High trait aggression in men is associated with low 5-HT levels, as indexed by 5-HT₄ receptor binding

Abstract

Impulsive aggression has commonly been associated with a dysfunction of the serotonin (5-HT) system: many, but not all, studies point to an inverse relationship between 5-HT and aggression. As cerebral 5-HT₄ receptor (5-HT₄R) binding has recently been recognized as a proxy for stable brain levels of 5-HT, we here test the hypothesis in healthy men and women that brain 5-HT levels, as indexed by cerebral 5-HT₄R, are inversely correlated with trait aggression and impulsivity. Sixty-one individuals (47 men) underwent positron emission tomography scanning with the radioligand [¹¹C]SB207145 for quantification of brain 5-HT₄R binding. The Buss–Perry Aggression Questionnaire (BPAQ) and the Barratt Impulsiveness Scale were used for assessment of trait aggression and trait impulsivity. Among male subjects, there was a positive correlation between global 5-HT₄R and BPAQ total score (P = 0.037) as well as BPAQ physical aggression (P = 0.025). No main effect of global 5-HT₄R on trait aggression or impulsivity was found in the mixed gender sample, but there was evidence for sex interaction effects in the relationship between global 5-HT₄R and BPAQ physical aggression. In conclusion we found that low cerebral 5-HT levels, as indexed by 5-HT₄R binding were associated with high trait aggression in males, but not in females.

Introduction

Aggression and impulsivity play a critical role in the manifestation of violent and criminal behaviours, thereby posing large costs to the victims and the society. Trait aggression identifies a personality with a propensity for hostile cognition, angry affect and readiness to engage in physical and verbal aggression (Buss and Perry, 1992). Impulsivity is characterized by behavioural disinhibition, attention deficits and lack of planning (Chamorro et al., 2012). It is well documented that men and women differ in terms of aggressive behaviour (de Almeida et al., 2015). Sex differences in aggression are often explained in the context of evolutionary and social role theories (Bettencourt and Miller, 1996; Cross, 2010), and appear to be especially pronounced in physical aggression (Archer, 2004). In humans, the biological underpinnings of this difference are only partly explained by the male sex hormone testosterone (Book et al., 2001), whereas additional possible neurobiological mechanisms are less well understood.

Serotonin (5-HT, 5-hydroxytryptamine) is a neurotransmitter known to be involved in the regulation of emotion and behaviour, including the regulation of aggressive impulses (Davidson et al., 2000). It was early shown that the primary 5-HT metabolite, 5-hydroxyindoleacetic acid (5-HIAA), was low in the cerebrospinal fluid of aggressive individuals, suggesting an inverse relationship between cerebral 5-HT levels and impulsive aggression (Brown et al., 1979; Linnoila et al., 1983). This was later backed up by numerous studies across a range of methodologies, leading to a ‘5-HT deficiency hypothesis’ of aggression (Duke et al., 2013). However, in a recent meta-analysis the combined weighted correlation between 5-HT, aggression, anger and hostility was found to be low, with only about 1.2% of the trait being explained by 5-HT (Duke et al., 2013). Importantly, this meta-analysis did not include any neuroimaging studies to assess 5-HT function, and the small effect size may reflect methodological limitations. The 5-HT deficiency hypothesis remains to be tested using improved methodologies, such as positron emission tomography (PET) for indexing features of the 5-HT system. It also remains to be elucidated which specific parts of the 5-HT transmitter system are involved in impulsive aggression, and whether sex differences in aggression can be explained by 5-HT function.

The 5-HT₄ receptor (5-HT₄R) is a component of the 5-HT system that is monotonically and inversely correlated with brain 5-HT levels. This was first observed in preclinical studies where cerebral 5-HT₄R binding reflected 5-HT tone (Licht et al., 2009), supported by a clinical study where carriers of the low-expressing variant of the 5-HT transporter (5-HTT)-gene had lower 5-HT₄R binding (Fisher et al., 2012). This inverse association between 5-HT₄R binding and stable extracellular levels of 5-HT was recently further corroborated in an intervention study where 5-HT₄R binding adjusted to changes in chronic 5-HT levels after a 3-week selective 5-HT reuptake inhibitor (SSRI) intervention in healthy volunteers (Haahr et al., 2014), whereas no changes in 5-HT₄R binding were observed after acute administration of SSRI (Marner et al., 2010).

