Abstract MP11: Systematic Discovery Of Thousands Of Gene-Environment Interactions For Cardiometabolic Serum Biomarkers

Abstract

Background: Gene-environment interaction (GEI) analysis enables us to understand how genetic variants modify the effects of environmental exposures on cardiometabolic risk factors, providing a foundation for genome-based precision medicine. Ideally, these interactions could be mapped comprehensively across all measured genetic variants, exposures, and outcomes, but this approach is computationally intensive and statistically underpowered. Recent studies have shown that variance-quantitative trait loci (vQTLs), or genetic variants that associate with differential variance of an outcome, are substantially enriched for underlying GEIs. Here, we sought to first identify vQTLs for cardiometabolic traits, then use this smaller genetic search space to uncover novel gene-environment interactions across thousands of environmental exposures. Methods: A two-stage, multi-ancestry analysis was conducted in 355,790 unrelated participants from the UK Biobank. First, we performed a genome-wide vQTL scan for each of 20 serum metabolic biomarkers, including but not limited to lipids, lipoproteins, and glycemic measures. This scan used Levene’s test to identify genetic markers whose genotypes are associated with the variance, rather than the mean, of the biomarker. Next, we collected over 2000 variables corresponding to socioeconomic, dietary, lifestyle, and clinical exposures, and conducted an interaction analysis for each combination of exposure and vQTL-biomarker pair. For each stage, the analysis was initially stratified by ancestry then meta-analyzed to generate the primary set of results. Results: vQTLs were identified at 514 independent regions in the genome, with most of these genetic variants already known to affect the mean biomarker level. In the subsequent gene-environment interaction analysis, we found 2,162 significant interactions passing a stringent significance threshold adjusted for multiple testing ( p < 0.05 / 578 vQTL-biomarker pairs / 2140 exposures = 4х10 -8 ). Some of these expanded on existing findings; for example, genetic marker rs2393775 in the HNF1A gene interacted with education level (as a proxy for socioeconomic status) to influence hsCRP ( p = 1.3х10 -10 ), building on a previous finding that HNF1A variants modify the effect of perceived stress on cardiovascular outcomes. Others highlighted novel biology, such as an interaction between variants near the fatty liver-associated gene TM6SF2 and oily fish intake on total and LDL-cholesterol levels ( p = 6.6х10 -9 ). Conclusions: Our systematic GEI discovery effort identified thousands of interactions that may impact cardiometabolic risk, both expanding on previous research and identifying novel biological mechanisms. This catalog of vQTLs and interactions can inform future mechanistic studies and provides a knowledge base for genome-centered precision approaches to cardiometabolic health.