Effects of childhood trauma on aggressive behaviors and hippocampal function: the modulation of COMT haplotypes

Abstract Background Aggression is a commonly hostile behavior linked to the hippocampal activity. Childhood trauma (CT) exposure has been associated with altered sensitization of the hypothalamic-pituitary-adrenal (HPA) axis and hippocampal volume，which could increase violent aggressive behaviors. Additionally, Catechol-O-methyltransferase (COMT), the major dopamine metabolism enzyme, is implicated in stress responsivity, including aggression. Hence, CT exposure may affect aggression through the effect on the hippocampal function, which might also be modulated by the COMT variations. Objectives This study examined whether both CT and haplotypes of COMT moderate hippocampal function and thus affect human aggressive behavior. Methods We obtained bilateral hippocampal functional connectivity maps using resting state functional magnetic resonance imaging (MRI) data. COMT haplotype estimation was performed using Haploview 4.2 and PHASE 2.1. Then we constructed a moderated mediation model to study the effect of the CTQ × COMT on aggressive behavior. Results Three major haplotypes were generated from thirteen single nucleotide polymorphisms (SNPs) within the COMT gene and formed three haplotypes corresponding to high, medium, and low enzymatic activity of COMT. The results showed interactive relationships between the Childhood Trauma Questionnaire (CTQ) and COMT with respect to the functional connectivity (FC) of the bilateral hippocampus (HIP)-orbital frontal cortex (OFC). Specifically, CT experience predicted lower negative HIP-OFC coupling in the APS and HPS haplotypes corresponding to the medium and high enzymatic activity of COMT, but greater FC in the LPS haplotypes corresponding to the low enzymatic activity. We also observed a conditional mediation effect of the right HIP-OFC coupling in the link between COMT and aggressive behavior that was moderated by CT experience. Conclusions These results suggest that CT and COMT have a combined effect on aggressive behavior through hippocampal function. This mediation analysis sheds light on the influence of childhood experience on aggressive behavior in different genetic backgrounds.


