Haplotype-Based Localization of an Alcohol Dependence Gene to the 5q34 γ-Aminobutyric Acid Type A Gene Cluster FREE

Alcohol dependence (“alcoholism”) is a common (10% in men, 4% in women),¹ genetically influenced disorder with heritability estimates ranging from 40% to 60%.^2- 4 The vulnerability/protective alleles for alcoholism represent both the pharmacokinetic (detected and widely replicated) and the pharmacodynamic (detected but still putative) domains. A substantial part of the genetic component of variance in alcoholism vulnerability is substance specific, that is, not cross-inherited with vulnerability to other drugs.⁵ Such genetic epidemiological data implicate genes that are relatively substance specific. Although a few substance-specific polymorphisms have already been isolated in the pharmacokinetic genetic risk domain (ie, in genes coding for aldehyde dehydrogenase type 2⁶ and for alcohol dehydrogenase type 2⁷), they are common in East Asians but not in most other populations. Pharmacodynamic-specific gene targets for predisposition to alcohol dependence include the genes that encode receptors at which ethanol (alcohol) first acts in the brain and that may act as gatekeepers in alcohol response.

γ-Aminobutyric acid (GABA) is the major inhibitory neurotransmitter in the mammalian central nervous system⁸ and exerts most of its actions at GABA type A (GABA_A) receptors, which are ligand-gated chloride-channel complexes. γ-Aminobutyric acid type A receptor complexes are generally heteropentamers composed of genetically distinct subunits; 16 related mammalian subunits have been reported.^9- 10 The human GABA_A receptor subunit genes are grouped into 7 classes (α, β, γ, δ, ε, π, and θ).¹⁰ Within these classes, protein sequence identity is approximately 70% (ie, α1-α6, β1-β3, or γ1-γ3) and protein sequence identity is approximately 30% between classes. Most of the human GABA_A receptor subunit genes have been assigned to chromosome regions, and subunit gene clusters have been identified on chromosomes 4, 5, 15, and X.¹⁰ Because of the number of subunits, the diversity of expressed GABA_A receptor pentamers is large and would be larger except that these receptors are usually assembled from defined proportions of α, β, and γ subunits and except that there is considerable variation in the localization and developmental timing of expression of the different subunits.^11- 14 The most abundant subunit combinations observed so far are α1β2γ2, α2β3γ2, and α3β3γ2, which make up about 80% of all GABA_A receptors. The receptors containing a α1β2γ2 combination, believed to assemble following the coordinated expression of GABRA1, GABRB2, and GABRG2 genes on 5q34, constitute the major GABA_A receptor subtype in the adult central nervous system (about 50%), have been identified in neurons at all levels of the neuraxis, and are believed to mediate the basic pharmacological spectrum of the classical, high-affinity BZ site ligands, except CL 218872.^12,15 A variety of central nervous system–depressant drugs that show cross-tolerance, including ethanol, benzodiazepines, and barbiturates, as well as inhalant anesthetics and some endogenous neuroactive steroids, positively modulate the GABA_A receptors.^8,16- 18 Many behavioral effects of ethanol (eg, anxiolytic, ataxic, and sedative/hypnotic) may be explained by allosteric enhancement of GABA_A receptor–mediated ionic influx and consequent hyperpolarization of the neuronal membrane.⁸ Agents that increase GABA_A receptor activity in the central nervous system by acting as GABA positive modulators (ie, benzodiazepines, barbiturates, and depressant steroids) enhance acute sensitivity to ethanol and maintain ethanol preference, whereas drugs that act as GABA antagonists at GABA_A receptors, such as picrotoxin, decrease many acute actions of ethanol and reduce ethanol preference.¹⁹ In addition, signs of ethanol withdrawal are diminished following treatments with those GABA agonists that increase GABA_A receptor function, whereas GABA antagonists at GABA_A receptors increase such signs.¹⁹ Effects of ethanol on GABA_A receptor function and expression make the GABA_A receptor subunit genes excellent candidates for vulnerability to alcohol dependence.^20- 21 In addition, genetic differences in the ethanol sensitivity of GABA_A receptors were observed in short-sleep and long-sleep mice that differ in the sedative response to ethanol. Differential γ subunit function was observed between the 2 lines, which has been proposed as a critical determinant of individual differences in ethanol sensitivity.²² In BxD RI strains and other rodent genetic stocks, quantitative trait loci for sensitization of locomotor activation by ethanol and for predisposition to acute ethanol withdrawal map to the region containing the gene cluster that includes the γ2 and α6 subunit genes (gabrg2 and gabra6), and all of the mouse GABA_A subunit gene clusters appear to have ethanol response–associated quantitative trait loci nearby.²³ In the mouse, the Ala11Thr variant of the γ2 subunit correlates with acute ethanol withdrawal severity,²⁴ and in the rat, an α6 Arg100Gln amino acid substitution that differentiates alcohol tolerant and nontolerant has been reported to alter response to benzodiazepines.²⁵

