Abstract P182: Large Scale Analyses Of Gene-by-smoking Interaction On Cardiovascular Disease

Abstract

Introduction: Cardiovascular disease risk is determined by both genetic susceptibility and lifestyle factors such as smoking. However, gene-by-smoking interactions have been understudied in this context. The objective of this study was to leverage large-scale individual level data to carry out a genome-wide gene-by-smoking interaction analysis on cardiovascular outcomes including coronary heart disease (CHD), stroke and peripheral artery disease (PAD). Methods: Our study sample included 401,384 unrelated, European-ancestry participants of the UK Biobank study. CHD, stroke, and PAD outcomes were based on data from baseline questionnaires and linked data from inpatient admissions and death registries. Smoking status (ever vs never) was based on self-report at baseline. Gene-environment interaction analyses were carried out using GEM (https://github.com/large-scale-gxe-methods/GEM), a new, computationally efficient software tool developed for biobank scale data. Using generalized linear models, we carried out a test of the genetic main effect adjusted for smoking status, a formal test for gene-by-smoking interaction, and a joint test of the main genetic and interaction effects. Single nucleotide polymorphisms with a minor allele frequency less than 1% were excluded from analyses. Results: The formal interaction test revealed a novel genome-wide significant (GWS) gene-by-smoking signal for PAD within the IREB2 gene (p=2.0x10 -8 ), which has been previously associated with smoking intensity and lung disease. The joint test has been shown to increase power for genetic discovery, which we observed here. The joint test for PAD identified 10 additional loci reaching genome-wide significance (p<5x10 -8 ) that were not revealed in the main effect analyses. Several of these loci are in genes linked with measures of body composition in earlier studies ( PNLIP, CDH4, STAT5A, PRKCE ). We did not detect any GWS gene-by-smoking interactions for CHD or stroke. However, for CHD, two novel loci within the ARSG and SARS1 genes were identified using the joint test but had no GWS main effects. Both genes have been associated with LDL-cholesterol levels in prior genome-wide studies. Although our analyses did not yield any signals that exceeded the GWS threshold for stroke, we did identify 25 and 47 independent loci that reached genome-wide suggestive significance (p<1x10 -5 ) for the formal interaction and joint tests, respectively. We are augmenting the current analyses to include additional data from approximately 120,000 trans-ethnic ancestry individuals. Conclusion: The analysis of gene-by-smoking interactions can provide novel insights into the genetic architecture and pathophysiology of cardiovascular disease.