ADH1B is associated with alcohol dependence and alcohol consumption in populations of European and African ancestry.

A coding variant in alcohol dehydrogenase 1B (ADH1B) (rs1229984) that leads to the replacement of Arg48 with His48 is common in Asian populations and reduces their risk for alcoholism, but because of very low allele frequencies the effects in European or African populations have been difficult to detect. We genotyped and analyzed this variant in three large European and African-American case-control studies in which alcohol dependence was defined by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria, and demonstrated a strong protective effect of the His48 variant (odds ratio (OR) 0.34, 95% confidence interval (CI) 0.24, 0.48) on alcohol dependence, with genome-wide significance (6.6 × 10(-10)). The hypothesized mechanism of action involves an increased aversive reaction to alcohol; in keeping with this hypothesis, the same allele is strongly associated with a lower maximum number of drinks in a 24-hour period (lifetime), with P=3 × 10(-13). We also tested the effects of this allele on the development of alcoholism in adolescents and young adults, and demonstrated a significantly protective effect. This variant has the strongest effect on risk for alcohol dependence compared with any other tested variant in European populations.

INTRODUCTION

Alcohol dehydrogenases catalyze the first step in the oxidative metabolism of ethanol (beverage alcohol). The Arg48His coding SNP (rs1229984) in ADH1B, encoding the β subunit of alcohol dehydrogenase, dramatically affects enzyme activity; enzymes with His48 oxidize ethanol approximately 70- to 80-fold faster than those with Arg48[1]. In East Asian populations, the His48 allele is common and lowers the risk for alcoholism, with an odds ratio of approximately 0.2 for a single allele[1-4]. Although it does not trigger a full flushing reaction, the protective mechanism is hypothesized to involve the aversive effects of transiently elevated acetaldehyde, conceptually similar to the protection accorded by the severe flushing experienced by individuals with the inactive variant of aldehyde dehydrogenase 2 (ALDH2)[1].

Recently, a meta-analysis of published studies on different populations, using varying definitions of alcohol dependence, abuse or alcohol-related diseases, reported a very robust association with rs1229984 in Asian populations[5] (47 studies, p value = 7 × 10−42). However, the results for alcohol dependence in European subjects were modest (19 studies, p value = 0.002) and African American subjects were not studied. This His48 allele has a low frequency in populations of European or African descent.

In contrast to the equivocal findings of rs1229984 with alcohol dependence in subjects of European descent, Arg48His has been strongly implicated in cancers of the upper aerodigestive tract, specifically esophageal cancer, with the protective influence of His48 being most pronounced in drinkers (N=15,841 OR=0.64 (95% CI 0.59, 0.71) p = 1 × 10−20)[6-7]. No evidence of association was seen in the relatively modest African American population with cancer (N=1,079 OR=0.99 (95% CI 0.46, 2.13) p = 0.98).

The purpose of this study was two-fold. We examined the association of rs1229984 in three large case-control studies of subjects of European and African descent in which alcohol dependence was directly diagnosed to test whether this SNP shows protective effects on alcohol dependence and alcohol consumption in non-Asian samples. We then confirmed the protective effect of the His48 allele in an adolescent and young adult population that is in the critical period of risk for the development of alcohol dependence.

SUBJECTS AND METHODS

Samples for this meta-analysis were drawn from three existing studies, the Collaborative Study on the Genetics of Alcoholism (COGA), the Collaborative Genetic Study of Nicotine Dependence (COGEND), and the Family Study of Cocaine Dependence (FSCD), each of which is described below. For purposes of this study, all case subjects met a lifetime diagnosis for DSM-IV alcohol dependence[8]. Control subjects had at least one drink of alcohol, but never met dependence for DSM-IV alcohol dependence. Characteristics of adult subjects from COGA, COGEND and FSCD are presented in Table 1.

Table 1

Characteristics of Adult Subjects

	Study*
	COGAN=1800	COGENDN=2666	FSCDN=1166
Sex, N (%)
Male	971 (53.9)	1,018 (38.2)	579 (49.7)
Female	829 (46.1)	1,648 (61.8)	587 (50.3)
Age, years
Mean ± sd	43.1 ± 11.4	36.5 ± 5.5	37.0 ± 8.8
Range	18.0–79.0	25.0–47.0	18.0–60.0
Race, N (%)
European American	1353 (75.2)	1981 (74.3)	558 (47.9)
African American	447 (24.8)	685 (25.7)	608 (52.1)
DSM-IV Alcohol Dependence
Diagnosis, N (%)	1139 (63.3)	612 (23.0)	547 (46.9)
FTND Nicotine Dependence
Never Smokers, N (%)	509 (30.3)	0	505 (43.3)
Nicotine Dependent, N (%)	803 (47.9)	1407 (52.8)	445 (38.2)
Not Nicotine Dependent, N (%)	366 (21.8)	1259 (47.2)	216 (18.5)
DSM-IV Cocaine Dependence
Diagnosis, N (%)	481 (26.7)	219 (8.2)	551 (47.3)
rs1229984**
AA	5 (0/5)	6 (0/6)	1 (0/1)
AG	96 (38/58)	162 (21/141)	51 (20/31)
GG	1699 (1101/598)	2498 (591/1907)	1114 (527/587)

