Case-control and within-family tests for an association between conduct disorder and 5HTTLPR.код для вставкиСкачать
American Journal of Medical Genetics Part B (Neuropsychiatric Genetics) 141B:825 –832 (2006) Case-Control and Within-Family Tests for an Association Between Conduct Disorder and 5HTTLPR Joseph T. Sakai,1* Susan E. Young,2 Michael C. Stallings,2 David Timberlake,2 Andrew Smolen,2 Gary L. Stetler,2 and Thomas J. Crowley1 1 Division of Substance Dependence, Department of Psychiatry, University of Colorado School of Medicine, Denver, Colorado Institute for Behavioral Genetics, University of Colorado, Boulder, Colorado 2 Several lines of research have suggested that serotonin dysfunction is associated with aggression, impulsivity, and antisocial behavior. A functional polymorphism in the promoter region (s, short and l, long allele variant) of the serotonin transporter gene (SLC6A4) that results in decreased transcription of the serotonin transporter gene has been linked with such serotonin dysfunction. To test for an association between 5HTTLPR genotype and conduct disorder diagnosis/aggression. Analysis for association between 5HTTLPR and conduct disorder/aggression using a case-control design and the transmission disequilibrium test. Conduct-disordered adolescents, who were drawn from admissions to a program that treats adolescents with serious substance and behavior problems, and conductdisordered siblings of these patients (n, 297) were compared with non-conduct-disordered control adolescents and non-conduct-disordered siblings of these controls (n, 93). Second, using patient families where parental DNA was available, transmission disequilibrium tests were conducted for two phenotypes: (1) conduct disorder (74 trios), and (2) conduct disorder with at least one aggressive symptom (57 trios). Case-control analyses suggested a strong association between the ss genotype and conduct disorder (x22 ¼ 14.3; P < 0.01). Within-family analyses for conduct disorder with at least one aggressive symptom significantly favored greater transmission of the s-allele to affected offspring (xtdt2 ¼ 4.13; P ¼ 0.04); for conduct disorder, without aggressive symptoms, however, results were non-significant (xtdt2 ¼ 1.61; P ¼ 0.20). These data suggest that the s-allele may confer some risk for aggressive behavior or may be in linkage disequilibrium with such an allele. ß 2006 Wiley-Liss, Inc. KEY WORDS: aggression; serotonin; serotonin transporter Please cite this article as follows: Sakai JT, Young SE, Stallings MC, Timberlake D, Smolen A, Stetler GL, and Crowley TJ. 2006. Case-Control and Within-Family Grant sponsor: NIDA; Grant numbers: DA 09842, DA 11015, DA 12845, K08DA016314; Grant sponsor: NIMH; Grant numbers: 5T32MH15442, MH01865. *Correspondence to: Dr. Joseph T. Sakai, 4200 East Ninth Ave., Box C268-35, Denver CO 80262. E-mail: firstname.lastname@example.org Received 7 April 2005; Accepted 27 October 2005 DOI 10.1002/ajmg.b.30278 ß 2006 Wiley-Liss, Inc. Tests for an Association Between Conduct Disorder and 5HTTLPR. Am J Med Genet Part B 141B:825–832. INTRODUCTION Conduct disorder is characterized by repetitive violation of the rights of others and violation of age-appropriate societal norms. It is common, associated with great morbidity [Crowley and Riggs, 1995], and better treatments for this disorder are needed. Because antisocial behavior and aggression are substantially heritable traits (accounting for about 40%–50% of the population variance) [Miles and Carey, 1997; Rhee and Wladman, 2002], genetic studies may offer one avenue to search for factors contributing to the liability toward this disorder. Serotonin dysfunction has long been implicated in impulsivity, aggression, and antisocial behavior. Low CSF (cerebrospinal fluid) levels of 5-HIAA (5-hydroxyindoleacetic acid), a serotonin metabolite, have been demonstrated in aggressive patients [Brown et al., 1979], violent offenders [Linnoila et al., 1983], oppositional defiant disordered or conduct-disordered adolescents versus controls [Van Goozen et al., 1999], and have predicted persistence of aggressive behavior in adolescent males with disruptive behavior disorders (62% had conduct disorder) [Kruesi et al., 1992]. Tests of serotonin responsiveness by fenfluramine administration have shown a blunted prolactin response in offenders with antisocial personality disorder [O’Keane et al., 1992]; similar results have been demonstrated in general population samples for males but not females [Manuck et al., 1998]. Measures of platelet serotonin binding offers one other peripheral measure of serotonin function; adolescents with conduct disorder have been shown to have significantly lower 3-H imipramine maximal binding sites on platelets compared with control children [Stoff et al., 1987]. The serotonin transporter is a high affinity transporter that removes serotonin from the presynaptic space, thus terminating serotonin-mediated neurotransmission. In humans a functional polymorphism in the promoter region of the serotonin transporter gene (SLC6A4) has been described (5HTTLPR) with a long (l) and short (s) variant. The short promoter variant is transcriptionally less active than the long allele resulting in a 30%–40% reduction of the serotonin transporter and a two-fold reduction in serotonin transport. In addition, imaging studies suggest that this promoter deletion alters brain activity in vivo [Heinz et al., 2000; Hariri et al., 2002]. Importantly, the s-allele has been associated with serotonin dysfunction, similar to that seen in impulsive-aggressive persons. The ss genotype has been associated with lower CSF 5-HIAA in Caucasians and males but not other groups [Williams et al., 2003]. The s-allele has been associated with blunted serotonin function as measured by prolactin release following fenfluramine administration among males [Reist et al., 2001] and with lower platelet serotonin binding availability [Stoltenberg et al., 2002]. Several case-control studies have tested for an association between 5HTTLPR genotype and