COMT but not serotonin-related genes modulates the influence of childhood abuse on anger traits

Authors

  • N. Perroud,

    Corresponding author
    1. Department of Psychiatry, University of Geneva, Switzerland
    2. King's College London, MRC Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, London, UK
    Search for more papers by this author
  • I. Jaussent,

    1. INSERM U888, Montpellier, F-34093, France
    Search for more papers by this author
  • S. Guillaume,

    1. INSERM U888, Montpellier, F-34093, France
    2. University of Montpellier I, F-34000, France
    3. Department of Psychological Medicine and Psychiatry, Lapeyronie Hospital, CHU Montpellier, France
    Search for more papers by this author
  • F. Bellivier,

    1. Department of Genetics, Psychiatry Genetics, INSERM, U 841, IMRB
    2. University Paris 12, Faculty of Medicine, IFR10
    3. Department of Psychiatry, AP-HP, Henri Mondor-Albert Chenevier Hospitals, Creteil, F-94000, France
    Search for more papers by this author
  • P. Baud,

    1. Department of Psychiatry, University of Geneva, Switzerland
    Search for more papers by this author
  • F. Jollant,

    1. INSERM U888, Montpellier, F-34093, France
    2. University of Montpellier I, F-34000, France
    3. Department of Psychological Medicine and Psychiatry, Lapeyronie Hospital, CHU Montpellier, France
    Search for more papers by this author
  • M. Leboyer,

    1. Department of Genetics, Psychiatry Genetics, INSERM, U 841, IMRB
    2. University Paris 12, Faculty of Medicine, IFR10
    3. Department of Psychiatry, AP-HP, Henri Mondor-Albert Chenevier Hospitals, Creteil, F-94000, France
    Search for more papers by this author
  • C. M. Lewis,

    1. King's College London, MRC Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, London, UK
    2. Department of Medical and Molecular Genetics, King's College London, London, UK
    Search for more papers by this author
  • A. Malafosse,

    1. Department of Psychiatry, University of Geneva, Switzerland
    2. INSERM U888, Montpellier, F-34093, France
    3. Department of Medical Genetics and Laboratories, University Hospital of Geneva, Switzerland
    Search for more papers by this author
  • P. Courtet

    1. INSERM U888, Montpellier, F-34093, France
    2. University of Montpellier I, F-34000, France
    3. Department of Psychological Medicine and Psychiatry, Lapeyronie Hospital, CHU Montpellier, France
    Search for more papers by this author

*N. Perroud, Department of Psychiatry, University of Geneva, Hôpital Belle-Idée, 2 ch. du Petit-Bel-Air, 1225 Chêne-Bourg, Geneva, Switzerland. E-mail:nader.perroud@hcuge.ch

Abstract

Anger-related traits are regulated by genes as well as early environmental factors. Both childhood maltreatment and genes underlie vulnerability to suicidal behaviors, possibly by affecting the constitution of intermediate phenotypes such as anger traits. The aim of this study was to test the interaction between nine candidate genes and childhood maltreatment in modulating anger-related traits in 875 adult suicide attempters. The State-Trait Anger Expression Inventory and the Childhood Trauma Questionnaire were used to examine anger traits and traumatic childhood experiences, respectively. The functional polymorphism of the catecholamine-O-methyl-transferase (COMT) gene Val158Met significantly modulated the association between sexual abuse and anger-trait level (P = 0.001). In the presence of sexual abuse, individuals carrying the Val high-activity allele displayed greater disposition toward anger than individuals homozygous for the Met allele (P = 0.0003). Notably, none of the serotonin-related genes influenced the effect of childhood abuse on anger traits. The results of the present study suggest that anger-trait level is influenced by the interaction between childhood abuse and functional polymorphism in the COMT gene. This study was carried out in a population with a high frequency of childhood abuse and a high disposition toward anger, and replication in healthy subjects is needed.

Anger is a basic, commonly experienced emotional state that consists of feelings of variable intensity from mild irritation to intense fury and rage. Anger has been related to several phenomena in medicine and several behavioral and psychiatric conditions (Chida & Steptoe 2009; Lara & Akiskal 2006). A high level of anger traits has indeed been associated with eating disorders (Fassino et al. 2001), borderline personality disorder (Zanarini et al. 2004), drug addiction (De Moja & Spielberger 1997) and suicidal behaviors (Baud 2005; Baud et al. 2007). For the latter, in comparison to controls suicide attempters have indeed been shown to express higher level of anger (Baud 2005; Baud et al. 2007).