There are currently no studies suggestive of 5-HT₄Rs being specifically involved in the regulation of aggression or impulsivity. Therefore, we used brain 5-HT₄R PET imaging data as an in vivo biomarker of central 5-HT tonus to test whether trait aggression and trait impulsivity are associated with cerebral 5-HT tonus in a cohort of healthy adults. In view of the ‘5-HT deficiency theory’ of aggression, we hypothesized that low cerebral levels of 5-HT, i.e. high 5-HT₄R binding, would be associated with high levels of trait aggression and impulsivity. We used 18 regions to calculate a global volume-weighted 5-HT₄R binding potential (BP) as a biomarker of central 5-HT tonus, which served as our primary predictor of trait aggression and trait impulsivity. In light of previous research indicating sex differences in both 5-HT function (Madsen et al., 2011a; Kaufman et al., 2015) and aggression (Archer, 2009), we also tested for sex differences in the relationship between 5-HT₄R and aggression. Finally, a post hoc voxel-based analysis was performed to determine whether there were any region-specific effects.

Materials and Methods

Participants

Subjects were included from a group of healthy volunteers that underwent 5-HT₄R PET imaging with the radioligand [¹¹C]SB207145. The subjects were healthy according to medical history, physical examination, blood biochemistry and magnetic resonance imaging (MRI) of the brain. The exclusion criteria were as follows: (i) primary psychiatric disease, (ii) substance or drug abuse, (iii) severe systemic or neurological disease and (iv) current use of antipsychotics or antidepressants. On the day of the PET scan, subjects completed the Symptom Checklist Revised (SCL-90) to assess symptoms of psychopathology. Two participants were excluded due to high [¹¹C]SB207145-injected mass (>6 μg) (Madsen et al., 2011b), and the final dataset consisted of 61 healthy subjects (47 men, mean age 33.8 ± 16.3 s.d., 20–86 years).

The subjects had been recruited for different research projects approved by the local ethical committee (H-1-2010-085, KF 01 274821, KF 01 2006-20, H-KF-2007-0028); these outcomes dealt with 5-HT₄R methodological developments (Marner et al., 2009), age and sex effects (Madsen et al., 2011a), effects of SSRI intervention (Haahr et al., 2014) and effects of genetic variants on 5-HT4 receptor binding (Fisher et al., 2015). The studies were carried out in accordance with the latest version of the Declaration of Helsinki and written informed consent was obtained from all subjects after the procedures had been fully explained.

Assessment of aggression and impulsivity

Danish versions (da Cunha-Bang et al., 2013) of the Buss–Perry Aggression Questionnaire (BPAQ) (Buss and Perry, 1992) and the Barratt Impulsiveness Scale version 11 (BIS) (Patton et al., 1995) were used for assessment of trait aggression and trait impulsivity. The BPAQ entails four dimensions of aggression; physical aggression, verbal aggression, anger and hostility. The BIS consists of three subscales including motor impulsivity, attentional impulsivity and non-planning impulsivity. The total scores represent a global measure of trait aggression and trait impulsivity. The Cronbach’s alpha coefficients of the BPAQ and BIS were 0.92 and 0.86, respectively, indicating good internal consistency and is in line with what we have previously reported (da Cunha-Bang et al., 2013).

Subjects that were scanned in the period from 2010 to 2013 (n = 32) completed the questionnaires around the time of scanning (median number of days between scanning and questionnaire responding: 12, interquartile range: 3; 27, range: 0–196). Subjects that were scanned in the period from 2006 to 2009 (n = 29) completed the questionnaires in 2010, upon invitation (median number of days between scanning and questionnaire responding: 795, interquartile range: 610; 1085, range: 217–1562). Participants who were scanned after 2010 completed the BPAQ and BIS as part of a battery of questionnaires assessing their personality.

