Association between aromatase in human brains and personality traits

Aromatase, an enzyme that converts androgens to estrogens, has been reported to be involved in several brain functions, including synaptic plasticity, neurogenesis, neuroprotection, and regulation of sexual and emotional behaviours in rodents, pathophysiology of Alzheimer’s disease and autism spectrum disorders in humans. Aromatase has been reported to be involved in aggressive behaviours in genetically modified mice and in personality traits by genotyping studies on humans. However, no study has investigated the relationship between aromatase in living brains and personality traits including aggression. We performed a positron emission tomography (PET) study in 21 healthy subjects using 11C-cetrozole, which has high selectivity and affinity for aromatase. Before performing PET scans, subjects answered the Buss-Perry Aggression Questionnaire and Temperament and Character Inventory to measure their aggression and personality traits, respectively. A strong accumulation of 11C-cetrozole was detected in the thalamus, hypothalamus, amygdala, and medulla. Females showed associations between aromatase levels in subcortical regions, such as the amygdala and supraoptic nucleus of the hypothalamus, and personality traits such as aggression, novelty seeking, and self-transcendence. In contrast, males exhibited associations between aromatase levels in the cortices and harm avoidance, persistence, and self-transcendence. The association of aromatase levels in the thalamus with cooperativeness was common to both sexes. The present study suggests that there might exist associations between aromatase in the brain and personality traits. Some of these associations may differ between sexes, while others are likely common to both.

To investigate molecular dynamics in the living body, positron emission tomography (PET) is a suitable technique, which allows quantitative analysis of compound accumulation in tissues. Previously, a PET study using 11 C-vorozole, which was developed as the first PET probe for aromatase imaging, was performed to image aromatase in the brain of healthy men and women 6,7 . In that study, the authors demonstrated a unique distribution of aromatase in living human brains for the first time. However, a certain amount of nonspecific signals was observed in 11 C-vorozole PET images probably caused by unintended reuptake of the radioactive metabolite into the brain.
We have developed a novel PET probe for aromatase imaging to overcome the disadvantages of 11 C-vorozole, 11 C-cetrozole 29 . Our previous study showed that 11 C-cetrozole had a higher signal-to-noise ratio than 11 C-vorozole since almost no radioactive metabolite of 11 C-cetrozole was not taken up into the brain, that is, 11 C-cetrozole is superior to 11 C-vorozole in terms of specificity and metabolic stability 29 . In the present study, we performed a 11 C-cetrozole PET study in healthy subjects to examine the association between aromatase in the brain and personality traits.

Results
Distribution of aromatase in human brain. The binding potential (BP ND ) values of 11 C-cetrozole, which are an index of aromatase concentration, were calculated in 21 healthy individuals (10 females and 11 males). High BP ND s were found in the thalamus with heterogeneous distribution among the medulla, hypothalamus, and amygdala in both sexes (Figs 1 and 2). Males had higher BP ND values in most of these regions than females, except for in the right hypothalamus; however, a significant sex difference was found only in the left hypothalamus (P = 0.005; Fig. 2). There were no regions which BP ND s depended on sex hormone levels in plasma (estradiol for females, and free testosterone for males).
Association between aromatase and personality traits. We assessed associations between BP ND images of 11 C-cetrozole and scores of BAQ. Given that earlier studies showed that the amygdala is implicated in aggression [18][19][20][21][22][23] , we focused on the amygdala as a volume of interest (VOI). Using Statistical Parametric Mapping 8 software (SPM8, Wellcome Department of Imaging Neuroscience), a voxel-wise analysis corrected by family-wise error rate for aggression scores on the VOI was performed. Apparently, region-specific differences among individuals are evident and the associations between personality traits and region-specific level of aromatase were also observed (Figs 3 and 4). In females, aggression scores were positively associated with BP ND in the left amygdala (P FWE-corr < 0.05, R 2 = 0.83, Fig. 3). In contrast, data from male and combined data from male and female did not exhibit a significant association between aggression scores and BP ND in the amygdala. For an evaluation of individuals' personality traits, subjects answered TCI. Associations between traits and brain regions analysed in each sex group are listed in Table 1. Concerning other traits, sex-specific associations were observed in the amygdala and supraoptic nucleus of hypothalamus and in the inferior parietal gyrus in females. On the other hand, males-specific associations were observed in the anterior cingulate gyrus, supramarginal gyrus, caudate nucleus, pons, and midbrain. When the data of female and male were combined, almost no significant associations were  found, except for cooperativeness. Cooperativeness scores were negatively associated with BP ND in the bilateral thalamus of females and males (R 2 = 0.71, Fig. 4). The subregions that associated with cooperativeness were localized to the ventral lateral and ventral posterior parts of the thalamus. There were 2 regions that exhibited associations between BP ND of 11 C-cetrozole and female traits, namely, the right supraoptic nucleus of the hypothalamus (SON; MNI: x = 8, y = 2, z = −14) and the right amygdala (x = 24, y = −6, z = −20), although several associations were present in small clusters (Tables 2 and 3, Fig. 5). The traits that associated with BP ND in the SON were novelty seeking (negative, R 2 = 0.998), harm avoidance (positive, R 2 = 0.97), reward dependence (negative, R 2 = 0.997), persistence (negative, R 2 = 0.98), cooperativeness (positive, R 2 = 0.995), and self-transcendence (positive, R 2 = 0.997; Table 2). The traits associated with BP ND in the right amygdala were novelty seeking (negative, R 2 = 0.999), persistence (negative, R 2 = 0.994), cooperativeness (positive, R 2 = 0.999), and self-transcendence (positive, R 2 = 0.999; Table 3).