Anger is part of the brain's fight-or-flight response to a perceived threat. This multidimensional structure is composed of physiological, behavioral and cognitive aspects that have been shown to correlate positively with aggression, hostility and impulsivity (Ramirez & Andreu 2006). From this perspective, particular attention has been paid to anger in the field of suicidal behaviors. Indeed, subjects with a history of suicide attempts have been shown to display a greater tendency toward anger or greater difficulty in outwardly expressing anger than control (Baud 2005; Baud et al. 2007).

Knowledge of the sources of individual differences in anger remains sparse but includes genetic and environmental factors. Twin and family studies have indeed suggested that anger-related traits are inheritable (Rebollo & Boomsma 2006; Sluyter et al. 2000). Several genes involved in serotoninergic and catecholaminergic neurotransmissions have recently been associated with anger traits and/or related measures of anger such as aggression or impulsivity, including tryptophan hydroxylase 1 (TPH1) (Baud et al. 2009; Manuck et al. 1999; Rujescu et al. 2002), several serotonin receptors (5-HTR) (Giegling et al. 2006; Serretti et al. 2007; Zouk et al. 2007), the monoamine oxidase A (MAOA) (Alia-Klein et al. 2008; Manuck et al. 2000) and the catecholamine-O-methyl transferase (COMT) genes (Alia-Klein et al. 2008; Baud et al. 2007, 2009; Giegling et al. 2006; Manuck et al. 1999; Rujescu et al. 2002; Serretti et al. 2007; Zouk et al. 2007).

Environmental factors such as childhood maltreatment, lack of social support and negative life events were also found to predict high levels of anger (Springer et al. 2007). Finally, the effects of gene–environment interaction (GxE) on measures of anger are also taken into account although this area has not been thoroughly investigated to date. Exactly how GxE modulates anger-related traits is not yet understood. Some studies such as those published by Caspi et al. (2002) provide some preliminary evidence pointing to the interaction of MAOA and early maltreatment on aggression and antisocial personality traits.

In this study, we intend to explore the extent to which some GxE underlies interindividual differences in the expression of anger in a population of suicide attempters. In view of the importance of childhood maltreatment in the development of suicidal behaviors (Dube et al. 2001), the close relationship between suicidal behaviors and anger (Horesh et al. 1997) and the existence of a genetic component in suicidal behaviors (Bellivier et al. 2004), a population composed of suicide attempters is ideal to unravel the environmental and genetic components of anger-related traits.

Materials and methods

Subjects

Suicide attempters (n = 875) were included after informed written consent was obtained. Suicide attempters were recruited from consecutive admissions to the psychiatric units of three university hospitals—Montpellier and Créteil (France), and Geneva (Switzerland)—between 1994 and 2006. Suicide attempt was defined as the occurrence of self-harming acts with an intent to end one's own life (Mann 2003). Suicide attempters were all of European ancestry for at least two generations. Suicide attempters were assessed for psychiatric diagnoses using either the French version of the Diagnostic Interview for Genetics Studies (DIGS) or the Mini International Neuropsychiatric Interview (MINI) (Nurnberger et al. 1994; Preisig et al. 1999; Sheehan et al. 1998). The study protocol was approved by research ethics committees in each center.

Behavioral and childhood trauma assessments

Anger-related traits were assessed with the State-Trait Anger Expression Inventory (STAXI) (Spielberger 1988). The STAXI is a 44-item self-report questionnaire measuring the experience and expression of anger in accordance with the state-trait personality theory (Norlander & Eckhardt 2005). The experience of anger comprises State Anger (the current feelings experienced by an individual) and Trait Anger (individual disposition to experience anger or the ease, frequency and intensity of becoming angered), whereas the expression of anger comprises the three components of Anger In (which measures the individual tendency to suppress angry feelings), Anger Out (the tendency to outwardly express anger toward people or objects) and Anger Control (the capacity of an individual to regulate or control his/her anger).

The Childhood Trauma Questionnaire (CTQ) (Bernstein & Fink 1998) is a retrospective self-report questionnaire that examines the traumatic childhood experiences of adults and adolescents. It assesses five types of childhood trauma: emotional abuse, emotional neglect, physical abuse, physical neglect and sexual abuse. Childhood Trauma Questionnaire has shown excellent test–retest reliability and convergent validity (Bernstein et al. 1994). It comprises 28 items and each item is rated from 1 (never) to 5 (very often). Scores range from 5 to 25 for each type of abuse. According to Bernstein and Fink, thresholds or cut scores have been set for each type of trauma at four levels of maltreatment: none, low, moderate and severe (see Table 1 for cut-off scores for each of the five CTQ subscales). The different cut-offs have been shown to have good specificity and sensitivity (Bernstein & Fink 1998).