In the human, preliminary linkage findings have also indicated that variation in GABA_A receptor subunit genes plays a role in differential vulnerability to alcohol dependence, and a specific polymorphism of the α6 subunit gene (ie, GABRA6) encoding an amino acid substitution (Pro385Ser) was implicated.²⁶ As measured by slowing of saccadic eye movement velocity, sons of alcoholics appear to have diminished sensitivity to benzodiazepines as compared with young men who are at lower risk for alcohol dependence, leading to the proposal that differences in the expression or function of GABA_A receptors alter vulnerability to alcohol dependence.²⁷ In these same subjects, the GABA_A α6 subunit Pro385Ser polymorphism was shown to predict benzodiazepine sensitivity.²⁸

When ethanol itself is administered as a drug challenge, decreased intoxication, motor effects, and hormonal responses are observed in sons of alcoholics,²⁹ which has been extensively replicated. In a 15-year follow-up study of 450 men, Schuckit^30- 31 showed that low level of response to alcohol in young adulthood is a predictor of later alcoholism. The risk attributable to the level of response is substantial (approximately 40% of the variance in vulnerability), largely independent of family history, and largely unshared with other drugs of abuse.³¹ By selective genotyping of a subsample of this same cohort, Schuckit et al²⁶ found preliminary evidence that the GABA_A α6 Pro385Ser amino acid substitution is associated both with alcohol sensitivity and with increased risk for alcohol dependence. Associations of markers at the 5q34 GABA_A gene cluster have also been reported in Scottish,³² German,³³ and Japanese³⁴ samples.

Here we report both locus-based linkage analyses in sib pairs and allele-based localization by association using GABA_A haplotypes in this Finnish population and in a second semi-isolated population of Southwestern Native Americans.

Subjects were selected from 2 semi-isolated populations: a Finnish population and a Southwestern Native American population.

A total of 511 psychiatrically interviewed Finns included 110 alcoholic offender probands, 277 relatives, and 124 unrelated controls.³⁵ Participants were studied under a human research protocol approved by the institutional review board of the National Institutes of Health, Bethesda, Md; the National Institute of Mental Health, Bethesda; Office for Protection from Research Risks, Bethesda; the University of Helsinki Department of Psychiatry institutional review board, Helsinki, Finland; and the University of Helsinki Central Hospital institutional review board. All subjects were 17 years or older and provided written informed consent.

The probands were male criminal offenders undergoing forensic psychiatric evaluation as inpatients in the Department of Psychiatry, University of Helsinki. Therefore, the sample is enriched for the early-onset form of alcoholism associated with impulsivity and antisocial behaviors, so-called type II.³⁶ Recently, it has become recognized that in addition to scoring high in novelty seeking (as hypothesized by Cloninger³⁶), such alcoholics also tend to have higher harm avoidance and anxiety,³⁷ and these particular alcoholics were above the norm for harm avoidance measured on the Tridimensional Personality Scale Questionnaire (data not shown).

