Studies of the potential role of the dopamine D 1 receptor gene in addictive behaviors

Transcription

1 Molecular Psychiatry (1997) 2, Stockton Press All rights reserved /97 $12.00 ORIGINAL RESEARCH ARTICLE Studies of the potential role of the dopamine D 1 receptor gene in addictive behaviors DE Comings 1, R Gade 1,SWu 1, C Chiu 1, G Dietz 1, D Muhleman 1, G Saucier 2, L Ferry 3, RJ Rosenthal 4, HR Lesieur 5, LJ Rugle 6 and P MacMurray 7 1 Department of Medical Genetics, City of Hope Medical Center, Duarte, CA 91010; 2 Department of Psychology, California State University, San Bernardino, CA; 3 Jerry L Pettis VA Hospital, Loma Linda, CA; 4 Department of Psychiatry, University of California, Los Angeles; 5 Dept of Criminal Justice Sciences, Illinois State, Normal, IL; 6 Veterans Addiction Recovery Center, Dept Vet Affairs Med Center, Brecksville, OH; 7 Department of Psychiatry, Loma Linda University School of Medicine, Loma Linda, CA, USA Abnormalities in the dopaminergic reward pathways have frequently been implicated in substance abuse and addictive behaviors. Recent studies by Self and coworkers have suggested an important interaction between the dopamine D 1 and D 2 receptors in cocaine abuse. To test the hypothesis that the DRD1 gene might play a role in addictive behaviors we examined the alleles of the Dde I polymorphism in three independent groups of subjects with varying types of compulsive, addictive behaviors Tourette syndrome probands, smokers and pathological gamblers. In all three groups there was a significant increase in the frequency of homozygosity for the DRD1 Dde I 1 or 2 alleles in subjects with addictive behaviors. The DRD1 11 or 22 genotype was present in 41.3% of 63 controls and 57.3% of 227 TS probands (P = 0.024). When 23 quantitative traits were examined by ANOVA those carrying the 11 genotype consistently had the highest scores. Based on these results, we examined the prevalence of the 11 genotype in controls, TS probands without a specific behavior, and TS probands with a specific behavior. There was a progressive, linear increase, significant at for scores for gambling, alcohol use and compulsive shopping. Problems with three additional behaviors, drug use, compulsive eating and smoking were significant at All six variables were related to addictive behaviors. In a totally separate group of controls and individuals attending a smoking cessation clinic, and smoking at least one pack per day, 39.3% of the controls versus 66.1% of the smokers carried the 11 or 22 genotype (P = ). In a third independent group of pathological gamblers, 55.8% carried the 11 or 22 genotype (P = vs the combined controls). In the TS group and smokers there was a significant additive effect of the DRD1 and DRD2 genes. The results for both the DRD1 and DRD2 genes, which have opposing effects on cyclic AMP, were consistent with negative and positive heterosis, respectively. These results support a role for genetic variants of the DRD1 gene in some addictive behaviors, and an interaction of genetic variants at the DRD1 and DRD2 genes. Keywords: alcoholism; drug abuse; smoking; dopamine; reward; Tourette syndrome; heterosis Addictive behaviors exact an enormous cost on society. have implicated a role of the DRD2 gene in a wide Dopaminergic neurons, especially of the mesolimbic range of addictive, impulsive, compulsive disorders system and reward pathways, have frequently been including drug abuse, pathological gambling, smoking, implicated in the etiology of alcoholism, drug abuse eating, ADHD, Tourette syndrome (TS) and conduct and other addictive behaviors. 1 6 In 1990 Blum et al 7 disorder. 13 reported a significant increase in the frequency of the In contrast to the extensive studies of the DRD2 gene, Taq I A1 allele of the dopamine D 2 receptor (DRD2) there have been no similar studies of the DRD1 gene. gene in severe alcoholics compared to controls. 7 While Sequencing of the DRD1 gene in controls and individ- a number of subsequent studies in ambulatory uals with schizophrenia, manic-depressive disorder alcoholics failed to confirm this association, 8,9 others and alcoholism, has failed to identify exon mutations have, 10,11 and a summary of many studies completed that produce an effect on the phenotype Linkage to date suggest an association between the D 2 A1 allele studies have been negative both for schizophrenia 17 and some forms of severe alcoholism. 12,13 Other studies and TS. 18 Studies by Goldman-Rakic and colleagues 19 have implicated a role of the D 1 receptors in frontal cortex memory, suggesting that among other effects, Correspondence: DE Comings, Department of Medical Genetics, genetic variants of the DRD1 receptors could play a role City of Hope Medical Center, Duarte, CA 91010, USA Received 12 August 1996; revised 30 September 1996; accepted in learning disorders. This is supported by the results 11 October 1996 of Chen et al 20 showing that D 1 receptor activation

2 enhanced long-term depression in rats while D 2 recep- the City of Hope Medical Center. All meet DSM-IV criteria tor activation inhibited LTD. Like long-term potentiation, for TS and all were personally interviewed by LTD, which is induced by low-frequency stimu- DEC. The controls for the TS group consisted of adopting lation, is also involved in memory storage. 21 and step parents of TS probands, subjects with nonlation, Self et al 22 reported on the interaction of the D 1 and psychiatric disorders from other clinics at the City of D 2 receptors in cocaine-seeking behavior in rats. Activation Hope, and professional and non-professional hospital of mesolimbic D 2 receptors enhanced the staff from the City of Hope Medical Center. Both the relapse into cocaine-seeking behavior, after exposure TS subjects and the controls have been described in to cocaine. By contrast, activation of mesolimbic D 1 detail elsewhere. 