COGA: Collaborative Study on the Genetics of Alcoholism

COGEND: Collaborative Genetic Study of Nicotine Dependence

FSCD: Family Study of Cocaine Dependence

Total N (Case n/Control n)

Collaborative Study on the Genetics of Alcoholism (COGA)

The Collaborative Study on the Genetics of Alcoholism (COGA) was initiated in 1989 with funding from the National Institute on Alcohol Abuse and Alcoholism (NIAAA)[9-10]. Subjects were recruited by seven sites across the United States. Alcohol dependent probands were systematically recruited from treatment facilities, and relatives were interviewed. Comparison families were drawn from the same communities. A case-control sample of biologically unrelated individuals was drawn from the sample[11]. All cases met DSM-IV criteria for alcohol dependence. Control subjects were defined as individuals who consumed alcohol, but did not meet any definition of alcohol dependence or alcohol abuse, or abuse or dependence on other drugs (except nicotine). For this study 1800 COGA case-control subjects were studied, including 1139 DSM-IV alcohol dependent cases and 661 non-alcohol dependent controls.

Collaborative Genetic Study of Nicotine Dependence (COGEND)

The Collaborative Genetic Study of Nicotine Dependence (COGEND) was initiated in 2001 as a three-part program project grant funded through the National Cancer Institute[12]. Nicotine dependent and non-dependent smoking subjects were recruited from communities in Detroit and St. Louis, using a Health Maintenance Organization in Detroit and the Missouri Family Registry in St. Louis maintained at Washington University for research purposes. Nicotine dependent subjects were current smokers with a Fagerström Test for Nicotine Dependence (FTND)[13] score of 4 or more. Non-nicotine dependent subjects were individuals who smoked at least 100 cigarettes lifetime, who never had any symptoms of dependence (lifetime FTND=0). For this study 2666 COGEND subjects were used, including 612 DSM-IV alcohol dependent cases and 2054 non-alcohol dependent controls. This sample includes 1610 subjects who were previously examined in Sherva et al., 2009[14].

Family Study of Cocaine Dependence (FSCD)

The Family Study of Cocaine Dependence (FSCD) was initiated in 2000 as a case-control family study of cocaine dependence funded through the National Institute on Drug Abuse[15]. Cocaine dependent individuals were systematically recruited from chemical dependency treatment facilities in the greater St. Louis metropolitan area. Community based comparison subjects with no substance dependence were recruited through the Missouri Family Registry and matched by age, race, gender, and residential zip code. For this study 1166 FSCD subjects were used, including 547 DSM-IV alcohol dependent cases and 619 non-alcohol dependent comparison subjects.

Adolescent and Young Adult Study as part of the Collaborative Study on the Genetics of Alcoholism (COGA)

The adolescent and young adults study is a sample of subjects from the COGA study who had at least one assessment between the years 1989–2008 when the subjects were between the ages of 12 and 25 years (Table 2). Subjects were recruited from families affected with alcoholism (either as a first- or second-degree relative) and from community-based comparison families. For this analysis, 2039 adolescents and young adults who ever had a drink of alcohol were studied. All subjects were of European descent. This sample is independent of the COGA case-control sample.

Table 2

Characteristics of Adolescent and Young Adult Sample

	Adolescents and Young AdultsN=2039
Sex, N (%)
Male	983 (48.2)
Female	1056 (51.8)
Age at first interview, years
Mean ± sd	15.9 ± 5.0
Range	6–25
DSM-IV Alcohol Dependence*
Diagnosis, N (%)	302 (14.8)

Phenotyping for All Studies

Alcohol related phenotypes were derived using the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) designed for genetic studies of alcoholism[16-18]. The SSAGA comprehensively assesses alcohol, nicotine and drug use and has high reliability and validity[16-18]. Alcohol and drug dependence were defined using DSM-IV criteria[8] and nicotine dependence was characterized using the Fagerström Test for Nicotine Dependence (FTND)[13].

