OUP user menu

A GABRA2 Variant Is Associated with Increased Stimulation and ‘High’ Following Alcohol Administration

Albert J. Arias, Jonathan Covault, Richard Feinn, Timothy Pond, Bao-Zhu Yang, Wenjing Ge, Cheryl Oncken, Henry R. Kranzler
DOI: http://dx.doi.org/10.1093/alcalc/agt163 1-9 First published online: 27 October 2013


Aims: Variation in genes encoding GABAA receptor subunits has been implicated in the risk of alcohol dependence (AD). We sought to replicate and extend previous findings of a moderating effect of single nucleotide polymorphisms (SNPs) in GABRA2 (which encodes the GABAA α-2 subunit) on the subjective effects of alcohol by examining SNPs in this and the adjacent GABRG1 gene on chromosome 4. Methods: Fifty-two European-Americans (22 males, 28 light drinkers and 24 heavy drinkers) completed 3 laboratory sessions, during which they drank low-dose, high-dose, or placebo alcohol prior to undergoing periodic assessments of stimulation, sedation and drug enjoyment. We genotyped subjects for three SNPs previously associated with AD: rs279858 in GABRA2, and rs7654165 and rs6447493 in GABRG1. Results: Two SNPs were associated with altered stimulatory effects of alcohol as measured on the Biphasic Alcohol Effects Scale, (rs279858: P = 0.0046; rs6447493: P = 0.0023); both effects were in the opposite direction of previous findings. Carriers of the rs279858 C allele experienced greater stimulation from alcohol. Further inspection of the rs6447493 interaction did not support a pharmacogenetic effect. The effects of rs279858 (but not the other two SNPs) on items from a secondary outcome measure, the Drug Effects Questionnaire (DEQ), were significant. Higher ratings by individuals with the C allele were observed on the DEQ items ‘feel the alcohol effect’ (P < 0.001), ‘like the alcohol effect’ (P < 0.001) and feel ‘high’ (P < 0.001). Conclusion: We did not find that the GABRG1 SNPs rs7654165 and rs6447493 moderated the effects of alcohol. Greater stimulatory and euphoric effects of alcohol in carriers of the rs279858 C allele may, in part, explain the previously reported association of this allele with AD.


Gamma-amino butyric acid (GABA) is the major inhibitory neurotransmitter in the human brain. GABAA receptors are pentameric, ligand-gated chloride channels that are expressed throughout the brain (Young and Chu, 1990). Ethanol acts on GABAA receptors to potentiate chloride ion flux by increasing the frequency and duration of channel opening (Young and Chu, 1990; Davies, 2003; Lobo and Harris, 2008). The modulation of GABA-ergic neurotransmission, and in particular GABAA receptors, by ethanol is thought to underlie many of alcohol's pharmacologic effects in the brain, and is a major contributor to the pathophysiology of alcoholism (Krystal et al., 2006). Thus, genes encoding GABA-related proteins are functional candidates that could influence the risk of alcohol dependence (AD).

Two genome-wide scans provided evidence of linkage of AD to chromosome 4p, in a region that includes a cluster of 4 GABAA receptor subunits (Long et al., 1998; Reich et al., 1998). Fine mapping of this region showed that single nucleotide polymorphisms (SNPs) throughout the 3′ region of the gene encoding the GABAA alpha-2 subunit (GABRA2) and in the intergenic region between GABRA2 and GABRG1 (which encodes the GABAA gamma-1 subunit) were associated with AD in a large family-based study (Edenberg et al., 2004), while other members of the GABAA subunit gene cluster were not. Specifically, 24 of 39 SNPs in GABRA2, 6 of 8 in the intergenic region and 1 of 5 in GABRG1 were associated with AD, and haplotypic analysis showed that all 3-SNP haplotypes were significantly associated with the AD phenotype (Edenberg et al., 2004). Covault et al. (2004) replicated this haplotypic association with GABRA2 in an independent sample. The most widely examined SNP in GABRA2, rs279858, was the only AD candidate marker that was at least modestly associated with AD in a genome-wide SNP analysis of AD (Olfson and Bierut, 2012). Taken together, these findings highlight a potential contribution of variation at the GABRA2 locus to the risk of AD.