Discussion
In this study, we demonstrated the distribution of aromatase in living human brains using our originally developed PET probe, 11 C-cetrozole, and suggested that aromatase levels in the brain may relate to personality traits. The first PET scan of brain aromatase using 11 C-vorozole was performed by Biegon et al. 6,7 . They demonstrated high levels of aromatase in the thalamus, amygdala, preoptic area, medulla, etc. 11 C-Vorozole had high specificity and affinity for aromatase; however, the metabolites of 11 C-vorozole were taken up into the brain with radiolabelling 29 . Thus, the measurements were less quantitative. Aiming for a more quantitative measurement of aromatase, we developed 11 C-cetrozole as a novel PET probe for aromatase imaging 29 .
In the present study, we performed PET scans with 11 C-cetrozole in 21 healthy human subjects. Approximately 50% of parental compound was intact 60 min after injection, indicating that this PET probe was suitable to measure aromatase in a living body (Supplemental Fig. S1). High BP ND s of 11 C-cetrozole were found in the thalamus with a heterogeneous distribution among the amygdala, hypothalamus, and medulla in both sexes (Figs 1 and 2). The distribution pattern of 11 C-cetrozole binding was consistent with that of 11 C-vorozole 7 and with immunohistochemistry in the postmortem human brain 5 , suggesting that considerable aromatase enzyme is expressed in the thalamus, hypothalamus, amygdala, and medulla in the human brain. Unlike other mammals that have more aromatase enzyme in male brains, such as rats and monkeys [2][3][4] , there was no distinct sex difference in aromatase levels in our human cohort, except for in the left hypothalamus (P = 0.005). However, males showed a tendency towards relatively higher aromatase expression in all brain regions than females, except for the right hypothalamus. The reason why males have more aromatase in the brain than females do is considered to compensate lower circulating estrogens since estrogen is an important hormone related to regulation of sexual behaviour and emotions, neural plasticity, neuroprotection, etc. in the brain [30][31][32] . In the present study we showed that BP ND of 11 C-cetrozole varied between individuals, especially in the thalamic subregions. This variability was associated with personality trait variability. Both females and males showed a negative association between BP ND of 11 C-cetrozole and cooperative scores in the thalamus region in this study. The subregions that associated with cooperativeness were localized to the ventral lateral and ventral posterior parts of the thalamus. The ventral lateral nucleus of the thalamus contains estrogen receptor β and is known to project to the primary motor cortex 33 . Diffusion-weighted imaging studies have segmented the thalamus on a connectivity basis and reported individual variations in segmentation 34 . Johansen-Berg et al. 34 reported that the ventral lateral nucleus primarily connected to the prefrontal cortex in some subjects and to the primary motor cortex in others. This individual variation may provide cause of different personality. High aromatase in the thalamus is unique to humans, while monkeys, baboons, and rats have high  amount of aromatase in the amygdala and hypothalamus 3,4,7,29,35,36 . The characteristically high social abilities of humans such as cooperativeness may be processed in the thalamus through the regulation of estrogens.
In addition to cooperativeness and the thalamus, the traits and associated regions were different between sexes (Table 1). In females, sex-specific associations were observed in the amygdala and supraoptic nucleus of hypothalamus and in the inferior parietal gyrus. On the other hand, males-specific associations were observed in the anterior cingulate gyrus, supramarginal gyrus, caudate nucleus, pons, and midbrain.
