Human genetic studies on alcohol-related phenotypes have used family-based linkage and population-based association analyses to identify quantitative trait loci (QTLs). Linkage studies are based on co-segregation between genetic markers and alcohol dependence in families with several affected members. By contrast, association studies evaluate the strength of association between genetic variants and alcohol phenotypes in samples of unrelated individuals; these can be attributable to a causal effect of the variant or linkage disequilibrium (LD) between the molecular variant and the true causal allele. Association analyses give more precise localization of QTLs than linkage studies; however, false-positive associations can arise from population stratification of cases and controls, and by chance in small samples. In both designs, large numbers of individuals are required to detect QTLs with small effects. Early efforts to dissect the genetic basis of alcohol consumption and addiction in humans were based on candidate genes. The main pathway of ethanol metabolism involves its conversion to acetaldehyde by alcohol dehydrogenase (ADH; Figure 1). Acetaldehyde is oxidized to acetate by aldehyde dehydrogenase (ALDH). The activated form of acetate, acetyl-CoA, can be metabolized into ketone bodies, fatty acids, amino acids and steroids, in addition to oxidation in the Krebs cycle. Cytochrome P450s (for example, those encoded by the gene CYP2E1) and catalase also metabolize a small fraction of ingested ethanol. Multiple ADH and ALDH enzymes are encoded by different genes 7 , and different ADH and ALDH alleles can differ in expression levels and in the rate at which their corresponding enzymes metabolize ethanol or acetaldehyde. The ADH1B His48Arg and ALDH2 Lys487 polymorphisms have long been associated with risk of alcoholism, and directly and predictably lead to alcohol-induced flushing through molecular mechanisms that include accumulation of acetaldehyde and release of histamine 8 . ADH1B, ALDH2 and ADH4 influence alcohol consumption and have been implicated as risk factors for developing alcohol abuse or dependence 9 10 11 .

The positive reinforcing effects of alcohol are mediated through the corticomesolimbic dopaminergic reward pathway, which extends from the ventral tegmental area to the nucleus accumbens and is modulated by a wide range of neurotransmitters. This pathway is indirectly activated by alcohol through the release of other neurotransmitters, including acetylcholine, dopamine, glutamate, gamma-aminobutyric acid (GABA), opioids and serotonin. Several candidate genes in neurotransmitter pathways associated with the ventral tegmental area and nucleus accumbens have been associated with alcohol dependence, including the genes encoding cholinergic receptor, muscarinic 2 (CHRM2) 12 ; cholinergic receptor, nicotinic, alpha 5 (CHRNA5) 13 ; catechol-O-methyltransferase (COMT) 9 ; GABA A receptor, alpha 2 (GABRA2) 14 ; glutamate receptor, metabotropic 8 (GRM8) 15 ; solute carrier family 6 (neurotransmitter transporter, serotonin), member 4 (5-HTT) 16 ; nuclear factor of kappa light polypeptide gene enhancer in B cells 1 (NFKB1) 17 ; monoamine oxidase A (MAOA) 18 ; neuropeptide Y receptor Y2 (NPY2R) 19 ; opioid receptor, kappa 1 (OPRK1) 20 ; opioid receptor, mu 1 (OPRM1) 21 ; prodynorphin (PDYN) 20 ; and tachykinin receptor 3 (TACR3) 22 .

More recently, several genome-wide association studies (GWASs) using 500,000 to 1 million SNPs spanning the entire genome have provided unbiased screens for variants affecting alcohol-related behaviors 23 24 25 26 27 28 29 30 31 32 (Table 1; Additional file 1). Many of these studies have used samples from large consortia, such as the Collaborative Studies of Genetics of Alcoholism (COGA), the Study of Addiction: Genetics and Environment (SAGE) and the Australian Twin Registry. Consistent with GWAS for other traits 33 , many novel loci have been implicated in alcohol dependence and alcohol consumption, but these loci have small effects and are thus difficult to detect with the available sample sizes, especially given the high significance threshold required to control for multiple tests. In one GWAS study, the gene encoding ACN9 homolog (ACN9), which is involved in gluconeogenesis and required for the assimilation of ethanol or acetate into carbohydrate 34 , has been associated with susceptibility to alcohol dependence 35 . Two SNPs in the 3' flanking region of the gene encoding peroxisomal trans-2-enoyl-CoA reductase, which is a key enzyme for the peroxisomal fatty acid chain elongation pathway, achieved genome-wide significance for alcohol dependence in a study of German males 31 . A GWAS on pooled DNA samples from individuals with a lifetime history of alcohol dependence, nicotine dependence and co-morbid alcohol/nicotine dependence in an Australian population 28 identified three SNPs that reached genome-wide significance for co-morbid alcohol/nicotine dependence. The implicated genes were: (1) the gene encoding MAP/microtubule affinity-regulating kinase 1 (MARK1), which is a kinase involved in the phosphorylation of microtubule-associated proteins; (2) the gene encoding DEAD box polypeptide 6 (DDX6), which is a putative RNA helicase; and (3) KIAA1409, which encodes a component of the NALCN Na⁺ channel complex. A family-based association analysis for alcohol dependence that utilized both COGA and the Australian twin-family samples implicated the genes encoding endothelin receptor type B (EDNRB), Usher syndrome 2A (USH2A), TCDD-inducible poly(ADP-ribose) polymerase (TIPARP), monoamine oxidase A, Na⁺/K⁺ transporting ATPase interacting 2 (NKAIN2), and Down syndrome cell adhesion molecule like 1 (DSCAML1) 32 , with four SNPs in DSCAML1 reaching genome-wide significance. A GWAS for alcohol consumption in Korean male drinkers 23 identified 12 SNPs in six genes (chromosome 12 ORF 51 (C12orf51), and the genes encoding coiled-coil domain containing 63 (CCDC63), myosin, light chain 2 (MYL2), 2'-5'-oligoadenylate synthetase 3 (OAS3), cut-like homeobox 2 (CUX2), and rabphilin 3A homolog (RPH3A)) on chromosome 12q24 associated with alcohol consumption at a genome-wide significance level. In contrast, two studies on alcohol dependence 25 26 , including one of the largest GWASs to date, with over 10,000 individuals from the Australian Twin Registry, failed to identify any SNPs with genome-wide significance. A GWAS meta-analysis of approximately 2.5 million SNPs with alcohol consumption among 12 population-based samples of European ancestry, comprising more than 20,000 individuals 30 , identified a single SNP, rs6943555, in the gene for autism susceptibility candidate 2 (AUTS2) associated with alcohol consumption at a genome-wide significance level.