The functional significance of this association is underscored by animal studies, which have identified the GABAA alpha-2 subunit as the primary alpha subunit in limbic regions (Fritschy and Mohler, 1995; McKernan and Whiting, 1996). Further, the alpha-2 subunit is key among the predominant alpha subunits mediating the anxiolytic (Low et al., 2000) and myorelaxant (Crestani et al., 2001) effects of benzodiazepines and the hypnotic, but not sedative, effects of combined exposure to alcohol and benzodiazepines (Tauber et al., 2003).

Molecular studies of the functional significance of these polymorphisms are lacking, but there is evidence supporting an association of these genetic variants with the acute subjective and physiologic responses to alcohol under laboratory conditions. In a study combining an alcohol challenge and administration of the medication finasteride or a placebo, the GABRA2 SNP rs279858 genotype moderated the subjective response to alcohol and interacted with the effect of finasteride in 27 healthy social drinkers (Pierucci-Lagha et al., 2005). This within-subject comparison did not include a placebo alcohol control. Subjects homozygous for the more common A allele showed significantly greater activation on central stimulant scores of the Alcohol Sensation Scale (SS) (Maisto et al., 1980) and greater stimulation on the Biphasic Alcohol Effects Scale (BAES) (Martin et al., 1993). Finasteride attenuated these effects to a greater extent in A-allele homozygotes than carriers of the G allele that had previously been associated with AD. (The GABRA2 gene has a reverse sense to the chromosome 4 reference forward strand. Some reports refer to the RNA base variation (A vs. G) while others refer to the chromosome 4 forward strand DNA sequence variation base (T vs. C) for rs279858. The RNA sense minor G allele is equivalent to the genomic forward strand minor C allele reported in this study, and both represent the alcohol dependence risk allele at rs279858.) Finasteride also attenuated effects on the anesthetic subscale of the SS more in A-allele homozygotes than G-allele carriers. Subjects homozygous for the A allele also had higher scores on the gastrointestinal subscale of the SS, and reported greater activation on the BAES stimulant subscale. Overall, G-allele carriers showed less sensitivity to the effects of alcohol, consistent with findings from population genetics studies showing this allele to be associated with AD.

Roh et al. (2011) studied the effects of SNPs in GABRA2 and ALDH2 on the subjective response to alcohol in 110 Japanese subjects, consisting mostly of social drinkers. For 3 GABRA2 SNPs (rs279858, rs279869 and rs279837), the major alleles were associated with a greater subjective response to alcohol. The effects of rs279858 and rs279869, but not rs279837, were moderated by ALDH2 genotype.

Uhart et al. (2012) examined the effect of GABRA2 variation on responses to an oral alcohol challenge in European-Americans. They found no association on their primary outcome measure (sedation on the BAES), but found that minor alleles of individual SNPs (including rs279858) associated with decreased effects of alcohol on secondary subjective outcomes. Further, a haplotype analysis showed significant moderating effects of GABRA2 on both primary and secondary outcome measures.

Kareken et al. (2012) used a functional imaging paradigm to study the effects of the GABRA2 SNP rs279871. In their study, 36 subjects underwent alcohol cue exposure during alcohol intoxication or placebo alcohol conditions. Subjects homozygous for the A allele that was previously associated with AD reported significantly less ‘high’ and ‘intoxication’ with alcohol on the Subjective High Assessment Scale. Subjects homozygous for the A allele showed greater activation with an alcohol cue in the frontal and medial frontal cortical areas of the brain than did heterozygotes, while heterozygotes showed greater activation in the ventral tegmentum. The findings were interpreted as showing that GABRA2 variation moderates inter-individual differences in alcohol-related reward processing.