In the case control samples, two phenotypes were selected for study: lifetime diagnosis of DSM-IV alcohol dependence and maximum number of alcoholic drinks consumed in a 24-hour period. A standard drink of alcohol was defined as a glass of wine, a bottle of beer, or a shot glass of hard liquor.

For the adolescent and young adult study, subjects 18 years and older were assessed with the SSAGA. Adolescent subjects 17 years of age or younger were assessed with the childhood version of the Semi-Structured Assessment for the Genetics of Alcoholism (C-SSAGA). DSM-IV defined alcohol dependence was the outcome under study.

Genotyping of rs1229984

DNA was extracted from blood samples or transformed lymphoblastoid cell lines.

COGA sample and FSCD sample

We used the PrimerPicker software (http://www.kbioscience.co.uk/primer-picker/) to design the assay and followed the protocol described in the KASPar SNP Genotyping System manual (http://www.kbioscience.co.uk), with detection on the ABI 7900 HT Fast Real-Time PCR system (Applied Biosystems, Foster City, California). Sixty-nine percent of samples were also genotyped using the Sequenom MassArray system. Samples with discrepant results (0.3%) were discarded.

COGEND sample

A large component of this study was previously genotyped at this variant as part of a previous genotyping project[19]. To complete the genotyping in the sample, we used Sequenom MassArray, homogenous MassEXTEND (hME) or iPLEX assays for genotyping, using Sequenom SpectroTYPER software v3.4 (see Wang et al., 2009)[20].

COGA adolescent and young adult sample

Rs1229984 was included in a set of 384 SNPs genotyped using the Illumina GoldenGate custom designed platform at the Microarray Core of the Genome Sequencing Center at Washington University. We used the default algorithm in BeadStudio (Illumina; San Diego, California) to cluster genotypes.

Statistical Analysis

Each study was analyzed separately. All studies have previously undergone extensive genotyping, and population structure has been reviewed with Eigenstrat[21]. Self reported race of “White” and “Black” corresponded to genetically determined groups.

Because of the low minor allele frequency for rs1229984 and the small number of subjects homozygous for the minor allele (less than 0.5% in each study), the G/A heterozygous subjects and A/A homozygous subjects were collapsed into a single group for analysis and compared to G/G homozygous subjects.

In the adult case control sample, alcohol dependence was analyzed using logistic regression to estimate odds ratios for the variant rs1229984. Maximum drinks consumed in a 24-hour period (MaxDrinks) was examined by linear regression of the log transformed number of drinks. Prior to log transformation, the few extreme values of maximum drinks in 24-hours (greater than 100) were set at 100. Each study was examined to determine which characteristics predicted alcohol dependence and maximum drinks consumed in a 24-hour period. Age, gender, nicotine dependence and cocaine dependence were significantly associated with alcohol dependence and maximum drinks consumed in a 24-hour period. Primary analyses were adjusted for sex, age (in quartiles), and nicotine and cocaine dependence.

To determine the sensitivity of analyses to the inclusion of covariates, analyses were repeated using SNP only as a predictor variable, and then with SNP, age (quartiles) and gender as predictor variables. In addition, analyses including principal components adjusting for population structure were performed. We combined results from the three studies to calculate a pooled estimate and meta-analysis p value using a fixed effects model for subjects of European and African descent.

In the adolescent and young adult sample, survival analysis was used to examine alcohol dependence. Only subjects who reported having at least one drink of alcohol were included. A total of 302 adolescents and young adults had developed DSM-IV alcohol dependence. Those subjects who had not developed alcohol dependence were designated as censored.

RESULTS AND DISCUSSION

The frequency of alcohol dependence and comorbid nicotine and cocaine dependence is consistent with the study design and ascertainment protocols for each study (Table 1). Rs1229984 is strongly associated with alcohol dependence in each dataset (Table 3). See Supplementary Table 1 for results for the full models. In the meta-analysis across the three adult studies, which included 2298 alcohol dependent cases and 3334 non-dependent controls, the allele encoding His48 is significantly associated with reduced risk for alcohol dependence, with an odds ratio of 0.34 (95% CI 0.24, 0.48) and p value = 6.6 × 10−10. The magnitude of effect is similar in both European American and African American populations. Analyses were repeated using principal components to ensure that the association is not an artifact of population stratification, and results are essentially unchanged (Supplementary Table 2). The findings are robust, not only with respect to ethnicity, but also with major covariates (age, sex, nicotine and cocaine dependence), as removing or limiting covariates did not markedly affect the results (Supplementary Tables 3 and 4). These results are consistent with the recent meta-analysis of Asian populations (47 studies, AG (Arg/His) and AA (His/His) versus GG (Arg/Arg), OR=0.24 (95% CI 0.19, 0.31), p value=2 × 10−31)[5].