Genotyping

Sixty-one datasets within the current study were previously included in a study relating genetic variants and 5-HT₄R binding (Fisher et al., 2012, 2015). Participants were genotyped for the tri-allelic 5-HTT-linked polymorphic region (5-HTTLPR) polymorphism and dichotomized as carriers of one of the two low-expressing alleles; the short allele (S-carriers) or the long Lg allele (Lg-carriers) (n = 38) or homozygotes of the high-expressing long La allele (La-homozygotes) (n = 23).

Sex hormones

Blood samples were collected on the day of the PET scan and extracted plasma was analysed using Cobas Testosterone II and Estradiol II-kits on a modular E170 (Roche Diagnostic) performed by the hospitals biochemical laboratory (LABKA, Rigshospitalet). Plasma from participants that were scanned after March 2011 (n = 24) was analysed immediately after sampling, and plasma from participants that were scanned before March 2011 (n = 28) was stored at−20°C until analysed using the same method in December 2011. Sex hormone measurements were available for 52 subjects (45 men, 7 women).

PET and magnetic resonance imaging

Acquisition and analysis of neuroimaging data have previously been described in details in Haahr et al. (2014) and Madsen et al. (2011b); therefore, a brief summary is provided below. After intravenous bolus injection of [¹¹C]SB207145 (mean radioactivity dose 530.5 ± 122.1 MBq), a 120 min dynamic PET scan was conducted on either an 18 ring GE-Advance scanner (n = 18) or a High Resolution Research Tomograph (HRRT) (n = 43). MRI was performed using a 3T Siemens Magnetom Trio scanner. Regions were automatically delineated on each subject’s MRI in a user-independent fashion with the Pvelab software package (Svarer et al., 2005). Regional time activity curves were constructed either with partial volume (PV) correction (GE-Advance scans) for a point-spread-function of 8 × 8 × 8 mm (Muller-Gartner et al., 1992), or without PV-correction (HRRT scans) because of the lower resolution of the GE Advance scanner. The quantitative analysis to obtain the BP of 5-HT₄R was performed with the simplified reference tissue model (SRTM), using cerebellum (excluding vermis) as reference region (Marner et al., 2009). This model estimates the non-displaceable BP (BPnd), which is defined as BPnd = f_nd * B_avail/K_d, where f_nd is the free fraction of tracer in the non-displaceable tissue compartment, B_avail is the number of receptors available for binding and K_d is the equilibrium dissociation constant for the tracer. The PMOD (version 3.0) software was used for kinetic modelling. A whole brain [¹¹C]SB207145 BPnd was calculated based on 18 brain regions (amygdala, anterior cingulate cortex, caudate, entorhinal cortex, hippocampus, hypothalamus, thalamus, insula, medial inferior frontal gyrus, medial inferior temporal gyrus, occipital cortex, orbitofrontal cortex, parietal cortex, posterior cingulate cortex, putamen, sensorymotor cortex, superior frontal gyrus and superior temporal gyrus) by volume weighting grey matter segmented brain region BPnd’s: Global 5-HT₄R binding = Σ [5-HT₄R BPnd(region_x) * volume(region_x)]/[Σvolume(region_x)]. We chose whole brain 5-HT₄R binding as our primary predictive variable for trait aggression and impulsivity because a global measure reflects cerebral 5-HT tonus, which we hypothesized to be inversely correlated with trait aggression and impulsivity. As we had no a priori hypotheses that the 5-HT₄R would be specifically involved in aggression, post hoc we instead used voxel-based analysis to evaluate possible regional effects.

