Genetics of two opioid receptor gene (OPRM1) exon I polymorphisms: population studies, and allele frequencies in alcohol- and drug-dependent subjects

Transcription

1 Molecular Psychiatry (1999) 4, Stockton Press All rights reserved /99 $15.00 ORIGINAL RESEARCH ARTICLE Genetics of two opioid receptor gene (OPRM1) exon I polymorphisms: population studies, and allele frequencies in alcohol- and drug-dependent subjects J Gelernter 1, H Kranzler 2 and J Cubells 1 1 Yale University School of Medicine, Department of Psychiatry, Division of Molecular Psychiatry and VA Connecticut Healthcare System, West Haven Campus, Department of Psychiatry; 2 University of Connecticut School of Medicine, Department of Psychiatry, Division of Addictive Disorders, Farmington, CT 06030, USA The gene encoding the opioid receptor, OPRM1, contains at least two polymorphisms affecting protein sequence in exon 1, Ala6Val and Asp40Asn. In previous studies, each variant has been reported to be associated with some form of drug dependence. Although past reports have not been consistent, they have also not considered comparable populations. The goals of the present study were to delineate allele and haplotype frequencies of these variants in a range of populations, and in drug- or alcohol-dependent subjects deriving from some of those populations. We developed new PCR-RFLP methods to detect both of these polymorphisms and studied them in control and substance-dependent populations of African American (AA), European American (EA) and Hispanic origin, and in a series of populations differing in geographic origin (Japanese, Ethiopians, Bedouins, and Ashkenazi Jews), 891 subjects overall. We designed primers flanking the DNA segment containing both polymorphisms, each primer creating a different artificial restriction site, such that a single PCR reaction can be completed, then divided, and the PCR product digested with either of two enzymes to reveal both polymorphisms. We found that allele frequencies for both polymorphic systems were significantly different between AA and EA subjects, and there was significant heterogeneity among the more extensive set of populations. Furthermore, there were no significant differences in allele frequency by diagnosis; that is, neither polymorphism appears to be a direct risk factor for substance dependence. Finally, we demonstrated linkage disequilibrium between the two exon 1 markers, and a previously described short tandem repeat (STR) marker. Keywords: opioid receptor; genetic polymorphism; association; linkage disequilibrium; substance dependence Introduction The opioid receptor plays a central role in mediating the effects of morphine and related opioid agonists. 1 The OPRM1 gene encodes the opioid receptor, OPRM1 maps to chromosome 6q24 q25. 2 Targeted disruption of the homologous gene in mice abolishes morphine-induced analgesia, place-preference activity, and morphine-induced physical dependence in null homozygotes. 3,4 Naltrexone, an approved pharmacological treatment for alcohol dependence, 5,6 acts as an antagonist of the three opioid receptor types; this medication was first developed for the treatment of opiate dependence, based on its opiate antagonist activity. A QTL mapping study 7 was consistent with polymorphic variation at the opioid receptor locus in mouse affecting morphine preference. A growing body of data supports a role for the Correspondence: Dr J Gelernter, Psychiatry 116A2, VA CT Healthcare System, West Haven, 950 Campbell Avenue, West Haven, CT 06516, USA. gelernter-joel cs.yale.edu Received 5 January 1999; revised and accepted 16 March 1999 endogenous opioid system in the reinforcing effects of a variety of drugs that are self-administered by animals and abused by humans, including nicotine, cocaine, amphetamines, alcohol, and cannabinoids. All of these drugs (as well as opioids) have been shown to release dopamine in the nucleus accumbens (NAc), 8,9 which is thought to provide a common pathway for reinforcement. 10 In turn, opioid receptors in the ventral tegmental area (VTA) are involved in the modulation of NAc dopaminergic activity. For example, infusion of a receptor agonist into the VTA increases dopamine release into the NAc, while infusion of a receptor antagonist decreases dopamine release. 11 Furthermore, alcohol preferring (AA) rats show a significantly higher density of opioid receptors in the VTA (and related limbic brain regions), in comparison to alcohol avoiding (ANA) rats. 12 Therefore, genetic variation at loci coding for opiatesystem proteins, including OPRM1, might be expected to affect risk for drug dependence and alcoholism, and possibly other addictive behaviors as well. Four recent studies examined the association between OPRM1 alleles and alcohol and/or drug