Although the available evidence supports GABRA2 variation as contributing to the risk of AD, findings from alcohol challenge laboratory studies are inconsistent, and the contribution of GABRA2 variants to the pathophysiology of the disorder remains unknown. The findings from some, but not all, alcohol administration studies show that the GABRA2 variant rs279858 is associated with altered stimulation from alcohol (possibly influencing the level of response to alcohol-induced conditioning) and alcohol-induced positive mood, i.e. the enjoyment of alcohol. In this study, we examined the moderating effects of rs279858 in GABRA2 on the subjective response to alcohol compared with a placebo drink. We hypothesized that carriers of the GABRA2 rs279858 allele previously associated with AD would show reduced subjective CNS stimulatory and sedative effects following alcohol administration. We also examined the effects of two SNPs in the adjacent GABRG1 gene that have been associated with AD (Covault et al., 2008, Enoch et al., 2009). Finally, we explored a secondary measure of subjective response, i.e. the hedonic value of alcohol, hypothesizing that subjects with AD-associated alleles would report greater hedonic effects of alcohol.


Study Design

The study utilized a double-blind, within-subject design. Each subject received three different doses of alcohol (placebo, low or high), one of each during three experimental sessions that took place, on average, 4 weeks apart. The high dose of alcohol was calculated by weight and sex to target a breath alcohol concentration (BrAC) of 100 mg% (0.10%), while the low dose targeted 40 mg% (0.04%). The volume of alcohol consumed at each session depended upon each subject's body weight and sex (for the low- and high-dose sessions, respectively, men received 0.5 and 1.0 g/kg, and women received 0.4 and 0.8 g/kg). The order of beverages for each individual was determined using a balanced Latin Square design to avoid an effect of session order.


Placebo drinks had a small volume of 1% ethanol floated on top to provide the odor of alcohol, and all drinks were served wrapped with alcohol-soaked gauze around the cup to disguise the placebo drinks. Drinks were mixed immediately before serving, so that subjects could taste and smell the alcohol floated on the placebo drink before it was diluted by the juice mixture. To control for expectancy effects, subjects were told that they might or might not be given alcohol (Martin and Sayette, 1993).

Among women, experimental sessions were held during the follicular phase of their menstrual cycle (i.e. 5–9 days following the onset of menstruation). To ensure comparability across sex, the laboratory sessions were also separated by 1 month for men.

The subjects fasted from solid foods beginning at midnight (for more consistent alcohol absorption), and were told to abstain from alcohol, nicotine and other drugs (caffeine not included) during the 24 h prior to each laboratory session. The subjects arrived at ∼9:00 AM on the day of the lab session. At 10:00 AM, they ate a light breakfast and at 10:30 they completed the pre-alcohol subjective ratings and baseline physiologic measures were obtained. The first standard drink of alcohol or placebo beverage was administered at ∼11:00 AM. Subjects had 36 min to consume the three beverages (i.e. one drink was administered every 12 min).


The subjects were recruited by advertisements in the Greater Hartford, CT region, including at nearby colleges and universities. The Institutional Review Board of the University of Connecticut Health Center approved the study protocol and informed consent form (study #06-162S-1). All subjects gave informed consent to participate and were paid to participate. Because of our primary interest in rs279858, we oversampled subjects homozygous at this locus by enrolling approximately half of the heterozygous subjects and 100% of homozygotes.

A total of 161 subjects were screened, of whom 53 were excluded based on GABRA2 genotype. An additional 42 subjects were screen failures or withdrew their participation. In total, 66 subjects completed at least 1 laboratory session and 58 subjects completed all 3 laboratory sessions.

We limited the analysis described here to the 52 European-American completers (2 Hispanics, 30 women). Representation of other population groups (1 Asian, 5 African Americans) in the sample was insufficient to analyze the genetic associations with alcohol response, as the markers examined are tag SNPs for unknown functional variants associated with AD.

Inclusion/exclusion criteria

Subjects were healthy volunteers, 21–45 years old, with a body mass index between 18.5 and 32.5, and body weight <225 lb. All subjects were non-treatment-seeking community drinkers. They were required to have had at least one occasion in the prior month on which they had consumed at least three drinks on a single day, and to have consumed an average of at least one to three drinks, one to three times per week. Heavy drinkers were not excluded, but individuals meeting DSM-IV criteria (American Psychiatric Association. Task Force on DSM-IV., 2000) for lifetime alcohol or drug dependence (except for a past history of nicotine dependence), current alcohol or drug abuse, or another major psychiatric disorder were excluded, as were those who showed evidence of liver dysfunction, were unable to read or write English at an 8th grade or higher level, were using psychotropic medications, or were pregnant or nursing. Women were included only if they reported having regular menstrual cycles and had no history of endocrine or reproductive abnormalities.