Statistical analysis

The effect of global 5-HT₄R binding on scale measures of aggression and impulsivity was evaluated in a multiple linear regression model, adjusting for age, sex, PET scanner type and 5-HTTLPR status (S or Lg carriers vs La homozygotes). Age and sex were included in the models as they are known to influence aggression (Archer, 2004) and age, sex, PET scanner type and 5-HTTLPR were included as they are known to influence 5-HT₄R binding (Madsen et al., 2011a; Fisher et al., 2012). Body mass index was not significantly associated with BPAQ and BIS scores or the global measure of 5-HT₄R, adjusted for age, gender, 5-HTTLPR and PET scanner type, and was therefore not included as a covariate.

In post hoc analyses, BPAQ and BIS subfacets were evaluated separately, adjusting for age, 5-HTTLPR-status and PET scanner type. Correlations between testosterone and BPAQ and BIS were also evaluated in men and women separately. As trait aggression and trait impulsivity decrease with age (effect of age on BPAQ: −0.16 scores per year, CI: −0.43; 0.12, P = 0.26, effect of age on BIS: −0.23 scores per year, CI: −0.37; −0.08, P = 0.002), we also confined the analyses to include only individuals younger than 50 years. We tested for sex-by-5-HT₄R binding interaction effects on BPAQ and BIS, adjusting for age, PET scanner type and 5-HTTLPR status.

Statistical analyses were performed using SAS software version 9.4 and R version 3.1. Model assumptions were tested graphically by examination of the distribution of the residuals, Q–Q plots (quantile probability plot) and predicted values plotted against residuals, none of which indicated substantial violation of assumptions. Significance level was set at a P-value of 0.05 and P-values are reported without multiple comparison adjustment.

Voxel-based analysis

To further explore the relationship between 5-HT₄R binding and trait aggression observed in men (results section), a voxel-based analysis was conducted in the male sample in order to detect possible regional effects. Only scans from the HRRT scanner were included in this analysis.

Parametric BPnd maps were generated similar to previously described procedures (Haahr et al., 2013). Briefly, voxel-level time-activity curves were smoothed with a 6 mm full-width half-maximum Gaussian kernel to account for spatial uncertainty. Voxel-level BPnd values were estimated using the basis function implementation of SRTM. Single-subject BPnd maps were warped to Montreal Neurological Institute (MNI) space within SPM8 (http://www.fil.ion.ucl.ac.uk/spm) based on normalization of the single-subject high-resolution T1-weighted structural image previously co-registered into PET space during Pvelab. Voxel size of normalized BPnd maps was 2 mm isotropic. Voxels with an average BPnd > 0.3 were analysed within a brain mask of 136 972 voxels. A voxel-wise multiple regression was performed with BPAQ as the predictive variable, adjusting for age and 5-HTTLPR status. Concerning the issue of multiple comparisons, 3dClustSim, a program within AFNI (http://afni.nimh.nih.gov/afni) that uses a Monte Carlo simulation method, was used to determine a cluster extent threshold unlikely to have occurred by chance (α < 0.01). The cluster extent threshold for a whole brain search volume given a voxel-level threshold of P < 0.005 uncorrected was k > 240 voxels. All voxel coordinates are given in MNI space.

Results

Participant characteristics, psychometric data and scan parameters are shown in Table 1. As expected, the total scores of the BIS and BPAQ were weakly positively correlated (Spearman’s rho = 0.34, P = 0.006). BIS was significantly higher in men than in women, but no sex difference was found for BPAQ. Age-adjusted plasma testosterone did not significantly predict neither BPAQ (males: P = 0.61, females: P = 0.30) nor BIS (males: P = 0.31, females: P = 0.40).

Table 1.