Table 3

Association of rs1229984 with DSM-IV alcohol dependence and maximum drinks consumed in a 24-hour period.

			Outcome
	Ethnicity	MAF	DSM-IV Alcohol Dependencea			Maximum Drinks in 24-hour periodb
			N	OR (95% CI)	p	N	β (95% CI)	p
COGA	EA	0.03	1250	0.25 (0.12, 0.48)	5.97 × 10−5	1246	−0.42 (−0.58, −0.25)	4.61 × 10−7
	AA	0.01	428	0.37 (0.06, 1.89)	2.48 × 10−1	425	−0.31 (−0.74, 0.12)	1.57 × 10−1
COGEND	EA	0.04	1981	0.38 (0.23, 0.63)	2.19 × 10−4	1981	−0.22 (−0.31, −0.12)	9.15 × 10−6
	AA	0.01	685	0.47 (0.10, 2.23)	3.42 × 10−1	685	−0.002 (−0.41, 0.41)	9.92 × 10−1
FSCD	EA	0.03	558	0.29 (0.10, 0.85)	2.32 × 10−2	550	−0.24 (−0.50, 0.01)	6.19 × 10−2
	AA	0.02	608	0.40 (0.13, 1.28)	1.24 × 10−1	588	−0.70 (−1.11, −0.28)	9.97 × 10−4
Meta-Analysis	EA		3789	0.32 (0.22, 0.47)	6.03 × 10−9	3777	−0.27 (−0.35, −0.19)	1.33 × 10−11
	AA		1721	0.41 (0.18, 0.93)	3.21 × 10−2	1698	−0.34 (−0.58, −0.10)	6.00 × 10−3
	Combined		5510	0.34 (0.24, 0.48)	6.57 × 10−10	5475	−0.28 (−0.35, −.020)	3.24 × 10−13

Logistic regression adjusted for sex, age (in quartiles), and nicotine and cocaine dependence.

Linear regression of the log transformed number of drinks adjusted for sex, age (in quartiles), and nicotine and cocaine dependence.

AA=African American; EA=European American; MAF=minor allele frequency; OR=odds ratio; CI=confidence interval.

These results are further supported by analyses of an independent sample of adolescents and young adults (aged 12–25 years) of European descent. Using survival analysis, the His48 allele is associated with a 0.51 reduced hazard of developing alcohol dependence (N=2039; Hazard Ratio=0.51 (95% CI 0.30, 0.88) and p value = 0.015).

The His48 allele is also associated with lower alcohol consumption as measured by the subjects’ lifetime maximum alcohol consumption in a 24-hour period (β = −0.28 (95% CI −0.35, − 0.20), p value = 3.24 × 10−13) (Table 3). A similar association is seen in both European American and African American subjects. These findings are consistent with, but exceed in magnitude, those from a prior report of association between His48 and maximum drinks consumed in a 24-hour period[22].

These results provide strong evidence that the His48 allele in β-ADH is strongly associated with a reduced likelihood of a lifetime diagnosis of alcohol dependence and a lower maximum number of drinks consumed in a 24-hour period in individuals of both European and African ancestry. The effect of this association is seen early in the course of illness and predicts the development of alcohol dependence in adolescents and young adults who are at the beginning of their drinking career. The direction of these effects is similar to that found in Asians, among whom the allele frequency is much higher.

Because rs1229984 is not on most GWAS arrays, is relatively uncommon (< 5% minor allele frequency) in populations of European and African descent, and the accuracy of imputation is poor, the association with alcohol dependence has not been detected in several recent genome-wide association studies[11, 23–24], including studies in which these subjects were included in this sample[11, 24]. For example, this SNP was imputed from the GWAS data using BEAGLE[25]. Though over 98% of the genotyped GG samples are correctly assigned, only 68% of the AG heterozygote and AA homozygote genotypes are accurately imputed. As a result, testing this association required direct genotyping of this SNP.

Although the population attributable risk of His48 differs because of the varying allele frequencies, at an individual level the effect of this ADH1B variant on amount of alcohol consumed and on the risk of developing alcohol dependence is the same regardless of ancestry. This polymorphism is an exception to the generally small effect sizes associated with other variants that affect complex traits including alcohol dependence; rs1229984 has a substantial effect on risk of developing alcohol dependency, reducing it by two-thirds.