2 dependence. Berrettini et al 13 identified two polymorphisms of OPRM1, one in a non-coding region and the second producing a change in the predicted protein sequence (Ala6Val). They compared the frequency of the variant alleles in a sample of 55 substance-dependent subjects (53% European American (EA), 47% African American (AA)) with a group of 51 controls (43% EA, 57% AA).) The substance-dependent group had a higher frequency of the Ala6Val variant (f = 0.22) than controls (f = 0.10) that was marginally significant (P = 0.05). However, allele frequencies were not significantly different between EA and AA subjects in this sample, and observations for EA and AA subjects were combined. Bergen et al 14 directly sequenced 91% of the OPRM1 coding sequence (1093 basepairs) and 1479 bases of intronic and untranslated sequence. They found four DNA sequence variants, including three that change protein sequence (Ala6Val, as reported by Berrettini et al; 13 also, Asn40Asp and Ser147Cys) and one intronic variant (IVS G/C). The variant forms of the exonic polymorphisms other than Asn40Asp were rare in the populations studied, in contrast to the report by Berrettini et al 13 for Ala6Val. Using OPRM1 genotypes and haplotypes from US Caucasian, Finnish Caucasian and Southwestern American Indian samples, these investigators conducted association and sib-pair linkage analyses with diagnoses of alcohol dependence and drug dependence. No significant associations were observed. Bond et al 15 described functional variation corresponding to Asn40Asp (A118G); the differing forms vary in their sensitivity to -endorphin (but not to other possible endogenous and exogenous ligands, including met-enkephalin, dynorphin A, morphine, and methadone; or to naloxone). The observed difference corresponded to about a three-fold alteration in affinity. These authors also described a statistically significant association of that variant with opioid dependence, but only in Hispanic subjects. In that population, frequency of the less common (Asp or G) allele was 0.10 in 58 opioid-dependent subjects, but 0.39 in nine non-dependent controls. Allele frequencies differed significantly by population in this sample as well. When they combined subjects from different populations (despite the observed difference in allele frequency by population), they also observed a significant difference in allele frequency at the Ala6Val system between opioid-dependent subjects and controls. We have previously reported a possible association of an STR polymorphism at this locus to substance dependence 16 and described its basic population genetics. 17 The reports by Berrettini et al, 13 Bergen et al, 14 and Bond et al 15 raise important possibilities regarding potential effects of OPRM1 exon 1 variants on risk for substance dependence and on function of the opioid receptor protein, but these reports are inconsistent. As discussed above, Berrettini et al 13 reported an association of Ala6Val with substance dependence, but also pooled EA and AA samples; they found no difference in allele frequencies between these populations. Bergen et al 14 identified an additional exon 1 polymorphism (Asn40Asp) and suggested that neither one is linked or associated with substance dependence in populations of European origin and Amerindians. Moreover, their observed Ala6Val allele frequencies were very different from those reported by Berrettini et al in similar populations: 1% for Bergen et al, but 10 22% for Berrettini et al depending not on population, but on diagnosis. Most recently, Bond et al 15 reported allele frequencies close to those of Bergen et al in comparable populations, found differences in allele frequency by population, and described a possible association of Asn40Asp with opioid dependence, but only in a small sample of Hispanic subjects. These studies differed methodologically also, as Berrettini et al 13 obtained genotypes by a single stranded conformational polymorphism (SSCP) technique. Bergen et al 14 and Bond et al 15 used either sequencing or (for the former research group, for Asn40Asp only) PCR-RFLP techniques. In order to address a possible relationship of polymorphic variation at this locus to alcohol and/or drug dependence, and to delineate the basic population genetics of these polymorphic systems, we studied both exon 1 polymorphisms in a total of 891 subjects: control populations of varying geographic origin (United States: EA, AA, and Hispanic; and Ethiopian Jews, Ashkenazi Jews, Bedouins, and Japanese) and varying diagnoses (EA, AA, and Hispanic drug-dependent subjects; EA alcohol-dependent subjects). We developed rapid PCR-RFLP assays for both polymorphisms, designing primers creating artificial restriction sites including both polymorphic bases. Genotyping both polymorphisms from the same PCR product also allowed molecular haplotyping in the few subjects heterozygous at both systems. We then addressed linkage disequilibrium (LD) relationships between the STR and the exon 1 polymorphisms. Materials and methods Clinical American control subjects: EA, AA, and Hispanic control subjects (see Table 1 for sample sizes) were collected at the VA CT Healthcare System, West Haven Campus, and at the University of Connecticut. The majority of these subjects were screened to exclude major psychiatric diagnoses, as described elsewhere. 