A total of 110 women and 167 male relatives were ascertained through the index cases. There were 275 sib pairs among whom 82 sib pairs were concordant for alcoholism (DSM-III-R³⁸ alcohol dependence), 64 were discordant, and 129 were unaffected. The 124 unaffected controls were unrelated healthy Finnish male volunteers recruited through local newspaper advertisements. Controls were also psychiatrically interviewed (Structured Clinical Interview for DSM-III-R³⁸) and were paid for their participation. Controls were in a good state of general health as established by physical examination, erythrocyte and lymphocyte indices, liver enzyme and thyroid hormone levels, and serum electrolyte and creatinine concentrations.

The Southwestern Native American sample (N = 433) was collected without proband ascertainment bias for a family-based study on alcoholism and related psychiatric disorders.³⁹ The human research protocol was approved by the tribal council and by the institutional review board of the National Institute on Alcohol Abuse and Alcoholism, National Institutes of Health. All subjects provided written informed consent. Subjects were ascertained based on familial relationship; both the original selection of the families and the later recruitment of subjects were performed blind to diagnoses. Participants were members of the same genealogy, older than 21 years, in general good health, and eligible for tribal enrollment (ie, with at least one fourth tribal heritage).

The 433 psychiatrically interviewed Southwestern Native Americans included 192 individuals affected with alcoholism (DSM-III-R³⁸ alcohol dependence) and 241 unaffected individuals. There were 419 sib pairs, which included 188 sib pairs concordant for alcoholism, 168 discordant, and 63 unaffected. Use of the tribal name and exact reservation location is avoided because these details are unnecessary for the analyses set out here.

The Finns were interviewed using the Structured Clinical Interview for DSM-III-R³⁸ by 2 psychiatrists in the Department of Psychiatry, University of Helsinki. Diagnoses were made independently and blindly by the 2 psychiatrists under the supervision of a senior research psychiatrist and according to criteria from the DSM-III-R.³⁸ The Southwestern Native Americans were interviewed using the Schedule for Affective Disorders and Schizophrenia–Lifetime Version by a psychologist experienced in psychiatric assessment with this and other Native American populations. Diagnoses for alcoholism and other psychiatric disorders according to criteria from the DSM-III-R³⁸ were based on the interview data; medical, educational, court, and other records; and corroborative information from family members. Diagnoses were made by 2 blind raters: a clinical social worker and a clinical psychologist. Diagnostic differences were resolved in a consensus conference that included a senior psychiatrist experienced in diagnosis in Native American people.

In the Finnish sample, 1 patient was excluded because of chronic schizophrenia. No patient was excluded from the Southwestern Native American population sample.

Six polymorphisms at 3 GABA_A receptor genes clustered on the long arm of human chromosome 5 were genotyped blind to diagnoses by polymerase chain reaction–restriction fragment length polymorphism analysis: GABRA6 1031G>C⁴⁰; GABRA6 1236C>T (Pro385Ser)²⁸; GABRA6 1519T>C³³; GABRB2 1412C>T³³; GABRG2 IVS9 + 99C>A⁴¹ (NCBI rs211014); and GABRG2 IVS10 + 3145A>G⁴¹ (NCBI rs211013). GenBank accession numbers for GABRA6, GABRB2, and GABRG2 are GABRA6, number S81944; GABRB2, number S67368; and GABRG2, number NM000816.

The primer pairs listed in Table 1 were used to amplify genomic DNA isolated from Epstein-Barr virus–immortalized lymphoblastoid cell lines. Each 25-μL polymerase chain reaction volume contained 50 to 100 ng of genomic DNA, 50 pmol of each primer, 0.125mM deoxyribonucleoside-5"-triphosphates (PerkinElmer, Fremont, Calif), and 1 unit of TaqGold (PerkinElmer) DNA polymerase. The final reaction volume also contained 10mM Tris base pH 9, 50mM potassium chloride (KCl), 1.5mM magnesium chloride (MgCl₂), 0.1% Triton X-100 (PerkinElmer), and 0.01% gelatin. The reactions were performed using a hot-start procedure: TaqGold DNA polymerase was activated only after a first denaturation step of 12 minutes at 95°C. Amplifications were carried out using a modified step-down thermocycling procedure.⁴² A 12-μL volume of each amplicon was digested with the appropriate restriction enzyme (New England Biolabs, Beverly, Mass) (Table 1) according to the manufacturer’s specifications. The digested products were electrophoresed in 10% 1X TBE polyacrylamide gels, at 100 V for 1.5 hours at room temperature. Gels were stained with ethidium bromide and photographed under UV light.