Sample characteristics

	All subjects Mean ± s.d.	Men Mean ± s.d.	Women Mean ± s.d.	Difference P-valuea
Number of subjects	61	47	14
Age	33.8 ± 16.3	31.7 ± 14.4	40.8 ± 20.5	0.14
SCL GSI	0.17 ± 0.18	0.17 ± 0.19	0.17 ± 0.13	0.84
Global 5-HT4R binding	0.75 ± 0.1	0.78 ± 0.09	0.69 ± 0.10	0.009
Testosterone (nmol/l)		16.7 ± 5.7b	0.51 ± 0.2	0.0001
BPAQ (total score)	57.3 ± 16.9	58.0 ± 17.2c	55.1 ± 16.2	0.86
Physical aggression	15.3 ± 5.6	16.1 ± 5.7	12.6 ± 4.6	0.03
Verbal aggression	12.7 ± 3.7	13.0 ± 3.9	11.7 ± 3.2	0.23
Anger	12.2 ± 5.3	12.0 ± 5.3	12.6 ± 5.5	0.72
Hostility	17.2 ± 6.0	16.9 ± 6.0	18.1 ± 6.0	0.50
BIS (total score)	61.1 ± 9.8	62.7 ± 9.9	55.5 ± 6.8	0.003
Motor impulsivity	22.0 ± 4.0	22.9 ± 4.0	19.1 ± 2.1	0.0001
Non-planning impulsivity	24.3 ± 4.4	25.0 ± 4.5	21.9 ± 2.8	0.004
Attentional impulsivity	14.7 ± 3.3	14.8 ± 3.2	14.4 ± 3.8	0.72

Note: SCL GSI: Symptom Checklist-Revised Global Severity Index score.

^aUnpaired t-test for sex differences.

^bn = 45.

^cn = 46.

Results from the multiple regression analyses are shown in Table 2. In the combined male and female sample, global 5-HT₄R binding was not significantly associated with BPAQ or BIS.

Table 2.

Main effect of global 5-HT₄R binding on trait aggression (BPAQ, n = 60) and trait impulsivity (BIS, n = 61), adjusted for age, sex, 5-HTTLPR-status and PET-scanner type

Outcome variable	Slope estimate	Standard error	95% CI	R2	P-value
BPAQ total score	33.6	18.4	[−15.8; 83.0]	0.07	0.18
Physical aggression	8.8	8.0	[−7.3; 24.9]	0.12	0.28
Verbal aggression	10.2	5.4	[−0.6; 21.1]	0.09	0.06
Anger	4.1	7.8	[−11.5; 19.8]	0.05	0.60
Hostility	10.4	8.7	[−7.1; 27.9]	0.07	0.24
BIS total score	−14.3	12.4	[−39.2; 10.5]	0.29	0.25
Motor impulsivity	−9.1	4.7	[−18.5; 0.3]	0.39	0.06
Attentional impulsivity	−0.5	4.8	[−10.1; 0.0]	0.09	0.91
Non-planning impulsivity	−4.7	5.9	[−16.4; 7.0]	0.21	0.42

Only among males, there was a significant positive correlation between global 5-HT₄R binding and BPAQ (parameter estimate: 58.7, SE: 28.9, CI: [0.3; 117.1], P = 0.048, Figure 1) in a regression model adjusting for age, PET scanner type and 5-HTTLPR status. Global 5-HT₄R binding was not correlated with BIS in males. We also tested whether addition of testosterone as a covariate changed the correlation between 5-HT₄R binding and BPAQ in males, but this did not change the parameter estimates substantially, nor significantly predict BPAQ in the model.

Fig. 1.

Scatter plot of the correlation between global 5-HT₄ receptor binding and trait aggression (BPAQ) in males, adjusted for age, PET scanner type and 5-HTTLPR status. Slope estimate: 58.7 BPAQ scores per unit BPnd, confidence limits: [0.3; 117.1], P = 0.048. Shades represent 95% confidence limits.

Post hoc analysis of the contributions from the four BPAQ subfacets showed that the correlation between global 5-HT₄R binding and BPAQ in males was primarily driven by physical aggression (P = 0.038) and verbal aggression (P = 0.05), as opposed to anger (P = 0.38) and hostility (P = 0.11). The interaction sex-by-5-HT₄R binding significantly predicted physical aggression (test for difference in slopes P = 0.024, effect in females:−22.5 physical aggression scores per unit BPnd, CI: [−52.3; 7.3] P = 0.12, effect in males: 20.7 physical aggression scores per unit BPnd, CI: [1.2; 40.3] P = 0.038), but not the total score or any of the other BPAQ subfacets.