18 Substance dependent (ie, alcohol-dependent and drugdependent) subjects All subjects (see Table 1) were recruited at the University of Connecticut (CT) School of Medicine (Farmington) or VA CT Healthcare System (West Haven). All substance-dependent subjects were diagnosed using DSM-III-R criteria. 19 The diagnosis was made with the Structured Clinical Interview for DSM-III-R (SCID), 20 the computerized Diagnostic Interview Schedule for DSM-III-R (C-DIS-R), 21 or a checklist comprised of DSM-III-R symptoms. Other populations DNA from unrelated Ashkenazi Jews, Ethiopian Jews, and Bedouins was obtained from 477

3 478 Table 1 Allele frequencies for each sample. Substance-dependent subjects are classified by primary diagnosis and each subject is counted once only Control Drug-dependent Alcohol-dependent 2n Ala6Val 2n Asn40Asp 2n Ala6Val 2n Asn40Asp 2n Ala6Val 2n Asn40Asp EA AA Hispanic Japanese Ashkenazi Bedouin Ethiopian the National Laboratory for the Genetics of Israeli Populations, Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, Israel. Samples from unrelated Japanese subjects were provided by Drs H Ichinose and T Nagatsu, as described previously. 22 Regression analysis A total of 519 subjects were included in the analysis that examined the relationship between OPRM1 genotypes and substance dependence by logistic regression. (Unscreened (random population) control subjects were included in allele frequency comparisons by diagnosis, but excluded from regression analysis.) Of this number, 77.6% were dependent on one or more of the following substances: alcohol (n = 316), cocaine (n = 202), or opioids (n = 79). The remaining subjects (n = 116) were without a major psychiatric disorder based on administration of the SCID (n = 84) or an unstructured clinical interview (n = 32). The majority (71.2%) of substance-dependent subjects were male. The mean age of this group was 37.9 (SD = 9.4) years. The racial/ethnic breakdown of the group was: 80.6% EA, 18.9% AA, and 6.9% Hispanic. The control group was 54.3% male, with a mean age of 31.8 (SD = 10.6) years. The racial/ethnic breakdown of the group was 62.1% EA, 29.3% AA, and 8.6% Hispanic. Laboratory PCR primers, designed based on reported sequence 14 and incorporating mismatched bases (capitalized below) to produce artificial restriction sites (bold), were ArtUp (5 -ccg tca gta cca tgg aca gca gcg Gtg-3 ) and R2-D2 (5 -gtt cgg acc gca tgg gtc gga cag AT-3 ), yielding a 154-bp PCR product. PCR cycling parameters consisted of a 5-min hold at 95 C, followed by 30 cycles of 95 C (30 s) and 72 C (60 s). From the 15- l reaction mixture, 5 l was digested in 10 u BanI (New England Biolabs, Beverly, MA, USA; recognition sequence, 5 -GG [C or T][G or A]CC-3 ), and 5 l digested in 10 u DpnII (NEB; recognition sequence, 5 - GATC-3 ), to yield genotypes for both systems from a single PCR reaction. Primer ArtUp is used to genotype the Ala6Val system; BanI cuts the C (Ala, or common) allele. Primer R2-D2 is used to genotype the Asn40Asp system; DpnII cuts the G allele (Asp, or rare) allele. Digestion results in 26-bp and 128-bp fragments for ArtUp (5 end) or 25-bp and 129-bp fragments for R2- D2 (3 end), when the respective recognition sequences are present. Fragments were resolved by MetaPhor (FMC, Rockland, ME, USA) agarose gel electrophoresis. Berrettini et al 13 used similar primer placement; some PCR reactions included one of the primers described above together with the opposing primer described by them, 13 for genotyping a single system. For subjects heterozygous at both systems, molecular haplotyping was accomplished by sequentially digesting in BanI, then DpnII. If the less common alleles are on the same chromosome (haplotype Ala 6 Asn 40 Val 6 Asp 40 ), one restriction site would be recognized on each chromosome, and predicted fragment sizes would then be 129, 128, 26 and 25 bp. If they are on opposite chromosomes (haplotype Ala 6 Asp 40 Val 6 Asn 40 ), the former chromosome would contain two restrictions sites and the latter none, so predicted fragment