32,33,37,39,40 receptors inhibited the relapse into cocaine-seeking behavior, after exposure to cocaine. This suggested D 1 Behavioral scores Each control and TS proband or receptor agonists could play a role in the treatment of relative was required to fill out a questionnaire based cocaine addiction. This priming of the brain s reward on the Diagnostic Interview Schedule 41 or DSM-III-R 42 system by D 2 agonists, and inhibition of this priming criteria for a range of disorders. The symptoms were by D 1 agonists, illustrates the important interaction of grouped into 23 different quantitative variables these two receptors in addictive behaviors. assessing the number of symptoms relating to attention All five of the dopamine receptor genes are part of deficit hyperactivity disorder (ADHD) (two scores), a superfamily of G-protein-coupled receptors, having alcohol, drugs, obsessive compulsive behaviors, learn- seven membrane-spanning domains. 23 The dopamine ing disorders, reading problems, gambling, manic receptors can be placed into two subgroups D 1 and symptoms, phobias, panic attacks, oppositional defiant D 5 that stimulate the production of cyclic AMP, and behavior, conduct disorder, academic problems in D 2, D 3 and D 4 that inhibit the production of cyclic grade school, smoking, sexual behaviors, schizoid AMP. Many reports have emphasized the interaction behavior, somatization, depression, sleep disorders, between the D 1 and D 2 receptors in a wide range of general anxiety, stuttering, and tics. The questions behaviors including schizophrenia, cataplexy, sub- used for these behavioral scores have been described stance abuse, and other behaviors These reports in in detail elsewhere. 32,35,37,39 41,43 The rationale for conjunction with the results of Self et al 22 suggest the examining comorbid behaviors is the prior observation need to test several hypotheses: 1) Are genetic variants that certain genes may be more strongly associated of the DRD1 gene associated with susceptibility to with some of the comorbid behaviors present in TS substance abuse? 2) Are genetic variants of the DRD1 than with the diagnosis per se. 39 This questionnaire is gene associated with addictive behaviors in general? not meant to provide DSM-IIIR or DSM-IV diagnoses 3) Are genetic variants of the DRD1 gene associated but rather to provide a highly structured method of with a broader spectrum of other behavioral disorders? assessing the presence or absence of symptoms in different and 4) In some or all of the above, is there a additive areas of behavior. The accuracy, utility and senand effect of genetic variants of the DRD1 and DRD2 genes? sitivity of a questionnaire-based approach to symptom To attempt to minimize the risk that positive findings evaluation has been demonstrated by others, 44,45 by might be due to a fortuitous random association, we comparing the use of such an instrument to an interviewer have examined three independent groups of subjects: administration of the same structured instruhave I) Tourette syndrome probands and their relatives; ment. Our review of the questionnaires with many II) a group of smokers attending a smoking cessation hundreds of subjects has indicated they accurately clinic; 30 and III) pathological gamblers. 31 TS probands reflect the information obtained by personal interview. were chosen based on our prior studies showing the Some of the symptoms that were especially relevant presence of a high frequency of addictive behaviors to the present study related to tobacco, alcohol and (alcohol and drug abuse, pathological gambling, sexual drug use, compulsive eating, and gambling. The alcobehaviors, compulsive eating, and others) in this hol score consisted of the summation of no or yes group answers to 18 questions derived from the MAST test Methods for alcohol use. 33,38 The drug score was based on no or yes answers to nine questions based on the Diagnostic Interview Schedule 32,41 concerning drug The three groups examined were the Tourette syndrome abuse/dependence. The variable for smoking was (TS) group, the Smoking Cessation group, and based on the question, Have you ever smoked ciga- the Pathological Gambling group. The subjects in all rettes, cigars or a pipe daily for more than a month or three groups were restricted to non-hispanic Cau- more where yes was scored as 1 and no as 0. The casians. variable for shopping score was based on the sum- mation of responses to the following questions: Have I) The TS group you ever bought more items than you really needed to This group included controls without alcohol or drug buy to meet your needs? Have you ever gotten into financial abuse, TS probands most of whom were severely affected trouble because of buying more things than you with multiple associated behavioral disorders, 37,38 could afford? Have you ever run up a total balance on and relatives of TS probands. The TS subjects and their all your credit cards that was greater than your net relatives came from the Tourette syndrome Clinic at monthly income? Have you ever shopped to fill a feel- 45

3 46 ing of emptiness? Do you ever shop to get a feeling of specific behaviors were present or absent in the controls. happiness? Have you ever taken things without paying For the TS groups, the presence or absence of a for them? The no responses were scored as 0, the behavior in both controls and subjects was based on occasionally responses as 1, and the often responses the dichotomous breakpoints as described in previous as 2. The gambling score was derived from nine yes studies. 37 The controls without a given behavior were or no questions relevant to the severity of involvement termed controls without. The same dichotomous break- in gambling described previously as the Gam- points allowed the TS probands and their relatives to bling Score. 31 The evaluation of compulsive eating was be divided into two groups, those without the specific based on yes or no responses to the question. Did behavior being examined, and those with that you ever consider yourself a compulsive eater? behavior. These groups were termed cases without and cases with and formed the second and third group. The II) Smoking cessation group a priori hypothesis was that if there was a significant The second group consisted of individuals attending a association between a given genotype and the behavior smoking cessation clinic at Loma Linda University. in question, there should be a progressive increase in These subjects and their own independent set of con- the frequency of this genotype across these three trols have been described in more detail in a previous groups. The use of the cases without and cases with study of the role of the DRD2 gene in smoking. 