Considering only SNPs in genes that achieve genome-wide significance reveals no overlap across the studies, with the exception of the large effects contributed by variation at ADH1B and ALDH2 in Asian populations. Among all SNPs that were significant at a nominal P-value in the studies described above, the gene encoding cadherin 13 (CDH13) was replicated in four independent studies, and eight genes were common across any three studies (Table 1). In addition, five differentially expressed genes in different areas of postmortem human brains of alcoholics were replicated in any of three transcriptional profiling studies (Table 1) 36 37 38 39 40 41 .

The lack of concordance across GWASs could be partially due to different measures of alcohol consumption used in different study populations or even across different samples from the same population. Common measures of alcohol consumption are frequency of drinking (weekly and annually), quantity by frequency, maximum drinks in a 24 hour period, frequency of heavy drinking and frequency of intoxication 5 ; if these measures are not perfectly correlated, they will be associated with different SNPs. Association studies in humans are limited in resolution by the structure of LD; to the extent that LD varies among populations, different genes may be implicated in different studies. Moreover, rare alleles that contribute to variation in alcohol consumption are essentially blind to detection by association studies using common variants, and many SNPs with small effects may contribute to risk for alcohol dependence.

In summary, GWASs have been limited by difficulties in quantifying alcohol-related phenotypes and in obtaining large sample sizes, together with co-morbidity of alcoholism with other behavioral and neuropsychiatric disorders, gender effects and population admixture. Furthermore, the diversity of mechanisms of vulnerability and resilience to alcohol pose challenges for human genetic studies on alcoholism or alcohol consumption. It has become increasingly clear that, in addition to a few common alleles, many different rare alleles may contribute to vulnerability in different populations.

One strategy that circumvents the limitations of human GWASs relies on comparisons with genes associated with ethanol-related behaviors in genetically amenable model organisms.

Given the evolutionary conservation of genes and pathways affecting key biological processes between vertebrates, invertebrates and humans, studies on model organisms (rats, mice, flies and nematodes) have played an important role in identifying potential candidate genes that contribute to alcohol intoxication. Invertebrate and vertebrate models show similar symptoms of alcohol intoxication, including loss of postural control, sedation, immobility and development of tolerance. After alcohol intoxication, mice and rats increase their alcohol consumption, develop tolerance and even alcohol dependence. Drosophila develops tolerance after a single exposure to ethanol 42 . In addition to rapid tolerance, flies develop chronic tolerance after prolonged exposure to a low concentration of ethanol 43 . Caenorhabditis elegans also exhibits tolerance after continuous ethanol exposure 44 and develops ethanol preference as a result of prolonged pre-exposure 45 .

In addition to the behavioral similarities between invertebrate and mammalian models, invertebrates use similar neurotransmitter systems, neuropeptides, synaptic proteins, channels and signaling processes to mediate ethanol-induced behaviors 46 . These include genes encoding Ca²⁺-sensitive adenylate cyclase and protein kinase A 47 48 49 , BK channels 50 51 52 , Homer 53 54 , genes encoding proteins involved in GABA neurotransmission 55 56 , the gene encoding protein kinase C 57 58 , and genes encoding proteins involved in dopamine and serotonin signaling 45 59 60 . In vertebrates, neuropeptide Y (NPY) signaling plays a role in alcohol intake and dependence 61 62 . Invertebrates have an ortholog to NPY, neuropeptide F (NPF), and signaling via NPF also influences ethanol-related behaviors 44 63 .