Introduction
Aggression is defined as any behavior to w ar d another individual through physical actions or w or ds, and so on, that is carried out with the intent to cause physical or psychological harm (Nelson & Trainor, 2007 ).The aggression will cause multiple undesirable pr oblems, suc h as inter personal tensions, incr eased suicides, and increased criminal activities (Lischinsky & Lin, 2020 ).The neural basis of aggression has been extensiv el y inv estigated, whic h is an imbalance between the "top-down" control of the pr efr ontal cortex and excessive "bottom-up" activity of the limbic regions (Yang et al., 2010 ).Specifically, both human and animal studies have indicated the correlation between the hippocampus (HIP) and the regulation of aggressive behaviors (Chang & Gean, 2019 ;Sala et al., 2011 ).The HIP, together with the amygdala and prefrontal cortex, is implicated in the processing of social emotions, fear, and a ggr essiv e r eaction (Cupaioli et al., 2021 ), the reduction in the gray matter volume of the HIP and amygdala were associated with ele v ated a ggr ession se v erity (Coccar o et al., 2015 ).Pr e vious studies also found that tumors or infections in the HIP could be considered as triggering causes for a ggr essiv e behaviors (Malamud, 1967 ).The higher activation in the HIP was also related to the ov err eactions to thr eatening stim uli (Lee et al., 2009 ).In addition, human a ggr essiv e behaviors ar e pr omoted by m ultiple envir onmental e v ents, especiall y c hildhood tr auma (CT).Studies hav e found that childhood abuse experience is related to violent aggr essiv e behavior, mor e pr ecisel y, psyc hological disorders, a ggr essive behavior, and behavior al pr oblems of c hildhood tr aumatized c hildr en ar e mor e than those of non-tr aumatized c hildr en (Dodge et al., 2003 ).
Childhood trauma refers to an adverse experience, including psychological, physical, or sexual abuse, and various forms of neglect during childhood (Bernstein et al., 2003 ).CT exposure contributes to the exhibit of a ggr essiv e beha vior in adolescence , whic h also pr edicts a ggr essiv e and anti-social behaviors of adults (Cupaioli et al., 2021 ).CT has also been associated with sensitization of the hypothalamic-pituitary-adrenal (HPA) axis (Heim et al., 2008 ) and volumetric abnormalities in the HIP (Janiri et al., 2017 ).The HPA axis is the major neur oendocrine str ess r esponse system, and hippocampal neuroplasticity is critically linked to the stressr elated r egulation of HPA axis activity (Sn yder et al., 2011 ).Evidence from animal and human studies suggested that hippocampal growth mainly occurs during childhood and that early stress exposure has lasting effects on the normal functioning of the HIP-HPA axis (Tottenham & Sheridan, 2009 ).Greater hippocampal activation has been linked to enhanced sensitivity while processing threatening information in people with a history of CT (Edmiston & Blackford, 2013 ).
Mor eov er, a ggr essiv eness is also linked to the dysregulation of the dopamine (DA) system.The DA system is involved in rew ar ding processing and goal-directed behavior, Overreaction and gr eater a ggr ession in the r esponse to pr ovocation ar e linked to the decr eased DA synthesis.Catec hol-O -methyltr ansfer ase (COMT) is a major enzyme responsible for degrading DA.T he COMT gene , mapped to 22q11.1-q11.2, is widely expressed in the prefrontal cortex and other regions that modulate a ggr essiv e behavior (Matsumoto et al., 2003 ).The most investigated functional single nucleotide pol ymor phism (SNP) of the COMT gene, which is in codon 158, codes for a substitution from valine (Bernstein et al., 2003 ) to methionine (Boec ker-Sc hlier et al., 2016), leading to a three-to four-times reduction in the activity of the COMT enzyme (Bilder et al., 2004 ).The low COMT activity leads to ele v ated dopamine le v el and thus leads to incr eased stim ulation of dopamine neural networks implicated in impulsivity and a ggr essiv eness suc h as pr efrontal cortex and limbic areas (Cupaioli et al., 2021 ;Drabant et al., 2006 ).There was a higher risk for a ggr essiv e behavior in healthy young people and sc hizophr enia patients carrying the Met allele r elativ e to the Val homozygotes (Albaugh et al., 2010 ;Bhakta et al., 2012 ).In addition, pr e vious studies also hav e concluded that COMT Val158Met is associated with hippocampal activity (Kr ac h et al., 2010 ;Wang et al., 2018 ) and its functional connectivity (FC) with the pr efr ontal cortex during human cognitive performance (Laatikainen et al., 2012 ;Meyer et al., 2016 ).During emotion processing and memory tasks, the Met 158 was associated with greater connectivity between HIP and pr efr ontal r egions (Kluc ken et al., 2015 ;Morris et al., 2020 ).The hippocampal reactivity to unpleasant stimuli is also correlated with the number of Met 158 , which might contribute to differences in emotional resilience under adverse conditions (Smolka et al., 2005 ).Considering the effect of COMT on the a ggr ession and hippocampal function, it is reasonable to infer that COMT might modulate the effect of environment adversity on the hippocampal function and aggressive beha viors .
The interaction between the environment and COMT Val158Met has been found to influence hippocampal volume in healthy adults (Rabl et al., 2014 ) and hippocampal activation in r esilient individuals (v an Rooij et al., 2016 ), but these interaction effects have not always been detected (Batalla et al., 2018 ).Differ ent measur ements of envir onment factors between samples may be a possible reason for this inconsistenc y.Accor ding to a pr e vious study, compar ed with single SNP anal ysis, ha plotype anal ysis impr ov ed the statistical po w er for the moderating effect of COMT haplotypes on the association between antenatal maternal anxiety and neonatal cortical morphology (Qiu et al., 2015 ).Hence, the modest effect of an individual SNP combined with neglecting associated SNPs might be another potential reason.Ho w e v er, COMT Val158Met cannot be the only explanation for the variability in the activity of the COMT enzyme .T his SNP, together with SNPs rs6269, rs4633, and rs4818, can be integr ated into thr ee major COMT haplotypes (HPS , APS , and LPS) corresponding to high, medium, and low COMT activity (Diatc henk o et al., 2006 ).Although based on different molecular genetic mechanisms, the APS and HPS ha plotypes, r espectiv el y, lead to a three times and a r oughl y 20 times reduction in COMT enzymatic activity compared with the LPS haplotype (Nackley et al., 2006 ).Hence, COMT haplotypes other than COMT Val158Met seem to have a very po w erful effect on brain function and behavior.
The interaction between adverse experience during childhood and COMT Val158Met has been found to affect adult antisocial behavior, suc h as a ggr essiv e and violent behavior (Hygen et al., 2015 ).But, to our knowledge, although altered hippocampal volume and activity have been shown to correlate with impulsive a ggr ession (Visser et al., 2014 ), little evidence has been presented about which neural substrate is involved in the way that this gene-envir onment inter action is r elated to a ggr essiv e behavior.
Because the potential gene-environment interaction predicts a ggr essiv e behavior and because both CT and a ggr ession hav e been associated with abnormal hippocampal activity, we speculated that the interaction between CT and COMT haplotypes would affect hippocampal function and that this region would mediate the gene-environment interaction on aggressive behavior.We first identified the three major COMT haplotypes and then investigated the resting-state FC of the bilateral HIP to reveal the neur al substr ates underl ying this mediation effect.