30 Here groups controlled for the possibility that the frequency the variable for smoking assessment was the average of that genotype might be increased in the TS subjects number of packs of cigarettes smoked per day. The con- because it was associated with a behavior other than trols were screened to exclude all types of substance the one being examined. Since we assumed there abuse including alcohol, tobacco, and other drugs. would be a progressive increase in the frequency of the genotype across these three groups, the linear chisquare III) Pathological gamblers test (Mantel Haenszel test in the SPSS Statisti- The third group consisted of pathological gamblers cal Package) was used. To assure that the results partially derived from a prior study of the role of the DRD2 gene match a linear increase across the three groups in pathological gambling. The details of patients ascertainment, and assessment have been described in detail elsewhere. 31 we also required that the frequency of the genotype being tested be at least 20% higher in the cases with than in the cases without group. Genotyping c) Regression analysis As in our study of dopami- To examine the DRD1 gene we utilized the Dde I poly- nergic genes in TS, 39 we found regression analysis to morphism consisting of an A to G change in the 5 UTR, be helpful in determining the percent of the variance tested by the PCR procedure described by Cichon et of the different behavior scores accounted for by a specific al. 46 The marker for the DRD2 gene was the Taq I gene. This provided r, and r 2 provided the fraction A1/A2 polymorphism. 47 of the variance of a given quantitative trait that was accounted for by that gene. Statistical analysis The allele groups were examined in the following Correction for multiple analyses Since 23 different ways: behaviors were examined in the TS group it was neces- sary to make an adjustment in the level of significance. a) ANOVA for means of behavioral scores versus genotype While an of 0.05 would be too liberal, since these In the TS groups, the means of the behavioral were exploratory studies, an of 0.05/23 or was scores were compared for subjects with different genotypes considered too conservative. We thus chose an inter- using the ANOVA statistical programs from the mediate of 0.05/10 or Since the study of the SPSS Statistical Package (SPSS, Inc, Chicago, IL, USA). Smoking Cessation group using the a priori hypotheses A Tukey analysis tested for significant individual dif- based on the results in the TS group, and involved the ferences between any of the individual groups when examination of a single variable, packs smoked per more than two groups were examined. In situations day, an of 0.05 was used. Finally, since the examination where a progressive increase in the means of different of the pathological gamblers only involved the scores across different genotypes were expected, the comparison of genotype frequencies, an of 0.05 was linear ANOVA was used by setting the subcommand used. of polynomial = 1 in the SPSS Statistical Package. b) Linear chi-square analysis Based on the results of Results the ANOVA analyses, chi-square analyses were carried I) The TS group out comparing the frequency of the genotype associated The allele frequency for the Dde I polymorphism for with the highest mean behavioral scores in the fol- the controls (n = 63) was 0.34 for the 1 allele, and 0.66 lowing three different groups. The first consisted of for the 2 allele. The allele frequency for the TS pro- controls without the behavior being examined. Since bands (n = 227) was 0.37 for the 1 allele and 0.63 for the controls were also required to fill out one or more the 2 allele. These were not significantly different questionnaires, it was possible to determine whether 2 = 0.43, d.f. = 1, P = The distribution of the geno-

4 types for the Dde I polymorphism for the controls and heterozygotes, intermediate scores for 22 homozygotes, TS group are shown in Table 1A. The 11 genotype was and the lowest scores for 11 homozygotes. Based on present in 4.9% of the controls and 17.5% of the TS these studies, in the present study we have examined group, 2 = 3.75, d.f. = 1, P = The differences in the percentage of 12 heterozygotes in the three subject the percent of those carrying either the 11 or the 22 groups. These results for the TS group are shown in genotype was 41.3% in controls and 57.3% in TS probands. Table 1B. Of 73 controls, 19.2% were 12 heterozygotes, This was significant, 2 = 5.08, d.f. = 1, while of 345 TS probands, 35.3% were heterozygotes, P = = 7.19, d.f. = 1, P = In prior studies of the Taq I A1/A2 alleles of the The interaction between the DRD1 and the DRD2 DRD2 gene, independent of the subjects examined genes was tested by examining the percentage of subhere, we have consistently observed that a wide range jects that were homozygous for the DRD1 Dde I11or of quantitative scores show the highest scores for alleles and heterozygous for the DRD2 Taq I A1/A2 47 Table 1 Allele and genotype frequencies in all three subject groups N P q % 11 % 11,22 (A) DRD1 Dde I I. Tourette syndrome group Controls TS probands a 57.3 b II. Smoking cessation group Controls Smokers c 66.1 d III. Gambling group Gamblers e 55.8 f Total controls Total subjects g 59.6 h (B) DRD2 Taq I I. Tourette syndrome group Controls TS probands i II. Smoking cessation group Controls Smokers j III. Gambling group Gamblers k Total controls Total subjects l N neither either both % both (C) DRD1 Dde I 11, 22 plus DRD2 Taq I12 I. Tourette syndrome group Controls TS probands m II. Smoking cessation group Controls Smokers n III. Gambling group Gamblers Total controls Total subjects o a 2 = 3.75; P = 0.053; b 2 = 5.08; P = 0.024; c 2 = 5.88; P = 0.015; d 2 = 13.45; P = ; e 2 = 5.39; P = (gamblers vs total controls); f 2 = 6.75; P = (gamblers vs total controls); g 2 = 8.81; P = 0.003; h 2 = 15.38; P = ; i 2 = 7.19; P = ; j 2 = 5.69; P = 0.017; k 2 = 18.61; P = (gamblers vs total controls); l 2 = 15.26; P = ; m 2 = 5.71; P = 0.016; n 2 = 8.25; P = ; o 2 = 8.63; P = Neither: DRD1 12 and DRD2 11, 22. Either: DRD1 12 and DRD2 12, or DRD1 11, 22 and DRD2 11, 12. Both: DRD1 11, 22 and DRD2 12.