Genotypes were read by 2 independent raters blind to diagnoses and were checked for mendelian segregation. Discrepant genotypes were regenotyped. The error rate was estimated to be less than 1%.

Differences in allele and genotype frequencies and deviation of genotypes from Hardy-Weinberg equilibrium were evaluated using the χ² test and the Fisher exact test for Hardy-Weinberg equilibrium, which is implemented in the computer program MENDEL version 5.0.⁴³ Extreme cases of population admixture and/or stratification can be detected by marked departure from Hardy-Weinberg equilibrium. Evidence for population heterogeneity or admixture was further evaluated using the computer programs LINKMAP from the LINKAGE package version 5.2⁴⁴ and HOMOG.⁴⁵ Population stratification was also evaluated using a variance-components approach that compares the between- and within-family components of association,^46- 47 which is implemented in the computer program QTDT version 2.2.1.⁴⁷

Pair-wise linkage disequilibrium coefficients (D), normalized linkage disequilibrium (D′, D divided by the maximum possible value of D), and correlation coefficients (r², the squared total linkage disequilibrium divided by the product of the allelic frequencies at both loci) were estimated with MLOCUS.^48- 49 Significance levels were estimated from the χ² distribution.

The only phenotype analyzed was DSM-III-R³⁸ alcohol dependence. Two-point linkage analysis was conducted using the nonparametric sib-pair regression method,⁵⁰ as programmed in the Sibpal module of the S.A.G.E. package.⁵¹ Haplotype association and localization was performed with TRIMHAP, a family-based haplotype test for linkage disequilibrium.⁵² First, the 6 5q34 marker genotypes were used to estimate marker haplotypes for all pedigree members using GENEHUNTER.⁵³ Afterwards, TRIMHAP was used to define a subset of markers that were feasible as ancestral haplotypes and to determine identity by descent within the pedigrees. For each 3-locus haplotype in the sample, a haplotype-sharing score was calculated. TRIMHAP determined the category of each haplotype, added it to the trimmed-haplotype table, and constructed the sum of haplotype-sharing scores over all categories. The empirical P values for the trimmed-haplotype statistic were estimated using 10 000 replicate samples generated by permutation bootstrapping. A feature of replicate samples constructed using a permutation scheme is that replicates are formed assuming the null hypothesis H₀ of linkage but linkage disequilibrium is true.

The use of this method, which is a family-based test, also addresses the issue of association analysis of affected and unaffected individuals who share some degree of relationship, which is the case in the Southwestern Native American population sample.

Reflecting the semi-isolated nature of both populations, the Finns and the Southwestern Native Americans differed in allele frequencies, in 6-locus haplotype frequencies, and in overall patterns of linkage disequilibrium across the 5q34 GABA_A cluster region (Figure 1). Within both populations, there were no significant deviations of genotypic distributions from Hardy-Weinberg equilibrium for all the polymorphisms in study. For both populations, there is no evidence of population heterogeneity or admixture.

Consistent with an autosomal linkage scan previously performed,⁵⁴ sib-pair linkage of GABRG2 to alcohol dependence was observed in Finns (P = .008), although not in Southwestern Native Americans. Association of the GABRB2 1412T allele with alcohol dependence was detected in both Finns (P = .01) (Table 2) and Southwestern Native Americans (P = .008) (Table 3), and the GABRA6 1519T allele was also associated with alcohol dependence in both Finns (P = .01) (Table 2) and Southwestern Native Americans (P = .03) (Table 3).