aggressive behavior. Positive 826 Sakai et al. associations between the l-allele and violence or aggression have been reported [Twitchell et al., 2001; Zalsman et al., 2001]. In contrast, some studies have found an excess of the s-allele among alcoholic, incarcerated, violent offenders [Hallikainen et al., 1999], Chinese males convicted of violent crime [Liao et al., 2004], violent heroin dependent males in Italy [Gerra et al., 2004], and physically violent males referred for forensic evaluations in Germany [Retz et al., 2004]. One study found an association between the s-allele and externalizing behaviors in males; but the opposite result was found for females [Cadoret et al., 2003]. Three studies found no association between 5HTTLPR genotype and aggression [Reist et al., 2001; Beitchman et al., 2003; Davidge et al., 2004]. These seemingly contradictory findings may have been due to confounding factors such as population stratification or small sample sizes in the studies reporting null results. In order to examine the relationship between 5HTTLPR and aggression, using larger samples than previous studies, we conducted case-control and within-family analyses of 5HTTLPR genotypes and measures of aggression (conduct disorder diagnosis (CD), and CD with at least one aggressive symptom). MATERIALS AND METHODS The research protocol was approved by the Colorado Multiple Institutional Review Board, and all participants gave informed written consent. discriminative validity for conduct disorder diagnosis [Crowley et al., 2001]. Patients also completed the CARI (Colorado Adolescent Rearing Inventory), which provides quantitative information about severity of physical/sexual abuse and neglect and has shown discriminative validity and clinical utility in this population [Crowley et al., 2003]. Pre-determined phenotypes were (1) conduct disorder (defined as at least three lifetime DSM-IV conduct disorder symptoms), and (2) conduct disorder with at least one aggressive symptom. The second phenotype was included because adolescents may meet criteria for conduct disorder by committing rule violations and not aggressive acts. Aggressive symptoms were defined by the DSM-IV’s first seven criteria for conduct disorder, which are grouped under the heading ‘‘aggression to people and animals.’’ Genotype Determination A DNA sample was obtained by cheek swab buccal cell collection. 5HTTLPR genotypes were determined by modification of the protocols [Gelernter et al., 1999; Anchordoquy et al., 2003]. The availability of genotypic data on nuclear families allowed for the confirmation of allele calls via Mendelian inheritance verification; all samples were genotyped twice. The recently described single nucleotide polymorphism in the lallele [Hu et al., 2005], which is present at low frequency, was not evaluated in this population. Sample Data Analysis Patient and control families. These patient and control samples have been described previously [Miles et al., 1998]. Briefly, patients 13–19 years of age were recruited from a University-based treatment program; most were referred by the criminal justice system or social services. Patients and siblings were included in the case-control analyses if they met lifetime criteria for conduct disorder, defined here as meeting three lifetime conduct disorder criteria (though the impairment criteria was not required only 3 of the 297 cases did not meet this criteria). Cases consisted of 297 individuals with conduct disorder; 261 patients and 36 siblings of patients. The within-patient-family analyses included those with at least one affected offspring and at least one participating parent (74 offspring-mother-father trios and 101 proband-parent duos). Control adolescents were recruited to be similar on average for age, gender, ethnicity, and zip code of residence to adolescent patients [Miles et al., 1998]. For inclusion in these case-control analyses, controls, or control siblings were required to have never met criteria for conduct disorder (at least three lifetime conduct disorder symptoms) (n, 93; 48 controls and 45 siblings of controls). Patients, controls, and siblings were administered the same study instruments and DNA samples were collected from consenting patients, siblings, and parents. Adolescents from patient families meeting lifetime diagnostic criteria for conduct disorder were compared with nonconduct-disordered adolescents from control families. Differences in demographic characteristics and selected diagnoses were compared using w2 test for categorical variables and independent t-tests for continuous, approximately normal variables. Adolescents from patient and control families were compared by genotype frequencies, using the w2 test. Withinpatient-family analyses were carried out using the transmission disequilibrium test (TDT) [Spielman et al., 1993; Sham, 1998] for our two pre-determined phenotypes. Because both parental genotypes were not available for all 297 adolescents, these analyses included 74 trios (offspring-mother-father). To avoid a potential bias, trios with heterozygous offspring and parents were included in all TDT tests [Sham et al., 2000]. TDT tests were also completed including informative duos and all TDT tests were rerun excluding siblings. Although not part of our pre-hoc hypotheses, because 95% of conduct disordered sample met lifetime criteria for at least one lifetime substance use disorder, we conducted post-hoc analyses to test for a possible association between 5HTTLPR genotype and measures of conduct disorder and substance dependence severity. We utilized two phenotypes: (1) conduct disorder symptom count and (2) dependence vulnerability (DV), defined as the lifetime substance dependence symptom count summed across 10 drug categories, divided by the number of substances ever used (using the CIDI-SAM minimum threshold for use definition, which generally requires that a substance has been tried at least five times). Both phenotypes were standardized for age trends and sex differences using large community samples from