5 48 Table 2 DRD1 gene in addictive behaviors DRD1 results using ANOVA in the TS group Behavior Allele groups F-ratio a P n = 43 n = 136 n = 125 M s.d. M s.d. M s.d. Alcohol a Smoking a Comp. eating Tics Gambling Maj.dep.epi Shopping a Significantly less than genotype 11 at = 0.05 by the Tukey test. alleles. The respective figures were 17.9% for the con- lence of the 11 genotype increased from 4.6% for the trols and 33.2% for the TS probands, 2 = 5.71, controls without gambling problems to 15.5% for the d.f. = 1, P = cases without gambling problems, to 33.3% for the The association between the Dde I genotype and 23 cases with gambling problems (P = ). When an quantitative trait variables was examined by ANOVA of 0.05 was used, the three additional variables, (Table 2). While none were significant at = 0.005, the drug use, compulsive eating, and smoking, were all P value for the alcohol score was Of the seven related to addictive behaviors. scores with a P value of 0.20, five were related to Univariate regression analysis (Table 4A) was sig- addictive behaviors alcohol use, smoking, compul- nificant at P for two behaviors gambling and sive eating, gambling, and shopping. Those with the 11 alcohol use. With an of 0.05, the four additional genotype had the highest scores. For example, for the variables were compulsive eating, smoking, tics and alcohol use variable, those with the 11 genotype had a reading. The next most significant behavior was shop- mean score of 1.30, compared to 0.23 for those with ping. Based on r 2 for the Dde polymorphism, the DRD1 the 12 genotype, and 0.42 for those with the 22 genotype. gene contributed to 3.6% of the variance of the gam- Of the total of 23 variables, those carrying the 11 bling score, 2.8% of the alcohol score, 1.9% of the com- genotype had the highest means for all variables except pulsive eating score, and 1.6% of the smoking score. learning disorders and somatization. Prior to examining the potential interaction of the On the basis of these results, we then examined DRD1 and DRD2 gene by multivariate regression analywhether there was progressive increase in the fre- sis, we first examined the effect of the DRD2 gene, quency of the 11 genotype for different behaviors based on the Taq I A1/A2 polymorphism, using univar- across the three groups of controls without, cases with- iate regression analysis. Subjects carrying the DRD2 11 out, and cases with (Table 3). There were three and 22 homozygotes were scored as 1, and 12 hetero- behaviors where the linear increase was significant at zygotes were scored as 2. At = the DRD2 gene 0.005, and where the frequency of the 11 genotype was significantly associated with oppositional defiant was at least 20% higher in the cases with than in the behavior, conduct disorder, compulsive eating, smoking, cases without group. These were, in order of significance, gambling and ADHD (Table 4B). At 0.05 gambling, alcohol use, and compulsive shop- additional variables were mania, stuttering, obsessivecompulsive, ping. For example, for the gambling score the preva- and schizoid behaviors. These were simi- Table 3 Chi-square analysis for the DRD1 Dde 11 genotype in the Tourette syndrome group Behavior Controls Cases Cases 2 P without without with n % n % n % Gambling Alcohol Shopping Drugs Eating Smoking

6 Table 4 Univariate regression analysis for the DRD1 Dde I significant for both the DRD1 and the DRD2 gene for 11 genotype and different behavioral scores in the Tourette gambling, compulsive eating, and smoking (Table 4C). syndrome group For each of these, the results were additive such that when combined the DRD1 and DRD2 genes accounted Behavior r r 2 T P for 5.9 to 4.8 percent of the variance of these scores. The alcohol score was included to show that the 11 (A) DRD1 Dde I (12,22 = 1, 11 = 2) genotype of the DRD1 gene had a significant effect, Gambling while in this group the effect of the DRD2 A1 allele Alcohol was not significant. Compulsive eating The additive effect of the DRD1 and DRD2 genes was Smoking also examined using univariate regression analysis Tics Reading where those with both the DRD1 Dde 11 and the DRD2 Shopping Taq A12 genotype were scored as 3, those with either (B) DRD2 Taq I (11,22 = 1, 12 = 2) genotype were scored as 2, and those with neither Oppositional defiant genotype were scored as 1 (Table 4D). Those variables Conduct disorder that were significant at were in order, gam- Compulsive eating bling, smoking, compulsive eating, oppositional Smoking defiant, ADHD, conduct disorder, obsessive-compul- Gambling sive, mania and alcohol use. ADHD The additive effect of the DRD1 and DRD2 genes was Mania also examined using linear ANOVA for three groups Stuttering Obsess-comp consisting of those that were DRD1 Dde I12or22and Schizoid DRD2 Taq I11or22(neither); DRD1 Dde I11orDRD2 (C) DRD1 and DRD2 Taq I12(either); or DRD1 Dde 11 and DRD2 Taq 12 Gambling (both) (Table 5). At 0.005, this showed a significant DRD progressive increase in mean scores from the neither, DRD to the either, to the both groups for the variables com- F pulsive eating, smoking, oppositional defiant, ADHD, Compulsive eating conduct disorder, obsessive compulsive, mania, and DRD alcohol behavioral variables. DRD The additive effect of the DRD1 and DRD2 gene was F Smoking also examined using linear chi-square for the frequency DRD of the presence of either the DRD1 Dde 11 genotype, or DRD the DRD2 Taq A12 genotype, or both (Table 6). Those F variables that were significant at 0.005, and where Alcohol the mean for the cases with group was at least 20% DRD higher than for the cases without group were in order: DRD mania, alcohol, obsessive-compulsive, conduct dis- F order, schizoid, sexual, compulsive eating and major (D) Both DRD1 11 and depressive episode. The results for the variables alco- DRD2 12 = 3; either = 2, hol use, compulsive eating and mania are diagrammed neither = 1(n= 293) Gambling in Figure 1. Smoking Since there was a significant difference between the Compulsive eating controls and TS probands in the percentage of subjects Oppositional def carrying the DRD1 Dde I 11 or 22 genotype, we also ADHD examined the percentage of subjects carrying the 11 or Conduct disorder genotype across the three groups controls without, Obsess-comp cases without, and cases with. The following variables Mania were significant at 0.05 and had a mean for the Alcohol use cases with that was at least 20% higher than for cases Tics without gambling, alcohol use, and grade school Schizoid problems. 