Using primers and TaqMan probes described previously (Covault et al., 2008), subjects were genotyped at one tag SNP for each of the two haplotype blocks in the GABRA2/GABRG1 gene region, which we previously found to be associated with AD (Covault et al., 2004, 2008). Rs279858 is a synonymous SNP in exon 5 of GABRA2; the C allele was associated with AD in European-American populations. Rs7654165 is 1 kb 5′ of the GABRG1 transcription start site; the T allele was associated with AD in European-American populations (Covault et al., 2008; Enoch et al., 2009). We also genotyped subjects for the rs6447493 (C/T) SNP, which is in the 3′ UTR region of GABRG1, the minor C allele at this locus was associated with AD (Enoch et al., 2009).

Physiological and subjective effects

Measurements of BrAC, heart rate, blood pressure and ratings of subjective alcohol effects began ∼40 min after the first drink was administered and were repeated (at ∼40-min intervals) throughout the lab session, Fig. 1. Subjective effects were measured using the self-report questionnaires listed below, which were administered on a computer using SPSS data Builder version 3.0.

Fig. 1.

Timeline of events in the laboratory sessions. VS, vital signs (heart rate and blood pressure); BrAC, breath alcohol concentration; SRM, subjective response measures.

The co-primary outcome variables (stimulation and sedation scores) were measured using the Biphasic Alcohol Effect Scale (BAES; Martin et al. 1993), a 14-item unipolar adjective rating scale. Before alcohol ingestion, subjects were asked to rate BAES items on a scale of 0 (not at all) to 10 (extremely) to reflect the extent to which they experienced ‘alcohol-like’ feelings. After alcohol ingestion, subjects were asked to rate the extent to which drinking alcohol produced these feelings.

The Drug Effects Questionnaire (DEQ), (Johanson and Uhlenhuth, 1980; Holdstock and de Wit, 1999) was used as a secondary measure. It consists of four items that measure current drug effects. These items were adapted to measure alcohol effects: ‘feel alcohol,’ ‘feel high,’ ‘like alcohol’ and ‘want more alcohol’.

Data analysis

Because we oversampled homozygotes, we did not test for Hardy–Weinberg equilibrium. Linear mixed-effects models were used to detect the effect of genotype on BrAC and subjective responses. Six time points were included in the analysis relative to the first alcohol drink: 40, 70, 110, 150, 210 and 270 min.

We examined the main effects of time, dose and genotype, and time by dose, time by genotype, genotype by dose and time by dose by genotype interactions. The order of the lab session (1–3) was included as a variable in the model also. The main interaction term of interest for testing the hypotheses is the two-way interaction of dose by genotype, which represents the pharmacogenetic interaction with alcohol.

The overall alpha level of significance was set at 0.05 and a method of correction for multiple comparisons for correlated genotypes described by Nyholt (2004) was employed to adjust for the examination of the three variants, which was reduced to 2.4 based on this method. The significance level was set at P < 0.021 (based on the Sidak correction with 2.4 independent tests and alpha of 0.05) for the co-primary outcome measures: the stimulation and sedation subscales of the BAES. For the secondary measure, the DEQ, we did not include the last item (which asks whether subjects want more alcohol) because we wanted to focus on alcohol's pleasurable effects, which are best represented by the first three items. Thus, for the DEQ, we applied the Sidak correction to set the significance level at P < 0.0053, accounting for 2.4 independent genotypes tested on the three DEQ items, in addition to the 2.4 comparisons on the co-primary aims (a total of 9.6 independent tests).



The subjects' demographics are shown in Table 1. The average study participant was young (mean age = 26.1 years, SD = 6.4), employed (73.1%), single (75.5%) and female (57.7%). Roughly half (28, 53.8%) of subjects had no heavy drinking days in the 30 days prior to participation in the study.