sizes would then be 154, 103, 26, and 25. Data analysis Deviation from the genotype counts predicted by Hardy Weinberg (HW) equilibrium expectations was tested using an exact test, as described by Weir 23 and implemented in software written by Lewis and Zaykin 24 (Genetic Data Analysis (GDA) version 1). Variation in allele frequencies among control populations was tested by contingency table analysis. We used the CLUMP program 25 to estimate empirical significance values of the 2 statistic (T1) using Monte Carlo simulations. Comparisons between sets of two populations used the Fisher exact test (FET). Hierarchical logistic regression was used to examine the association of race, sex and OPRM1 genotypes with diagnosis (substance-dependent vs control). The following independent variables were entered in the order shown: race, sex, Ala6Val genotype, Asn40Asp genotype and all two and three-way interactions involving these variables. The three-level race variable was entered using two orthogonal contrasts (EA vs AA or Hispanic, and AA vs Hispanic). Because of the absence of homozygotes for the rare allele, the Ala6Val genotype was dichotomized (ie, presence/absence of

4 the rare allele). The Asn40Asp genotype was trichotomized (homozygous for the common or rare alleles and heterozygous). Since there is no a priori basis to determine the order of entry of genotypes in the regression equation, the analysis was repeated with the Asn40Asp genotype being entered first. Reversing the order of entry of the genotypes, however, yielded results that were comparable to those obtained using the order of independent variables described above. The use of hierarchical regression made it possible to control for the effects of race and sex prior to examining the effect of the two genotypes on diagnosis (controls vs alcohol- or drug-dependent subjects). This approach also made it possible to examine the contribution to diagnosis of all two-way and three-way interactions involving these variables. To maximize the degrees of freedom available and the parsimony of the model, all variables that were nonsignificant and accounted for less than 1% of the variance were removed from the analysis using a backward elimination approach. Subsequent to the analysis in which all 519 subjects were included, similar analyses were conducted to examine predictors of specific substance use diagnoses. In these three secondary analyses, only control subjects and those with a specific substance use diagnosis (eg, alcohol dependence) were included. We also studied linkage disequilibrium (LD) between three polymorphisms at this locus (Ala6Val, Asn40Asp, and the STR) in a sample of 96 AA subjects and 274 EA subjects ( LD sample ); the LD sample was composed of drug-dependent, alcohol-dependent, and control subjects, and overlapped considerably but not completely with that used for the clinical analyses. LD was evaluated using an exact test as implemented in GDA version 1.0 software; probability levels reflect estimates of significance based on a shuffling test. 24 Results There were no deviations from HW expectations in any populations (control, substance-dependent, or alcohol-dependent). Allele frequencies and sample sizes are presented in Table 1. There was significant heterogeneity among the set of control populations for each polymorphism; for Ala6Val, 2 = 48.7; P ; for Asn40Asp, 2 = 69.3; P (Figure 1). Specific allele frequency comparisons between EA and AA control populations showed significant differences in allele frequency for both systems. For Ala6Val, P 10 6, for Asn40Asp, P (FET). Within-population comparisons of allele frequency by diagnosis showed no significant differences. For AA subjects, control vs SD groups, Asn40Asp, P = 0.08; Ala6Val, P = For EA subjects, control vs SD groups, Asn40Asp, P = 0.32; Ala6Val, P = For EA subjects, control vs alcohol dependence groups, Asn40Asp, P = 0.27; Ala6Val, P = For Hispanic subjects, control vs SD groups, Asn40Asp, P = 0.42; Ala6Val, P = 0.61 (all FET) (Figure 2). Figure 1 OPRM1 exon 1 allele frequencies for both polymorphisms; controls. Figure 2 OPRM1 exon 1 allele frequencies for both polymorphisms; comparisons by diagnosis. Thus, we found, in summary: (1) large differences in allele frequency for both systems by population (note, eg, the Asn40Asp frequency in the Japanese compared to all other populations; Table 1); (2) no significant differences, within population, by diagnosis. We used regression analysis to further evaluate possible relationships between genotype and substance (drug or alcohol) dependence. Race was a significant predictor of diagnosis [ 2 (2) = 6.54, P = 0.038; Wald (2) = 11.26, P = 0.004]; this simply reflects the relative sizes of