In the sample younger than 50 years (n = 49, nine women of whom all were younger than 39 years), the interaction sex-by-5-HT₄R binding significantly predicted BPAQ (test for difference in slopes P = 0.021, effect in females:−125.9 BPAQ scores per unit BPnd, CI: [−241.5; 10.3] P = 0.039, effect in males: 60.5 BPAQ scores per unit BPnd, CI: [−9.7; 130.8] P = 0.088, Figure 2) and physical aggression (test for difference in slopes P = 0.012). The interaction sex-by-5-HT₄R binding did not significantly predict BIS in the full sample or in the sample younger than 50 years.

Fig. 2.

Plot showing that the association between 5-HT₄ receptor binding and trait aggression (BPAQ) depends on sex, in the sample younger than 50 years (n = 49, nine women), adjusted for age, PET scanner type and 5-HTTLPR status. In women, the slope estimate was −125.9 BPAQ scores per unit BPnd, confidence limits: [−241.5; 10.3], and in men 60.5 BPAQ scores per unit BPnd, confidence limits: [−9.7; 130.8]. The difference in slopes is significant at P = 0.02. Gray triangles: men; orange circles: women. Coloured shades represent 95% confidence limits.

The finding observed only in men was then further explored in a voxel-based analysis, including a subsample scanned on the HRRT scanner, which consisted of 36 men (mean age 27.9 ± 9.1 years, range 20.2–56.4). In these subjects, there was a minimal time lag (<1 year) between PET scan and psychometric assessment. Whole brain voxel-based analysis revealed two clusters where BPAQ showed a significant positive correlation with 5-HT₄R binding; a cluster of 522 voxels covering parts of the left anterior cingulate cortex (ACC) and the left middle cingulate gyrus (x = −8, y = 28, z = 22, P < 0.01, corrected, Figure 3), and a cluster of 252 voxels covering parts of the right anterior insula and inferior frontal gyrus (x = 30, y = 16, z = 12, P < 0.01, corrected, Figure 3). Associations between 5-HT₄R binding and BIS revealed no significant clusters.

Fig. 3.

Whole-brain voxel-based analysis, adjusted for age and 5-HTTLPR status in 36 men, showing two clusters of k = 522 voxels and k = 252 voxels mapped onto a magnetic resonance template image, covering parts of the left anterior cingulate cortex, the left middle cingulate gyrus and the right anterior insula where trait aggression (BPAQ) shows a significant positive correlation with 5-HT₄R binding. Colour bars indicate t-scores.

Discussion

For the first time, we here demonstrate that low 5-HT levels in the brain are associated with high trait aggression in males. Until now, the concept of low cerebral 5-HT levels augmenting aggressive behaviour has mostly been based on experimental studies and a recent meta-analysis of clinical studies using more indirect measures of central 5-HT suggested that less than 2% of the trait is explained by 5-HT (Duke et al., 2013). The idea of low cerebral 5-HT levels being associated with impulsive aggression is substantially bolstered by our data, although this seemed only to be the case in men. When we restricted the analysis to age below 50 years, the interaction of sex-by-global 5-HT₄R binding significantly predicted BPAQ and its subfacet physical aggression. In the full sample, there was evidence for sex interaction effects in the association between 5-HT₄R and physical aggression but not BPAQ. We believe that the interaction effect may have come across more clearly in a sample more homogenous with respect to age. The finding of a negative association between 5-HT₄R binding and aggression in women in the reproductive age was, however, unexpected and would need to be replicated in a larger sample.