Participants
A total of 360 healthy Han Chinese participants (186 males and 174 females, mean age 19.41 ± 1.09 years) were recruited.This study was a ppr ov ed by the Ethics Committee of the School of Life Science and Tec hnology, Univ ersity of Electr onic Science and Technology of China.Written informed consent was obtained fr om eac h participant, and eac h was scr eened using the structured clinical interview for DSM-IV-TR Axis I disorders, Chinese edition (Phillips & Liu, 2011 ) to ensure that they had no current or lifetime diagnosis of Axis I mental illnesses or family history of psychiatric disorders .T he exclusion criteria also included contraindications to magnetic resonance imaging (MRI) screening, tr aumatic br ain injury, or br ain lesions.
COMT haplotype estimation was performed using Haploview v .4.2 and PHASE v .2.1 (Stephens & Scheet, 2005 ).Using the linka ge disequilibrium (LD) r elationship betw een SNPs, w e calculated the major haplotypes and their frequency in our participant group using Ha plovie w softwar e. PHASE v.2.1 was further used to estimate the most likel y ha plotype combination for each participant with a degree of confidence > 90%.
The LD analysis sho w ed that the significant haplotype block comprised six SNPs (rs4646312, rs6269, rs4633, rs4818, rs4680, rs4646316) in the strong LD (mean pairwise r 2 = 0.976, see Fig. 1 ).Two major haplotypes, TACCGC and TATCAC, r espectiv el y, corr esponded to the HPS and APS ha plotypes (Nac kley et al., 2006 ).The frequencies were 34.8% for HPS and 28.0% for APS.Another two haplotypes (31.9% for CGCGGT and 1.30% for CGCGGC) corresponded to the LPS and the above four haplotypes accounted for 96% of all the detected ha plotypes.Eac h pair of haplotypes from the pair of homologous c hr omosomes was combined to form a diplotype that could provide more complete genetic information.These three major haplotypes formed six diplotypes (LPS/LPS , LPS/APS , LPS/HPS , APS/APS , APS/HPS , and HPS/HPS).We excluded 25 participants who carried none of the six diplotypes from the subsequent imaging genetic analyses.Due to the relativ el y small sample size of the haplotype homozygotes of LPS , APS , and HPS, we combined them with LPS/APS, LPS/HPS, and APS/HPS, r espectiv el y, and defined them as LPS haplotypes (52 males and 44 females), APS haplotypes (54 males and 57 females), and HPS haplotypes (49 males and 54 females), corresponding to the high, medium, and low enzymatic activity of COMT (Diatc henk o et al., 2006 ;Nackley et al., 2006 ).

Childhood trauma and aggression assessments
Childhood trauma was assessed with the Childhood Trauma Questionnaire (CTQ) (Bernstein et al., 1994 ), which is a self-report instrument that assesses CT in five areas: emotional, physical, and sexual abuse, and emotional and physical neglect.The CT score was calculated by the sum of all of these categories.As in the pr e vious study (Fr odl et al., 2014 ), we used a broad definition of CT according to the established cut-off score for each type of CT and classified participants as having a history of CT if they had scores greater than the cut-off value in at least one type of CT.Cut-off scor es ar e as follows: ≥10 for physical abuse, ≥13 for emotional abuse, ≥8 for sexual abuse, ≥15 for emotional neglect, and ≥10 for physical neglect.Aggr essiv e attitudes and behaviors were assessed using the Aggression Questionnaire (Buss & Warr en, 2000 ), whic h consists of 34 items assessing a ggr essiv e pr oneness on five subscales: physical aggression, verbal aggression, indir ect a ggr ession, anger, and hostility.Also, according Buss and Perry ( 1992 ), physical and verbal aggression subscales are highly correlated with each other, both of which represent the individual engagement in aggressive behaviors; whereas anger and hostility subscales r epr esent the tendency to hav e a ggr essiv e thoughts or feelings (Karlsgodt et al., 2015 ).