the SD and control groups, which differ by population. Of the orthogonal contrasts involving race, only that comparing EAs with AAs and Hispanics was significant [Wald (1) = 6.45, P = 0.011]. This reflected the fact that 80.6% of EA subjects were substancedependent compared with 70.3% of AAs and Hispanics. The comparison of AAs and Hispanics was not significant [Wald (1) = 0.28, P = 0.59]. Sex was a significant predictor when entered [ 2 (2) = 9.50, P = 0.002], though in the final equation it was not significant [Wald (1) = 0.002, P = 0.97]. Ala6Val genotype was not a significant predictor [ 2 (2) = 0.53; Wald (1) = 0.41, P = 0.52]. Asn40Asp was also nonsignificant [ 2 (2) = 2.66, P = 0.26; Wald (2) = 4.10, P = 0.13], as were both of the orthogonal contrasts involving that genotype: presence of the rare allele vs absence of the rare 479

5 480 allele [Wald (1) = 2.08, P = 0.15]; one copy vs two copies of the rare allele [Wald (1) = 0.03, P = 0.87]. Of the interactions, only that involving race sex was significant [ 2 (2) = 11.18, P = 0.004; Wald (2) = 9.13, P = 0.010]. The first orthogonal contrast (EAs vs AAs and Hispanics) did not interact significantly with sex [Wald (1) = 0.003, P = 0.96]. However, the second contrast (AAs vs Hispanics) was significant [Wald (1) = 7.88, P = 0.005], which reflected the fact that among AAs a greater percentage of substancedependent subjects were males (65.8%) compared with Hispanics (42.9%). Conversely, among AAs, the percentage of control subjects who were male (44.1%) was smaller than that among Hispanics (80.0%). The three secondary regression analyses (ie, focusing on specific diagnoses of alcohol, cocaine, or opioid dependence) yielded results that were generally comparable to the overall analysis. That is, in no case did either polymorphism (either as a main or interaction effect) account for unique variance in diagnosis. Allele frequencies for both polymorphisms by specific diagnosis are given in Table 2. Some subjects received more than one specific diagnosis and were counted for the purposes of this tabulation in more than one category (which was not the case for Table 1). Haplotype results We observed a total of four subjects heterozygous at both systems: an AA cocaine-dependent subject, an AA alcohol-dependent subject, an Ethiopian, and a Bedouin. (The AA alcohol-dependent subject is not discussed elsewhere or included in the analyses because of insufficient numbers in that diagnostic group for AAs.) All these subjects had the Ala 6 Asp 40 -Val 6 Asn 40 genotype, ie the rarer alleles were on opposite chromosomes (example, Figure 3). All other haplotypes could be resolved unambiguously because all other subjects were homozygous for at least one of the systems; the Val 6 Asp 40 haplotype was never observed. Linkage disequilibrium analysis For each population, all four possible LD relationships were evaluated, ie, each pairwise set of markers and the three markers taken together. In the AA sample, there was a trend for LD between Ala6Val and the STR (P = 0.08) and there was significant LD across all three systems taken together (P = 0.04). The other LD measures did not Figure 3 Molecular haplotype analysis. Lane 1 (counting from the left), BanI; lane 2, DpnII; lane 3, double digest; lane 4, marker. reach significance. In the EA sample, there was significant LD between Asn40Asp and the STR (P = 0.04) and a trend for LD across the three systems taken together (P = 0.07). The other LD measures did not reach significance. Discussion The rapid genotyping method presented here will facilitate future studies of this important gene. The present data document a large range in allele frequency (from a low of in the AAs to a high of in the Japanese) for the functional Asn40Asp polymorphisms in the seven populations studied, and a more narrow range (from a low of 0 in the Japanese and Hispanics to a high of in the AAs) for the Ala6Val polymorphism. For both polymorphisms, there is significant heterogeneity between populations, and for both there are, specifically, significant differences between African and European Americans. The observation that in six of the seven populations studied either one variant or the other, but not both, has frequency 5%, coupled with the molecular haplotype data showing that the two less common alleles never occur on the same chromosome in this sample, suggest that the Ala 6 Asn 40 chromosome represents the ancestral haplotype. (The exception is the Ethiopian sample, where allele frequencies were (Val6) and (Asp40), respectively.) The existence of important Table 2 Allele frequencies for substance-dependent subjects by specific substance of abuse. Some subjects were dependent on more than one substance and are included in more than one diagnostic group. Only subjects genotyped for both systems are included Alcohol-dependent Cocaine-dependent Opioid-dependent 2n Ala6Val Asn40Asp 2n Ala6Val Asn40Asp 2n Ala6Val Asn40Asp EA AA Hispanic