49 II) Smoking cessation group lar to prior results. 39 Based on the r 2 values, heterozygosity The frequency of the Dde I 1 allele in the 61 smoking for the Taq A1 allele accounted for 4.2% of controls was 0.35, and in the 177 smokers it was 0.34 the variance of the oppositional defiant score, 3.8% of (Table 1A). Among controls, 4.9% carried the 11 geno- the variance of the conduct disorder score, 4.1% of the type vs 17.5% of the smokers, 2 = 5.88, d.f. = 1, eating score, 3.3% of the smoking score and 2.9% of P = Among the controls 39.3% carried the 11 or the gambling score. the 22 genotype vs 66.1% of the smokers, 2 = 13.45, Multivariate regression analysis showed that r was d.f. = 1, P =

7 50 Table 5 DRD1 + DRD2 results using ANOVA in the TS group. (neither = DRD1 12,22 and DRD2 11,22; either = DRD1 11 or DRD2 12; both = DRD1 11 and DRD2 12) Behavior Allele groups F- P ratio a neither either both n = 168 n = 112 n = 15 M s.d. M s.d. M s.d. Comp. eating 0.87 d Smoking 0.73 d Oppositional def b ADHD b Conduct dis b Obsess-comp b Mania Alcohol 0.24 c c a linear ANOVA b Significantly lower than either at = 0.05 by Tukey test. c Significantly lower than both at = 0.05 by Tukey test. d Significantly lower than either or both at = 0.05 by Tukey test. Table 6 Chi-square analysis for the percent of subjects with either DRD1 11 or DRD2 12 genotype, or both, in the Tourette syndrome group Behavior Controls Cases Cases with 2 P without without n % n % n % Mania Alcohol Obsess-comp Conduct dis Schizoid Sexual Comp. eating MDE Only subjects where cases with was at least 20% greater than cases without are shown. For the DRD2 gene, among the controls 26.2% car- 18.0% across these three groups, 2 = 5.14, P = ried the 12 genotype vs 42.8% of the smokers, 2 = 5.69, The percentage of subjects carrying either the 11 or the d.f. = 1, P = For the controls, 24.1% carried both 22 genotype increased from 39.3%, to 61.8%, to 68.0% the DRD1 11 or 22 genotype and the DRD2 12 genotype, across these three groups, 2 = 12.87, P = The vs 45.5% for the smokers, 2 = 8.25, d.f. = 1, P = percent of subjects heterozygous for the DRD2 Taq I The variable examined in the smoking cessation A1/A2 polymorphism increased from 26.2%, to 34.5%, group was packs smoked per day. Since all the controls to 46.3% across these three groups, 2 = 7.99, had completed the behavioral questionnaire, which P = The percent of subjects that were both included the same questions about smoking used in the homozygous for the DRD1 11 or 22 and heterozygous TS group, it was possible to exclude those controls that for the DRD2 Taq I A1/A2 alleles, increased from 24.1, had ever smoked cigarettes, cigars or a pipe, as well as to 34.5, to 50.4 across these three groups, 2 = 23.48, those with drug or alcohol abuse. These individuals P = (see Figure 2). constituted the 0 packs/day group. The subjects in the The interaction of the DRD1 and DRD2 genes for smoking cessation group were divided into those who smoking was also examined using multivariate linear smoked 1 to 1. and 2 to 2. packs per day. (Since only regression analysis (Table 8). This showed that the those smoking at least 1 pack per day were admitted DRD1 and DRD2 gene, as marked by these polymorto the study, there were no subjects smoking less than phisms, each contributed about equally. Combined 1 pack per day.) The results for the DRD1 gene are they accounted for 10.5% of the variance of the shown in Table 7. The percentage of subjects carrying packs/day variable. the 11 genotype increased from 4.9% to 16.4% to

8 51 Figure 1 Percent of TS group subjects with either the DRD1 Dde I 11 genotype or the DRD2 Taq I 12 genotype, or both, in controls without the behavior (controls without), TS probands or relatives without the behavior (cases without), and TS probands or relatives with the behavior (cases with). Figure 2 Percent of smokers with both the DRD1 Dde I11 or 22 genotype and the DRD2 Taq I 12 genotype by packs of cigarettes smoked per day. Table 7 DRD1 and DRD2 in smoking cessation group Number (%) Packs smoked per day Genotype Total 2 P DRD (4.9) 9 (16.4) 22 (18.0) , Total a DRD1 11,22 24 (39.3) 34 (61.8) 83 (68.0) Total a (26.2) 20 (34.5) 63 (46.3) , DRD2 Total a Neither Either DRD1 and Both 14 (24.1) 19 (34.5) 61 (50.4) 94 DRD2 Total b a Linear chi-square d.f. = 1. b Pearson chi-square d.f. = 4. III) Pathological gamblers Unlike the TS and the smoking cessation group, the Table 8 Multivariate linear regression analysis of DRD1 and pathological gambling group did not have its own set DRD2 genes in smoking cessation group of controls. Thus, for comparative purposes the TS and smoking cessation controls were combined to form the r r 2 T P Total control group, and this was used for the pathological gamblers. For the gamblers, 14.1% were homo- DRD1 Dde I zygous for the DRD1 Dde I 1 allele, 2 = 5.39, P = 0.020, DRD2 Taq I and 55.8% were homozygous for either the 11 or the 22 genotype, 2 = 6.75, P = For the DRD2 gene, F

