Proteome-wide mendelian randomization identifies circulating proteins influencing cardiac structure and function

Background Combining high-throughput plasma proteomics with cardiac magnetic resonance (CMR) imaging-derived phenotypes (IDPs) offers a scalable framework to uncover biological mechanisms and therapeutic targets. However, conventional observational analyses are vulnerable to confounding and reverse causation. Mendelian randomization (MR) strengthens causal inference by leveraging genetic variants as instrumental variables. Purpose To evaluate causal associations of circulating plasma proteins with prognostically and mechanistically relevant left ventricular (LV) phenotypes derived from CMR. Methods We performed proteome-wide MR integrating genome-wide association studies (GWAS) of plasma proteomics (n=2,923 proteins from n=54,219 participants) and seven CMR phenotypes (n=36,041 to 43,230 participants) from UK Biobank (Figure 1). CMR IDPs were LV ejection fraction (LVEF), indexed end-diastolic volume (LVEDVi), indexed end-systolic volume (LVESVi), indexed stroke volume (LVSVi), indexed mass (LVMi), trabeculation, and native T1. Genome-wide significant protein quantitative trait loci (pQTLs; p<5x10-8) underwent linkage disequilibrium clumping (r2<0.001; 10,000 kb window) to retain independent variants. Both cis- and trans-pQTLs were included. Exposure and outcome summary statistics were harmonized to align effect alleles. Inverse-variance weighted (IVW) MR was the primary estimator for each protein-CMR pair. Prespecified sensitivity analyses included weighted median/mode estimators, MR-Egger, MR-Pleiotropy RESidual Sum and Outlier (MR-PRESSO), and Steiger filtering to evaluate heterogeneity, horizontal pleiotropy, and directionality. IVW p-values were Bonferroni-corrected across all protein-CMR analyses. Results Five proteins showed robust associations with LV phenotypes after multiple testing correction and sensitivity testing (Table 1). Higher genetically predicted Heart of glass homolog 1 (HEG1) and Ephrin type-B receptor 4 (EPHB4) levels were causally associated with lower LVSVi (HEG1 IVW β=-0.106; EPHB4 IVW β=-0.105) and LVEDVi (HEG1 IVW β=-0.096; EPHB4 IVW β=-0.088). Higher Semaphorin-4C (SEMA4C), Interleukin-3 receptor subunit alpha (IL3RA) and Roundabout homolog 4 (ROBO4) levels were causally associated with lower LVSVi (SEMA4C IVW β=-0.179; IL3RA IVW β=-0.099; ROBO4 IVW β=-0.081). Effect directions were concordant across sensitivity estimators, Steiger filtering supported directionality, and Egger intercept/MR-PRESSO tests did not indicate pleiotropy or outlier-driven results. No robust causal associations were identified for LVEF, LVESVi, LVMi, trabeculation or T1. Conclusions We identified five circulating proteins with robust MR evidence of causal association with reduced LVSVi and/or LVEDVi. These proteins implicate endothelial integrity (HEG1, EPHB4, ROBO4) and immune signaling (IL3RA) as candidate upstream pathways influencing LV remodelling, and may represent therapeutic targets for experimental validation.Figure 1Table 1