MRI data acquisition
MRI scans were performed using an MR750 3.0 Tesla MR scanner (GE Healthcar e, Mil waukee, WI, USA).Resting-state functional ima ging data wer e acquir ed using a gr adient-ec ho ec ho-planar imaging sequence with the following parameters: repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, field of view = 240 × 240 mm 2 , matrix = 64 × 64, flip angle (Cupaioli et al., 2021) = 90 • , voxel size = 3.75 × 3.75 × 4.00 mm 3 , 39 slices, and 255 volumes.Before the scanning, all the participants were asked to relax, think of nothing in particular, k ee p their ey es closed, and av oid sleeping.After scanning, the participants were asked whether they fell asleep to try to ensure that the y k e pt awak e during the entire scanning.The bilateral hippocampi were extracted as regions of interest from the Harvar d-Oxfor d subcortical atlas (Desikan et al., 2006 ).The regions of interest masks were resampled to 3 × 3 × 3 mm 3 .Then, the Pearson correlation coefficients were calculated between the av er a ge time series of the bilater al hippocampi and those of all the voxels throughout the whole brain.The correlations were then converted to z values following Fisher's r -to-z transformation to improve normality.From these procedures we obtained bilateral hippocampal FC maps for each participant.

Sta tistical anal ysis
The normality of CTQ and a ggr ession scor es wer e c hec ked by means of skewness and kurtosis and the K olmogoro v-Smirno v test.Tw o-w ay analyses of variance (ANOVA) were performed with CTQ and COMT haplotypes as two independent variables and with age and sex as covariates to examine the effect of the CTQ × COMT haplotypes interaction on hippocampal function.These statistical steps were completed using Statistical P ar ametric Mapping 12 ( http:// www.fil.ion.ucl.ac.uk/ spm ).Multiple comparison corr ections wer e performed using the Gaussian random field (GRF) method implemented in the Data Processing & Analysis for Brain Imaging ( http:// rfmri.org/DPABI ).The significance le v el was defined as those clusters that survived a threshold of P < 0.01 (cluster-forming threshold at voxel-level P < 0.001, and cluster le v el P < 0.01).The significant region, as identified by the tw o-w ay ANOVAs based on hippocampal FC, was extracted and the z values of the voxels in the r egion wer e av er a ged for eac h participant.Tw o-w ay ANOVAs and simple effect analyses were performed using SPSS v.20.0.Then, to determine the relationship between FC and a ggr ession, Pearson corr elation coefficients wer e measur ed between the mean z values and the a ggr essiv e scor es.Next, a moder ated mediation model was constructed to test whether the effect of the CTQ × COMT haplotypes interaction on a ggr essiv e behavior could be mediated by the hippocampal whole brain connectivity.We used the SPSS version of the PR OCESS macro ( www.afhay es.com ) (Hay es, 2012 ) to test the mediation model (model 8), with the CTQ results as the predictor, COMT haplotypes as the moderator, aggression score as the outcome variables, the av er a ge hippocampal FC as the mediator, and age and sex as co variates .T he significance le v el was established by 95% confidence intervals (Ansell et al., 2012) with 50 000 bootstr a pped iter ations.

Demographics and genetic characteristics
Although the K olmogoro v-Smirno v test sho w ed that the a ggr ession scores was violated the assumption of normality ( z = 0.05, P = 0.04), skewness and kurtosis were within the acceptable range of −1 to + 1 (skewness 0.61, kurtosis 0.46).The CTQ results was not normall y distributed (ske wness 1.50; kurtosis 3.77; K olmogoro v-Smirnov test: z = 0.11, P < 0.001), and we tr eated it as a dic hotomous variable to distinguish between the presence and absence of significant CT experience .T he details of the demogr a phics and genetic c har acteristics ar e sho wn in Table 1 .Tw o-w ay ANOVAs and c hi-squar e tests did not show any significant main effects of CTQ and COMT, or an y inter activ e effect of CTQ × COMT across all groups in age, gender, and aggression thoughts ( P > 0.05), except for the main effect of the CTQ on a ggr essiv e behaviors ( P < 0.05).