6 population variation for both of these polymorphic sites is consistent with our previously reported population data for another polymorphism at this locus. 17 For that short tandem repeat (STR) polymorphism, there was not, however, a significant difference in allele frequency between African and European Americans. There was no significant difference for allele frequency for either system in our Hispanic subjects by phenotype; that is, we could not confirm the results of Bond et al. 15 We observed an allele frequency in affected subjects identical to what they observed in their affected subjects, but our controls showed a lower allele frequency than theirs. We speculate that the difference in allele frequency for Hispanic controls (0.14 vs 0.39) may be attributable to the very small sample size (only nine subjects) available to Bond et al. 15 However, our sample size is also very small (36 chromosomes, or 18 subjects) and other explanations are possible. Nonetheless, if the hypothesis is accepted that the difference in allele frequency shown by Bond et al 15 is attributable to functional variation in the protein, it is then difficult to explain why this was observed initially only in Hispanic subjects. Parsimony suggests that this is unlikely, taking into account the observed association only in a group with a very small sample; moreover, this group represents the most admixed and heterogeneous population studied (membership defined by language rather than geographical region of origin or any physical features). It is also possible that these observations reflect a minor effect of this variant, and with a sufficiently large sample, such an effect would be observed in other populations. For each of the three populations we studied, the frequency of the Asp40 variant was nonsignificantly higher in control samples, which could be viewed as consistent with the Bond et al 15 suggestion of a protective effect for this allele with respect to phenotype. Power analysis suggested that a sample size of 40 subjects (divided equally between affected and unaffected) would have been sufficient to replicate the finding of Bond et al, 15 and 162 subjects (also divided equally between ill and well) to replicate the finding of Berrettini et al, 13 both for = 0.05 and = 0.20 (80% power). Since the Berrettini et al finding was not reported as population sensitive, we clearly had sufficient power to replicate this finding. The Bond et al result was, however, confined to Hispanic subjects; on that basis the present study had limited, but probably adequate, power to replicate (as our 46 Hispanic subjects were not divided evenly by diagnosis). Although there were between-population differences in allele frequency for both polymorphisms, there was no significant within-population difference by diagnosis. Furthermore, after controlling for population differences, no significant effect of genotype on diagnosis was observed, despite a sample size in excess of 500 subjects. Similarly, when the sample was subdivided based on the presence of a specific substance use diagnosis, no effect of genotype was observed. Parsing out the results for drug dependence separately by specific drugs of abuse for inspection of allele frequencies (Table 2), did not provide any consistent patterns suggestive of association with opioid rather than cocaine dependence, or vice versa. Of the previous reports of exon 1 polymorphism at this locus, our results are most consistent with those of Bergen et al 14 who reported no linkage or association with substance dependence. We could not replicate the positive findings of Berrettini et al, 13 regarding an increased frequency of the 6Val variant in American substancedependent subjects, or of Bond et al, 15 who found a decreased frequency of the 40Asp variant in American Hispanic substance-dependent subjects. Our results differed from those of Berrettini et al, 13 in several respects: (1) although they did not identify differences in Ala6Val frequency between EA and AA subjects, in our sample the difference was highly significant; P 10 6 (FET); (2) we did not identify significant differences in allele frequency in either EAs or AAs by phenotype. Although Berrettini et al 13 identified only the Ala6Val polymorphism in exon 1, their SSCP amplicon may also have contained the Asn40Asp site. If their assay actually detected both polymorphisms, this methodological difference could explain the difference in results. In a previous study of an STR