Table 1.  Clinical and demographic characteristics of the 875 suicide attempters
  n%
SexMale25629.3
 Female61970.7
Recruitment centerMontpellier63872.9
 Geneva12314.1
 Paris11413.3
DiagnosesMajor depressive disorder60268.8
 Bipolar disorder18621.3
 Schizophrenia and related disorders242.7
 Anxiety disorders333.8
 Alcohol/cocaine/heroin dependence80.9
 Unknown192.2
 No psychiatric disorder30.3
Emotional neglectYes74184.7
 Low (10–14)27531.4
 Moderate (15–17)15918.2
 Severe (18–25)30735.1
 No13415.3
Emotional abuseYes59467.9
 Low (9–12)18220.8
 Moderate (13–15)12113.8
 Severe (16–25)29133.7
 No28131.1
Physical neglectYes44751.1
 Low (8–9)17820.3
 Moderate (10–12)13615.5
 Severe (13–25)13315.2
 No42848.9
Physical abuseYes33438.2
 Low (8–9)10512
 Moderate (10–12)819.3
 Severe (13–25)14816.9
 No54161.8
Sexual abuseYes34939.9
 Low (6–7)687.8
 Moderate (8–12)11713.4
 Severe (13–25)16418.7
 No52660.1
  MeanSD
Age 39.612.9
STAXIState Anger20.97.6
 Trait Anger24.16
 Anger In21.35
 Anger Out16.64.9
 Anger Control21.24.6

Genotyping

The following polymorphisms previously associated with related measures of anger were chosen for the analyses: the TPH1 rs1800532 (A218C) polymorphism (Jollant et al. 2007; Manuck et al. 1999; Rujescu et al. 2002); the MAOA variable number of tandem-repeat polymorphisms in the promoter region (MAOA u-VNTR) (Jollant et al. 2007; Shih & Chen 1999); the functional rs6295 (C-1019G) promoter polymorphism in the 5-HTR1A gene (Keltikangas-Jarvinen et al. 2008; Serretti et al. 2007); the 5-HTR2A rs6311 (A-1438G) and rs6313 (C102T) polymorphisms (Keltikangas-Jarvinen et al. 2008; Turecki et al. 1999); the 5-HTR1B promoter rs130058 (A-161T) and rs6296 (G861C) polymorphisms (Zouk et al. 2007); and COMT rs4680 (Val158Met) polymorphism (Rujescu et al. 2003) and with suicidal behaviors: three TPH2 polymorphisms: rs11179000 and rs11179001 in the intron 4 and rs7305115 in exon 7 (Jollant et al. 2007); the 44 base pair insertion/deletion polymorphism within the serotonin transporter gene (5-HTTLPR) (Jollant et al. 2007; Roy et al. 2007) and the brain-derived neurotrophic factor (BDNF) rs6265 (Val66Met) polymorphism (Perroud et al. 2008). All primers and genotyping conditions were carried out as previously published (Baud et al. 2007; Etain et al. 2004; Jollant et al. 2007; Paoloni-Giacobino et al. 2000; Perroud et al. 2008) or are available upon request.

Statistical analyses

Demographic and clinical characteristics of the study population were described using mean and standard deviation for quantitative variables and proportions for categorical variables. We used a linear regression model to test for the main effect of genes. P-values were two tailed, and P < 0.05 was considered to indicate statistical significance. The statistical package stata V.10 and plink V.1.06 software (Purcell et al. 2007) were used for the analyses.

For statistical purposes and to enhance the power of tests, all cases of childhood trauma were pooled in two categories: not abused or not neglected vs. abused or neglected individuals (by pooling low, moderate and severe abused/neglected subjects) (Perroud et al. 2008). Sex, age at interview, diagnosis and the recruitment center were added as confounding variables because there were significant associations for each of these variables with at least one of the STAXI subscales, the childhood trauma items and/or the genetic polymorphisms. When a significant main effect for a genetic polymorphism was detected on a STAXI subscore, we tested for an interaction between polymorphism and childhood maltreatment on this STAXI subscale.

The one-sample Kolmogornov–Smirnov test was used to investigate whether STAXI subscale distributions were normally distributed. The Trait Anger scale was normally distributed and therefore used without any transformation in our analyses. With regard to other subscales, since no transformation was able to normalize them, we firstly used the robust command (provided with stata) in order to take into account skewed distributions. Secondly, we used a permutation test that randomly assigned STAXI subscales to subjects while keeping each subject's genotype and environmental variable fixed. These permutation procedures have been shown to relax assumptions about the normality of continuous phenotypes and to be robust against abnormal distribution patterns (Epstein & Satten 2003; Purcell et al. 2007; Zhao et al. 2000). For each analysis, the empirical P value was based on 10 000 permutations. Results were then compared with those obtained with the original distribution to verify the accuracy of our model.

In a second step and if an interaction was observed for a polymorphism, we tested for a possible evocative correlation association (evocative rGE) by determining with a chi-square test if the genetic polymorphism could be involved in evoking or eliciting maltreatment exposure. If so, we used logistic regression to adjust on sex, age at interview, recruitment center and diagnosis.

As we compared the genotypic and allelic distributions of 13 polymorphic markers between non-maltreated and maltreated individuals on five STAXI subscales, a correction for multiple testing was required. It is difficult to correct for multiple testing in a multistage analysis. The primary purpose of our research was to detect interactions between the STAXI anger subscales and the childhood abuse and neglect measures in association with the genetic polymorphisms tested. We initially tested for the main effects of association with the five STAXI subscales. For genetic variants with a liberal P-value threshold of P = 0.05, we then proceeded to test for interaction with the five childhood trauma items. We used a more stringent P-value threshold for the interactions. We tested nine independent polymorphisms [as there is strong linkage disequilibrium (LD) among the polymorphisms within TPH2, 5-HRT2A and 5-HTR1B]; the childhood trauma items are all highly correlated, and we therefore consider this to be equivalent to two independent measures. For a Bonferroni correction on the P-values for interaction, we therefore used P = 0.05/(9 × 2) = 0.0028 as a threshold for significance.

Power calculation

Power analysis was performed using the quanto V.1.2.3 program (Gauderman & Morrison 2006). We aimed to detect the genetic effects accounting for at least 2% of variance in the various STAXI subscales under a dominant genetic model for a polymorphism with a minor allele frequency of 0.2. The power for detecting such an effect was calculated for the nominal significance level (α = 0.05). The sample of 875 individuals had 0.99 power to detect an effect of a genetic marker accounting for 2% of the variance for the studied trait at an α level of 0.05.

Assuming that 40% of the subjects are exposed to childhood maltreatment and that the latter accounts for at least 1% of the variance of the trait, the GxE analysis had a power of 0.88 to detect an interaction effect accounting for more than 3% of the variance at an α level of 0.0028. In short, our study is sufficiently powered to detect not only the clinically significant effects of genes on STAXI subscales but also GxE.

Results

Genotype and allele frequencies of genetic polymorphisms in suicide attempters were in Hardy–Weinberg equilibrium and similar to the frequencies reported in other clinical samples (Table 2) (Roy et al. 2007; Rujescu et al. 2002, 2003; Serretti et al. 2007; Spurlock et al. 1998; Thorisson et al. 2005; Zalsman et al. 2005; Zouk et al. 2007).

Table 2.  Genotypes in all suicide attempters
  n%  n%
COMT Val158MetVal/Val25230.665-HTR1A C-1019GCC20925.61
 Val/Met38747.08 CG43052.7
 Met/Met18322.26 GG17721.69
5-HTTLPRLL26632.565-HRT2A A-1438GAA12523.11
 LS39948.84 AG26749.35
 SS15218.6 GG14927.54
TPH1 rs1800532AA15218.515-HRT2A C102TCC5725.91
 AC39347.87 CT11853.64
 CC27633.62 TT4520.45
TPH2 rs11179000AA51763.055-HRT1B A-161TAA40048.9
 AT25931.59 AT33140.46
 TT445.37 TT8710.64
TPH2 rs11179001AA101.225-HT1B C861GCC495.98
 AG8910.84 CG32139.19
 GG72287.94 GG44954.82
TPH2 rs7305115AA12415.14MAOA u-VNTR maleL8234.6
 AG39848.6 H15565.4
 GG29736.26    
    MAOA u-VNTR femaleLL7012.05
BDNF Val66MetMet/Met414.98 LH23740.79
 Val/Met27032.81 HH27447.16
 Val/Val51262.21    

Table 1 displays the demographic and clinical characteristics and rates of childhood abuse and neglect in suicide attempters. Most of the suicide attempters suffered from major depressive disorder (68.8%), were female (70.7%), and displayed non-violent suicide attempt (75.4%). The majority of suicide attempters showed that they had suffered from childhood abuse of at least minor severity: 38.2% of suicide attempters reported a history of physical abuse, 39.9% sexual abuse, 51.1% physical neglect, 67.9% emotional abuse and 84.7% emotional neglect. Women scored significantly higher on the Trait Anger subscale [24.5 (6.1) vs. 23.4 (5.8); b = 0.89;t = 2.02;95%CI from 0.01 to 1.79;P = 0.044] and lower on the Anger Control subscale [21 (4.7) vs. 21.8 (4.4); b = −0.76;t = −2.22;95%CI from −1.42 to −0.09;P = 0.027] than men. Moreover, age was negatively correlated to Trait Anger and Anger Out scores (b = −0.05;t = −3.26;95%CI from −0.08 to −0.02;P = 0.001 and b = −0.05;t = −4.13; 95%CI from −0.07 to −0.02; P < 0.0001, respectively).

There were more females (n = 471,73.8%) in Montpellier than in Creteil (n = 70,61.4%) and Geneva (n = 78,63.4%) (X 2 = 10.9;df = 2;P = 0.004). Moreover, subjects recruited in Geneva scored significantly higher on Anger Control subscale than individuals from Creteil [21.8 (5.0) vs. 20.5 (4.7); b = 1.33;t = 2.17;95%CI from 0.12to 2.53;P = 0.031]. For these reasons, recruitment center was added as confounding variable in all the analyses. No other variables distinguished individuals from the different center of recruitment.

Effect of childhood abuse on anger traits

Results of the regression analyses estimating the effect of childhood trauma on STAXI subscales are shown in Table 3. Interestingly, neither sexual abuse nor emotional neglect had an effect on any of the STAXI subscales. However, other forms of maltreatment had a principal effect on most of the different STAXI subscales (Table 3).

Table 3.  STAXI by childhood trauma
 VariableMeanStandard deviationMeanStandard deviationb; t; (95%CI)P
  NoYes  
Sexual abuseState Anger20.617.5921.437.620.59; 1.05; (−0.51 to 1.68)0.295
 Trait Anger23.765.9524.766.130.76; 1.73; (−0.1 to 1.61)0.083
 Anger In20.924.8421.765.250.72; 0.37; (−0.01 to 1.44)0.052
 Anger Out16.394.8417.025.030.43; 1.20; (−0.27 to 1.13)0.232
 Anger Control21.124.5121.214.80.19, 0.56; (−0.48 to 0.85)0.578
  NoYes  
Emotional neglectState Anger20.318.2121.097.520.76; 1.05; (−0.66 to 2.19)0.295
 Trait Anger23.755.9624.246.060.39; 0.70; (−0.72 to 1.51)0.484
 Anger In21.075.3121.294.960.24; 0.50; (−0.70 to 1.18)0.617
 Anger Out16.435.3516.694.840.17; 0.36; (−0.74 to 1.07)0.718
 Anger Control21.754.5721.044.63−0.63; −1.43; (−1.49 to 0.23)0.152
  NoYes  
Emotional abuseState Anger19.837.3621.57.71.44; 2.51; (0.32 to 2.58)0.012
 Trait Anger23.035.724.76.131.43; 3.19; (0.55 to 2.31)0.001
 Anger In20.385.121.684.931.23; 3.27; (0.49 to 1.98)0.001
 Anger Out15.794.6417.065.011.12; 3.08; (0.41 to 1.84)0.002
 Anger Control21.294.3321.094.76−0.15; −0.44; (−0.84 to 0.53)0.657
  NoYes  
Physical neglectState Anger20.197.4221.717.761.52; 2.87; (0.48 to 2.57)0.004
 Trait Anger23.785.6224.536.410.82; 1.98; (0.01 to 1.64)0.048
 Anger In20.795.0921.714.90.95; 2.72; (0.27 to 1.64)0.007
 Anger Out16.354.6916.945.130.70; 2.05; (0.03 to 1.36)0.041
 Anger Control21.64.620.714.61−0.91; −2.86; (−1.55 to −0.29)0.004
  NoYes  
Physical abuseState Anger20.547.6521.657.560.88; 1.59; (−0.21 to 1.96)0.112
 Trait Anger23.745.7424.866.460.95; 2.20; (0.11 to 1.80)0.028
 Anger In20.944.9221.785.130.80; 2.10; (0.05 to 1.47)0.037
 Anger Out16.094.6517.575.221.35; 3.88; (0.67 to 2.03)0.0001
 Anger Control21.274.4720.944.85−0.24; −0.72; (−0.89 to 0.41)0.471

Effects of genotypes and GxE on anger traits

COMT Val158Met

As previously described (Baud et al. 2007), we found that COMT Val158Met had a principal effect on the Trait Anger subscale. Individuals carrying the high-activity Val allele scored significantly higher on the Trait Anger scale than individuals homozygous for the low-activity Met allele [Val /Val = 24.9(5.6); Val/Met = 24.1(6.2); Met /Met = 23.4(6.1); b = −0.85; t = −2.96; 95%CI from −1.35 to −0.19; P = 0.003] (Table 4).

Table 4.  Significant associations between SNPs and STAXI subscales
  nMean (SD)b; t; adjusted P for allele comparisons
COMT Val158Met  Trait Anger
 Val/Val24024.9 (5.6)−0.85;−2.96;0.003
 Val/Met36824.1 (6.2) 
 Met/Met17123.4 (6.1) 
   Anger Out
 Val/Val23917 (4.8)−0.48;−2.05;0.041
 Val/Met36716.8 (5.1) 
 Met/Met17116 (4.7) 
TPH1 A218C  Anger Control
 AA14420.3 (4.4)0.51; 2.19; 0.029
 AC36921.5 (4.4) 
 CC26521.3 (4.9) 
   State Anger
MAOA u-VNTR maleL7818.9 (6.1)1.89; 2.05; 0.041
 H15021 (7.4) 

On the same lines, we also noted that COMT Val158Met had a main effect on Anger Out scores, [Val /Val = 17(4.8); Val /Met = 16.8(5.1); Met /Met = 16(4.7); b = −0.48; t = −2.05;95%CI from −0.98 to −0.03; P = 0.041] (Table 4). COMT genotypes did not affect other STAXI subscales.

We found a significant interaction between COMT Val158Met and sexual abuse on Trait Anger (F2,762 = 6.96;P = 0.001) (Fig. 1). This interaction was highly significant when looking at individuals with a Met/Met genotype compared with those carrying a Val allele (F1,764 = 13.13;P = 0.0003). This result was significant even after correction for multiple testing. In patients reporting sexual abuse, this interaction was explained by a significantly lower Trait Anger score among individuals carrying a Met/Met genotype [22.3 (5.7)] compared with carriers of a Val allele [Val/Val = 25.9 (6.1) or Val/Met = 25.4 (6.3); b = −3.75;t = −3.79; 95%CI from −5.70 to −1.81; P < 0.0001 and b = −3.61;t = −4.00; 95%CI from −5.38 to −1.83; P < 0.0001]. We also found a significant three-way interaction with sex (F5,757 = 2.81;P = 0.016), showing that the above interaction was mainly explained by the female population of suicide attempters (Fig. 2).

Figure 1.

Trait Anger scores by sexual abuse and COMT Val158Met polymorphism in 822 suicide attempters.ns : Val /Val = 159, Val /Met = 228, Met/ Met = 106 for individuals without history of sexual abuse; ns : Val /Val = 93, Val/Met = 159, Met /Met = 77 for individuals with history of sexual abuse.

Figure 2.

Trait Anger scores by sexual abuse and COMT Val158Met polymorphism stratified by gender. For males (n = 239;Val /Val = 67, Val/Met = 74, Met /Met = 34 for males without history of sexual abuse; Val/Val = 11, Val /Met = 38, Met /Met = 15 for males with history of sexual abuse) and females (n = 583;Val /Val = 92, Val /Met = 154, Met /Met = 72 for females without history of sexual abuse; Val /Val = 82, Val /Met = 121, Met /Met = 62 for females with history of sexual abuse), respectively.

To less of an extent but along the same lines as the results obtained for Trait Anger, COMT Val158Met also interacted with sexual abuse when considering Anger Out scores (F2,760 = 3.38;P = 0.035). However, this interaction was no longer significant following adjustment for multiple testing.

We did not observe any interaction between other types of maltreatment and COMT Val158Met for Trait Anger and Anger Out.

TPH1 A218C

TPH1 A218C was found to have a main effect on Anger Control scores (Table 3). Subjects carrying the AA genotype displayed lower Anger Control scores than the AC and CC carriers [20.3 (4.4) vs. 21.5 (4.4) and 21.3 (4.9), respectively; b = 0.51; t = 2.19;95%CI from 0.14 to 2.00; P = 0.029] (see Table 4). Other STAXI subscales were not influenced by TPH1 A218C.

There was no significant interaction between childhood maltreatment and TPH1 A218C polymorphism on the Anger Control subscale.

MAOA u-VNTR

Monoamine oxidase A u-VNTR was found to have a significant effect on the Anger State STAXI subscale among male suicide attempters. Individuals carrying the H allele scored significantly higher on this subscale (b = 1.89;t = 2.05;95%CI from 0.05 to 4.11; P = 0.041) (Table 4). An interaction between MAOA u-VNTR and sexual abuse on Anger State scores was observed, but did not survive correction for multiple testing (data not shown).

5-HTR1B A-161T, 5-HTR1B G861C, 5-HTR2A A-1438G, 5-HTR2A C102T, TPH2 rs11179000, TPH2 rs11179001, TPH2 rs7305115, 5-HTR1A, BDNF Val66Met and 5-HTTLPR

These polymorphisms were not found to have any principal effect on STAXI scores.

rGE

Finally, we tested for a possible evocative genotype–environment correlation (evocative rGE) between COMT Val158Met and sexual abuse. No evocative rGE was found.

Discussion

We found that COMT Val158Met polymorphism modulates the association between childhood sexual abuse and adulthood anger-trait level. In the presence of sexual abuse, individuals carrying the Val allele display a higher disposition toward anger than individuals homozygous for the Met allele. Our results corroborate the findings of three recent investigations analyzing the modulating effect of COMT Val158Met on different outcomes. Savitz et al. (2008) found that Val/Val genotype, and not Met/Met genotype, was associated with increasing levels of dissociation in subjects exposed to higher levels of childhood trauma. Thapar et al. (2005) found that the Val/Val genotypes are particularly susceptible to the effects of lower birth weight in developing antisocial behavior. And finally, Stefanis et al. (2007) showed that the carriers of the Val allele were more sensitive to the effect of stress on the development of psychosis than those with the Met/Met genotype. Overall, these findings support the involvement of COMT Val158Met in mediating the relationship between early traumas and psychopathology during adulthood, with the Val allele being more sensitive to environmental influences.

The current findings partly contrast with those from other studies showing increased sensitivity to stress in carriers of the Met as opposed to the Val allele (Drabant et al. 2006; Smolka et al. 2005). Moreover, as the Met allele has been associated with violent suicide attempts and has a greater tendency toward external expression of anger (Rujescu et al. 2003), one would have expected to find this allele associated with a high level of anger. The results of direct association studies between COMT Val158Met and psychiatric disorders have been inconsistent with some involving the Met allele (Park et al. 2002; Rujescu et al. 2003) and others the Val allele (Glatt et al. 2003; Wonodi et al. 2003). Val158Met polymorphism certainly appears to have a pleiotropic effect on human behavior and various cognitive functions (Mier et al. 2009). Partly for this reason, several recent studies have focused on GxE rather than the principal genotypic effect. Although contrasting evidences have emerged from these findings, most of the interactions (ours included) conducted with this polymorphism provided evidence of synergism between the Val allele and environmental exposure (Caspi et al. 2005; Henquet et al. 2006). It can be assumed that both the Val and Met alleles are involved in the development of psychopathology through different interactions with specific environmental factors, but with more environmental susceptibility in Val carriers.

The inconsistencies in studies analyzing the effect of Val158Met polymorphism have been the subject of many debates. Some authors proposed that COMT Val158Met may modulate the balance of tonic and phasic dopamine function in different areas of the brain depending on specific environment (Bilder et al. 2004). According to this hypothesis, the Val allele is associated with decreased tonic dopamine but increased phasic dopamine neurotransmission subcortically. The predominance of phasic over tonic dopamine in Val allele carriers may explain reduced stability of neural networks but an increase in cognitive flexibility. The opposite effect is put forward for the Met allele (Nolan et al. 2004; Rosa et al. 2004). From this perspective, the Val allele, which is associated with increased COMT activity (Shield et al. 2004), may result in reduced dopamine neurotransmission in the prefrontal cortex associated with deficits in working memory, attention and executive functioning (Bilder et al. 2004; Goldberg et al. 2003). Our finding confirms the hypothesis that lower levels of tonic dopamine, associated with the Val high-activity allele, have less control over phasic (subcortical) dopamine release resulting in an abnormally high phasic dopamine response in the nucleus accumbens (Bilder et al. 2004; Meyer-Lindenberg et al. 2005). It is therefore tempting to presume that, with equal environmental exposure, stress-induced phasic dopamine release will be greater in Val allele carriers than in Met/Met individuals. This would also explain why most of the interactions found the Val allele to be associated with vulnerability to psychopathologies when exposed to adverse environmental conditions.

An inverted ‘U’-shape relationship between dopamine levels and prefrontal cortex function could provide another related explanation for the paradoxical findings (Mattay et al. 2003). In this model, it is postulated that both excessive and insufficient dopamine levels impair working memory performance and that only a narrow range of dopamine levels offer optimal functioning. The pleiotropic effect of Val158Met polymorphism could be explained by the position of each allele on this particular ‘U’-shaped curve, where the Val/Val genotype is positioned in a less favorable position compared with the Met allele carriers (Meyer-Lindenberg & Weinberger 2006; Mier et al. 2009).

In Goldman's ‘warrior/worrier’ model (Goldman et al. 2005), each allele is supposed to be maintained in the population because each confers an environmental-specific advantage. In Goldman's view, the Val allele is useful in threatening environments where maximal performance is required despite threat and pain (a warrior strategy), whereas the Met allele may be useful in complex environments where maximal performance is required in terms of memory and attention tasks (a worrier strategy). The high level of anger observed in Val allele carriers sexually abused in childhood could be the surviving response of this warrior strategy to this particularly threatening environment.

Val158Met polymorphism is not the only variation in the COMT gene. Considerable complexity in haplotypes and LD patterns exist across this gene and vary among populations around the world (Mukherjee et al. 2008). Inconsistencies regarding association studies may be because of the haplotypic combination of alleles comprising the Val158Met and several other single nucleotide polymorphisms (SNPs) across the gene. Several studies also showed that substratification in the investigated population could also be responsible for discrepancies between studies (Bray et al. 2003; Mukherjee et al. 2008). Given the recent investigations showing moderate stratification in even closely related populations, this hypothesis is not to be ruled out (Novembre et al. 2008).

Finally and as evidenced in our previous finding (Baud et al. 2007), the moderating effects of the COMT Val158Met may be sex specific. We have also highlighted that the observed interaction was present only in the female sample, which corroborates the idea of the sex specificity of this polymorphism. This could also partly explain the discrepancies noted in studies concerning Val158Met polymorphism.

Limitations

The measurement of childhood maltreatment was retrospective—hence recall bias may have influenced the reports. Sample size is another issue in GxE studies (Hunter 2005). In the present study, the power was enhanced by the accurate measurement of the phenotype and childhood maltreatment and the fact that we explored biologically plausible candidates. Moreover, our study is sufficiently powerful to detect such an interaction. Another limitation is that we only considered SNPs that reached a specific level of significance for the interaction analyses, excluding potential significant interactions.

Catecholamine-O-methyl-transferase Val158Met does not account for all the genetic variation in the COMT activity (Nackley et al. 2006) and, future studies genotyping further SNPs within the gene could help to disentangle the complex relationship between COMT gene and outcome. We however did not genotype other SNPs in this gene and were therefore not able to perform such analysis.

Although the direct association between COMT Val158Met and Anger Trait was significant, this association was not as strong as in our previous article (Baud et al. 2007). Indeed, when looking only at the new sample of suicide attempters (448 subjects), the results were not significant. In the present study, the two samples were pooled as there was no heterogeneity between them. However, in both samples, the interaction was significant and in the same direction. This suggests that it is crucial to take into account environmental factors to consistently detect a significant association.

Interestingly, sexual abuse did not significantly influence anger dimensions on itself. One explanation for this intriguing result could be that we are looking at an enriched population for history of childhood sexual abuse and anger-related traits, and therefore at the right end of the distribution for both variables. It could be therefore more difficult to highlight a potential association. The other possible explanation is that the retrospective investigation of sexual abuse could have increased either false positives or false negatives and secondarily led to a loss of significant association. See Appendix S1 for discussion on other polymorphisms.

Conclusions

To the best of our knowledge, this is the first study showing an interaction between COMT Val158Met polymorphism and childhood sexual abuse in modulating anger-trait levels in adulthood. These are important findings as they point toward sexual abuse as main environmental factors over other types of maltreatment, interacting with susceptibility genes. Recent studies suggest that the early environment acts through epigenetic modifications to modify behavior (McGowan et al. 2009; Weaver et al. 2004). It could be hypothesized that early and repeated exposure to adverse environmental factors may elicit permanent changes in gene expression patterns via epigenetic modifications. From this perspective, Val allele carriers could be more susceptible to these epigenetic modifications, which would explain why most of the studies found this allele to be associated with increased psychopathology when exposed to an adverse environment.

Acknowledgments

This study received financial support from the CHU Montpellier (PHRC UF 7653) and Agence Nationale de la Recherche (NEURO 2007—GENESIS). This study was also funded by the Swiss National Fund for Scientific Research grant 320000-112084 awarded to A.M. N.P. was funded by the Swiss National Science foundation: PASMA-118605. We would like to thank Camille Laurent and Claire Belloc for collecting the data. The authors have no conflicting financial interests.

Ancillary