View this table:
Table 1.


n = 52
Education years16.12.2
Recent Drinking
 Mean HDD in the 30 days prior to study1.42.1
 Mean HDD in the 90 days prior to study4.66.5
Subjects without any HDD 30 days prior2853.8
Subjects without any HDD 90 days prior1834.6
Marital Status
 Administrative, clerical2344.2
 Skilled/Semi-skilled worker1019.3
Current Employment
 <$60 0002751.9
 >$60 0002548.1
  • The descriptive statistics on the 52 European-American subjects. The average study participant was young (mean age 26.1 years, SD 6.4), employed (73.1%), single (75.5%) and female (57.7%). HDD, heavy drinking days.

Subject genotypes and allele frequencies

Table 2 shows the number of subjects with each genotype tested, and the allele frequencies. Genotypes at rs279858 and rs7654165 were moderately correlated (Pearson r = 0.6), consistent with the correlation of markers between the two adjacent haplotype blocks [r = 0.51 to 0.59 for a sample of 535 European-Americans (Covault et al., 2008)].

View this table:
Table 2.

Genotypes, allele frequencies

Single nucleotide polymorphismGenotype countAllele frequencies
  • The genotype counts and allele frequencies for the three SNPs studied in the sample of n = 52 European-American subjects. Because we selected subjects based on genotype, we did not consider Hardy–Weinberg equilibrium statistics.

Breath alcohol concentration

The mean BrAC values achieved by subjects are illustrated in Fig. 2.

Fig. 2.

Mean breath alcohol concentration (BrAC) over time for women and men. The targeted BrAC was achieved, but with men averaging slightly higher BrACs than women.


As expected, there was a graded response across all measures to placebo, low and high doses of alcohol. Although sex and light vs. heavy drinking were significant predictors in the models, their inclusion did not influence the significance level of the interaction of dose by genotype, which was the primary analysis for hypothesis testing.

Co-primary outcome BAES (stimulation)

GABRA2 SNP rs279858

Significant main effects were observed for time (F(1,404) = 236.40, P < 0.0001), alcohol dose (F(2,213) = 54.41, P < 0.0001) and genotype (F(2,145) = 7.54, P = 0.0008), time by dose (F(2,470) = 45.41, P < 0.0001), time by genotype (F(2,404) = 6.78, P = 0.0013) and dose by genotype (F(4,251) = 3.86, P = 0.0046). There was also a nominally significant interaction of time by alcohol dose by genotype (F(4,528) = 2.40, P = 0.049). Figure 3 shows BAES stimulation over time: subjects homozygous for the C allele that has been associated with AD showed a similar response following the low dose of alcohol as for placebo, but a greater stimulatory response than the T-allele homozygotes following the high dose of alcohol. When the analysis was run using the two-level C-allele carrier genotype, the dose by genotype interaction was not significant. However, to better understand the significant interaction effect from the three-level genotype model, we applied a simpler model with two-level genotype (C-allele carriers vs. TT genotype subjects) and two-level dose (high dose, placebo), main and two-way interactions, which showed a significant effect for C-allele carriers for the dose by genotype interaction (F(1,564) = 7.09, P = 0.008) on the BAES stimulation subscale. When we tested the interaction effect of dose by sex by genotype on BAES stimulation, it was not significant (P = 0.29). Additionally, when we examined the dose by genotype interaction separately in males and females, the result was similar across sexes with the dose by genotype interaction P = 0.011 for males and P = 0.022 for females. Inclusion of the interaction of heavy drinking status by genotype by dose on BAES stimulation was not significant (P = 0.73).

Fig. 3.

Influence of the GABRA2 SNP rs279858 on Biphasic Alcohol Effects Scale (BAES) stimulation. A significant interaction was seen for alcohol dose by genotype (F(4,251) = 3.86, P = 0.0046). Subjects heterozygous and homozygous for the C allele showed a higher stimulatory response than the T-allele homozygotes following the high dose of alcohol. The results were also significant for a simplified model with the two-level genotype of ‘C-allele carrier.’

GABRG1 upstream SNP rs7654165

There were significant main effects of time (F(1,411) = 236.80, P < 0.0001), alcohol dose (F(2,203) = 55.62, P < 0.0001) and genotype (F(2,144) = 8.35, P = 0.0004), but the interaction of dose by genotype was not significant (F(4,240) = 1.33, P = 0.26). The results were similar and also non-significant for the model with the two-level T-allele carrier genotype.