Linkage disequilibrium mapping with haplotypes yielded evidence for an alcohol dependence locus in the GABA_A gene cluster region in both populations (Table 4 and Figure 2). For Finns, the most highly significant signals were at 3-locus haplotypes that included GABRA6 polymorphisms, with the peak signal at a 3-locus haplotype with the 3 GABRA6 polymorphisms (empirical P = .004). Results were also significant at haplotypes that included combinations of GABRA6 1236C>T, GABRA6 1519T>C, GABRG2 IVS9 + 99C>A, and GABRG2 IVS10 + 3415A>G polymorphisms (empirical P = .01 to .03). For Southwestern Native Americans, the most highly significant signal was also at a haplotype that included the 3 GABRA6 polymorphisms (empirical P = .02). Results were also significant at a haplotype that included the GABRB2 1412C>T, GABRA6 1031G>C, and GABRA6 1236C>T polymorphisms (empirical P = .01).

Alcohol dependence has recently been the object of both positional cloning, or “reverse genetics,” and candidate gene, or “forward genetics,” approaches for gene mapping. Roles in alcoholism vulnerability for alcohol metabolic gene polymorphisms have been validated and reviewed,^21,55 but these loci contribute a substantial portion of the variance in risk only in the Southeast Asian populations in which they are abundant. These findings, achieved by forward genetics, were made by relating functional alleles to alcohol pharmacokinetics to phenotype. On the other hand, positional cloning in humans^56- 57 and rodents^58- 60 has yielded loci, several at least partially replicated, but not determinant alleles. Successes in other complex diseases, such as breast cancer⁶¹ and Alzheimer disease,⁶² clearly indicate that a combination of mapping methods can lead to the discovery of vulnerability alleles.

Both the positional cloning and the candidate gene paradigms that have been applied to the genetics of ethanol-related behaviors in humans^20- 21,55 and rodent genetic models^58- 60 have produced convergent evidence implicating the region (5q34 in humans) containing a GABA_A receptor gene cluster.^20- 21 These studies have focused further interest on genetic variation in GABA_A receptors, which had already been implicated by pharmacobehavioral and receptor-pharmacology data as potential gatekeepers in ethanol response.^9- 10 Furthermore, locus-based linkage evidence in the Finnish sample studied here positionally implicated the chromosome 5q34 region in alcohol dependence.⁵⁴ The allele-based associations we observed both in Finns and in Southwestern Native Americans narrow the region, so that it seems the linkage signals emanate from the GABA_A gene cluster itself. Since the GABA_A cluster spans more than 700 kilobases and includes 4 receptor subunit genes, we applied haplotype-based localization to further narrow the location of the risk gene, and we found the α6 Pro385Ser amino acid polymorphism positioned within the signal peak.

Semi-isolated populations with relatively smaller sizes, low ethnic diversity, and more homogeneous environments are invaluable in the genetic analysis of complex disorders because reduction in genetic and environmental heterogeneity facilitates the identification of specific factors that affect vulnerability.⁶³ Since alcohol dependence is a complex disorder with diverse genetic and environmental contributions, we decided to study samples from 2 semi-isolated populations with well-defined genetic compositions and relatively homogeneous environments (ie, Finns and members of a Southwestern Native American population). In addition, because of founder effects and reduced population size, linkage disequilibrium in such populations is enhanced, leading to conservation of haplotypes containing disease alleles.^64- 65 The strategy was to increase the odds of finding out whether genetic changes in the 5q34 GABA_A gene cluster have a role in the susceptibility to alcohol dependence. In addition to the testing for significant departure from Hardy-Weinberg equilibrium, which occurs in more extreme cases of population admixture and/or stratification and which was not detected in either population, the results from the tests that were used for detection of more subtle population heterogeneity or admixture and for population stratification confirmed the low genetic heterogeneity for both populations.