Colorado as has been done previously [Stallings et al., 2005]. We tested whether these phenotypes varied by genotype within patients-only (n, 261) using ANOVA (ss vs. sl vs. ll) and t-tests (ss/sl vs. ll). To test for a possible gene-environment interaction, we used general linear modeling to test whether the slopes of the association between standardized, age-and-sex corrected conduct disorder symptom count and physical/sexual abuse and Assessments and Phenotype Selection Trained interviewers administered structured diagnostic instruments to all consenting subjects. Patients, siblings of patients, control adolescents, and their siblings were administered the CIDI-SAM (Composite International Diagnostic Interview–Substance Abuse Module) [Cottler et al., 1989] and the DISC (Diagnostic Interview Schedule for Children). The CIDI-SAM provides DSM-IV [American Psychiatric Association, 1994] diagnoses of substance abuse and dependence for 10 drug categories: alcohol, nicotine, hallucinogens, cocaine, amphetamines, opioids, PCP, sedatives, cannabis, and inhalants. Adolescent self-reporting on the DISC has shown Conduct Disorder and 5HTTLPR neglect differed across the three 5HTTLPR genotypes. If a 5HTTLPR by maltreatment interaction were present we would expect the slopes to differ significantly. 827 cents from control families (examining probands and siblings, and probands only) were in HWE; adolescents from patient families (examining probands and siblings, and probands only) were not. Relative Risk Required for About 80% Power We estimated the relative risk for disease conferred by possession of one or two s-alleles, assuming about 80% power to detect an association [Purcell et al., 2003]. Relative risks were estimated (see Table I) assuming an s-allele frequency of 0.40, disease prevalence 0.10, setting the linkage disequilibrium parameter to one, alpha to 0.05, a power of approximately 80%, and using 74 trios. RESULTS Demographic and Diagnostic Comparisons Table II shows females and Caucasians were over-represented in the adolescent controls. As would be expected from studies of comorbidity with conduct disorder [Crowley and Riggs, 1995], adolescents from patient families were significantly more likely to have met lifetime criteria for major depression, attention-deficit hyperactivity disorder, and oppositional defiant disorder. Interestingly, the two groups did not differ in the lifetime prevalence of generalized anxiety disorder. About 95% of conduct-disordered adolescents and 13% of controls met lifetime criteria for at least one substance use disorder. Case Control Analyses Table III shows a significant association of the ss genotype in patients with conduct disorder and conduct disorder with aggression. This association was found to be significant for the entire sample (w2 ¼ 14.3, 2df, P < 0.01), or when analyzed separately for Caucasians (w2 ¼ 7.0, 2df, P ¼ 0.03), Hispanics (w2 ¼ 8.3, 2df, P ¼ 0.02), and males only (w2 ¼ 10.5, 2df, P < 0.01). For females the association of the ss genotypes was found only in those with the aggressive form of conduct disorder (w2 ¼ 6.1, 2df, P ¼ 0.05). No gender by genotype frequency differences were seen in cases (w2 ¼ 0.22, 2df, P ¼ 0.89) or controls (w2 ¼ 0.19, 2df, P ¼ 0.91). A significant association was also seen within Caucasian-males for conduct disorder (124 cases, 39 controls; w2 ¼ 6.5, 1df, P ¼ 0.04) and conduct disorder with at least one aggressive symptom (106 cases; w2 ¼ 7.5, 1df, P ¼ 0.02). Excluding siblings, ss genotype frequency differences were similar to those seen in the overall sample but the difference was non-significant (w2 ¼ 4.1, 2df, P ¼ 0.13). Adolescent genotype frequencies were tested for Hardy– Weinberg equilibrium (HWE) within control and patient families. Regarding 5HTTLPR genotype frequencies, adolesTABLE I. Power Versus Relative Risk Dominant versus additive versus recessive Power s-allele dominant 83% s-allele additive 83% s-allele recessive 81% Relative risk sl ¼ 6 ss ¼ 6 sl ¼ 2 ss ¼ 4 sl ¼ 1 ss ¼ 3.4 Assumptions include: s-allele frequency of 0.40, disease prevalence 0.10, linkage disequilibrium parameter of one, alpha to 0.05. Using 74 trios, the table presents the relative risk for disease given genotype with approximately 80% power. Within Family Analyses Table IV shows transmission disequilibrium tests for 5HTTLPR and conduct disorder phenotypes. Although the s-allele was more often transmitted from heterozygous parents, the difference was non-significant (w2 ¼ 1.61, 1 df, P ¼ 0.20) for the 74 conduct disorder trios; however, for the 57 aggressive-conduct-disorder trios, the s-allele was significantly more likely to be transmitted to offspring from heterozygous parents (w2 ¼ 4.1, 1df, P ¼ 0.04). Examination of Informative Duos (Parent-Offspring) and Sibling Effects Because of a large number of missing fathers, both TDT analyses were also calculated using informative duos (parentoffspring). For all comparisons (Table IV, secondary analyses) examining conduct disorder with at least one aggressive symptom, the s-allele was found to be significantly more likely to be transmitted to affected adolescents. There is a potential bias from including several children from a single family in these analyses. However, TDT analyses excluding siblings also showed significantly greater transmission of the s-allele to cases from heterozygous parents (Table IV). Because we also had genotypic information on parents of controls we undertook secondary analyses utilizing these families (Table IV). We conducted TDT analyses for control adolescents (controls and their siblings) who had never met criteria for conduct disorder. Using the 63 available probandfather-mother trios, we found that the l-allele was significantly more likely to be transmitted from heterozygous parents to offspring; including informative duos did not affect the level of significance. Excluding siblings of controls we had small samples (28 trios and 11 duos) and within-family tests were non-significant. Post Hoc Analyses—Conduct Disorder Symptom Count