0.0442, P = 0.0185).

To further examine shared genetic architecture across cardiovascular phenotypes, we estimated pairwise genetic correlations among MI, HF, AF, and arrhythmia (Table 1). MI showed a strong genetic correlation with HF (rg = 0.43, SE = 0.10, P = 1.76 × 10⁻⁵), and a moderate correlation with AF (rg = 0.27, SE = 0.11, P = 0.0165). However, the correlation between MI and arrhythmia was not significant (rg = 0.07, SE = 0.07, P = 0.2953). In contrast, HF exhibited significant genetic overlap with both AF (rg = 0.48, SE = 0.14, P = 0.0005) and arrhythmia (rg = 0.55, SE = 0.087, P = 3.78 × 10⁻¹⁰), indicating common underlying etiological pathways likely related to electrical and structural remodeling. As expected, AF and arrhythmia were highly correlated (rg = 1.0, SE = 0.08, P = 1.73 × 10⁻⁴³), reflecting substantial phenotypic and mechanistic overlap.

s-LDSC analysis demonstrated significant heritability enrichment for cardiovascular and cardiometabolic traits in cardiac and vascular tissues, particularly the left ventricle, right atrium, aorta, coronary artery, and tibial artery (Supplementary Figure S1). Among these, MI and SBP exhibited the highest enrichment in coronary and aortic tissues, respectively, consistent with their biological roles. Genomic regions bearing active chromatin marks—such as H3K27ac and H3K4me3—in these tissues were preferentially enriched, indicating that regulatory elements, including enhancers and promoters active in cardiomyocytes and endothelial cells, are likely to harbor causal variants. Notably, the enrichment profiles were broadly consistent across MI and related cardiometabolic traits, pointing to shared regulatory architecture and suggesting that pleiotropic effects may be mediated by tissue-specific regulatory programs. These results underscore the importance of cardiovascular tissues in shaping the genetic basis of MI and its metabolic risk factors, and emphasize the value of investigating the functional roles of these tissue-specific regulatory elements.

To further explore the genetic interplay between cardiovascular diseases and cardiometabolic traits, we conducted a regional analysis of local genetic correlations. This analysis revealed several loci with statistically significant genetic overlap between CVD phenotypes and metabolic risk traits. For instance, a positive local genetic correlation was observed between MI and HDL cholesterol on chromosome 12 (Supplementary Table S3), a region enriched with genes implicated in lipid metabolism, including HNF1A (Ai-Ghalayini et al., 2020), which is known to affect HDL levels and coronary artery disease risk. In contrast, we identified negative local correlations between MI and LDL cholesterol at multiple loci. On chromosome 9 (Supplementary Table S3), the region includes ABO, a gene involved in coagulation and lipid regulation (Li et al., 2015), and SURF4, which facilitates the secretion of lipoproteins such as VLDL and LDL (Wang et al., 2021). Two additional regions on chromosome 19 were noteworthy: the first (positions 9,383,877–11,849,449; Supplementary Table S3) encompasses LDLR, a gene central to LDL clearance and cardiovascular risk (Franceschini et al., 2009); the second (positions 43,862,455–45,579,043; Supplementary Table S3) contains the APOE gene cluster (APOE, APOC1, APOC2, and APOC4), all of which play essential roles in lipid transport and homeostasis. Notably, genetic variants in APOE have been consistently associated with LDL levels and predisposition to cardiovascular disease (Chaudhary et al., 2012).

To identify pleiotropic loci contributing to CVDs and related cardiometabolic traits, we conducted a multi-trait GWAS using MTAG across MI, HF, SBP, DBP, and other relevant traits. This analysis revealed several new genome-wide significant loci (P < 5 × 10⁻⁸) exhibiting consistent, albeit modest, associations across traits (Table 2). Additionally, we identified several previously unreported East Asian-specific loci characterized by substantially higher minor allele frequencies in East Asian populations compared to other populations (Supplementary Table S4). Notably, the locus tagged by rs13414987 (APOB) was jointly associated with HF and MI (P = 1.35 × 10⁻⁸), implicating a gene central to lipid metabolism and atherosclerosis (Supplementary Figure S2). Similarly, rs1942867 (MC4R) showed pleiotropic effects on HF and MI (P = 4.24 × 10⁻⁸), consistent with MC4R’s known role in energy homeostasis and obesity, both critical contributors to cardiac risk (Supplementary Figure S2). Of particular interest, rs3824359 in QSOX2 was significantly associated with HF and SBP (P = 7.06 × 10⁻¹⁰), a novel pleiotropic locus not previously reported in cardiometabolic GWAS (Supplementary Figure S2). To explore the functional relevance of rs3824359, we examined eQTL data from GTEx and found that this variant significantly influences QSOX2 expression across multiple tissues. The lead allele was associated with increased QSOX2 expression in cultured fibroblasts (P = 2.00 × 10⁻²⁰), testis, tibial artery (P = 1.60 × 10⁻¹²), and subcutaneous adipose tissue, as well as in cardiovascular-relevant tissues including the heart (left ventricle and atrial appendage), skeletal muscle, and arterial vasculature (Supplementary Table S5). Notably, the variant also modulated gene expression in esophageal and brain tissues, suggesting a broad regulatory footprint. Given QSOX2’s role in oxidative protein folding and redox homeostasis, its upregulation in vascular and cardiac tissues could implicate reactive oxygen species modulation as a shared pathway in HF and blood pressure regulation. These findings nominate QSOX2 as a previously unrecognized contributor to cardiovascular pathophysiology and highlight the potential of pleiotropic loci to reveal convergent molecular mechanisms underlying complex traits.

Moreover, the region encompassing GUCY1A1 and GUCY1B1—encoding key subunits of soluble guanylate cyclase (sGC), the intracellular receptor for nitric oxide—emerged as a candidate with dual associations to both MI and blood pressure traits (Table 3). This pathway mediates cGMP synthesis, promoting vasodilation and vascular tone regulation. Regional association analyses revealed evidence of allelic heterogeneity, with at least two statistically independent association peaks, implying the presence of multiple functional variants or distinct regulatory elements within this locus (Figure 2). Prior studies have linked this region to blood pressure phenotypes, and the cumulative burden of risk alleles in this region has been associated with heightened susceptibility to stroke and coronary disease. Our eQTL investigation using GTEx data further supports the functional relevance of this locus, as risk alleles were associated with reduced expression of GUCY1A1 and GUCY1B1 in arterial tissues, potentially impairing NO-sGC signaling and contributing to vascular dysfunction (Supplementary Table S5). These findings underscore the mechanistic convergence between endothelial signaling, hemodynamic regulation, and ischemic cardiac events.