Interacti v e effects of CTQ and COMT on hippocampal FC
We found significant interactions of CTQ × COMT on the FC between the right HIP and the right orbital frontal cortex (OFC, MNI coordinates: 15, 39, −24; Z -score = 3.94, GRF correction, voxel le v el P < 0.001, cluster le v el P < 0.01, Fig. 2 A).The FCs between the left HIP and right OFC were also affected by the CTQ × COMT inter action (GRF corr ection, voxel le v el P < 0.001, cluster le v el P < 0.01, Fig. 2 C).No main effect for either the CTQ or the COMT results alone was identified by ANOVAs (GRF correction, voxel le v el P < 0.001, cluster le v el P < 0.01).
We further calculated mean value of z scores in the above regions that had significant inter gr oup differ ences.Specificall y, the

Moder a ted media tion anal ysis
Before the mediation analysis, Pearson correlation coefficients were computed to examine the relationship between the FC and a ggr ession.We onl y observ ed a significant and positiv e corr elation between the z values of the right HIP-OFC and the a ggr essive behaviors measure ( r = 0.122, P = 0.032).Considering the co-occurr ent r elationships between the CTQ × COMT interaction, hippocampal function, and the a ggr essiv e beha viors , a moderated mediation model was constructed to examine whether the FC of the right HIP-OFC mediated the effect of the CTQ × COMT interaction on the a ggr essiv e beha viors .T he details of the moderated mediated model are presented in Table 2 , and the overall model is illustrated in Fig. 3 .The FC of the right HIP-OFC was significantly associated with the CTQ results ( β = −0.188;P < 0.001) as well as with COMT ( β = −0.122;P < 0.001) and with CTQ × COMT ( β = 0.088; P < 0.001).Participants who had a high enzymatic activ-