and Dependence Vulnerability Means for standardized, age-and-sex corrected conduct disorder symptom count were calculated within each 5HTTLPR genotype for patients (n, 261). The means for ss, sl, ss/sl, and ll genotypes were 3.49, 3.40, 3.44, and 3.01, respectively. Comparisons across three genotypes (ss vs. sl vs. ll) were non-significant (F ¼ 1.70; P ¼ 0.18) as were comparisons between the ss/sl and ll genotypes (t259 ¼ 1.82; P ¼ 0.07). Means for standardized, age-and-sex corrected dependence vulnerability (DV) was calculated within each 5HTTLPR genotype for patients (n, 261). The means for ss, sl, ss/sl, and ll genotypes were 2.20, 2.00, 2.08, and 2.35, respectively. Comparisons across three genotypes were non-significant (F ¼ 1.74; P ¼ 0.18) as were comparisons between the ss/sl and ll genotypes (t259 ¼ 1.56; P ¼ 0.12). Gene-Environment Interaction Using the subsample of patients with available CARI scores (n, 197), we used general linear modeling to model standardized age-and-sex corrected conduct disorder symptom count by the total number of abuse and neglect items endorsed on the CARI within 5HTTLPR genotype. Slopes of the individual lines (within ss, sl, and ll genotypes) did not significantly differ (F ¼ 0.37; P ¼ 0.69); 5HTTLPR genotype was not associated with differences in abuse and neglect (F ¼ 0.08; P ¼ 0.93). 828 Sakai et al. TABLE II. Adolescents From Patient and Control Families: Demographic Variables and Psychiatric Diagnoses Patients and siblings with CD (n, 297) Age (standard deviation) Gender (male) Ethnicity White Hispanic Other Generalized anxietya disorder Major Depressiona ADHDa Conduct disorder symptom count Sum of aggressive CD symptoms Any substance use disorder Substance dependencea Any substance dependence Cannabis Nicotine Alcohol Amphetamine Hallucinogen Cocaine Controls and their siblings without CD (n, 93) 15.69 (1.35) 252 (84.8%) 15.24 (1.85) 64 (68.8%) 143 (48.1%) 118 (39.7%) 36 (12.1%) 18 (6.1%) 46 (15.5%) 59 (19.9%) 6.3 (2.48) 2.1 (1.54) 281 (94.6%) 61 (65.6%) 23 (24.7%) 9 (9.7%) 3 (3.2%) 6 (6.5%) 0 (0%) 0.7 (0.74) 0.08 (0.28) 12 (12.9%) 216 (72.7%) 162 (54.5%) 134 (45.1%) 83 (27.9%) 40 (13.5%) 40 (13.5%) 26 (8.8%) 2 (2.2%) 1 (1.1%) 1 (1.1%) 1 (1.1%) 0 (0) 0 (0) 0 (0) Statistic t124.3 ¼ 2.2* w2 ¼ 11.8* w2 ¼ 8.8* w2 ¼ 1.1 w2 ¼ 5.0* w2 ¼ 21.8* t387.6 ¼ 34.2* t349.7 ¼ 21.2* w2 ¼ 253.1* w2 ¼ 143.1* w2 ¼ 83.3* w2 ¼ 60.7* w2 ¼ 30.3* w2 ¼ 14.0* w2 ¼ 14.0* w2 ¼ 8.7* a DSM-IV diagnosis (lifetime). *P < 0.05; , CD, at least three whole life conduct disorder symptoms. DISCUSSION This study using two methods (case-control and withinfamily analyses) suggests that 5HTTLPR (s-allele) contributes to the liability to conduct disorder and aggressive behavior. This work is consistent with other studies indicating that this variant in the promoter region is a functional variant that alters serotonin-mediated behavior. Although our case-control analyses (even with ethnic subgroup analyses) leave open the possibility of bias from population stratification, the withinfamily analyses, which eliminate the possibility of this bias, support the case-control results. Interestingly, some of our TABLE III. Case Versus Control Association: 5HTTLPR Genotype Primary analyses Control (no CD) Case (CD) Case (CD þ 1) Secondary analyses Caucasians-only Control (no CD) (n ¼ 61) Case (CD) (n ¼ 143) Case (CD þ 1) (n ¼ 114) Hispanics-only Control (no CD) (n ¼ 23) Case (CD) (n ¼ 118) Case (CD þ 1) (n ¼ 102) Females-only Control (no CD) (n ¼ 29) Case (CD) (n ¼ 45) Case (CD þ 1) (n ¼ 31) Males-only Control (no CD) (n ¼ 64) Case (CD) (n ¼ 252) Case (CD þ 1) (n ¼ 215) Probands-only Control (no CD) (n ¼ 48) Case (CD) (n ¼ 261) Case (CD þ 1) (n ¼ 219) ss sl ll Statistic 8 (8.6%) 72 (24.2%) 62 (25.2%) 37 (39.8%) 124 (41.8%) 104 (42.3%) 48 (51.6%) 101 (34.0%) 80 (32.5%) w22 ¼ 14.3; P < 0.01 w22 ¼ 15.6; P < 0.01 8.2% 23.8% 26.3% 42.6% 38.5% 38.6% 49.2% 37.8% 35.1% w22 ¼ 7.0; P ¼ 0.03 w22 ¼ 8.7; P ¼ 0.01 13.0% 27.1% 27.5% 30.4% 46.6% 47.1% 56.5% 26.3% 25.5% w22 ¼ 8.3; P ¼ 0.02 w22 ¼ 8.5; P ¼ 0.01 10.3% 24.4% 25.8% 37.9% 44.4% 51.6% 51.7% 31.1% 22.6% w22 ¼ 3.9; P ¼ 0.14 w22 ¼ 6.1; P ¼ 0.05 7.8% 24.2% 25.1% 40.6% 41.3% 40.9% 51.6% 34.5% 34.0% w22 ¼ 10.5; P < 0.01 w22 ¼ 11.0; P < 0.01 12.5% 26.1% 26.5% 47.9% 41.0% 41.1% 39.6% 33.0% 32.4% w22 ¼ 4.1; P ¼ 0.13 w22 ¼ 4.2; P ¼ 0.12 Control (no CD), controls and their siblings never meeting criteria for conduct disorder (n, 93); Case (CD), adolescent patients and their siblings who met at least three lifetime criteria for conduct disorder (n, 297); Case (CD þ 1), adolescent patients and their siblings who met at least three lifetime criteria for conduct disorder and endorsed at least one aggressive symptom (n, 246); s, short variant of 5HTTLPR (484); l, long variant of the 5HTTLPR (528). Conduct Disorder and 5HTTLPR 829 TABLE IV. Transmission Disequilibrium Tests For 5HTTLPR Adolescents with conduct disorder from patient families and both parents (74 trios)a Not transmitted Transmitted s l Total s l Totalb 23 32 55 43 50 93 66 82 148 TDT ¼ (43–32)2/(43 þ 32) wtdt2 ¼ 1.61; P ¼ 0.20 Adolescents with whole-life conduct disorder with at least one aggressive symptom from patient families and both parents (57 trios) Transmitted s 16 39 55 TDT ¼ (39–23)2/(39 þ 23) l 23 36 59 wtdt2 ¼ 4.13; P ¼ 0.04 Total 39 75 114 Secondary analyses including informative duosc and/or excluding siblings, for both conduct disorder and conduct disorder with at least one aggressive symptom Sample Sample Phenotype Trios Duos Statistic Patient families 2 Patients and sibs Case (CD) 74 101 wtdt ¼ 3.06; P ¼ 0.08 Patients and sibs Case (CD þ 1) 57 85 wtdt2 ¼ 5.94; P ¼ 0.02 Patients only Case (CD) 55 0 wtdt2 ¼ 4.74; P ¼ 0.03 Patients only Case (CD þ 1) 45 0 wtdt2 ¼ 5.33; P ¼ 0.02 Patients only Case (CD) 55 88 wtdt2 ¼ 8.24; P < 0.01 Patients only Case (CD þ 1) 45 74 wtdt2 ¼ 9.39; P < 0.01 Control families Controls and sibs Control (no CD) 63 0 wtdt2 ¼ 4.57*; P ¼ 0.03 Controls and sibs Control (no CD) 63 20 wtdt2 ¼ 6.25*; P ¼ 0.01 a One trio is two parents and one offspring. Each