9 % of the gamblers carried the Taq I 12 group, studies of numerous quantitative traits in over 500 subjects, 2 = 18.61, P = Of the gamblers 23.3% carried independent of those shown here, have repeat- the DRD1 11 or 22 genotype and the DRD2 Taq I 12 edly shown that DRD2 Taq I A1/A2 heterozygotes have genotype. higher or more abnormal scores, than either the A1/A1 or A2/A2 homozygotes. Based on these results, in the Discussion present study we have examined the percentage of A1/A2. In the subject groups of TS probands, smokers, Despite the many studies indicating an interaction gamblers, and totals, the results were 35.3, 42.8, 45.7 between the dopamine D 1 and D 2 receptors in a range and 40.0 percent respectively. For the TS, smoker and of psychiatric disorders, there has been a paucity of total controls the results were 19.2, 26.2 and 22.4 per- studies examining the potential association between cent respectively. In all four groups the prevalence of genetic variants of the DRD1 gene and behavior, or the the A1/A2 heterozygotes was significantly higher in the potential interaction of genetic variants of the DRD1 subjects than in the controls with P values of , and DRD2 genes. Based in part on the recent studies 0.017, and In three of the four groups, of Self et al, 22 and on the numerous studies implicating the results were still significant with a Bonferroni corrected a role of the DRD2 gene in addictive, impulsive 0.05/4 or behaviors, 13,48 we were interested in the possibility that The above results indicated a significant negative variants of the DRD1 gene, or an additive effect of the association with heterozygosity for the DRD1 alleles, DRD1 and DRD2 genes, might also play a role in these and a significant positive association with heterozygosity behaviors. We chose the DRD1 Dde I polymorphism 46 for the DRD2 alleles. To examine the potential interbehaviors. because it was a PCR-based test and the minor allele action of the DRD1 and DRD2 genes, we divided the was common in the general population. Since our previous cases into those who were not heterozygous for the studies of the DRD2 gene have shown the value DRD1 alleles (ie were 11 or 22 homozygotes) and were of examining the interactions of more than one gene, heterozygous for the DRD2 alleles. Since these geno- we examined the association between the genetic vari- types were optimized for both genes, this group was ants of the DRD1 gene alone, and the potential additive termed both. The second group, termed either, con- effects of the Taq I A polymorphism of the DRD2 gene. sisted of those that were either DRD1 11 or 22 homozygotes To minimize the effect of race we examined only non- or DRD2 heterozygotes. The third group, termed Hispanic Caucasians. To minimize the effects of neither, were heterozygous at the DRD1 alleles, and 11 chance, we sought to cross replicate any findings by or 22 homozygotes for the DRD2 alleles. These results examining three different groups of subjects, two of are shown in Table 1C. The percentage of subjects that which had their own set of controls. were in the both group was significantly higher for the TS probands (33.2) vs the TS controls (17.9), P = 0.016; Allele and genotype frequencies The results of the significantly higher for the smokers (45.5) vs the smokallele and genotype frequencies are shown in Table 1A. ers controls (24.1), P = , and significantly higher For the DRD1 variant, the frequencies of the Dde I1 for the total subjects (34.3) vs the total controls (20.8), allele in controls and subjects was virtually identical. P = The percentage of subjects in the both For the three groups, TS probands, smokers and gamblers, group was not significantly increased for the gam- the frequency of the 1 allele was 0.37, 0.34 and blers (23.3) respectively. For the TS and smokers control groups the frequency was 0.34 and 0.35 respectively. TS group Since the initial exploratory studies of the Many reports of association studies in psychiatric disorders potential role of the DRD1 gene in behavior was perforquencies limit themselves to a comparison of gene fre- med on the TS group, and since more than one in controls versus subjects with a specific disorder. behavior was studied, these results will be presented A similar limitation would have suggested that in more detail. We first compared the means of the 23 the DRD1 gene played no role in any of these disorders. different groups of behaviors in the different DRD1 However, since the distribution of genotypes may be genotypes using ANOVA. Of the seven behaviors that different, despite a similarity of gene frequencies, we gave a P value of less than 0.2, five were addictive also examined genotype frequencies. This showed a behaviors alcohol use, smoking, compulsive eating, significant increase in the prevalence of the 11 genotype gambling and shopping. While the alcohol use variable in the smokers and gamblers, and borderline sig- was the most significant (P = ), none were sig- nificant (P = 0.053) increase in the TS probands. There nificant at (see Methods). All of the quanti- was a significant increase in homozygosity for either tative traits shown in Table 2, and 14 of the remaining allele for all four groups with P values of 0.024, , 16 variables not shown, had the highest means for the and for the TS probands, smokers, gam- 11 homozygotes. While the mean scores in the 22 blers and totals, respectively. The latter three were still homozygotes were often higher than in the 12 hetero- significant if a Bonferroni corrected of 0.5/4 or zygotes, the relative magnitude of the scores for the was used. versus the 22 heterozygotes indicated that in the TS The allele and genotype frequencies for the Taq A1 group of subjects, homozygosity for the 11 allele allele of the DRD2 gene are shown in Table 1B. As showed a greater association with elevated scores than stated in the Methods section, our prior unpublished homozygosity for the 22 allele.