The right SON and the right amygdala showed associations between BP ND s of 11 C-cetrozole and female traits (Fig. 5). The localization of aromatase in the SON was demonstrated previously via immunohistochemistry 13 . Further, the SON is known to be a region in which oxytocin is synthesized 37 . Oxytocin production is regulated by estradiol 38 . Regulation of oxytocin by estradiol may affect personality traits given that oxytocin is implicated in stress susceptibility, emotion, memory, and social interaction 39 . The traits that associated with BP ND in the SON were novelty seeking (negative), harm avoidance (positive), reward dependence (negative), persistence (negative), cooperativeness (positive), and self-transcendence (positive; Table 2). The traits associated with BP ND in the right amygdala were novelty seeking (negative), persistence (negative), cooperativeness (positive), and self-transcendence (positive; Table 3). These 4 traits are consistent with the traits that associated with BP ND in the SON, which suggests functional connectivity between the right amygdala and SON in females.
Since animal and human studies suggested that the amygdala is involved in aggression 18-23 , the VOI was drawn on the amygdala, and a regression analysis was performed between BP ND of 11 C-cetrozole in the amygdala and aggression score. There was a positive association between aggression scores and BP ND in the left amygdala in females. Earlier studies found a relationship between aggression and reactivity in the left amygdala, although one of the studies combined female and male data 21,22 . Another study reported that lower amygdala volume was correlated with more aggression in healthy females 23 . Animal studies have demonstrated that aromatase neurons in the medial amygdala regulate male aggression and maternal aggression 20 . Our results show that aromatase in the female left amygdala is associated with aggression, which indicates that estrogen synthesis in the left amygdala may induce aggression in females.
Furthermore, polymorphisms in the aromatase and estrogen receptor genes have been reported to be associated with harm-avoidance traits 24,40 . Although we did not perform genetic analyses, gene polymorphisms or DNA methylation may affect the expression of aromatase or hormone receptors. Multi-disciplinary studies that consider gene polymorphisms, epigenetic changes, and imaging are needed in the future.
Our results showed that there may be associations between aromatase in the brain and personality traits, and some of the associations may differ between sexes, while others are likely common to both. Whether sex differences exist in the brain has long been contentious and remains controversial. Classical differences in brain structure, e.g., greater corpus callosum volume in females or enlarged cortical language regions in females, have been dismissed via meta-analysis 41 . However, neurochemistry suggests tendencies for females to have higher activity in serotonergic (5-HT transporter, 5-HT 1A and 5-HT 2A receptors), dopaminergic (dopamine transporter), and GABAergic (neurotransmitter level) systems 42 , which are involved in mood, emotions, and personality traits. An animal study revealed that exogenous estrogen treatments increased the mRNA and protein levels of tryptophan hydroxylase, a rate-limiting enzyme in the production of a serotonin precursor, in the raphe nucleus of spayed rhesus monkeys 43 . Further studies are needed to clarify the differences and commonalities between sexes.
Previously the results between monoamine oxidase (MAO) level measured with PET and mood/personality disorders were reported 44,45 . Our present results also show that the aromatase level in different brain sub-regions measured with PET is associated with a variety of personality traits. The dynamics of interaction between sex hormone and monoaminergic neurotransmitter systems are partly regulated by both enzyme levels. The PET studies in combination of both enzymes may be an interesting target in the future study.
There are several limitations of the present study. The associations between TCI scores and BP ND did not survive a family-wise error correction, then were analysed with a significance threshold of P < 0.001, uncorrected, k ≥ 10 voxels. As for the SON and the right amygdala of females, the associations with TCI scores were discussed even though the criteria were not fulfilled (k < 10 voxels). When we could increase the number of subjects, the results might become clearer. As regards sex hormone in plasma, we measured testosterone, free testosterone, estradiol, and progesterone in both males and females, however, free testosterone in female (N = 2), estradiol in male (N = 7) were too little to be quantified. Thus we tested the association between aromatase level and free testosterone in male or estradiol in female.
Although the subjects in this study were all healthy, there were associations between aromatase levels in the brain and personality traits, which suggests that regulation of estrogens might affect personality. Further PET studies that include patients with personality disorders using other PET tracers for estrogen and androgen receptors would help clarify the association between sex hormone systems and personality traits.