Rats meet all the criteria for animal models of alcoholism, including: the ability to orally self-administer ethanol; elevation of blood ethanol concentration after alcohol consumption; willingness to work for ethanol access; development of functional tolerance; and, after a deprivation period, relapse-like behavior 64 . A study on recombinant inbred rat strains identified several genomic regions on chromosomes 1, 6 and 12 that harbor candidate genes for alcohol consumption, including the genes encoding actin filament associated protein, cholecystokinin 2 receptor, melanocortin 4 receptor; protein tyrosine phosphatase receptor type E and tubulin B6 65 . Furthermore, several microarray studies have identified differential expression of genes between alcohol-preferring and alcohol-non-preferring rat strains, including the genes encoding alpha-adducin (Add1), retinal dehydrogenase 1 (Aldh1a1), adenylate cyclase type 3 (Adcy3), alpha-crystallin B chain (Cryab), glutamate decarboxylase 1 (Gad1) and NPY (Npy) 66 67 . However, most studies on the genetic underpinnings of alcohol-related phenotypes have focused on mice, because, in contrast to rats, they can be more easily genetically manipulated. Mice are amenable to complete elimination of the gene of interest, gene silencing by RNA interference (RNAi), overexpression and mutagenic technologies 68 . Numerous genetic models have been developed to investigate specific aspects of alcoholism in mice, including tolerance, withdrawal, motivational effects and high-dose sensitivity 68 . Over 90% of the mouse and human genomes can be partitioned into regions of synteny 69 .

Among invertebrate models Drosophila is advantageous because large numbers of genetically identical individuals can be reared at relatively low cost and without regulatory restrictions, and many community resources are available for sophisticated genetic manipulations. Drosophila is also readily amenable to neuroanatomical studies.

C. elegans presents a useful model system for examining the effects of ethanol on development 70 . The lineage of each of its 302 neurons and their chemical synapses has been characterized. Nematodes have a short (approximately 3 days) reproductive cycle, enabling large-scale mutagenesis screens within a relatively short time, and they can be cryopreserved.

Evolutionary conservation of pathways offers opportunities for comparative cross-species analyses (Table 2). For example, AUTS2 was identified in human GWASs for alcohol consumption and verified by genotype-specific expression in human prefrontal cortex samples. Differences in expression of Auts2 were also observed in whole-brain extracts of mice selected for differences in voluntary alcohol consumption, and downregulation of an Auts2 homolog was causally associated with reduced alcohol sensitivity in Drosophila. Thus, evidence for the involvement of AUTS2 in alcohol drinking or sensitivity is corroborated across three different species 30 .

Because at least 60% of Drosophila genes have conserved human orthologs, the latter can be identified and superimposed on computationally predicted networks from Drosophila. This allows identification of candidate genes for subsequent human association studies based on a previous unbiased genome-wide approach in Drosophila. Not only can this strategy empower human association analysis by reducing the prohibitive multiple testing correction of a GWAS, but it provides also functional contexts to the candidate genes as they form part of defined networks.

To provide proof of principle for the potential of this translational approach, the human ortholog of the Drosophila Men gene, which encodes malic enzyme, was targeted as a candidate gene based on artificial selection, mutational and transcriptional profiling studies 86 93 94 . The gene encoding malic enzyme is also differentially expressed in mice upon acute alcohol treatment 95 . Malic enzyme represents a metabolic switch, converting malate into pyruvate while generating NADPH, an essential co-factor for fatty acid biosynthesis (Figure 2). Thus, the malic enzyme reaction enables the development of alcohol-induced fatty liver syndrome. Association studies on the Framingham Offspring cohort showed that intronic SNPs of the gene encoding malic enzyme 1 (ME1) were associated with amount of cocktail drinking, indicating that variation in expression of cytoplasmic malic enzyme contributes to variation in alcohol consumption. Thus, translational approaches from model organisms to humans can identify SNPs that are associated with drinking behavior, with an effect size that could not have been resolved with large-scale unbiased GWASs 92 .

During the past decade a wealth of information on alcohol consumption has been obtained from human and model organism studies, but rarely have data from different studies been integrated to form a comprehensive blueprint of the genetic networks that contribute to alcohol drinking. In future, studies integrating data on alcohol-related phenotypes from GWASs and transcriptional profiling studies on both humans and model organisms will make it possible to construct biologically meaningful networks of genes that contribute to alcohol consumption and dependence, and generate a deeper understanding of the genetic susceptibility for alcoholism.

ADH: alcohol dehydrogenase; ALDH: aldehyde dehydrogenase; COGA: Collaborative Studies of Genetics of Alcoholism; GABA: gamma-aminobutyric acid; GWAS: genome-wide association study; LD: linkage disequilibrium; Mb: megabase; NPF: neuropeptide F; NPY: neuropeptide Y; QTL: quantitative trait locus; RNAi: RNA interference; SNP: single nucleotide polymorphism.

The authors declare that they have no competing interests.