parent has two alleles (i.e., ss, sl, or ll), one of which is transmitted to an offspring. Cells in the table indicate which allele was and was not transmitted from parent to offspring, for example, in the upper right cell, there were 43 parents with the sl genotype who transmitted their s-allele to an offspring but not their l-allele. c One duo is one parent and one offspring. *Favoring greater transmission of the l-allele; Control (no CD), controls and their siblings never meeting criteria for conduct disorder (defined here as meeting at least three lifetime conduct disorder criteria); Case (CD), adolescent patients and their siblings who met at least three criteria for conduct disorder; Case (CD þ 1), adolescent patients and their siblings who met at least three lifetime criteria for conduct disorder and endorsed at least one aggressive symptom; s, short variant of 5HTTLPR (484); l, long variant of the 5HTTLPR (528). b findings also are consistent with recently published animal work on an analogous polymorphism found in non-human primates [Suomi, 2003]. Case-Control Analyses Our case-control analyses showed significant 5HTTLPR genotype differences between cases and controls but the casecontrol method has certain limitations that may make the results difficult to interpret. Family structure, ethnicity, and gender can be confounding variables in these comparisons. For example, the control sample included more siblings than did the cases, which could be a source of bias. Compared to the overall analysis, an analysis of probands only was nonsignificant even though the ss genotype frequencies were similar to those seen in the overall sample. The ethnic makeup of the cases and controls can be a source of bias since significant ethnic differences in 5HTTLPR allele frequencies have been reported [Gelernter et al., 1999]. The ethnic makeup in this study did differ between conduct-disordered and non-conductdisordered adolescents. Despite these differences, case-control analyses within our two largest ethnic subgroups were consistent with our overall results. It has been suggested that the link between serotonin dysfunction and aggression may differ by gender [Manuck et al., 1998]. We found no gender-bygenotype differences within cases or controls. Although gender differences are not likely to have contributed to our results, they cannot be ruled out completely since our adolescent sample from patient families was mostly male. One concern is that the control ss genotype frequencies are lower than might be expected from some studies of general population Cauca- sians [Munafo et al., 2003]. It is important to note that these control adolescents were not a homogeneous ethnic group, and this was not a general population sample; instead controls were selected from similar neighborhoods to patients and then were further selected for never having met criteria for conduct disorder. It is somewhat reassuring that our ss-genotype frequencies are similar to a non-violent control group from a previous study [Retz et al., 2004], but potential confounds to our case-control results cannot be ruled out. Within-Family Analyses We conducted TDT analyses, which control for population stratification by using within family tests. Our analyses for conduct disorder were non-significant. The analyses of conduct disorder with at least one aggressive symptom demonstrated significantly greater transmission of the s-allele in these families. Additional tests that excluded siblings (restriction to simplex families allows a test of association/linkage disequilibrium) were significant, again favoring greater transmission of the s-allele from heterozygous parents. Many patient families had only one parent available for interview (fathers with antisocial personality disorder often are absent) [Sakai et al., 2004]. To control for this potential bias TDT analyses were repeated using informative parent-offspring duos. Analyses using duos may introduce a bias since on average, transmission of the common allele from a known heterozygous parent is more likely to be included as informative, because the missing parent is most likely to have also contributed a common allele [Curtis and Sham, 1995]. In our sample, the opposite was true; when duos were included in the 830 Sakai et al. analyses, support for the s-allele (usually the less common allele) [Munafo et al., 2003] being preferentially transmitted in patient families increased. Although this may have occurred simply by chance, it is likely that the s-allele was more commonly transmitted to the offspring from missing parents. If true, this result adds support to the observed association between 5HTTLPR and conduct disorder with at least one aggressive symptom. The TDT was also conducted for control adolescents who had never met criteria for conduct disorder. In contrast to the cases, the l-allele was preferentially transmitted from heterozygous parents to unaffected offspring. This result is entirely consistent with our overall results. Hardy–Weinberg Equilibrium The lack of HWE in the clinical sample at first glance is somewhat concerning. However, this population is a selected one in which non-random mating occurs for the phenotype of interest [Sakai et al., 2004]. We can examine a simple model of selection, where the s-allele acts recessively, the ss genotype confers a relative risk of 3.4 (Table I), disease prevalence is 0.10, and the s-allele frequency is 0.40. The expected ss, sl, and ll genotypes in the general population would be 16%, 48%, and 36%, respectively. Because of differences in relative risk, prevalence of conduct disorder among those with the ss genotype would be about 24.6%, and 7.2% among those with the sl or ll genotype. Within those with conduct disorder we would expect the ss, sl, and ll genotype frequencies to be 39%, 35%, and 26%, respectively; with a sample the size examined in this report (n, 297), such genotype frequencies would be out of HWE as we have found. Alternatively, those never meeting criteria for conduct disorder would be less