10 Based on the ANOVA results, we examined the per- smoking. As discussed above, when taken as a group centage of subjects that carried the 11 genotype in the the smokers showed a significant increase in the preva- controls without, versus the cases without versus the lence of the DRD1 Dde I 11 genotype and the 11 or 22 cases with groups (Table 3). At 0.005, three genotype. To explore the relationship between the behaviors, gambling, alcohol use and shopping, were DRD1 gene and smoking in more detail, we examined all significant. For gambling, this percentage increased the quantitative trait of packs smoked per day (Table from 4.5 to 15.2 to 35.2 percent (P = ). All six 7). There was a progressive and significant increase in of the traits significant at 0.05 were addictive the percentage of subjects with the DRD1 11 genotype behaviors. of 4.9 to 16.4 to 18.0 percent across three groups of Similar results were obtained using univariate controls smoking 0 packs per day, and smokers using regression analysis where those who were homozygous 1 1. packs per day, and smokers using 2 2. packs per for the DRD1 Dde I 1 allele were scored as 2, and those day, P = In contrast to the TS group, for smokers with the 12 or 22 genotypes were scored as 1. The gam- the 11 or 22 homozygote group gave more significant bling and alcohol use variables were significant at results. Thus, the percentage of homozygous subjects (Table 4). With univariate regression analysis increased from 39.3 for the 0 packs per day controls, for the DRD2 gene, where A1/A2 heterozygotes were to 61.8 for the smokers using 1 1. packs per day, to scored as 2, and homozygotes as 1, seven different 68.0 for the smokers using 2 2. packs per day, traits were significant at (Table 4B). These P = The percentage of subjects heterozygous included three addictive behaviors, compulsive eating, for the DRD2 A1/A2 alleles increased from 26.2 to 34.5 gambling and smoking. Each of these behaviors has to 46.3 percent across these three groups, P = As previously been reported to be associated with the with the TS group, the effect of the DRD1 and DRD2 DRD2 gene 12,31,49 53 in subjects distinct from the TS genes was additive. Thus, the percentage of subjects in group. the both group increased from 24.1 to 34.5 to 50.4 Using multivariate regression analysis there were across these three groups, P = only three traits where both the DRD1 and the DRD2 To further examine the additive effect of the DRD1 gene gave significant results: gambling, compulsive eating, and DRD2 genes in smokers, we examined the packs and smoking (Table 4C). The same results were per day variable using multivariate regression analysis obtained using univariate regression analysis where (Table 8). There was a comparable effect of both genes, those in the both group were scored as 3, those in the and combined they accounted for 10.5 percent of the either group as 2, and those in the neither group as 1 variance of the packs per day variable. (Table 4D). Here the P values for gambling, smoking and compulsive eating were Gamblers As discussed above, there was a significant To evaluate the magnitude of the scores, we examined increase in the percentage of subjects that were DRD1 the means for those in the neither, either or both 11 homozygotes, or 11 or 22 homozygotes in gamblers groups by ANOVA (Table 5). Here, eight traits, including compared to total controls, and a significant increase compulsive eating, smoking, and alcohol use were in the percentage of subjects that were DRD2 A1/A2 significant at The means were consistently heterozygotes. Unlike the TS and the smokers groups, highest in the both group. The final test was an examin- these effects were not additive in the gamblers since ation of the percentage of subjects that were either the percentage of subjects in the both group was not DRD1 11 homozygotes, or DRD1 A1/A2 heterozygotes, increased over the total controls. or both, across the controls without, cases without and cases with groups (Table 6). Alcohol use and compul- Heterozygosity The concept that heterosis or heterozygosity sive eating were among the eight traits significant at vs homozygosity (12 vs ) for specific For alcohol use the percentage increased from alleles is more relevant than recessive inheritance ( to 46.9 to 70.6 across the three groups, vs ) or dominant inheritance ( vs 22) P = is not new. In a paper entitled Association between Combined these results were consistent with a role schizophrenia and homozygosity at the dopamine D3 of the DRD1 gene in a number of addictive and other receptor gene Crocq et al 54 reported a decrease in 12 behaviors, and with an additive effect of genetic variants heterozygosity for the Bal I(Msc I) polymorphism, and at the DRD1 and DRD2 genes. While both homo- an increase in 11 and 22 homozygosity in subjects with zygosity for both the DRD1 1 and the 2 alleles gave schizophrenia vs controls. Some studies have sup- higher scores than 12 heterozygosity, in the TS group, homozygosity for the 1 allele gave the strongest associations with a number of traits. Smoking cessation group To determine if we could replicate any of these findings in a totally different group of subjects, we utilized individuals from a prior study of the role of the DRD2 gene in smokers. 30 All subjects in this group smoked at least one pack of cigarettes per day, and had tried unsuccessfully to stop ported this finding, 55,56 while others have not. 57,58 We have observed a significant decrease in DRD3 Msc I12 heterozygosity in Tourette syndrome 59,60 and pathological gambling. 31 Our observations of the mean scores for a range of behaviors in a number of different subject groups, suggest that genetic variants at the DRD1 and DRD2 genes may also show heterosis. It was of interest that the effect was opposite in the two genes. Thus, the DRD2 gene Taq I A1/A2 heterozygotes had more abnormal scores than most variables, while for the 53