GABRG1 3′ UTR SNP rs6447493

Significant main effects were seen for BAES stimulation over time (F(1,402) = 234.80, P < 0.0001) and alcohol dose (F(2,214) = 52.79, P < 0.0001), but the main effect of genotype was not significant (F(2,141) = 2.20, P = 0.11). The interactions of time by dose (F(2,469) = 45.14, P < 0.0001) and genotype by dose (F(4,253) = 4.29, P = 0.0023) were significant. A nominally significant time by dose by genotype interaction was observed (F(4,527) = 2.64, P = 0.033). Though the dose by genotype interaction suggested a pharmacogenetic effect, upon visual inspection of the graphs of BAES response over time, the TT subjects had a strong stimulatory response to placebo, and heterozygous subjects had a strong stimulatory response to low-dose alcohol, while there was virtually no observable difference in stimulatory response between genotypes at the high dose (not shown).

Co-primary outcome BAES (sedation)

BAES sedation scores did not show an ordinal decrease over time, similar to BAES stimulation scores, so we could not treat time as continuous. Thus, we analyzed this outcome by averaging time over the six time points and including it as a main effect in the mixed models analysis, examining the two-way interaction of genotype by dose. No significant genotype by dose interaction was observed.

Secondary outcome: DEQ

We further evaluated the effect of the GABRA2 SNP rs279858 (as a two-level C variable; C allele carriers and TT subjects) on the response to alcohol by testing the interactive effects on DEQ measures in a mixed model comparing the high-dose alcohol to placebo (two-level dose) with main and two-way interactions. The item ‘feel the alcohol effect’ showed significant main effects of time (F(1,563) = 176.93, P < 0.001), dose (F(1,320) = 332.08, P < 0.0001), but not main effect of genotype (F(1,50) = 3.44, P = 0.069), with a significant dose by genotype interaction (F(1,563) = 16.085, P < 0.001). Figure 4 shows the alcohol-induced ‘feel the alcohol effect’ over time; C-allele carriers reported greater ‘feel the alcohol effect’ following alcohol ingestion in the high-dose condition.

Fig. 4.

Influence of the GABRA2 SNP rs279858 on the Drug Effects Questionnaire (DEQ) item ‘feel the alcohol effect’. The interaction of dose by genotype was significant for the two-level genotype of C allele carriers (F(1,563) = 16.085, P < 0.001). Subjects having at least one of the C alleles reported greater ‘feel the alcohol effect’ following high-dose alcohol ingestion.

A similar pattern of response was seen for the other two DEQ items: ‘like the alcohol effect’ and feel ‘high’ from the alcohol. There was a significant dose by genotype response for rs279858 on ‘like the alcohol effect’ (F(1,563) = 21.25, P < 0.001). Figure 5 shows the response to ‘like the alcohol effect’ over time; C-allele carriers reported greater ‘like the alcohol effect’ following alcohol ingestion. There was a significant dose by genotype response for rs279858 on alcohol-induced ‘high’ (F(1,563) = 13.08, P < 0.001). Figure 6 shows the response to alcohol-induced ‘high’ over time; subjects having at least one of the C allele reported a greater ‘high’ following high-dose alcohol ingestion.

Fig. 5.

Influence of the GABRA2 SNP rs279858 on the Drug Effects Questionnaire (DEQ) item ‘like the alcohol effect’. A significant dose by genotype interaction (F(1,563) = 21.25, P < 0.001) was found for the two-level genotype variable of C allele carriers. Subjects having at least 1 of the C alleles reported greater ‘like the alcohol effect’ following high-dose alcohol ingestion.

Fig. 6.

Influence of the GABRA2 SNP rs279858 on the Drug Effects Questionnaire (DEQ) item ‘feel high’. The dose by genotype interaction was significant for the two-level genotype C allele carriers (F(1,563) = 13.08, P < 0.001). Subjects having at least 1 of the C alleles reported greater ‘feel high’ following high-dose alcohol ingestion.