Discussion
This study r e v ealed the associations between c hildhood tr auma, COMT, hippocampal function, and a ggr essiv e beha vior.T he ge-netic effect was c har acterized by v ariations in the combination of the haplotypes composed of individual SNPs.We found that the CTQ results and the COMT ha plotypes hav e inter activ e effects on the FC of the HIP-OFC.Our r esults wer e, to the best of our knowledge, the first to indicate that the regulation of childhood trauma experience on a ggr ession thr ough hippocampal FCs is moder ated by COMT ha plotypes.Suc h gene-envir onment inter actions may partially explain how individual differences in environmental experience can modulate genetic influences on brain function and, consequently, on behavior.
Our study highlighted an interaction between the CTQ and COMT results on the FC between the bilateral HIP and the OFC.Consistent with a pr e vious study (Cheng et al., 2017 ), the OFC was negativ el y connected to the bilater al hippocampi.The OFC is a k e y r egion involv ed in ada ptation to envir onmental c hange and r e w ar d processing, as w ell as to emotional and social regulation (Boec ker-Sc hlier et al., 2016 ).The HIP is involved in processing stress and regulating arousal and affective states.Animal studies suggested that stress can lead to alterations in dendritic arborization in the OFC (McEwen & Morrison, 2013 ), as well as attenuated long-term potentiation in the HIP (Alfarez et al., 2003 ).Human studies also found that adv ersity-r elated r eductions in the volume and functional activity of these two regions may act as vulnerability markers that precede the presence of psychiatric symptoms, such as loss of self-control or fear extinction deficits (Ansell et al., 2012 ).These results indicate that the FC of the HIP-OFC might be affected by the envir onment, especiall y by adverse and stress conditions.A recent study reviewed the available researches that focused on the effect of COMT Val158Met on FC, they found that the Met 158 was associated with an increased FC between HIP and OFC during resting state in healthy adults (Morris et al., 2020 ).To our knowledge, pr e vious studies also have investigated the effect of COMT the influence of COMT Val158Met on PFC connectivity at rest and found a dose-dependent relationship between COMT Val 158 allele and FC of PFC regions (Lee et al., 2011 ;Liu et al., 2010 ).Meanwhile, in light of the high le v el of expr ession of COMT in both the HIP and the frontal cortex, it is reasonable to find an interaction effect of the CTQ and COMT on the negative HIP-OFC couplings.
Specifically, the CT experience predicted lo w er negative HIP-OFC coupling in the APS and HPS ha plotypes, wher eas the CT experience pr edicted gr eater FC in the LPS haplotypes.Cell studies r e v ealed that the LPS haplotype is associated with about three times higher enzymatic activity than the APS ha plotype, whic h is most likely due to the substitution from the Val 158 to the Met 158 allele (Nackley et al., 2006 ).Hence, from this perspective, our findings were consistent with a pr e vious study in which environmental adversity was associated with increased hippocampal volume in Val 158 homozygotes, and this correlation exhibited a gradual c hange fr om positiv e to negativ e with the dose of the Met 158 allele (Rabl et al., 2014 ).In addition, the moderating effect of COMT Val158Met on the relationship between antenatal maternal anxiety and neonatal cortical thickness sho w ed a similar pattern (Qiu et al., 2015 ).COMT Val158Met also has a pleiotropic effect on cognitive and emotional processing, with the Val allele related to poor memory performance and with the Met allele related to decreased emotional regulation (Mier et al., 2010 ).Heightened connectivity in the LPS haplotype may indicate increased aversive memory formation and consolidation caused by CT experience (Mutso et al., 2014 ).Given the involvement of the OFC in the down-regulation of negative emotion, weakened connectivity in the APS haplotype with CT experience may indicate diminished control of negative emotions .Moreo ver, the imaging study found that the Met 158 allele is associated with a r elativ el y smaller hippocampal and prefr ontal volume compar ed with Val 158 allele carriers (Honea et al., 2009 ).According to pr e vious findings, the pre-existing smaller hippocampal volume might predispose individuals to w ar d the de v elopment of malada ptiv e str ess r esponses when exposed to a se v er e and sustained stressor (Vachon-Presseau et al., 2013 ).The association between CT experience and decreased FC in the APS haplotype may also indicate a vulnerability to stress that is related to the Met 158 allele.
Like the LPS haplotypes, the HPS haplotype codes for the Val 158 v ariant but hav e differ ent mRNA secondary-structur es that lead to almost a 20-fold reduction in enzymatic activity (Nackley et al., 2006 ).The limited metabolic capacity of the HPS haplotypes could result in a pronounced, prolonged elevation of the synaptic dopamine le v el.According to the inverted U-shaped modulation of the relationship between brain structures and functioning of the dopamine system, the excessive dopamine level might impair neuronal integrity and survival.Hence, the LPS haplotypes (LPS/LPS, LPS/APS) may be also vulnerable to childhood adversity, and the reduction in hippocampal FCs with CT experience, as found for APS haplotypes, may contribute to an inability to suppress behaviors with negative consequences, such as a ggr ession.
We also found a conditional mediation effect of the hippocampal FC in the link between CT experience and the a ggr essiv e behaviors that were moderated by COMT.Consistent with pr e vious studies, adverse experience and COMT have an interactive effect on impulsiv e a ggr ession in earl y adolescents (Zhang et al., 2016 ).Sie v er ( 2008 ) posited that a ggr essiv e behavior might be a consequence of an imbalance between the top-down control or "brakes" provided by the OFC and excessive bottom-up "driv es" trigger ed or signaled by the limbic r egions.OFC dama ge or dysfunction could cause a lack of emotional contr ol, whic h, in turn, could result in impulsiv e and a ggr essiv e behaviors (Gansler et al., 2009 ).Activation in the HIP is associated with the extinction of conditioned fear (Maheu et al., 2010 ).Considering the differences in the effects of CT experience in the various COMT haplotype individuals in our study and considering the involvement of the HIP in emotion regulation and adverse stimuli processing, we speculate that the CTQ × COMT interaction influences a ggr essiv e behavior through the HIP-OFC FC.Other shreds of evidence have also been found to support this suggestion.First, the two genotypes of COMT Val158Met have different effects on emotional reactivity to av ersiv e stim uli (Monta g et al., 2008 ).Mor eov er, the inter action between this pol ymor phism and early life stress contribute to susceptibility to neur opsyc hiatric disorders r elated to emotional dysregulation (Hosang et al., 2017 ).Second, the interaction effect between CT experience and COMT Val158Met influences fear conditioning and extinction (Deslauriers et al., 2018 ).
In our study, we found that only the right HIP-right OFC connectivity mediated the association between gene-environment inter action and a ggr essiv e behavior, a finding that was consistent with pr e vious studies that sho w ed a right laterality effect on the association between dopamine release and impulsivity in the limbic regions (Oswald et al., 2007 ) as well as a specific relationship between the right OFC and a ggr ession in sc hizophr enia patients (Hoptman et al., 2002 ).Mor eov er, dopamine le v els in the right hemispher e ar e higher than in the left side in rats (Afonso et al., 1993 ).Ho w e v er, the rightw ar d OFC asymmetry in a ggr ession is still up for debate, and one pr e vious study reported a correlation between the left OFC gray matter volume and a ggr ession r atings (Gansler et al., 2009 ).Hence, further studies ar e warr anted to investigate the specific effect of the HIP-OFC coupling on aggressive beha viors .
The strengths of this study include that its design provided an opportunity to examine the mediating effects of hippocampal function on the relationship between gene , en vir onment, and a ggr essiv e behavior.In addition, the haplotypes we considered are gr oups of alleles, whic h can show str onger linka ge equilibrium and carry more information about the underlying functional variants; in turn, this could show a stronger association with behavior than individual variants and improve the statistical po w er in subsequent imaging genetic analyses (Qiu et al., 2015 ).A weakness of our study is that the childhood trauma questionnaire was a retr ospectiv e and self-r eport measur ement, whic h might hav e led to inaccuracies caused by reporting bias or forgetfulness.Ho w ever, a pr e vious study suggests that r etr ospectiv e self-r eports tend to under estimate r ather than ov er estimate actual situations (Hardt & Rutter, 2004 ).This might also be a possible reason for the undetected inter activ e effects on a ggr essiv e behavior.
In conclusion, these findings provide the first evidence that hippocampal function is the underlying mechanism for the relationship between the CTQ-COMT interaction and a ggr ession.This study contributes to a better understanding of the influence of childhood experience on aggressive behavior in different genetic bac kgr ounds.