11 54 DRD1 gene heterozygotes had more normal scores. We dopamine, serotonin, cannabinoid, nitric oxide, nicotinic assume that both polymorphisms are in linkage muscarinic, GABA, and others. The diagnosis of disequilibrium with other mutations that affect the Tourette syndrome is dependent upon the presence of function of the DRD1 and DRD2 genes. 61 The mechanism motor tics, and dopamine plays a major role in the of action of this apparent heterozygous regulation of muscle movement. Thus, it would be advantage/disadvantage is unknown. It is also expected that comorbid substance abuse or other unknown whether the opposite effect in the two genes addictive behaviors in TS would be more likely to is due to the fact they have opposite effects on cyclic involve genetic defects of dopamine receptors, than in AMP, or simply due to chance variations in the types a group of subjects ascertained on the basis of any type of other mutations they are associated with, by linkage of substance abuse. disequilibrium. Examples of both positive and negative heterosis in humans have been described by Stern. 62 Importance of testing at a symptom rather than diagnostic level These and previous studies 39 suggest that Polygenic nature of addictive behaviors In a previous single dichotomous diagnostic categories may be so study we demonstrated the additive nature of variants broad that they result in a significant loss of power in of the DRD2, D H and DAT1 genes in TS, ADHD, conduct association studies. For example, TS subjects can range disorders, stuttering, and related behaviors. 39 The from individuals with a few mild tics and no other present results extend that principle to the role of the problems, to individuals with a devastating combi- DRD1 plus the DRD2 genes in addictive behaviors. We nation of tics, stuttering, ADHD, obsessive-compulsive, believe that most psychiatric disorders are polygenic 63 conduct, anxiety, mood, substance abuse and learning and that each gene accounts for less than 10% and usu- disorders. If a given gene contributes to TS but is ally less than 5% of the variance of a given behavioral especially associated with stuttering, and stuttering is variable. In both studies, the strength of the associations was increased by the examination of the additive effect of more than one gene. present in only 20% of the cases, the role of that gene could be missed in a comparison of controls vs all TS probands, but would probably be detected in a comparison of controls without stuttering vs TS probands with stuttering. This concept was of particular impor- The role of allele combinations While an integral part of polygenic inheritance is the critical role of combi- tance in the present study. In the TS group, the nations of genes, this study also illustrates the critical behavioral variables that were significantly associated effect of combinations of alleles. Thus, when only the with the DRD1 gene were those involving addictive frequency of the DRD1 Dde I 1 and 2 alleles was examined, behaviors. This was unlikely to be a chance finding there were no differences between the controls because animal studies also suggested a role of the and any of the subject groups. However, there were differences DRD1 in addictive traits, and because the finding was in how the two alleles were combined. Thus, replicated in three different groups of subjects. in all three subject groups, there was a significant tendency for the same alleles to occur together, ie increased Biological mechanisms Since the DRD1 Dde I polymorphism homozygosity (11 or 22), while in the controls the two is a neutral base change, we assume it is in different alleles tended to occur together, ie increased linkage disequilibrium with mutations in regions at or heterozygosity (12). The inverse trends occurred for the near the DRD1 locus that affect the dopamine D 1 recep- DRD2 Taq I A alleles with increased heterozygosity in tor density. the subjects and increased homozygosity in the controls. There are several caveats concerning the present study. A frequent concern of association studies is the possibility that the results may be due to a hidden racial or ethnic stratification, rather than the gene itself. The importance of subject ascertainment We were able to confirm a role of the DRD1 gene across three While we attempted to minimize this by restricting the independent sets of subjects, and the importance of the studies to non-hispanic Caucasians, and by replicating additive effect of the DRD1 and DRD2 genes in two the findings in three independent groups of subjects, independent sets of subjects. Our preliminary studies this remains a possible explanation. A second potential of subjects ascertained because the primary diagnosis problem, especially in the TS group, is the examination was alcoholism or drug addiction, have supported the of 23 different quantitative traits. This was compen- role of the DRD1 gene, and the additive effect of the sated for by the use of a very conservative of DRD1 and DRD2 genes in some, but not all types of When the combined effects of the DRD1 and DRD2 substance abuse. This is a not unexpected aspect of genes were examined, many of the P values were less polygenic inheritance. Since the DRD1 and DRD2 than 0.001, well below a full Bonferroni correction of genes individually accounted for less than 6% of the 0.05/23 or In addition, the clustering of the significant variance of a given trait, and combined accounted for results around addictive behaviors less than 11% of the variance of a trait, these moderate (alcoholism, compulsive eating, gambling, shopping, effects could be easily overwhelmed by differences in and smoking), provided some internal consistency. A subject ascertainment. For example, it is very likely final caveat is that homozygosity for the DRD1 11 allele that abnormalities of a number of receptors are was emphasized in the tables for the TS group, while involved in various types of substance abuse including 11 or 22 homozygosity was emphasized in the tables