In this study, there were several instances where we found allele-based association in the absence of sib-pair linkage of the same phenotype to the same locus, and we detected linkage of GABRG2 to alcohol dependence in the Finnish sample but not in the Southwestern Native American sample. The explanation for this disparity need not be a biological mechanism whereby allele-based associations could occur in populations in the absence of locus linkage in families.⁶⁶ Sib-pair linkage may fail to detect linkage to loci of minor effect,⁶⁷ whereas allele effects at these loci may be readily detectable by association.⁶⁸

The results we obtained are convergent with previously published data, despite the ethnic diversity of the populations represented. The GABRA6 1519T and the GABRB2 1412T alleles were associated with alcohol dependence in both Finns and Southwestern Native Americans. The GABRA6 1519T associations are consistent with previous studies in a Scottish sample³² and a German sample.³³ The GABRB2 1412T associations in Finns and Southwestern Native Americans are consistent with the Scottish population findings but not the German sample findings. Furthermore, although we found sib-pair linkage to GABRG2 in 1 population (Finns) and neither population showed association of alcohol dependence with GABRG2 alleles, association between alcoholism and GABRG2 IVS10 + 3145 G has been reported in a Japanese sample.³⁴

In addition to sib-pair linkage and association analyses, we also applied a family-based haplotype localization strategy, which exploits the power of linkage disequilibrium to identify ancestrally related chromosomes that carry an allele that influences disease.⁵² The trimmed-haplotype statistic allows detecting linkage disequilibrium due to ancestral haplotypes, and the algorithm is designed to be repeated with the disease-susceptibility locus located at a grid of positions covering a chromosomal region in study. This method addressed the need of a family-based association test for related individuals in the Southwestern Native American sample and allowed the localization of a putative alcohol-dependence susceptibility locus within the 5q34 GABA_A cluster, with the lowest P values implicating GABRA6 polymorphisms in both Finnish and Southwestern Native American population samples (Figure 2).

Several disease-vulnerability alleles have been shown to reside on characteristic haplotype backgrounds so that their detection would have been possible through association to ancestrally related haplotypes.^69- 70 Linkage disequilibrium coefficient estimates between alleles at the 6 loci in the 5q34 GABA_A gene cluster did not clearly vary as a function of distance in either Finns or Southwestern Native Americans, nor did the correlation coefficients (Figure 1). Decoupling between the extent of linkage disequilibrium and physical distance can reflect a variety of complexities in the histories (eg, mutation, selection, genetic drift, and migration) of populations^{64- 65,71- 72} and can vary even within the same population from one chromosome region to the other.^64- 65

Given that our findings in these 2 semi-isolated populations, in conjunction with data in the literature, strongly implicate the 5q34 GABA_A cluster in alcohol dependence, the next critical steps in evaluating the genes in this region are likely to be single nucleotide polymorphism genotyping across the 5q34 GABA_A gene cluster and the identification of functional alleles that may influence gene expression and of alleles that alter receptor structure. Among the 5q34 GABA_A receptor genes, GABRA6, which encodes the α6 subunit, is peculiar in showing selective expression in the cerebellum. It is implicated in alcohol dependence both by haplotype localization and by association with Ser385, a relatively abundant nonconservative missense variant, which may alter phosphorylation of the receptor protein.⁴⁰ However, functionality of Ser385 or any other human GABA_A subunit variant has yet to be demonstrated. For GABA_A receptors, this challenge is complicated by the heteropentameric nature of the receptor channel (creating a potential need to express subunit alleles in different molecular contexts) and the possibility that differences in cellular environment can affect the ability of an allele to alter expression or protein posttranslational modification. Isolation of the vulnerability allele(s) located within the chromosome 5q GABA_A gene complex will enable us to better understand interindividual variation in vulnerability to alcohol dependence and would lead to improvements in detection and treatment of at least a subset of patients with this common, debilitating disorder.

Correspondence: Marta Radel, MD, PhD, Bioconsult, 17702 Calabar Dr, Gaithersburg, MD 20877 (martaradel@verizon.net).

Submitted for Publication: December 12, 2003; final revision received August 3, 2004; accepted August 19, 2004.

Funding/Support: This study was supported by grants from the National Institute on Alcohol Abuse and Alcoholism, National Institutes of Health, Bethesda, Md (Dr Goldman).

Acknowledgment: We would like to thank Longina Akhtar, MS, and Lisa Moore for their excellent technical assistance.