Common and Rare Variants in Alcohol Dependence

Abstract

c h a b c M any lines of evidence converge to show that genetics makes a substantial contribution to the risk for alcohol dependence, explaining about 60% of the variance. However, alcohol dependence is not a Mendelian trait with a simple pattern of inheritance, nor is it deterministic. It is a complex trait like most psychiatric diseases and other common diseases, with both genetic and environmental factors affecting risk. It is difficult to identify genes that affect the risk for complex diseases, but the huge impact of these diseases makes the attempt important. Early studies of alcohol dependence were limited to examining the role of small numbers of genetic variants in candidate genes, genes with a hypothesized role in alcoholism. Some were successful, particularly the identification of coding variations in ADH1B and ALDH2, genes that encode key enzymes in the metabolism of alcohol, with protective variants reducing risk by 2to 8-fold (1). Although other genes have also been implicated, no other variants with comparable effects are known. Some candidate genes have remained controversial. A strategy of linkage analysis followed up by genotyping of candidate genes within linked regions has also had success. A handful of genes have been confirmed to have variants affecting risk, including GABRA2 and ADH4 (2). Effect sizes are, however, small. Technological progress in genotyping single-nucleotide polymorphisms (SNPs) triggered genome-wide association studies (GWAS), which have the potential to discover genes not previously thought to be involved. After an initial striking success for agerelated macular degeneration (3), a wave of studies on many diseases was carried out with disappointing initial results: very few genes passed the accepted level of genome-wide significance, set at a high level of stringency because of the massive multiple testing these studies entail. Frequently, different studies on the same disease did not find the same results. Although this caused some to write off GWAS as a failure, a different take-home message began to emerge recently, as groups combined their data in metaanalyses. As the number of cases and controls grew dramatically, more and more genes have been identified in many diseases (4). It is clear that real effect sizes are much smaller than initially thought (or than calculated from the initial study, the so-called winners curse), so very large studies are needed to reliably detect them. The results of several initial GWAS on alcohol dependence and related traits, including one in this issue (5), fit this pattern. The studies differ in populations and ascertainment. A study of German male alcoholics with early onset of the disease, ascertained from hospitals, did not yield any SNPs at genome-wide significance, but follow-up of the top SNPs and selected candidate genes showed nominal significance, and a combined analysis reported that two SNPs on chromosome 2q35 (in strong linkage disequilibrium) reached genome-wide significance (6). Edenberg et al. (7) carried out a GWAS on subjects from the Collaborative Studies on the Genetics of Alcoholism (COGA) study, with no SNPs reaching ge-