There was no significant dose by genotype interactions for either of the GABRG1 SNPs. There was a trend for an effect of rs7654165 on the DEQ item ‘like the alcohol effect’ (F(4,312) = 3.38, P = 0.01), but it was not significant after correcting for multiple comparisons.


We found significant moderating effects of rs279858 on the stimulation scale of the BAES, with carriers of the AD-associated C allele experiencing more stimulation from alcohol than TT subjects. Although there was a statistically significant finding for rs6447493 on the stimulation scale of the BAES, further inspection of the interaction did not support a pharmacogenetic effect. There was no significant pharmacogenetic interaction for the other GABRG1 SNP rs7654165 on primary or secondary measures of alcohol response. We did not replicate the previous findings of less stimulation on the BAES among carriers of the rs279858C allele (Pierucci-Lagha et al., 2005). In the present study, the finding was in the opposite direction, showing less stimulation in the protective T-allele homozygotes than in individuals with one or two copies of the C allele. A significant moderating effect of the GABRA2 SNP rs279858 was also seen on a secondary measure of subjective alcohol effects, the DEQ. Specifically, carriers of the C allele that has been previously associated with AD reported feeling more alcohol effect, liking alcohol more and feeling ‘higher’ from alcohol. We did not find convincing evidence of a pharmacogenetic effect of rs6447493 on primary or secondary measures of alcohol response.

The significant effect on BAES stimulation for rs279858 was in the direction opposite to that of previous findings by our group and others, such that individuals with the C allele reported greater alcohol-induced stimulation. This raises the possibility that other genetic variants in GABRA2 that are in linkage disequilibrium with rs279858 are responsible for the moderating effects on the response to alcohol. Alternatively, different GABRA2 SNPs could be driving different components of the alcohol response. The direction of the observed trend in this study for rs279858 on BAES stimulation is opposite that expected under the low level of response to alcohol model (Schuckit, 1994). Chromosome 4 regions were not identified in genome-wide linkage studies of the low level of alcohol response phenotype (Schuckit et al., 2001).

The finding that carriers of the C allele reported a greater alcohol-induced ‘high’ (i.e. rewarding effects) than TT subjects suggests that these subjects may be at greater risk of alcoholism because they find alcohol more rewarding. The heterozygotes at this locus had the highest peak euphoria scores. This unexpected finding is dissimilar to those of some previous studies, but an examination of the contrasting and convergent findings among all studies reveals a somewhat consistent pattern of association of the rs279858 C allele to the positively reinforcing effects of alcohol.

It is unclear why the present genotype findings do not replicate those from the earlier study by Pierucci-Lagha et al. (2005), but it may relate to different alcohol doses examined in the two studies. Furthermore, the present study has several methodological advantages over the former. Most importantly, in contrast to our (and other) prior studies of this SNP, the present study included a placebo control for alcohol, as well as a low and a high dose of alcohol. Second, oversampling of homozygotes may have resulted in a more accurate assessment of the pharmacogenetic effects of rs279858. Third, we examined subjective effects over a longer time period than in our prior study.

The results of this study also contrast with those of Roh et al. (2011), who found a greater response to alcohol on the Alcohol Sensation Scale among individuals with the T allele of rs279858. However, the study by Roh et al. did not find an association between rs279858 and the response to alcohol on the BAES. Population-specific effects of rs279858 could also explain the differences between the Roh study, conducted in Japanese subjects, and the present one, which was conducted in European-Americans.

Our findings for the GABRA2 SNP rs279858 relative to liking alcohol effects are in part consistent with those of Uhart et al. (2012), who found a significant interaction of alcohol and rs279858 on a secondary outcome measure (a visual analog scale measure of positive and negative alcohol effects), but not on the BAES. They found an association between the C allele at rs279858 and a less negative experience of alcohol, while we found greater stimulating and rewarding effects of the same allele.

Our results are also consistent with those of Haughey et al. (2008), who found an association between the AD-associated rs279858 C allele and greater alcohol enjoyment. In their study, individuals with the C allele reported experiencing more vigor, more positive mood, more stimulation and greater ‘liking’ of alcohol.