affected by such selection and would have ss, sl, and ll genotype frequencies of 13%, 50%, and 37%, respectively; with a sample the size examined in this report (n, 93), such genotype frequencies would be in HWE as in our control sample. Consistent with this simple model and our results, two studies of violent offenders have been published with 5HTTLPR genotype frequencies out of HWE [Hallikainen et al., 1999; Liao et al., 2004]. Comparison of Results With the Existing Literature The observed association of the s-allele with conduct disorder with at least one aggressive symptom is consistent with expectations from the literature. The s-allele has been associated with serotonin dysfunction (low CSF 5-HIAA, blunted prolactin secretion, and low platelet binding) and such serotonin dysfunction has been associated with aggressive behavior. An analogous polymorphism in the regulatory region of the 5HTT gene in monkeys has been described and the short allele has been associated in vivo with reductions in serotonin (reduced CSF 5-HIAA). This association in non-human primates between genotype, biochemical changes (measures of serotonin), and behaviors (aggression) appears to be moderated by environmental factors [Suomi, 2003]. Monkeys with the s-allele who experience adverse rearing environments (peer-reared) have significantly reduced CSF 5-HIAA and exhibit greater aggression when compared with monkeys with the ll genotype reared in the same environment; in less stressful rearing environments the sl and ll monkeys do not differ in CSF 5-HIAA or aggression [Bennett et al., 2002; Suomi, 2003]. Our results differ in that they suggest an association between the s-allele and conduct disorder with aggression, one that is not mediated by physical/sexual abuse and neglect. Two previous studies have demonstrated an association of the ll genotype with violence or aggression [Twitchell et al., 2001; Zalsman et al., 2001], and three found no association [Reist et al., 2001; Beitchman et al., 2003; Davidge et al., 2004]. However, none of these studies employed a within-family study design, all utilized small samples relative to the current study, and none used conduct disorder with aggression as their phenotype. Finally, we utilized two tests (case-control and TDT) and found consistent results. Thus, there are clear differences between the current study and those of previous samples perhaps explaining the discrepancies in results. Comment on Comorbity Antisocial behavior and substance use disorders are highly comorbid. For example, in one national epidemiological sample 84% of those with antisocial personality disorder also met criteria for a lifetime substance use disorder [Regier et al., 1990] and in another alcohol dependence (odds ratio 7.1) and drug dependence (odds ratio 18.5) were strongly associated with antisocial personality disorder [Grant et al., 2004]. Our sample is consistent with these studies. Although with great effort it may be possible to collect samples of conduct disordered adolescents who have never met criteria for a substance use disorder, this would, in fact, be a very atypical, non-representative sample, one which would likely also be less severely affected [Krueger, 1999]. This high rate of comorbidity raises the concern about whether co-occurring conduct and substance problems in most cases represent independent disorders or alternatively, whether the conduct symptoms are more likely substanceinduced. Although use of substances can lead to aggressive antisocial behavior, we argue that it was highly unlikely that many subjects in our study had substance-induced conduct disorder. First, examination of an available dataset [NESARC (National Epidemiologic Survey on Alcohol and Related Conditions), 2001–2002, 2005] suggests that among those with conduct disorder, most report that none of their antisocial symptoms before age 15 were associated with substance use. Second, conduct disorder symptoms tend to predate the use of drugs and alcohol by many years [Kuperman et al., 2001], again supporting that for most adolescents conduct disorder symptoms are not simply induced during periods of intoxication. Third, our patients while substance-free in treatment often continue to exhibit impulsive, aggressive behavior and show continued risk-taking behavior on psychological testing when substance free for at least 7 days [Crowley et al., 2005, under review]. Thus, we do not believe that many of our subjects suffered from substance-induced conduct disorder. We did undertake analyses in an attempt to clarify whether the association was with conduct disorder or substance dependence. Unfortunately, our post-hoc analyses could not differentiate which disorder was driving our results (conduct disorder and substance dependence symptom counts did not differ significantly by genotype within patients). Therefore, this must simply remain an important limitation in our analyses. This distinction may, however, be of relatively limited value as several investigators have shown that there are substantial common genetic influences on both disorders [Hicks et al., 2004]. Thus it is conceivable that 5HTTLPR may ultimately be associated with both disorders, perhaps through an intermediate phenotype such as behavioral dysinhibition [Young et al., 2000]. Comment on Power Power estimates for the sample size used for within-family analyses in this report vary based on the mode of action of the risk increasing allele and other assumptions. Some studies suggest that the s-allele operates in a dominant fashion [Heils et al., 1996] but others do not [Williams et al., 2003]. If the sallele operates in a dominant fashion, to have had about 80% power to detect the association in our TDT analyses would require the s-allele to confer a relative risk of about six (Table I). Even assuming an additive or recessive mode of action, the Conduct Disorder and 5HTTLPR relative risk conferred to those possessing the s-allele is higher than expected given the assumptions that most complex traits are influenced by a number