Meta-analysis of genome-wide association study of 1.2 million people finds novel signals in alcohol and nicotine use

Alcohol and nicotine use are leading causes of preventable deaths in the US. Past twin and family studies have shown moderate heritability and genome-wide analyses have found a handful of replicable genes (eg. CYP2A6 and... [ view full abstract ]

Alcohol and nicotine use are leading causes of preventable deaths in the US. Past twin and family studies have shown moderate heritability and genome-wide analyses have found a handful of replicable genes (eg. CYP2A6 and cigarettes per day). However, there are still many variants that may be important to addiction but not yet found due to their small effect sizes. To discover and better understand the biological mechanisms of addiction, GWAS & Sequencing Consortium of Alcohol and Nicotine use (GSCAN) conducted a large meta-analysis across 29 cohorts in over 1 million people for 5 substance use related phenotypes. We targeted four phenotypes related to nicotine use: age of initiation of regular smoking (AgeSmk, N = 234,398), initiation of regular smoking (SmkInit, N = 1,232,091), cigarettes per day (CigDay, N = 337,334), smoking cessation (SmkCess, N = 547,219) and one related to alcohol use: drinks per week (DrnkWk, N = 941,280). We report to have found 564 independently associated variants in 405 loci for these 5 phenotypes. Amongst those, we replicated many previous findings such as the aforementioned CYP2A6 for CigDay and ADH1B for DrnkWk. Pleiotropy was also observed with ~47% of our tested loci implicated in 2 or more phenotypes and 3 of those implicated in all 5 phenotypes. Single nucleotide polymorphism (SNP) based heritability estimated ranged from 4% (DrnkWk) to 8% (CigDay). Genetic correlation analysis shows the four smoking phenotypes are more genetically correlated with each other and most behavioral, psychiatric, substance use and medical phenotypes. DrnkWk is only correlated with SmkInit as compared to the other smoking phenotypes and weakly correlated with most of the other phenotypes. In conclusion, we conducted the largest addiction genome-wide meta-analysis to date and have found numerous loci in different biological pathways. By studying all 5 phenotypes simultaneously, we hope to shed light on the biological mechanisms that are not only specific to each phenotype, but also reflect a general risk for substance use and addiction.