polymorphism at the OPRM1 locus, Kranzler et al 16 reported an association between OPRM1 alleles and substance dependence (including alcohol and drug dependence) that reached nominal, but not Bonferroni-corrected, significance. The results reported here suggest that if there really is an association, it must be explained by a polymorphism or polymorphisms other than the exon 1 variants studied here. A mutational analysis of most of the gene has been reported by Bergen et al, 14 but some regions not yet studied (9% of the coding region, plus much of the promoter) could harbor such variants. In the EA sample, there was significant LD between the exon 1 polymorphism more common in that population, Asn40Asp, and the STR, and in the smaller AA sample, there was a trend for LD between the exon 1 polymorphism more common in that population, Ala6- Val, and the STR, and significant LD across the three markers. Thus, there is significant LD involving the STR and the other markers in each population. However, our LD analysis showed no significant LD between the two exon 1 polymorphisms in either population studied. The most likely explanation for this is lack of informativeness, because one of the polymorphisms was rare in each population, and the molecular haplotype analysis showed that the rare alleles were never observed on the same chromosome in our sample. In this situation, a very large sample is required to demonstrate significant LD. Given the LD shown with exon 1 markers, since the STR is much more informative in both populations than either exon 1 marker (with observed heterozygosity of 0.76 in AAs and 0.63 in EAs, compared to observed heterozygosities of 0.27 and lower for the other polymorphisms observed in the LD sample), this STR represents a useful marker for other (as yet undescribed) 481

7 482 polymorphic variation at this locus that could be responsible for phenotypic variation. The results of Bond et al 15 demonstrate that functional variation is tolerated at this locus. As such, LD with another polymorphism could provide an explanation for our previous positive results with this STR 16 in the presence of the negative results presented here. Understanding genetic overlap between alcohol and drug dependence, and different forms of drug dependence, is very important for studies addressing the determination of the molecular genetic basis of each. Several studies addressing this issue have been published recently, and their results have bolstered the view that the genetic factors leading to alcohol and drug dependence are largely (but not entirely) separate; moreover, both distinct and common genetic factors affect liability to abuse particular substances That is, it should be possible to identify genes that predispose to the range of substance dependence disorders, as well as genes that contribute to drug dependence (in the aggregate), alcohol dependence, and certain different substance dependence disorders, including opioid and cocaine dependence, individually. We found that in the present sample, neither OPRM1 exon 1 coding region polymorphism influences risk for substance dependence. A study using a family-controlled design could yield more definitive results; use of a family-controlled design would be important for this locus, as documented by the results presented here showing significant differences in allele frequency by population. A study with a much larger sample size would also be of interest in testing for a small protective effect of the Asp40 allele with respect to substance dependence. Acknowledgements A Wantroba, R Capper and P Fall provided excellent technical assistance. We appreciate having received DNA samples of healthy subjects from Yale s Neurochemical Brain Imaging Program, and DNA samples of Japanese subjects from Drs H Ichinose and T Nagatsu (of the Institute for Comprehensive Medical Science, Fujita Health University School of Medicine, Toyoake, Japan). Dr W Berrettini kindly provided primer sequences prior to publication. This work was supported in part by funds from the US Department of Veterans Affairs (the VA Medical Research Program, the VA Connecticut-Massachusetts Mental Illness Research, Education and Clinical Center (MIRECC), and the VA-Yale Alcoholism Research Center), NIMH grants K02-MH01387 and P30-MH30929, NIAAA grants R01-AA11330, K02-AA00239, and P50- AA03510, NIDA grant K12-DA-00167, and NCRR grant M01-RR06192 (General Clinical Research Center of the University of Connecticut Health Center). References 1 Yu L. The opioid receptor: from molecular cloning to functional studies. Addiction Biol 1996; 1: Wang J-B, Johnson PS, Persico AM, Hawkins AL, Griffin CA, Uhl GR. Human mu opiate receptor: cdna and genomic clones, pharmacologic