of genes of small to moderate effect, and that, at most, such genes would double the liability for such a trait [Munafo et al., 2003]. Our results suggest that within the population of study, that we have found an allelic association of at least moderate influence, though we cannot rule out that we have made a chance finding. Although the findings in this sample are consistent with recent animal literature and within the dataset two different methods of analysis show convergent results, these power calculations emphasize the importance of replication with larger samples. Conclusions and Limitations Our results add support to the evidence that the s-variant of the SLC6A4 promoter contributes to the liability toward aggressive behavior in humans. These results require careful interpretation and further replication across multiple samples. Although cases were selected to meet diagnostic criteria for conduct disorder, adolescent patients were drawn from a treatment program for youth with serious substance and behavior problems. Nearly all patients also meet criteria for at least one substance dependence diagnosis on standardized testing [Miles et al., 1998]. Therefore, we cannot determine whether the association is with conduct disorder, substance dependence, or some unmeasured but highly correlated latent factor. Still conduct disorder and substance dependence are commonly comorbid, in keeping with recent research suggesting strong, common genetic influences on both disorders [Hicks et al., 2004]. ACKNOWLEDGMENTS Portions of this study were previously presented at annual meeting of the College on Problems of Drug Dependence (2005). REFERENCES American Psychiatric Association. 1994. Diagnostic and statistical manual of mental disorders, (DSM IV), 4th edn. Washington, DC: American Psychiatric Association. Anchordoquy HC, McGeary C, Liu L, Krauter KS, Smolen A. 2003. Genotyping of three candidate genes after whole-genome preamplification of DNA collected from buccal cells. Behav Genet 33:73–78. Beitchman JH, Davidge KM, Kennedy JL, Atkinson L, Lee V, Shapiro S, Douglas L. 2003. The serotonin transporter gene in aggressive children with and without ADHD and nonaggressive matched controls. Ann NY Acad Sci 1008:248–251. Bennett AJ, Lesch KP, Heils A, Long JC, Lorenz JG, Shoaf SE, Champoux M, Suomi SJ, Linnoila MV, Higley JD. 2002. Early experience and serotonin transporter gene variation interact to influence primate CNS function. Mol Psychiatry 7:118–122. Brown GL, Goodwin FK, Ballenger JC, Goyer PF, Major LF. 1979. Aggression in humans correlates with cerebrospinal fluid amine metabolites. Psychiatry Res 1:131–139. Cadoret RJ, Langbehn D, Caspers K, Troughton EP, Yucuis R, Sandhu HK, Philibert R. 2003. Associations of the serotonin transporter promoter polymorphism with aggressivity, attention deficit, and conduct disorder in an adoptee population. Compr Psychiatry 44:88–101. Cottler L, Robins L, Helzer J. 1989. The reliability of the CIDI-SAM. Br J Addictions 84:801–814. Crowley TJ, Riggs PD. 1995. Adolescent substance use disorder with conduct disorder and comorbid conditions. NIDA Res Monogr 156:49–111. Crowley TJ, Mikulich SK, Ehlers KM, Whitmore EA, MacDonald MJ. 2001. Validity of structured clinical evaluations in adolescents with conduct and substance problems. J Am Acad Child Adolesc Psychiatry 40:265–273. Crowley TJ, Mikulich SK, Ehlers KM, Hall SK, Whitmore EA. 2003. Discriminative validity and clinical utility of an abuse-neglect interview for adolescents with conduct and substance use problems. Am J Psychiatry 160:1461–1469. 831 Crowley TJ, Raymond KE, Mikulich-Gilbertson SK, Thompson LL, Lejuez CW. 2005. A risk-taking ‘‘set’’ in a novel task among adolescents with serious conduct and substance problems. J Am Acad Child Adolesc Psychiatry 45:175–183. Curtis D, Sham PC. 1995. A note on the application of the transmission disequilibrium test when a parent is missing. Am J Hum Genet 56:811– 812. Davidge KM, Atkinson L, Douglas L, Lee V, Shapiro S, Kennedy JL, Beitchman JH. 2004. Association of the serotonin transporter and 5HT1Dbeta receptor genes with extreme, persistent and pervasive aggressive behaviour in children. Psychiatry Genet 14:143–146. Gelernter J, Cubells JF, Kidd JR, Pakstis AJ, Kidd KK. 1999. Population studies of polymorphisms of the serotonin transporter protein gene. Am J Med Genet 88:61–66. Gerra G, Garofano L, Santoro G, Bosari S, Pellegrini C, Zaimovic A, Moi G, Bussandri M, Moi A, Brambilla F, Donnini C. 2004. Association between low-activity serotonin transporter genotype and heroin dependence: Behavioral and personality correlates. Am J Med Genet (Neuropsychiatric Genetics) PartB 126B:37–42. Grant BF, Stinson FS, Dawson DA, Chou SP, Ruan WJ, Pickering RP. 2004. Co-occurrence of 12-month alcohol and drug use disorders and personality disorders: Results from the National Epidemiological Survey on Alcohol and Related Conditions. Arch Gen Psychiatry 61:361–368. Hallikainen T, Saito T, Lachman HM, Volavka J, Pohjalainen T, Ryynanen OP, Kauhanen J, Syvalahti E, Hietala J, Tiihonen J. 1999. Association between low activity serotonin transporter promoter genotype and early onset alcoholism with habitual impulsive violent behavior. Mol Psychiatry 4:385–388. Hariri AR, Mattay VS, Tessitore A, Kolachana B, Fera F, Goldman D, Egan MF, Weinberger DR. 2002. Serotonin transporter genetic variation and the response of the human amygdala. Science 297:400–403. Heils A, Teufel A, Petri S, Stober G, Riederer P, Bengel D, Lesch KP. 1996. Allelic variation of human serotonin transporter gene expression. J Neurochem 66:2621–2624. Heinz A, Jones DW, Mazzanti C, Goldman D, Ragan P, Hommer D, Linnoila M, Weinberger DR. 2000. A relationship between serotonin transporter genotype and in vivo protein expression and alcohol neurotoxicity. Biol Psychiatry 47:643–649. Hicks BM, Krueger RF, Iacono WG, McGue M, Patrick CJ. 2004. Family transmission and heritability of externalizing disorders: A twin-family study. Arch Gen Psychiatry 61:922–928. Hu X, Oroszi G, Chun J, Smith