Figure 1 :
Figure 1: LD map of the 13 SNPs in our sample.Using varying shades of red and the D ' statistic value, the map shows the LD level of the corresponding two SNPs.

Figure 2 :
Figure 2: Interaction of COMT haplotypes and the CTQ results on bilateral hippocampal FC. ( A ) The significant region shows the effect of the CTQ × COMT interaction on the FC between the right HIP and right OFC (GRF correction, voxel level P < 0.001, cluster level P < 0.01).(B) Mean ± SE FC of R.HIP-OFC in (A) across all participants ( post hoc t -test, Bonferroni corrected).( C ) The significant region shows the effect of the CTQ × COMT interaction on the FC between the left HIP and right OFC (GRF correction, voxel level P < 0.001, cluster le v el P < 0.01).(D) Mean ± SE FC of R.HIP-OFC in (C) across all participants ( post hoc t -test, Bonferroni corrected).R., right, L., left, SE, * * P < 0.01, * * * P < 0.001.

Table 1 :
Demogr a phics: and genetic c har acteristics of the differ ent gr oups.Tw o-w ay ANOVA (main effect of CTQ and COMT : P > 0.05; CTQ × COMT : P > 0.05); c Tw o-w ay ANOVA (main effect of CTQ: P < 0.05; main effect of COMT: P > 0.05; CTQ × COMT : P > 0.05).All the raw functional MRI (fMRI) data were reviewed independently b y tw o experienced radiologists who knew nothing about the genotype information.In total, 25 participants were excluded because of bad ima ging quality, suc h as a ppar ent signal loss and inter-slice motion artifacts.All the images were preprocessed using DPARSFA (Data Processing Assistant for Resting-State fMRI Adv anced Edition, http://www.restfmri.net/forum/DPARSF), with the pr ocedur es as follo ws: (i) discar ding the first 10 volumes; (ii) SD: standard deviation; LL: LPS/LPS; LA: LPS/APS; LH: LPS/HPS; AA: APS/APS; AH: APS/HPS; HH: HPS/HPS; CT + : the presence of significant CT experience; CT −: the absence of significant CT experience a Chi-square test ( P > 0.05) across the six groups; b slice timing correction; (iii) head motion correction; (iv) spatial normalization to the EPI template in MNI space; (v) resampling to 3 × 3 × 3 mm 3 ; (vi) smoothing with a 6 mm Gaussian kernel; (vii) r egr essing out nuisance signals including six motion parameters, white matter, cer ebr ospinal fluid signals, and global signals; and (viii) temporal bandpass filtering (0.01-0.8 Hz).A further 25 participants were excluded due to a maximum displacement in any of the cardinal directions ( x , y , z ) of > 2 mm or a maximum spin ( x , y , z ) of > 2 • .The remaining 310 participants were included in the subsequent analyses.

Table 2 :
Results: of moderated mediation model.