Excessive alcohol use is a leading contributor to morbidity and mortality. Of those who drink alcohol, 12% meet criteria for alcohol dependence (AD), a serious psychiatric disorder characterized by tolerance, loss of control over drinking, and excessive consumption despite negative consequences. AD is heritable and previous studies have implicated functional variants in ADH1B [1,2], but relatively little is known about the full polygenic architecture of AD.
We performed genome-wide meta-analysis of 14,904 individuals with AD and 37,944 controls from 28 case/control and family-based studies, stratified by genetic ancestry (European, N = 46,568; African; N = 6,280). AD cases were defined using DSM-IV criteria, primarily from structured interviews. Controls were screened for AD and alcohol abuse where possible. Strict quality control, imputation, and GWAS were performed for each cohort, with family-based studies analyzed using generalized estimating equations or logistic mixed models to control for family structure as appropriate.
One locus, covering ADH1B, shows significant association with AD, with independent variants observed in European (rs1229984; p=9.8E-13) and African ancestries (rs2066702; p=2.2E-9), replicating previous findings [1] and providing the strongest evidence to date for rs2066702. A variety of trans-ethnic meta-analysis models yield nearly identical results. The different lead variants are consistent with differences in allele frequency between ancestries; both are functional missense variants in ADH1B that affect the efficiency of oxidizing ethanol. Conditional analysis, along with GTEx gene expression data, suggests that there may be additional independent effects in the locus similarly associated with ADH1B activity.
We observe substantial genome-wide polygenicity in both African and European ancestry samples (European h2g=.09, p=8.0E-7). There was no substantial heterogeneity within or between ancestry, by study design (case/control vs. family-based), or across the full meta-analysis. Using polygenic scores, we evaluate the variance explained for AD diagnosis and symptom counts in external cohorts.
Comparing the meta-analysis of AD in European ancestry samples to related traits identified significant genetic correlation with several psychiatric disorders, including depression (rg=.57, p=3.0E-4), ADHD (rg=.44, p=4.2E-6), and schizophrenia (rg=.36, p=3.2E-11). We also observe noteworthy genetic correlations with use of cannabis (rg=.79, p=2.5E-4) and cigarettes (rg=.71, p=1.3E-7), as well as the age of having a first child (rg=-.63, p=2.0E-9) and educational attainment (rg=-.42, p=6.8E-9). We then evaluate whether these correlations are reflected by associations of AD with the transmission of polygenic risk within families.
Notably, genetic correlations are meaningfully less than 1 with GWAS of alcohol consumption from UK Biobank [3] (rg=.37; p=5.2E-5) and GWAS of the Alcohol Use Disorders Identification Test in 23andMe [4] (rg=.08, p=.65). This suggests genetics of AD does not solely reflect very high risk for alcohol consumption, but also involves other factors, such as loss of control over intake. Potential differences between the genetic etiology of AD and the genetics of alcohol consumption in the population will be an important focus for future research.
References:
[1] Polimanti et al. (2018). Am J Med Genet, doi:10.1002/ajmg.b.32523
[2] Bierut et al. (2012). Mol Psychiatry. doi:10.1038/mp.2011.124
[3] Clarke et al. (2017). Mol Psychiatry, doi:10.1038/mp.2017.153
[4] Sanchez-Roige et al. (2017). Addict Biol., doi:10.1111/adb.12574