TL, Goldman D, Schuckit MA. 2005. An expanded evaluation of the relationship of four alleles to the level of response to alcohol and the alcoholism risk. Alcohol Clin Exp Res 29: 8–16. Krueger RF. 1999. The structure of common mental disorders. Arch Gen Psychiatry 56:921–926. Kruesi MJ, Hibbs ED, Zahn TP, Keysor CS, Hamburger SD, Bartko JJ, Rapoport JL. 1992. A 2-year prospective follow-up study of children and adolescents with disruptive behavior disorders. Prediction by cerebrospinal fluid 5-hydroxyindoleacetic acid, homovanillic acid, and autonomic measures? Arch Gen Psychiatry 49:429–435. Kuperman S, Schlosser SS, Kramer JR, Bucholz K, Hesselbrock V, Reich T, Reich W. 2001. Developmental sequence from disruptive behavior diagnosis to adolescent alcohol dependence. Am J Psychaitry 158: 2022–2026. Liao DL, Hong CJ, Shih HL, Tsai SJ. 2004. Possible association between serotonin transporter promoter region polymorphism and extremely violent crime in Chinese males. Biological Psychiatry 50:284–287. Linnoila M, Virkkunen M, Scheinin M, Nuutila A, Rimon R, Goodwin FK. 1983. Low cerebrospinal fluid 5-hydroxyindoleacetic acid concentration differentiates impulsive from nonimpulsive violent behavior. Life Sci 33:2609–2614. Manuck SB, Flory JD, McCaffery JM, Matthews KA, Mann JJ, Muldoon MF. 1998. Aggression, impulsivity, and central nervous system serotonergic responsivity in a nonpatient sample. Neuropsychopharmacology 19: 287–299. Miles DR, Carey G. 1997. Genetic and environmental architecture of human aggression. J Pers Soc Psychol 72:207–217. Miles DR, Stallings MC, Young SE, Hewitt JK, Crowley TJ, Fulker DW. 1998. A family history and direct interview study of the familial aggregation of substance abuse: The adolescent substance abuse study. Drug Alcohol Depend 49:105–114. 832 Sakai et al. Munafo MR, Clark TG, Moore LR, Payne E, Walton R, Flint J. 2003. Genetic polymorphisms and personality in healthy adults: A systematic review and meta-analysis. Mol Psychiatry 8:471–484. NESARC (National Epidemiologic Survey on Alcohol and Related Conditions), 2001–2002. 2005. http://niaaa.census.gov/index.html. Accessed July 27. O’Keane V, Moloney E, O’Neill H, O’Connor A, Smith C, Dinan TG. 1992. Blunted prolactin responses to d-fenfluramine in sociopathy. Evidence for subsensitivity of central serotonergic function. Br J Psychiatry 160:643–646. Purcell S, Cherny SS, Sham PC. 2003. Genetic power calculator: Design of linkage and association genetic mapping studies of complex traits. Bioinformatics 19:149–150. Regier DA, Farmer ME, Rae DS, Locke BZ, Keith SJ, Judd LL, Goodwin FK. 1990. Comorbidity of mental disorders with alcohol and other drug abuse: Results from the epidemiologic catchment area (ECA) study. JAMA 264:2511–2518. Reist C, Mazzanti C, Vu R, Tran D, Goldman D. 2001. Serotonin transporter promoter polymorphism is associated with attenuated prolactin response to fenfluramine. Am J Med Genet 105:363–368. Retz W, Retz-Junginger P, Supprian T, Thome J, Rosler M. 2004. Association of serotonin transporter promoter gene polymorphism with violence: Relation with personality disorders, impulsivity, and childhood ADHD psychopathology. Behav Sci Law 22:415–425. Rhee SH, Wladman ID. 2002. Genetic and environmental influences on antisocial behavior: A meta-analysis of twin and adoption studies. Pscyhol Bull 128:490–529. Sakai JT, Stallings MC, Mikulich-Gilbertson SK, Corley RP, Young SE, Hopfer CJ, Crowley TJ. 2004. Mate similarity for substance dependence and antisocial personality disorder symptoms among parents of patients and controls. Drug Alcohol Depend 75:165–175. Sham P. 1998. Statistics in human genetics. New York: Oxford University Press. Sham PC, Zhao JH, Waldman I, Curtis D. 2000. Should ambiguous trios for the TDT be discarded? Ann Hum Genet 64:575–576. Spielman RS, McGinnis RE, Ewens WJ. 1993. Transmission test for linkage disequilibrium: The insulin gene region and insulin-dependent diabetes mellitus (IDDM). Am J Hum Genet 52:506–516. Stallings MC, Corley RP, Dennehey B, Hewitt JK, Krauter KS, Lessem JM, Mikulich-Gilbertson SK, Rhee SH, Smolen A, Young SE, Crowley TJ. 2005. A genome-wide search for quantitative trait loci influencing antisocial drug dependence in adolescence. Arch Gen Psychiatry 62:1042–1051. Stoff DM, Pollock L, Vitiello B, Behar D, Bridger WH. 1987. Reduction of (3H)-imipramine binding sites on platelets of conduct-disordered children. Neuropsychopharmacology 1:55–62. Stoltenberg SF, Twitchell GR, Hanna GL, Cook EH, Fitzgerald HE, Zucker RA, Little KY. 2002. Serotonin transporter promoter polymorphism, peripheral indexes of serotonin function, and personality measures in families with alcoholism. Am J Med Genet 114:230– 234. Suomi SJ. 2003. Gene-environment interactions and the neurobiology of social conflict. Ann NY Acad Sci 1008:132–139. Twitchell GR, Hanna GL, Cook EH, Stoltenberg SF, Fitzgerald HE, Zucker RA. 2001. Serotonin transporter promoter polymorphism genotype is associated with behavioral disinhibition and negative affect in children of alcoholics. Alcohol Clin Exp Res 25:953– 959. Van Goozen SH, Matthys W, Cohen-Kettenis PT, Westenberg H, van Engeland H. 1999. Plasma monoamine metabolites and aggression: Two studies of normal and oppositional defiant disorder children. Eur Neuropsychopharmacol 9:141–147. Williams RB, Marchuk DA, Gadde KM, Barefoot JC, Grichnik K, Helms MJ, Kuhn CM, Lewis JG, Schanberg SM, StaffordSmith M, Suarez EC, Clary GL, Svenson IK, Siegler IC. 2003. Serotoning-related gene polymorphisms and central nervous system serotonin function. Neuropsychopharmacology 28:533– 541. Young SE, Stallings MC, Corley RP, Krauter KS, Hewitt JK. 2000. Genetic and environmental influences on behavioral disinhibition. Am J Med Genet 96:684–695. Zalsman G, Frisch A, Bromberg M, Gelernter J, Michaelovsky E, Campino A, Erlich Z, Tyano S, Apter A, Weizman A. 2001. Family-based association study of serotonin transporter promoter in suicidal adolescents: No association with suicidality but possible role in violence traits. Am J Med Genet 105:239– 245.