The GABRA2 rs279858 genotype effect reported here and in the published literature on subjective alcohol effects is somewhat consistent in terms of its effect on the positively reinforcing effects of alcohol. Recent efforts to obtain a better understanding of the phenomenology of the subjective alcohol response may be helpful in integrating the results from different alcohol challenge studies. Ray et al. (2009) performed an exploratory factor analysis on available alcohol challenge studies and found a three-factor solution for the subjective response. Factor 1 represented positive reinforcement, and consisted of a positive correlation with measures of stimulation, vigor and happiness (positive mood). Factor 2 represented punishment (or aversive) effects of alcohol, and consisted of a positive correlation with the sedating and unpleasant effects of alcohol. Factor 3 represented negative reinforcement, and consisted of a positive correlation with decreases (changes) on measures of tension and depression with alcohol, suggesting that tension and low mood were alleviated by alcohol. When we consider all of the GABRA2 studies (including the present one) that have examined the association with the alcohol response in light of the 3-factor construct, the most consistent findings for rs279858 SNP are that the C allele is associated with increased positive reinforcement (factor 1, via increased stimulation, ‘liking’ and ‘high’), and not with factor 2 (i.e. the sedative/unpleasant effects of alcohol). Unfortunately, the DEQ was not included in the exploratory factor analysis by Ray and colleagues (2009), so there are limits to the application of this approach to interpret the existing findings.

We did not measure or examine the effects of individual subjects' alcohol expectancies (i.e. subjects' belief about the effects of alcohol) on their subjective response to alcohol. Variation in subjects' alcohol expectancies in this study could potentially have played some role in the observed discrepancies across studies, and may not have been completely controlled for by the use of a placebo (Sher, 1985). Although we obtained data on alcoholism family history, and personal history of conduct disorder and antisocial personality symptoms, the number of subjects with a family history or who endorsed such symptoms was too low (10 of 52 and 6 of 52 subjects, respectively) to make meaningful evaluation of the potential moderation by these characteristics. The subjective response to alcohol may differ by sex and whether the subject is a heavy or a light drinker, so we explored these potential moderating factors in this analysis (King et al., 2002, 2011; Quinn and Fromme, 2011). Sex and heavy drinking status did not appear to moderate the effect of genotype or alcohol dose in this study, and thus may not explain discrepancies among studies, though the small sample size prohibits a firm conclusion regarding these potential moderators.

Taken together, the present study and the extant literature suggest that the positively reinforcing responses to alcohol may be moderated by genetic variation linked to rs279858. The possibility that GABRA2 variation would affect the hedonic and rewarding aspects of the alcohol response is biologically plausible given findings that the alpha-2 subunit is highly expressed in the striatum and nucleus accumbens—areas that are involved in hedonic and reward processing (Sieghart and Sperk, 2002). Although rs279858 is potentially functional, and also is a ‘tag’ SNP based on previous studies, we cannot rule out that the possibility that other variants in linkage disequilibrium with rs279858 moderate the ‘high’ and positively reinforcing effects of alcohol.

Further research evaluating the moderating effects of GABRA2 variants on the response to alcohol and their implications for AD risk is needed. Neuroimaging studies, alcohol self-administration studies and pharmacogenetic clinical trials would be particularly useful in better understanding the functional effects of these variants. Recent studies associating GABRA2 genotypes with impulsive phenotypes suggest that this relationship should be further explored in the context of the alcohol response (Bauer et al., 2012; Villafuerte et al., 2012). Additionally, the effect of other GABAA receptor subunit gene variants (e.g. those of GABRA1) on the risk of AD and as pharmacogenetic candidates has been largely unexplored, and is another important focus for research on the risk and pharmacogenetics of AD.


This work was supported by NIH grants R01 AA015606 (to J.C.), K24 AA13736 (to H.R.K.), K23 AA017689 (to A.J.A.), P60 AA03510 (Alcohol Research Center) and M01 RR06192 (University of Connecticut General Clinical Research Center).

Conflict of interest statement

H.R.K. has been a consultant or advisory board member for Alkermes, Lilly, Lundbeck, Pfizer and Roche. He is also a member of the American Society of Clinical Psychopharmacology's Alcohol Clinical Trials Initiative, which is supported by Lilly, Lundbeck, Abbott and Pfizer. The other authors have no disclosures to report.


View Abstract