A common categorical approach to aggression is the distinction between reactive and instrumental aggression (Vicario, 2014). Reactive aggression is triggered by a provocative, frustrating or threatening stimulus and involves unplanned aggressive behaviour, whereas instrumental aggression is deliberate and goal-oriented. In one of the early studies assessing 5-HT function in aggressive individuals, a lower concentration of the 5-HT metabolite 5-HIAA was found in impulsive violent offenders compared with violent offenders with premeditated acts (Linnoila et al., 1983). Several studies have hereafter investigated 5-HT function in individuals with impulsive aggression, but in comparison with a healthy control group and not with an instrumentally aggressive control group. PET studies of 5-HT function both in healthy participants and in a range of diagnostic categories that involve impulsive aggression have been inconsistent. The 5-HT_2A receptor (5-HT_2AR) and the 5-HTT have predominantly been studied. An increased brainstem 5-HTT-availability and reduced cortical 5-HT_2AR binding were reported in 14 subjects with high levels of impulsive aggression compared with control subjects (Rylands et al., 2012). This is in contrast to a study showing increased 5-HT_2AR availability in the orbitofrontal cortex (OFC) of 14 patients with intermittent explosive disorder (IED) compared with both IED patients without current aggression and healthy controls, suggesting that 5-HT_2AR function is related to state and not trait impulsive aggression (Rosell et al., 2010). We recently reported that cerebral 5-HT_2AR binding, using [¹⁸F]-altanserin PET, was not associated with trait aggression and trait impulsivity in 94 healthy individuals (da Cunha-Bang et al., 2013). Conflicting results in case–control studies of 5-HTT-availability have also been reported. 5-HTT availability in the ACC was initially reported to be decreased in 10 patients with IED compared with control subjects (Frankle et al., 2005). However, this finding could not be reproduced in a recent study where there was no difference in pregenual ACC (pgACC) 5-HTT availability between 29 impulsive aggressive patients and healthy controls (van de Giessen et al., 2014). In the latter study, the authors instead found a positive correlation between trait callousness and pgACC 5-HTT-availability, and they propose that presynaptic 5-HT function might be augmented in the psychopathy domain. Also, Rylands et al. (2012) were careful not to include subjects with high callous-unemotional (CU) traits in their study, as it has been suggested that opposing 5-HT abnormalities might exist such that 5-HT function is positively correlated to CU traits (Dolan and Anderson, 2003). Due to lack of psychometric measures assessing psychopathic traits in our study, we were not able to assess the impact of psychopathy and CU traits in our healthy volunteers.

In males we found a positive correlation between global 5-HT₄R binding and BPAQ, putatively reflecting an association wherein lower 5-HT levels were associated with higher trait aggression. Certain domains of aggression differ between men and women and in particular, sex differences are present when it comes to physical, but not anger and verbal aggression (Archer, 2004). Consistent with this research, we observed that men scored significantly higher than women in BPAQ physical aggression, but not in any of the other BPAQ subfacets. In terms of psychopathology, men more frequently display externalizing disorders such as antisocial behaviour (Moffitt and Caspi, 2001) whereas there is a higher prevalence of internalizing disorders such as depression among women (Cyranowski et al., 2000). The neurobiological underpinnings of this difference might involve the 5-HT system, but few studies have directly investigated whether differences in 5-HT are related to differences in aggression between men and women. In two studies, prolactin response to fenfluramine was inversely correlated with self-reported aggression in a sample of male patients with various personality disorders (New et al., 2004), and in male borderline personality disorder (BPD) patients (Soloff et al., 2003), but this relationship was not present in female patients. A recent PET study observed differences in 5-HT_2AR binding between BPD patients and healthy control subjects, but only among females, who had higher 5-HT_2AR binding in several regions (Soloff et al., 2014). In the latter study, sex differences were also found in the relationship between personality traits and 5-HT_2AR binding, but only in the BPD patients, where 5-HT_2AR binding in prefrontal regions was related to impulsivity and aggression in male, but not in female patients. Trait impulsivity and trait aggression were not related to 5-HT_2AR binding in either male or female control subjects (Soloff et al., 2014). A possible link that comes to mind is that the 5-HT system might act as a mediator of testosterone in the regulation of aggression. For example, in a study using an S-citalopram challenge test to index 5-HT activity, an interaction of 5-HT and testosterone in measures of aggression was observed in healthy males, where low 5-HT activity and high testosterone levels predicted high trait aggression (Kuepper et al., 2010). Therefore we tested for correlations between testosterone and trait aggression as well as for the effect of testosterone on the relationship between 5-HT₄R binding and aggression, but could not reproduce this finding.