characterization and chromosomal assignment. FEBS Lett 1994; 338: Matthes HWD, Maldonado R, Simonin F, Valverde O, Slowe S, Kitchen I et al. Loss of morphine-induced analgesia, reward effect and withdrawal symptoms in mice lacking the -opioid-receptor gene. Nature 1996; 383: Sora I, Takahashi N, Funada M, Ujike H, Revay RS, Donovan DM et al. Opiate receptor knockout mice define mu receptor roles in endogenous nociceptive responses and morphine-induced analgesia. Proc Natl Acad Sci USA 1997; 94: Volpicelli JR, Alterman AI, Hayashida M, O Brien CP. Naltrexone in the treatment of alcohol dependence. Arch Gen Psychiatry 1992; 49: O Malley SS, Jaffe AJ, Chang G, Schottenfeld RS, Meyer RE, Rounsaville B. Naltrexone and coping skills therapy for alcohol dependence. A controlled study. Arch Gen Psychiatry 1992; 49: Berrettini WH, Ferraro TN, Alexander RC, Buchberg AM, Vogel WH. Quantitative trait loci mapping of three loci controlling morphine preference using inbred mouse strains. Nature Genet 1994; 7: Di Chiara G, Imperato A. Drugs abused by humans preferentially increase synaptic dopamine concentrations in the mesolimbic system of freely moving rats. Proc Natl Acad Sci USA 1988; 85: Tanda G, Pontieri FE, DiChiara G. Cannabinoid and heroin activation of mesolimbic dopamine transmission by a common opioid receptor mechanism. Science 1997; 276: Koob GF, LeMoal M. Drug abuse: hedonic homeostatic dysregulation. Science 1997; 278: Spanagel R, Herz A, Shippenberg TS. Opposing tonically active endogenous opioid systems modulate the mesolimbic dopaminergic pathway. Proc Natl Acad Sci USA 1992; 89: De Waele JP, Kiianmaa K, Gianoulakis C. Distribution of the and opioid binding sites in the brain of the alcohol-preferring AA and alcohol-avoiding ANA lines of rats. J Pharmacol Exp Ther 1995; 275: Berrettini WH, Hoehe MR, Ferraro TN, DeMaria PA, Gottheil E. Human mu opioid receptor gene polymorphisms and vulnerability to substance abuse. Addiction Biol 1997; 2: Bergen AW, Kokoszka J, Peterson R, Long JC, Virkkunen M, Linnoila M et al. opioid receptor gene variants: lack of association with alcohol dependence. Mol Psychiatry 1997; 2: Bond C, LaForge KS, Tian M, Melia D, Zhang S, Borg L et al. Singlenucleotide polymorphism in the human mu opioid receptor gene alters -endorphin binding and activity: possible implications for opiate addiction. Proc Natl Acad Sci USA 1998; 95: Kranzler HA, Gelernter J, O Malley S, Hernandez-Avila CA, Kaufman D. Association of alcohol or other drug dependence with alleles of the opioid receptor (OPRM1). Alc Clin Exp Res 1998; 22: Gelernter J, Kranzler H, Lacobelle J. Population studies of polymorphisms at loci of neuropsychiatric interest (tryptophan hydroxylase (TPH), dopamine transporter protein (SLC6A3), D3 dopamine receptor (DRD3), apolipoprotein E (APOE), opioid receptor (OPRM1), and ciliary neurotrophic factor (CNTF)). Genomics 1998; 52: Gelernter J, Kranzler H, Satel S. No association between D 2 dopamine receptor (DRD2) alleles or haplotypes and cocaine dependence or severity of cocaine dependence in European and African Americans. Biol Psychiatry 1999; 45: American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 3rd edn rev. American Psychiatric Press: Washington, DC, Spitzer RL, Williams JBW, Gibbon M, First MB. The structured Clinical Interview for DSM-III-R (SCID). I: History, rationale and description. Arch Gen Psychiatry 1992; 49: Blouin AG, Perez EL, Blouin JH. Computerized administration of the diagnostic interview schedule. Psychiatry Res 1988; 23: Cubells JF, Kobayashi K, Nagatsu T, Kidd KK, Kidd JR, Calafell F et al. Population genetics of a functional variant of the dopamine

8 -hydroxylase gene (DBH). Am J Med Genet (Neuropsych Genet) 1997; 74: Weir BS. Genetic Data Analysis II. Methods for Discrete Population Genetic Data. Sinauer Associates: Sunderland, MA, Lewis PO, Zaykin D. Genetic Data Analysis: Computer program for the analysis of allelic data. Version Free program distributed by the authors over the internet from the GDA Home Page at 25 Sham PC, Curtis D. Monte Carlo tests for associations between disease and alleles at highly polymorphic loci. Ann Hum Genet 1995; 59: Tsuang MT, Lyons MJ, Meyer JM, Doyle T, Eisen SA, Goldberg J et al. Co-occurrence of abuse of different drugs in men; the role of drug-specific and shared vulnerabilities. Arch Gen Psychiatry 1998; 55: Merikangas KR, Stolar M, Stevens DE, Goulet J, Preisig MA, Fenton B et al. Familial transmission of substance use disorders. Arch Gen Psychiatry 1998; 55: Bierut LJ, Dinwiddie SH, Begleiter H, Crowe RR, Hesselbrock V, Nurnberger JI et al. Familial transmission of substance dependence: alcohol, marijuana, and habitual smoking. Arch Gen Psychiatry 1998; 55: