Investigation of 17 candidate genes for personality traits confirms effects of the HTR2A gene on novelty seeking

Authors


A. Heck, Max Planck Institute of Psychiatry, Kraepelinstrasse 2-10, 80804 Munich, Germany. E-mail:heck@mpipsykl.mpg.de

Abstract

Genes involved in serotonergic and dopaminergic neurotransmission have been hypothesized to affect different aspects of personality, but findings from genetic association studies did not provide conclusive results so far. In previous studies, however, only one or a few polymorphisms within single genes were investigated neglecting the possibility that the genetic associations might be more complex comprising several genes or gene regions. To overcome this limitation, we performed an extended genetic association study analyzing 17 serotonergic (SLC6A4, HTR1A, HTR1B, HTR2A, HTR2C, HTR3A, HTR6, MAOA, TPH1, TPH2) and dopaminergic genes (SLC6A3, DRD2, DRD3, DRD4, COMT, MAOA, TH, DBH), which have been previously reported to be implicated with personality traits.

One hundred and ninety-five single nucleotide polymorphisms (SNPs) within these genes were genotyped with the Illumina BeadChip technology (HumanHap300, Human-1) in a sample of 366 mentally healthy Caucasians. Additionally, we tried to replicate our results in an independent sample of further 335 Caucasians. Personality traits in both samples were assessed with the German version of Cloninger's Tridimensional Personality Questionnaire.

From 30 SNPs showing associations at a nominal level of significance, two intronic SNPs, rs2770296 and rs927544, both located in the HTR2A gene, withstood correction for multiple testing. These SNPs were associated with the personality trait novelty seeking. The effect of rs927544 could be replicated for the novelty seeking subscale extravagance, and the same SNP was also associated with extravagance inthe combined samples.

Our results show that HTR2A polymorphisms modulate facets of novelty seeking behaviour in healthy adults suggesting that serotonergic neurotransmission is involved in this phenotype.

Dissecting the molecular genetic basis of personality dimensions has been in the focus of a large number of association studies. Especially genes involved in serotonergic and dopaminergic transmission have been hypothesized to affect different aspects of behaviour underlying personality traits. Cloninger suggested in his biosocial model of personality a connection between personality traits and monoaminergic neurotransmitter systems (Cloninger 1987), especially between the personality trait of novelty seeking (NS) and the dopaminergic system as well as harm avoidance (HA) and the serotonergic system. By using endocrine challenge test, several groups found support for such links (impairment of dopaminergic transmission and NS: Gerra et al.2000; serotonergic activity and HA: Gerra et al.2000; Hennig et al.2000; Ruegg et al.1997). However, the genetic validation of Cloninger's biosocial theory is lacking confirmation. First reports regarding the genetics of personality focussed on a restriction fragment length polymorphism in the promoter region to the serotonin transporter (SLC6A4) gene (HTTLPR) and on a 7-repeat allele of the variable number of tandem repeat (VNTR) polymorphism in the dopamine receptor 4 (DRD4) gene. These two polymorphisms were associated with anxiety-related traits (e.g. HA, neuroticism) and outward behaviour (e.g. NS, extraversion), respectively (Benjamin et al.1996; Lesch et al.1996) and stimulated a large number of replication studies which could not unambiguously confirm the initial findings (for meta-analysis, see Kluger et al.2002; Munafo et al.2003; Munafo et al.2005; Schinka et al.2002; Schinka et al.2004; Sen et al.2004).

Besides SLC6A4 and DRD4, growing evidence for associations with personality traits emerges for further candidate genes involved in serotonergic and dopaminergic neurotransmission (for reviews, see Ebstein 2006; Noblett & Coccaro 2005; Reif & Lesch 2003; Savitz & Ramesar 2004; Van Gestel & Van Broeckhoven 2003). These include genes involved in the metabolic inactivation of monoamines, the catechol-o-methyltransferase (COMT) gene and the monoamine oxidase A (MAOA) gene (Eley et al.2003; Samochowiec et al.2004; Stein et al.2005). Positive evidence also exists for the dopamine receptor 2 (DRD2) gene and for the serotonin receptor 2A (HTR2A) gene (Jonsson et al.2003; Serretti et al.2007). However, there are also negative results, e.g. for MAOA, DRD2, COMT and HTR2A (Garpenstrand et al.2002; Gebhardt et al.2000; Henderson et al.2000; Jonsson et al.2001).

Most studies in personality genetics examined so far only one or several variants within single genes neglecting the complex nature of genetic associations with personality traits. Furthermore, investigating only one or several polymorphisms is usually not sufficient to provide a complete coverage of the variations within a gene, which could result in non-replication because of population differences (Neale & Sham 2004).

To overcome these shortcomings, we performed an extended genetic association study in two independent samples of mentally healthy Caucasians. We included 195 single nucleotide polymorphisms (SNPs) from 17 candidate genes involved in serotonergic and dopaminergic neurotransmission, which have been previously reported as associated with personality traits, comprising genes related to synthesis, transport or degradation of dopamine and serotonin as well as corresponding receptor genes.

Methods

Screening sample

A total of 366 subjects were recruited at the Max Planck Institute of Psychiatry, Munich, as a healthy control group for the Munich Antidepressant Response Signature (MARS) (Binder et al.2004; Künzel et al.2003) project. The recruitment procedure is described in detail elsewhere (Heck et al.2009). Briefly, controls were selected randomly from a Munich population-based community sample and screened for the presence of alcohol dependence, drug abuse or dependence, possible psychotic disorder, mood disorder, anxiety disorder including obsessive-compulsive disorder (OCD) and post-traumatic stress disorder (PTSD), somatoform disorder, dissociative disorder not otherwise specified (NOS) and eating disorder using a modified version of the Munich Composite International Diagnostic Interview (DIA-X/M-CIDI) (Wittchen & Pfister 1997), an updated version of the World Health Organization's CIDI version 1.2 (World Health Organization 1992), which allows standardized assessment of symptoms, syndromes and diagnoses of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) (American Psychiatric Association 1994) disorders. Only subjects with a negative lifetime history for the above-mentioned disorders were included.

Ethnicity was assessed with a self-report questionnaire asking for nationality, first language and ethnicity of the subject and all four grandparents. All subjects are of Caucasian descent, 91.5% of them of German origin.

Replication sample

An independent replication sample (n = 335) was recruited at the Max Planck Institute of Psychiatry. All subjects were randomly selected from a Munich population-based community sample and were screened with the Composite International Diagnostic Screener (Wittchen et al.1999). Subjects negative for an affective or anxiety disorder were included in the sample. Less stringent inclusion criteria were therefore applied compared to the screening sample. In the screening sample 56.0% were female with a mean age of 48.59 (±11.02), whereas in the replication sample 65.4% were female with a mean age of 56.16 (±11.02).

Ethnicity was recorded using the same questionnaire as for the screening sample. Again, all participants were of Caucasian descent, and 93.1% of them were of German origin.

The protocol was approved by the ethics committee of the Ludwig Maximilians University Munich, and the study was conducted following the guidelines of the Declaration of Helsinki; written informed consent was obtained from all participants before study inclusion.

Candidate genes

Using the GeneSelectAssist search tool of the HuGE Navigator (http://www.hugenavigator.net), we selected 17 candidate genes with attributable function in the dopamine and serotonin neurotransmitter system, for which associations with personality traits were published according to the HuGE Navigator database and PubMed (Yu et al.2008). Our literature search resulted in six serotonin receptor genes (HTR1A, HTR1B, HTR2A, HTR2C, HTR3A and HTR6), three dopamine receptor genes (DRD2, DRD3, DRD4), the dopamine and serotonin transporter genes (SLC6A3, SLC6A4) and genes involved in the synthesis of dopamine [dopamine β-hydroxylase (DBH), tyrosine hydroxylase (TH)] and serotonin (TPH1, TPH2) as well as in the degradation of catecholamines (COMT, MAOA).

DNA extraction

At the time of enrolment, 40 ml of ethylenediaminetetraacetic acid (EDTA) blood was drawn from every subject. DNA was extracted from fresh blood using the Puregene® whole blood DNA extraction kit (Gentra Systems Inc., Minneapolis, MN, USA).

SNP genotyping

One hundred and ninety-five SNPs in 17 candidate genes including the coding regions and flanking sequence (±5 kb) were genotyped in the screening sample using Illumina BeadChip technology (HumanHap300, 300k, and Human-1, 100k), according to the manufacturer's standard protocols (Illumina Inc., San Diego, CA, USA).

SNPs on HumanHap300 had been chosen to tag haplotype blocks identified by Phase I HapMap covering genetic variation with r2≥ 0.80 for bins within 10 kb of genes or in evolutionarily conserved regions whereas SNPs on Illumina BeadChip Human-1 are located in exons or within 10 kb of transcripts (70%) or in highly conserved regions (15%), whereas only 15% are selected to provide gene coverage.

To control for gene coverage we compared a tag SNP selection for the same loci of the HapMap CEU (Central European Population). We used the Tagger software implemented in HapMap (http://www.hapmap.org) and applied a forced choice method by including the Illumina SNPs as tagging SNPs using the pairwise tagging method and tested which fraction of structural variation was covered by the Illumina SNPs (de Bakker et al.2005). We chose a minor allele frequency (MAF) threshold of 0.05 and an r2 threshold of 0.75 and used the HapMap data from the latest HapMap release, Phase II Release 23a.

Genotyping in the replication sample was performed on MALDI-TOF mass spectrometer (MassArray® system) using the Spectrodesigner software (Sequenom Inc., San Diego, CA, USA) for primer selection and the homogeneous mass extension (hMe) process for producing primer extension products. All primer sequences are available on request.

As quality criteria we chose a MAF threshold of 10% and a call rate of at least 99%, i.e. a SNP was excluded if more than 1% of all individuals failed to have genotypic information for this SNP. For all polymorphisms, exact tests for Hardy-Weinberg Equilibrium (HWE) were performed (Wigginton et al.2005). For all SNPs, P values for deviation from HWE higher than 2.6 e-04 (0.05/195, Bonferroni corrected P value) were obtained. Therefore, no SNPs were discarded from analysis because of Hardy-Weinberg deviance.

To test the possibility of potential population stratification effects, we applied the eigenstrat program (http://genepath.med.harvard.edu/~reich/EIGENSTRAT.htm) using all available genotype data both in the screening and in the replication sample. eigenstrat performs a principal component analysis on the genotypic data yielding continuous axes of variation. Each individual has a specific position on each of these axes. The positions on the top 10 axes of variation, i.e. the 10 axes that explain most of the genotypic variability, were compared. For each axis, mean position and standard deviation for the position, based on all individuals, were calculated. We classified individuals as outliers whose position was at least 6 SDs from the mean position on at least one of these axes. However, no such individual was identified and hence no evidence for population stratification was detected (data not shown).

Phenotypes

In the screening and in the replication sample, personality was assessed with the German version of Cloninger's Tridimensional Personality Questionnaire (TPQ) (Cloninger 1987; Krebs & Weyers 1998) The TPQ is a widely used self-report questionnaire assessing personality with 100 items each answered with Yes or No. Cloninger's TPQ personality concept is based on a biosocial theory of personality and including three temperament dimensions of personality: novelty seeking (NS), harm avoidance (HA) and reward dependence (RD) which can be further divided into four subscales each: exploratory excitability (NS1), impulsiveness (NS2), extravagance (NS3), disorderliness (NS4), anticipatory worries and pessimism (HA1), fear of uncertainty (HA2), shyness with strangers (HA3), fatigability and asthenia (HA4), sentimentality (RD1), persistence (RD2), attachment (RD3) and dependence (RD4).

Statistical analysis

All genetic association analysis were performed using WG PERMER software (http://www.wg-permer.org/) that calculates analysis of variance (anova) for assessing genetic associations between genotypes and phenotypes and allows for testing different models of inheritance.

To encounter potential violations of the anova requirement of normal distribution, the phenotypes were ranked by applying the rank transformation procedure as implemented in the statistical software package SPSS (release 12.0.1 SPSS Inc., Chicago, IL, USA). Rank scores were corrected for the effects of gender and age by multiple regression analysis using the corrected residuals for the final analysis. This procedure has been shown to have the same power, but superior robustness comparable to a standard analysis of covariance (ancova) (Conover & Iman 1982).

The genotypic and the allelic models of inheritance were calculated in a first step, with the factors of the TPQ NS, HA and RD as phenotypes. For TPQ factors showing an association we also analyzed the corresponding subscales to identify the special facets of the second-order factor explaining the association. In case of a significant genotype effect, we tested for the following genetic models of inheritance: recessive/dominant model and the heterozygote advantage/disadvantage model. Uncorrected P values in the screening and in the replication sample were corrected for multiple testing with the permutation-based Westfall and Young method using 10 000 permutations, which is also implemented in the WG PERMER software. This correction method takes the Linkage Disequilibrium (LD) structure between single SNPs into account (Westfall & Young 1993).

We additionally performed a gene-wise multivariate analysis treating the gene as a functional entity (Neale & Sham 2004). For this we applied the permutation-based Fisher product method (FPM) using 100 000 permutations, which calculates the empirical P value for a gene as the proportion of cases in which the product of permuted P values within a gene is less or equal to the product of observed P values These FPM P values were then corrected for multiple testing applying Bonferroni correction for the number of genes tested.

For SNPs of genes mapping on the X chromosome (MAOA, HTR2C), males and females were analyzed separately: the allelic model was calculated in both genders whereas the genotypic model was calculated for females only.

Polymorphisms that showed an association in the initial analysis withstanding correction for multiple testing were selected for replication in an independent sample of unrelated healthy controls and these SNPs were also analyzed after combining both samples. As a measure of effect size for quantitative phenotypes Cohen's f was calculated (Cohen 1988).

Differences in potentially confounding variables between both samples were evaluated with the Pearson's Chi-square statistics for qualitative data (gender) and with anova for quantitative data (age). ancova controlled for age and sex was used to investigate differences in personality factors between both samples. All non-genetic analyses were performed with SPSS (release 12.0.1 SPSS Inc., Chicago, IL, USA).

Results

Associations with personality traits in the screening sample

Of the 195 SNPs genotyped in the screening sample, four SNPs were monomorphic and excluded from further analyses. Forty-nine SNPs did not fulfill our quality criteria as they had either a MAF of less than 10% (19 SNPs) or a genotyping call rate of less than 99% (30 SNPs). This resulted in 142 SNPs entering the analyses. For two genes (DRD4, HTR1B) no SNP fulfilled our quality criteria. SNP IDs, map positions (according to University of California Santa Cruz (UCSC) genome built version hg18), genes, chromosomes, alleles, P values of the test for HWE deviation, MAFs, call rates and gene coverage are reported in Supplementary Table S1.

Single marker analysis

Thirty SNPs in eight genes showed nominally significant (Puncorrected< 0.05) associations with personality traits. Two intronic SNPs, rs2770296 and rs927544, located in the HTR2A gene showed an association with NS (Pcorrected = 0.03 and 0.049; effect size f = 0.21 and 0.21, respectively) withstanding Westfall-Young correction for multiple testing (see Table 1).

Table 1. F values (anova) and nominal P values for SNPs in the HTR2A gene in the screening sample
Screening sample (n = 366)rs2770296rs927544rs731779
Genotypic modelAllelic modelGenotypic modelAllelic modelGenotypic modelAllelic model
F*P valueF*P valueFP valueFP valueFP valueFP value
  1. Bold letters: significant after Westfall-Young correction for multiple testing.

  2. *F2,363 genotypic model; F1,730 allelic model.

Novelty seeking (NS)8.300.00031.080.307.690.00050.730.396.170.00070.020.90
Exploratory excitability (NS1)2.370.091.650.203.960.021.900.173.400.020.0010.98
Impulsiveness (NS2)4.080.022.140.143.610.032.230.142.800.071.260.26
Extravagance (NS3)5.090.0072.990.086.900.0013.870.054.560.0084.790.03
Disorderliness (NS4)4.740.0091.210.272.200.110.840.361.980.100.230.63
Harm avoidance (HA)4.100.020.710.403.280.110.940.331.310.251.780.20
Reward dependence (RD)0.100.90.180.700.530.660.100.751.490.330.030.85

For both SNPs, the heterozygote advantage/disadvantage model provided the best fit for the data with heterozygous subjects showing lower NS scores than both homozygous groups (Pcorrected = 0.01 and 0.03; f = 0.20 and 0.19, respectively; see Fig. 1).

Figure 1.

Novelty seeking scores of rs2770296 and rs927544 genotypes in the screening sample. Means and standard errors of the means are shown.

When considering the subscales of the NS factor, the strongest effects were observed for the subscale extravagance, which, however, failed to withstand correction for multiple testing (Pcorrected = 0.5 and 0.10; f = 0.17 and 0.20, respectively). The two SNPs, rs2770296 and rs927544, were found to be in moderate linkage disequilibrium (r2 = 0.65). Another intronic SNP, rs731779, also mapping on the HTRH2 gene, showed a trend for significance after correction for multiple testing for an association with NS (Pcorrected = 0.06,f = 0.20). Rs731779 lies in one block with the best SNP rs2770296 (r2 = 0.57), and is also in moderate LD with rs927544 (r2 = 0.57). Demographic variables (age, gender) did not differ between the genotype groups of the above-mentioned SNPs (P > 0.05).

The FPM representing the combined effect of all SNPs within one gene showed several nominal significant associations, sc. HTR2A with NS (PFPM = 0.002), TH with HA (PFPM = 0.02) as well as DRD2 with RD (PFPM = 0.04). After Bonferroni correction for multiple testing, only the association with HTR2A and NS remained significant (PFPMcorrected = 0.03). The FPM thus supports the results of the single marker analysis suggesting a general association between genetic variants in the HTR2A gene and NS.

Replication analysis sample and combined results

Screening and replication samples significantly differed in both age (F = 65.8, df = 1,P < 0.001) and in gender distribution (X2 = 6.4, df = 1,P = 0.013). Considering these differences, we corrected for the effects of age and gender statistically using adjusted residuals for the final analysis.

Two intronic SNPs in the HTR2A gene, rs927544 and rs2770296 fulfilled our criterion for entering replication (Pcorrected< 0.05). As these SNPs were associated with NS in the screening sample we analyzed the effects of these genotypes on NS and corresponding subscales. Although the two samples differed in the scores of some subscales of the TPQ, no differences emerged for the NS scales (Table 2).

Table 2.  Scores of the TPQ factors and TPQ subscales for the screening and the replication sample. Data are given as mean (SD). P values are calculated by analysis of covariance (ancova), controlled for age and sex
 Screening sample (n = 366)Replication sample (n = 335)F1,697P value
TPQ scales /subscales    
Novelty seeking (NS)14.96 (4.84)14.25 (5.13)0.000.994
  Exploratory excitability (NS1)4.24 (1.77)3.90 (1.86)0.2050.651
  Impulsiveness (NS2)3.52 (1.94)3.45 (1.93)0.0150.901
  Extravagance (NS3)3.77 (1.74)3.54 (1.80)0.000.989
  Disorderliness (NS4)3.43 (1.79)3.37 (1.91)0.1600.689
Harm avoidance (HA)10.86 (5.22)12.30 (5.69)8.120.005
  Anticipatory worries and pessimism (HA1)3.28 (2.11)3.63 (2.19)3.400.046
  Fear of uncertainty (HA2)3.50 (1.79)4.11 (1.90)9.250.002
  Shyness with strangers (HA3)2.24 (1.73)2.23 (1.75)0.510.821
  Fatigability and asthenia (HA4)1.83 (1.70)2.33 (1.90)11.80<0.001
Reward dependence (RD)16.27 (4.53)15.60 (4.21)0.2530.615
  Sentimentality (RD1)3.69 (1.26)3.94 (1.18)1.000.315
  Persistence (RD2)3.91 (2.12)3.39 (2.07)1.420.234
  Attachment (RD3)6.17 (2.61)6.20 (2.48)1.380.241
  Dependence (RD4)2.49 (1.38)2.43 (1.36)0.1360.712

We applied the same genetic model (heterozygous disadvantage/advantage model) since it had provided the best fit for our data in the screening sample. Even though we could not replicate the associations with the factor NS (Pcorrected> 0.8 and 0.9, respectively), we found a significant association between rs927544 and the NS subscale extravagance (NS3; Pcorrected = 0.04;f = 0.12; see Table 3).

Table 3. F values (anova) and nominal P values for SNPs rs927544 and rs2770296 in the replication and in the combined sample (heterozygote advantage/disadvantage model)
VariableReplication sampleCombined sample
rs927544rs2770296rs927544rs2770296
F 1,327P valueF1,327P valueF1,693P valueF1,693P value
  1. Bold letters: significant after Westfall-Young correction for multiple testing.

Novelty seeking (NS)0.260.610.190.678.660.0038.960.003
Exploratory excitability (NS1)0.400.530.100.752.670.101.760.19
Impulsiveness (NS2)0.750.390.650.421.690.196.950.009
Extravagance (NS3)5.380.020.030.868.850.0030.910.35
Disorderliness (NS4)0.100.750.100.752.680.105.790.02

This is the same subscale for which the strongest association in the screening sample could be observed. The direction of the effects was the same in both samples with the heterozygous group having lowest scores on this subscale (see Fig. 2).

Figure 2.

Extravagance scores of rs927544 genotypes in the screening sample (A) and in the replication sample (B). Means and standard errors of the means are shown.

When both samples were combined, associations between the NS phenotypes and rs927544 were observed (see Table 3) with a similar effect size for the factor NS (Pcorrected = 0.007;f = 0.11) and for the subscale extravagance (NS3; Pcorrected = 0.006;f = 0.11). Additionally, the combined data showed an association for rs2770296 and the factor NS (Pcorrected = 0.005;f = 0.11) as well as for the NS subscales impulsiveness (NS2; Pcorrected = 0.01;f = 0.10) and disorderliness (NS4; Pcorrected = 0.04;f = 0.10).

Discussion

We presented the results of an extended candidate gene association study for personality traits assessed with Cloninger's TPQ, a widely established self-report questionnaire. We investigated 195 SNPs from 17 candidate genes related to serotonergic and dopaminergic neurotransmission that were selected because of prior reports of associations with personality traits. Genetics associations were evaluated in a sample of healthy Caucasians, and associations were analyzed in an independent replication sample.

We found two SNPs of the HTR2A gene, rs2770296 and rs927544, to be associated with NS withstanding correction for multiple testing. One further SNP, rs731779, located in the same gene, showed an association with the same trait with a trend for significance. Effect size measures suggested a small to medium effect according to the guidelines by Cohen (Cohen 1988). The link between HTR2A and NS is further supported by the gene-wise analysis, as the association between HTR2A and NS was the only association that remained significant after correction for multiple testing.

We investigated the two significant SNPs in a second sample of healthy Caucasians. Even though we could not replicate the associations with the factor NS, we found a significant association between rs927544 and the NS subscale extravagance, for which the strongest subscale association also emerged in the initial sample. Combining both samples further supported the associations between rs927544 with extravagance as well as with the second-order factor NS with medium effects according to the guidelines by Cohen. Subjects scoring high on the extravagance scale are prone to be extravagant in spending money, their energy and their emotions and are described as chivalric characters. Also rs2770296, which showed no association in the replication sample, was associated with the factor NS as well as with its two corresponding subscales, impulsiveness and disorderliness in the combined sample. Subjects scoring high on the impulsivity scale are described as excitable individuals coming quick to decisions, whereas subjects scoring high on the disorderliness subscale are described as capricious characters that prefer activities without rules and regulations. It is notable, that our results are contradictory to Cloninger's hypothesis of a link between NS and the dopaminergic system but indeed are in line with several reports of associations between NS and other serotonergic genes, particularly with the serotonin transporter SLC6A4 (Serretti et al.2006; Suzuki et al.2008). No association was found with HA or RD. As the extended version of the TPQ, the Temperament and Character Inventory (TCI, Cloninger et al.1993) employs a reduced version of reward dependence scale excluding the RD2 subscale persistence, we reanalyzed our data according to the TCI scale structure, but did not find any effects neither for the reduced reward dependence scale nor for the subscale persistence (data not shown).

Few studies have reported associations between HTR2A polymorphisms and NS or related personality dimensions, and those studies are mainly restricted to patient samples thus not directly comparable to our findings in healthy individuals. Ni et al.(2006) report for a small subsample of 71 patients with borderline personality disorder that the C-allele of rs6313 and the A-allele of rs4941573 were associated with higher extraversion scores as measured with the revised NEO Personality Inventory (Ni et al.2006). Serretti et al.(2007) report on a weak association between rs6313 and NS measured with the Tridimensional Character Inventory in a small group (n = 60) of Italian mood disorder patients. Patients homozygous for the A-allele show higher NS scores than the other groups (Serretti et al.2007). HapMap data show that SNP rs6313 is not in strong LD to the SNPs rs2770296 (r2 = 0.30) and rs927544 (r2 = 0.26) associated in our study. For SNP rs4941573 associated with higher extraversion in Ni et al.(2006), no associations emerged in our sample with the related phenotype NS (Puncorrected = 0.6; data not shown). In a sample of panic patients, Unschuld and colleagues observed associations between several HTR2A SNPs and reward dependence but not with NS (Unschuld et al.2007). According to HapMap data, the LD structure between the best SNP in this study (rs2770304) and the associated SNPs in our study is at best weak. Using statistical imputation methods the genotypes of two functional HTR2A SNPs, rs6311 and rs6314, that have been shown to be associated with schizophrenia and memory function (de Quervain et al.2003; Spurlock et al.1998), did not show any association with NS in our study (data not shown).

To our knowledge, only one study investigating HTR2A polymorphisms in healthy subjects was performed reporting an association between rs6313 and HA in a Russian sample (Golimbet et al.2004). Noteworthy, the model of inheritance is in line with the model of inheritance we found for our associated SNPs, with the heterozygous genotype significantly differing from both homozygous genotypes. At this point it is important to note that our associated SNPs are not functional (contrary to the SNPs reported by Golimbet et al.), and that the heterozygote advantage/disadvantage model is a rather counterintuitive model for genetic associations. Nevertheless, evidence exists that this phenomenon of molecular heterosis exists for several genes that are reported for associations with personality traits, including the HTR2A gene showing associations with neuroticism (Costa & McCrae 1996). A possible explanation could be that an independent third factor might operate on this phenotype–genotype correlation (Comings & MacMurray 2000).

Although HTR2A is an important candidate gene for mental disorders (review: Norton & Owen 2005) and has been reported to affect antidepressant treatment outcome (McMahon et al.2006; Peters et al.2004), our findings suggest that polymorphisms of HTR2A also play a role in NS behaviour in healthy subjects. A recent animal study showed that disruption of HTR2A receptor signalling in mice enhances the drive to explore novelty in a behavioural test incorporating conflict anxiety but not in emotion-related behavioural tests (Weisstaub et al.2006), thus supporting that HTR2A exerts direct effects on NS behaviour. Given the fact that the exploration of a novel environment is partly mediated by serotonin mediated neurotransmission (Ray et al.2006) one could speculate that the HTR2A polymorphisms in humans lead to a downregulation in HTR2A expression thus dampening receptor function which in turn could facilitate the development of NS like traits.

A few limitations of the present study should be mentioned: all associated SNPs are located in intronic regions. Since we do not have expression or protein data with respect to these genotypes, the functional relevance of these SNPs is not clear. Bioinformatic data (http://www.genome.ucsc.edu/) did not provide evidence for conserved transcription factor binding sites for rs927544 and rs2770296. The LD structure in this gene region, however, suggests that these SNPs might be linked with a yet unknown functional variant requiring thorough resequencing of this gene region in future studies.

Another potential shortcoming is the failure to screen for personality disorders as further exclusion criterion besides presence of mental axis I disorders. However, inspection of the distribution of the evaluated personality traits did not provide any evidence for outliers potentially resulting from subjects with a personality disorder. In addition, a recent study suggested no effects of personality disorders on Extraversion (Bunce et al.2005) that is closely related to NS, for which we have observed genetic associations.

One further limitation of the actual study is that because of limitations in genotyping resources we did not genotype non-SNP polymorphisms, e.g. 5-HTTLPR in the SLC6A4 gene, and did not include several coding polymorphisms within the investigated gene regions. However, it is important to mention that the scope of the present study rather was to achieve a sufficient coverage of the genetic region, regardless of functionality of the investigated polymorphisms, which was achieved with the SNPs selected for the screening analysis. Nevertheless, we used imputation methods to estimate the genotypes of additional non-synonymous and synonymous SNPs within the investigated genes, for which we could not detect any further significant findings withstanding correction for multiple testing (data not shown).

One should also have in mind that, although we used permutation methods to control the experiment-wise error rate and investigated a limited number of hypotheses, our results still harbour the possibility of a false-positive finding and should be regarded as tentative knowledge as suggested by Sullivan (2007).

In conclusion, we presented the results of the first extended candidate gene association study searching for serotonergic and dopaminergic genes associated with personality in healthy adults. Our findings show that variants of the HTR2A gene modulate facets of NS behaviour suggesting that serotonergic neurotransmission is involved in this phenotype.

Acknowledgments

The authors would like to thank Elisabeth Binder, Sonja Horstmann, Stefan Kloiber, Angelika Erhardt and Hubertus Himmerich for their valuable help in performing the study; Sabine Damast, Gertrud Ernst-Jansen, Gisela Gajewsky, Melanie Hartung, Johannes Huber and Susanne Sauer for their excellent technical assistance. Genotyping was supported by the Excellence Foundation of the Max Planck Society.

Ancillary