Alcohol abuse is a global problem, constituting the seventh leading risk factor for death and disability. Worldwide, over 100 million people had an alcohol use disorder in 2016. Statistics from the National Survey on Drug Use and Health show that >85% of adults in the United States report ever having consumed alcohol, with >25% reporting binge drinking in the past month. The proportion of adults in the United States with an AUD is estimated to be 6.2%. Alcohol use behaviors are complex, and how and why people drink is partially influenced by genetic factors. However, identifying the genetic factors that increase the risk for harmful drinking has been challenging, partially because patterns of alcohol use are dynamic across the lifespan. The terms used to describe alcohol use and abuse are as diverse as the behaviors themselves. Hazardous drinking describes heavy drinking that places an individual at risk for future harm. Harmful drinking and alcohol abuse are defined as drinking that causes mental or physical damage to the individual. These descriptive terms were devised to identify individuals who would benefit from brief interventions and are assessed using screening questionnaires such as the Alcohol Use Disorders Identification Test. Alcohol dependence was, until recently, defined according to the DSM-IV and required the presence of 3 or more of 7 criteria in a 12-month period. The DSM-IV made a distinction between alcohol abuse and dependence that was removed under DSM-V and replaced with “mild”to “severe” definitions of AUD. Genetic studies encompass the wide range of alcohol use phenotypes; in this review we mirror the language used in the original studies. AUD can be viewed as the end point of a series of transitions , which begin with the initiation of use,grow rack continue with the escalation to hazardous drinking and culminate in compulsive harmful use that persists despite negative consequences.
Genome-wide association studies have been instrumental in discovering novel genetic loci associated with multiple psychiatric conditions. In the field of AUD genetics, studies have mostly focused on either levels of consumption or AUD diagnosis. Recent GWAS have now begun to identify hundreds of genome-wide significant variants, and provide evidence that the components of alcohol use behavior have a distinct genetic architecture. In this review, we provide an overview of recent molecular genetic findings of alcohol use behaviors from the largest GWAS performed to date. Other reviews have elegantly summarized findings from twin and family studies of heritability, linkage, candidate gene and GWAS [e.g. ], and we extend on recent reviews of the molecular genetics of AUD by including additional GWAS of alcohol use behaviors that identify genome-wide significant hits. In addition, we discuss the application of polygenic methods, which provide mounting evidence that alcohol use and misuse are partially distinct. Finally, we delineate future directions to investigate the different etiologic sources that underlie the life course of alcohol use behaviors. For decades, candidate gene studies were used to determine the contribution of specific genes that increase risk for AUD. Candidate gene studies tended to focus on genes that influenced pharmacokinetic and pharmacodynamic factors.One exception to this are the genes encoding ethanol metabolizing enzymes, particularly alcohol dehydrogenase and aldehyde dehydrogenase , which have repeatedly been shown to have the largest impact on alcohol consumption and risk for AUD. As study designs have evolved to incorporate GWAS, researchers have been able to scan the whole genome without any hypotheses about the underlying biology of alcohol use behaviors. Initial efforts focused on collecting clinically defined cases of AUD, but these ascertainment strategies could not amass the large sample sizes required for GWAS.
Accordingly, multi-ethnic and clinically defined samples have been combined through the Psychiatric Genomic Consortium of Substance Use Disorders working group. The efforts of the PGC-SUD have led to a trans-ancestral meta-analysis consisting of almost 15,000 AD cases and almost 38,000 controls from 28 independent cohorts , identifying a single locus , which was robustly associated with AD. More recently, using information from electronic health records to infer AUD status, a GWAS of 274,424 multi-ethnic individuals from the Million Veterans Program cohort identified 10 loci associated with AUD. Kranzler et al showed that alcohol consumption and AUD were genetically correlated but distinct, thus allowing them to adjust for consumption in the AUD GWAS and for AUD in the GWAS of consumption. In parallel with these efforts, which have focused on clinical diagnoses, other GWAS have incorporated continuous measures of alcohol use. These include self reported weekly alcohol intake or the scores from screening questionnaires such as the AUDIT. The AUDIT can be decomposed to provide a measure of alcohol use from the first 3 questions and misuse from questions 4-10. These quantitative measures are available in large population-based cohorts such as the UK Biobank , MVP and 23andMe. The GWAS meta-analysis of AUDIT identified 10 associated risk loci. Large consortia were also formed to collate quantitative measures of alcohol use, including AlcGen and the GWAS & Sequencing Consortium of Alcohol and Nicotine Use. GSCAN have recently identified nearly 100 loci associated with alcohol consumption. The MVP study also examined alcohol consumption,greenhouse grow tables allowing for an explicit comparison between AUD and consumption in a single population; of the 18 loci detected in that study, 5 were common to both AUD diagnosis and alcohol consumption.
