Shared genetic loci underlying smoking consumption, alcohol use, and body mass index: evidence from large multi-ethnic genome-wide association studies

Smoking, alcohol abuse, and obesity are major global public health concerns, causing extensive burdens on the healthcare system. Previous studies have characterized all three traits as addictive behaviors with central nervous system (CNS) regulation. There is increased interest in this area due to reported changes in addictive behavior in some individuals taking the weight loss medication semaglutide. Twin studies and genome-wide association studies (GWAS) have demonstrated pleiotropy among smoking, alcohol use, and body mass index (BMI). To date no study has identified the actual pleiotropic variants affecting these phenotypes. We investigated pleiotropic underpinnings of smoking, alcohol use, and obesity using data from three large GWAS of individuals of European (EUR), East Asian (EAS), African (AFR), and Hispanic/Latino (AMR) populations. We used data from the GSCAN GWAS for smoking (i.e., cigarettes smoked per day; N=449,744; 73% EUR, 23% EAS, 3% AFR, and 1% AMR) and alcohol use (i.e., drinks per week; N=771,070; 87% EUR, 12% EAS, 1% AFR, < 1% AMR). Data for BMI were from the ongoing GWAS from the GIANT consortium (N=1,924,645; 81% EUR, 13% EAS, 5% AFR, and 1% AMR). We calculated genetic correlations (rg) using Linkage Disequilibrium Score Regression and applied the multi-trait Adaptive Sum of Powered Score (aSPU) method to identify pleiotropic loci across all three traits. Pleiotropic loci were defined as loci that: 1) exceeded significance threshold for the multi-trait test at PaSPU < 5E–9; and b) nominally significant for more than one trait in univariate GWAS results. Novel loci were defined as loci that: 1) exceeded significance threshold of PaSPU < 5E–9; and 2) were not significantly associated with any trait for any ancestry in univariate GWAS (p ≥ 5E-9) and previous GWASs. In the EUR subpopulation, all trait pairs showed significant genetic correlations, with a positive rg between smoking and alcohol use (rg=0.07, SE=0.03, p=0.01) and smoking and BMI (rg=0.29, SE=0.03, p=1.56E-20). A negative rg between alcohol use and BMI was noted (rg= -0.09, SE=0.02, p=1.08E-5). Across all ancestries, single nucleotide polymorphism (SNPs) representing 1,752 independent loci exceeded the significance threshold of PaSPU < 5E–9. A total of eight loci showed potential pleiotropy for smoking and alcohol use, 262 for smoking and BMI, 326 for alcohol use and BMI, and 150 for all three traits. The remaining loci were only significant for one trait in univariate GWASs (four for smoking, nine for alcohol use, and 993 for BMI). We also identified three novel variants, two of which were potentially pleiotropic for smoking and BMI: KDM5B (lead SNP rs4333896, PaSPU =1.03E-9) that encodes a lysine-specific histone demethylase and has been significantly associated with mitral valve prolapse, and RFWD3 (lead SNP rs11859900, PaSPU =2.11E-9) that has been associated with telomere length, testicular germ cell tumor, and multiple myeloma. The other novel variant, CAMK2D (lead SNP rs4834346, PaSPU =2.13E-9), was potentially pleiotropic for all three traits and has been associated with cardiovascular traits (e.g., electrocardiogram morphology and atrial fibrillation). Our study identified novel variants underlying smoking, alcohol use, and BMI using data from three large GWASs. These results provide further evidence on the shared genetic risks underlying these behaviors and could help clarify molecular functions and identify key biologic pathways.