We used a voxel-based analysis to evaluate whether the relationship between 5-HT₄R binding and trait aggression among males was driven by specific regional effects, revealing a cluster covering parts of the ACC. However, this should be interpreted cautiously, as the BPnd for the ACC in the region of interest (ROI) analysis did not significantly predict trait aggression in these males (results not shown). This might be due to the regions in the ROI and voxel-based analysis only partly overlaps. That is, the ACC ROI is somewhat larger than the cluster in the left dorsal ACC. Previous neuroimaging studies have reported the ACC to be involved in the regulation of emotion and aggression (Kramer et al., 2007, 2011; Beyer et al., 2014). The Taylor Aggression Paradigm (TAP) is a competitive reaction time task where the winner gets to punish the loser with aversive stimuli (e.g. an electric shock) that can be used to elicit aggression during functional MRI. Using the TAP in healthy subjects, Kramer et al. (2007) found activations in the anterior insula and the ACC when the participants had to select the intensity for the punishment following high compared with low provocation. The activation of dorsal ACC during punishment in the TAP was later reproduced in a study where participants also received acute tryptophan depletion (ATD), an intervention that transiently lowers brain 5-HT levels. Kramer et al. (2011) found no main effects of ATD on neural response in the dorsal ACC, which is inconsistent with the observation in the present study that 5-HT levels in the left ACC showed particularly strong association with trait aggression.

The present study should be interpreted in the light of the following limitations. The results of this study are interpreted based on the assumption that 5-HT₄R binding reflects 5-HT levels, an association supported by both animal and human studies. However, even though no previous studies are suggestive of 5-HT₄Rs being involved in aggression and impulsivity, we cannot discard that the effect of 5-HT₄R binding on aggression might be specific for this receptor. Another limitation is the delay between PET scan and BIS and BPAQ assessment, with up to 4 years in some participants. Although trait aggression and trait impulsivity generally decline with age, cerebral 5-HT₄R binding is age-independent, except for the striatum that shows about 1% decrease per decade (Madsen et al., 2011a). That is, if anything, the time delay could have the opposite effect on the association between aggression score and cerebral 5-HT₄R binding (Madsen et al., 2011a). In any instance, adding the time delay as a covariate in the analyses resulted in the same outcomes regardless if we included the full sample, the sample confined to age below 50, or considered males and females separately. If we instead of adding time delay as a covariate split the full sample in those with and without a significant time delay, no significant correlations between BPAQ and global 5-HT₄R binding were seen anymore in the two separate groups, possibly because of loss of statistical power.

Impulsivity and aggression are different psychological constructs, measured using different psychometric instruments. Yet an overlap exists, but with only the BPAQ and the BIS available we were not able to tap impulsive aggression specifically. Also, the use of only self-report measures of aggression and impulsivity can induce bias from factors such as social desirability, which might influence questionnaire responding. The sample consisted of healthy volunteers, with relatively low scores of trait aggression and impulsivity. Including a group of individuals with higher levels of aggression would increase the variability in aggression and impulsivity, and the correlation between 5-HT and impulsive aggression might be more pronounced in aggressive individuals. The female sample size was small compared with the male sample, and the women’s menstrual cycle was not accounted for.

In conclusion we found that low cerebral 5-HT levels, as indexed by 5-HT₄R binding, were associated with high trait aggression in males but not in females. 5-HT might play a role in the difference in expression of aggression between men and women.

Funding

This study was funded in part by an unrestricted grant from GlaxoSmithKline and by a centre grant to the Center for Integrated Molecular Brain Imaging from the Lundbeck Foundation. S.C.B. was funded by Rigshospitalet Research Council. The Funding Sources had no involvement in the study design or in the collection, analysis, writing or publication of data.

Conflict of interest. G.M.K. received honorarium from H. Lundbeck A/S as a consultant. The other authors declare no competing financial interests.