Methods
Subjects. Subjects were recruited by advertisements at Osaka City University and RIKEN. Twenty-four healthy adults (11 females and 13 males) participated in the present study. Participants were excluded if they had past or current serious medical illnesses and/or organic brain diseases or if they took drugs actin on the central nervous system. All females were not taking oral contraceptives and had regular menstrual cycle. Two females (35 and 45 yrs old) and 2 males (36 and 35 yrs old) received a whole-body PET scan to measure their radiation exposure, and 10 females and 11 males (average age of 34.7 ± 6.4 and 31.7 ± 8.1 yrs, mean ± SD, respectively) received a brain PET scan. Each subject completed the validated Japanese version of BAQ 25  Positron emission tomography. 11 C-Cetrozole was synthesized at Osaka City University Hospital according to the previously published procedure 29 . The desired compound was dissolved in a mixture of polysorbate 80 (0.1 ml), propylene glycol (0.9 ml), and saline (9 ml). The identity and concentration of 11 C-cetrozole were assessed using high-performance liquid chromatography. The specific activity of 11 C-cetrozole was 81.2 ± 26.0 GBq/μmol (mean ± SD) at administration. The radiochemical purity was greater than 99.5%. The dose of 11  PET image processing. Brain PET images were reconstructed by Fourier rebinning and 2-dimensional filtered backprojection without additional smoothing filters. For quantitative image analyses, PMOD software (PMOD Technologies Ltd., Zurich, Switzerland) was used. VOIs were delineated in the thalamus, amygdala, hypothalamus, and medulla, which are structures known to contain a rich supply of aromatase enzyme [5][6][7] , and in the cerebellar lobules on the same individual's MR images and transferred to PET images. Decay-corrected time-activity curves were generated for each brain region, arterial blood plasma, and parent unchanged compound, as measured by thin-layer chromatography (Supplemental Fig. S1). The time-activity curves for plasma and parent unchanged fraction were fitted to a 3-exponential model and a Hill function, respectively. The data with arterial blood sampling were analysed with a Logan plot 46 , and the total distribution volume (Vt) in each brain region was calculated. The data without arterial blood sampling were analysed with a Logan reference tissue model based on average k2' 47 using the cerebellum as a reference. Nondisplaceable binding potential (BP ND ) and distribution volume ratio (DVR), which are linear functions of enzyme availability, were calculated (DVR = BP ND + 1). Since no arterial blood sampling is preferable because of possible intense pain caused by arterial puncture, we compared the data with or without arterial blood sampling, that is, the data calculated by Logan plot or by Logan reference tissue model. Vt values of all examined brain regions were divided by the Vt value of the cerebellum and were compared with BP ND values. Since the difference between normalized Vt values and BP ND was 4 ± 1% (mean ± SD, N = 4), we decided to employ  the Logan reference tissue model to analyse the remainder of the data. Then, BP ND images were generated by model fitting. After co-registration of PET and MR images, whole-head structural images were normalized to the Montréal Neurological Institute (MNI) T1 image template, with the same parameters applied to the BP ND images. BP ND images were resampled to a voxel size of 2.0 × 2.0 × 2.0 mm using SPM8.
Quantification of aromatase in the rich regions. For measurements of the aromatase level in the brain, VOIs of the amygdala, hypothalamus, thalamus, and medulla were superimposed on BP ND images. Because 11 C-cetrozole binding was heterogeneously distributed in the thalamus, the thalamus was further divided into 5 subregions, namely, anterior medial, anterior lateral, central medial, central lateral, and posterior parts, consistent with a prior study 48 . BP ND values were extracted from BP ND images. The difference between sexes in each VOI was analysed by a t-test. All P-values were two-tailed, and P values less than 0.05 were considered significant. These analyses were performed with the GraphPad PRISM 5.0 software package (GraphPad Software, Inc., La Jolla, CA).

Statistical analysis of PET images.
For aggression, we focused on the amygdala since earlier studies showed that this region is implicated in aggression [18][19][20][21][22][23] . A VOI of the amygdala was delineated using the WFU-Pickatlas SPM tool 49 . Using SPM8, voxel-wise analysis was performed on the amygdala of BP ND images by applying the score of BAQ as a covariate. A family-wise error corrected significance threshold was set at P < 0.05 in the amygdala. The associations between scores of TCI and BP ND were analysed as whole-brain analyses with a significance threshold of P < 0.001, uncorrected, k ≥ 10 voxels (=80 mm 3 ). Voxel-wise analyses were performed on the BP ND images using the general linear model in SPM8, with covariates of 7 traits of TCI. Although the criteria were not fulfilled (k < 10 voxels), 6 and 4 traits of TCI showed the associations in the identical coordinates (x = 8, y = 2, z = −14 and x = 24, y = −6, z = −20, respectively), thus these 2 regions were discussed separately.