As the prior two paragraphs make clear, population based cohorts have provided larger sample sizes, which are critical for obtaining adequate power for GWAS. Their use can come at the cost of missing more severe alcohol use phenotypes. For example, the frequency of AUD in the UKB is lower than the population average [7% ], indicating that certain population studies may be under powered to detect genetic effects specific to dependence. The frequency of AUD in the MVP, on the contrary, was much higher [20%, ]. Despite these limitations, population based cohorts provide a cost-effective strategy for obtaining very large samples, compared to traditional study designs that require obtaining a diagnosis from clinically trained staff. The 4q23 region, which contains the genes for several alcohol metabolizing enzymes, has been associated with multiple alcohol use behaviors. This association is one of the most consistently replicated findings in the field of psychiatric genetics, although the effects are clearly ancestry-specific. There appear to be multiple signals in this region, including ADH1C , ADH4 , ADH5 and the METAP1/EIF4E region. The GSCAN consortium recently showed that there are at least 13 independent signals with minor allele frequencies over 0.001 at 4q23. Intriguingly, several of those loci are also strongly implicated in non-psychiatric, somatic traits. Beyond the alcohol metabolizing genes, the region containing the genes beta-klotho and the Fibroblast growth factor 21 has been robustly associated with alcohol consumption. The AlcGen consortium was the first to show that the A allele of rs11940694 , located in the intron of KLB, was associated with reduced alcohol consumption. This finding has since been replicated – the same SNP was associated with alcohol consumption and alcohol misuse. Beta-klotho is a transmembrane protein that acts as a cofactor for the circulating hormone fibroblast growth factor 21 by facilitating its binding to FGF receptors. Interestingly the FGF21 gene, which is located on chromosome 19, was also associated with AUDIT scores at the gene-level in humans. Beta-klotho is primarily expressed in the liver, adipose tissue and pancreas , and recent studies have shown that it regulates brain specific functions related to alcohol consumption in mice. For example, mice lacking brain expressed Klb showed increased ethanol preference. Furthermore, FGF21 was found to suppress ethanol consumption in wild-type mice but had no effect on mice lacking Klb in the brain. Previous studies have shown that FGF21 and KLB are involved in sweet and alcohol preference in mice , and a recent study in humans found increased FGF21 expression in blood after binge drinking. These findings suggest that KLB and FGF21 act as part of a brain-liver endocrine axis that regulates alcohol consumption. Future studies could explore the effects of analogues of FGF21 on alcohol consumption, which are currently being tested in clinical trials for the treatment of type 2 diabetes and obesity. Although KLB and FGF21 seem to be promising avenues for translational research, it is worth noting that while SNPs in KLB are associated with alcohol consumption, they have not yet shown any association with AUD. This implies that this system might only be relevant for the regulation of normative consumption, although studies of larger AUD populations may yet reveal a role for these loci in AUD.
Furthermore, although the locus probably impacts KLB, rs11940694 was found to be an expression quantitative trait locus for RFC1 gene expression in the cerebellum and hemisphere. Another well-replicated locus associated with both alcohol consumption and AUD is the region containing the glucokinase receptor gene, whose product is a regulatory protein that is produced by hepatocytes and is involved in the cellular trafficking of glucokinase. A non-synonymous SNP in GCKR, rs1260326, was robustly associated with alcohol consumption in the MVP, UKB and 23andMe samples. Intriguingly, rs1260326 has also been previously associated with multiple metabolic traits, including diabetes, obesity and liver disease. Given that alcohol consumption is strongly associated with both metabolic and lipid profiles , it is not clear whether the association with rs1260326 pinpoints a pleiotropic process central to metabolic traits, or whether alcohol causally impacts glucose metabolism and lipid levels, in part via GCKR. A recent study characterized the effects of alcohol in neural cell cultures derived from induced pluripotent stem cells and found that genes down-regulated upon alcohol exposure were involved in cholesterol homeostasis in the brain. These findings could suggest that AUD has both psychiatric and metabolic components, a theme that has also been suggested for other psychiatric disorders, such as anorexia nervosa. Additional evidence supporting this provocative hypothesis is the fact that several genes associated with alcohol use and dependence involve brain-endocrine-metabolic mechanisms. KLB is part of a brain-liver feedback loop, acetaldehyde modulates a number of ethanol effects in the brain, and enrichment analyses of alcohol-associated genes found glutamatergic enrichment not only in the brain but also in glucose and carbohydrate processing pathways. The ability to process caloric alcoholic beverages may be linked to individual differences in alcohol consumption. In general, the ‘candidate genes’ for AUD that were examined in smaller cohorts have not been replicated by larger and better powered GWAS. One exception is the corticotropin releasing hormone receptor 1 , a candidate gene extensively studied in humans and rodents before the advent of large-scale GWAS studies. CRHR1 is central to the cortisol stress response as part of the hypothalamic-pituitary-axis. Extensive preclinical literature has shown that CRHR1 is associated with relapse to drug taking in mice [e.g. ] and there is some evidence that variation in CRHR1 modulates the role of psychological stress on alcohol intake. Encouragingly, the genomic region surrounding CRHR1 has been associated with alcohol consumption and misuse in several recent GWAS studies. However, CRHR1 is located in an inversion polymorphism of roughly 900kb that is common in Europeans and induces extensive LD spanning many genes , including CRHR1 and MAPT. MAPT encodes the protein tau, is involved in Parkinson’s and Alzheimer’s disease. Further work is therefore required to determine which variant are causal, as the inversion in this region complicates the ability of GWAS to fully address this question. Recent GWAS have identified several regions containing a set of genes that have pleiotropic effects on many psychiatric disorders and related traits; these genes may be tagging a latent factor. For example, the largest GWAS of alcohol and smoking, which used over 1 million individuals, performed a multivariate GWAS approach to show that 150 loci were associated with multiple substance use phenotypes; variation at PDE4B and CUL3 were associated with both smoking and drinks per week. Similarly, CADM2 has been recently associated with alcohol and cannabis use. CADM2 is a cell adhesion molecule that influences brain wiring and appears to have a role in multiple neuropsychiatric disorders. There is now mounting evidence from independent GWAS showing an association between common genetic variants at CADM2 and risky or impulsive behaviors including risk tolerance, automobile speeding propensity, number of sexual partners , sensation seeking and drug experimentation , cannabis initiation , and obesity and body mass index. CADM2 has also been associated with cognitive phenotypes, including educational attainment.