Association of monoamine oxidase-A genetic variants and amygdala morphology in violent offenders with antisocial personality disorder and high psychopathic traits

Violent offending is elevated among individuals with antisocial personality disorder (ASPD) and high psychopathic traits (PP). Morphological abnormalities of the amygdala and orbitofrontal cortex (OFC) are present in violent offenders, which may relate to the violence enacted by ASPD + PP. Among healthy males, monoamine oxidase-A (MAO-A) genetic variants linked to low in vitro transcription (MAOA-L) are associated with structural abnormalities of the amygdala and OFC. However, it is currently unknown whether amygdala and OFC morphology in ASPD relate to MAO-A genetic polymorphisms. We studied 18 ASPD males with a history of violent offending and 20 healthy male controls. Genomic DNA was extracted from peripheral leukocytes to determine MAO-A genetic polymorphisms. Subjects underwent a T1-weighted MRI anatomical brain scan that provided vertex-wise measures of amygdala shape and surface area and OFC cortical thickness. We found that ASPD + PP subjects with MAOA-L exhibited decreased surface area in the right basolateral amygdala nucleus and increased surface area in the right anterior cortical amygdaloid nucleus versus healthy MAOA-L carriers. This study is the first to describe genotype-related morphological differences of the amygdala in a population marked by high aggression. Deficits in emotional regulation that contribute to the violence of ASPD + PP may relate to morphological changes of the amygdala under genetic control.

abnormal amygdala morphology in samples with high levels of psychopathy 11,12 . One investigation described tissue enlargement of the lateral and central nuclei in psychopathy, which was accompanied by reduction effects in the basolateral nucleus 11 . The other study reported bilateral amygdala volume loss and atrophy of cortical, central, basolateral, and lateral nuclei that showed a relation with increased psychopathic traits in a predominantly male sample with high psychopathy features 12 . These studies highlight abnormal amygdala morphology as a potential biomarker of elevated psychopathic traits, but they do not provide information on possible genetic factors influencing this brain phenotype. Another brain region that has been implicated in the pathogenesis of ASPD + PP is the orbitofrontal cortex (OFC) 13 . The OFC contributes to emotion processing, moulds personality structure, and regulates social conduct and behavior 14 . One study found increased bilateral OFC grey matter volumes in male carriers of MAOA-L versus MAOA-H 15 , whereas another investigation reported that the MAOA-L genotype was associated with grey matter loss in bilateral OFC 16 . Importantly, both of these studies were conducted in healthy individuals. In addition, OFC volume loss in a sample of individuals with cocaine use disorders was attributed to the low-activity MAO-A allele in combination with lifetime cocaine use 17 . Tissue reductions have also been reported in the OFC of psychopathy 11 . The OFC is a logical structure to study in tandem with the amygdala, as reciprocal connections exist between the two regions in primates 18 . Moreover, altered white matter connections between the amygdala and OFC have also been described in psychopathy 19 .
Monoamine oxidase-A (MAO-A) is an enzyme that degrades neurotransmitters, such as serotonin, norepinephrine, and dopamine, that are known to play a role in aggressive behavior 20 . Important to the current study is that MAO-A may influence amygdala and OFC morphology. The MAO-A gene includes a 30-base pair variable nucleotide tandem repeat (VNTR) polymorphism that yields variants associated with low (MAOA-L) or high (MAOA-H) transcription efficiency 21 . In some studies, individuals with the MAOA-L genotype exhibit elevated antisocial traits 22 and, in offender samples, show evidence of severe violent behavior. The MAOA-L allele may also influence amygdala reactivity to the expression of anger 23,24 . Interestingly, a positron emission tomography investigation of ASPD found that MAO-A activity was lower in the OFC and regions that make direct connections with the amygdala, such as the hippocampus and thalamus 25 .
Previous findings linking MAO-A with alterations in amygdala/OFC volume and morphology have yielded mixed results. At least one study found that MAOA-L carriers display reduced amygdala volumes bilaterally relative to MAOA-H carriers 15 . By contrast, another investigation reported no difference in amygdala volume by MAO-A genotype 26 . The wide age range (18-80 years) of the second sample compared with the younger age of the participants tested in the first study was suggested as one reason for the discrepant results. One study found increased bilateral OFC grey matter volumes in male carriers of MAOA-L versus MAOA-H 15 , whereas another investigation reported that the MAOA-L genotype was associated with grey matter loss in bilateral OFC 16 . Importantly, both of these studies were conducted in healthy individuals. In addition, OFC volume loss in a sample of individuals with cocaine use disorders was attributed to the low-activity MAO-A allele in combination with lifetime cocaine use 17 . Tissue reductions have also been reported in the OFC of psychopathy. However, to the best of our knowledge, no study has ever investigated amygdala or OFC morphology in relation to the MAO-A VNTR polymorphism among a clinical group (ASPD) presenting severe symptoms (PP).
Since MAOA-L shows a relationship with violent behavior and has also been linked to amygdala/OFC structure, the principal aim of the study was to investigate whether the low-activity MAO-A variant was associated with amygdala/OFC volume and morphology in ASPD + PP with a history of violence. We hypothesized that a gene-diagnosis interaction would emerge, such that ASPD + PP with MAOA-L would exhibit brain changes not seen in healthy subjects with the same genotype. We did not expect to find group differences amongst the MAOA-H carriers.

Results
Subject Characteristics. Participants ranged in age from 18 to 50 years. ASPD + PP subjects had higher rates of past substance use disorders ( Table 1). As expected, ASPD + PP participants endorsed greater PCL-R total and facet scores and presented significantly more conduct disorder and ASPD symptoms. Psychopathy Checklist-Revised Scores and Amygdala Surface Area. In an exploratory analysis among subjects with MAOA-L, Psychopathy Checklist-Revised (PCL-R) scores were found to have a significant effect on right amygdala surface area (F(1, 13) = 6.55, corrected at 15% FDR). Specifically, PCL-R scores showed an inverse correlation with surface area in the right BLA nucleus (Fig. 3).
OFC Cortical Thickness. The two-way interaction was not significant for the middle orbitofrontal gyrus     MAOA-L. Two vertices were selected for each region. At each vertex, the mean surface area of the two groups and a range of variances were used to simulate the required sample size for a study with power = 0.8 and α = 0.05. These simulated required sample sizes were then plotted against the actual sample size and variance of surface area at the given vertex. In three of four vertices, the sample size of our study was sufficient to meet the requirements, suggesting that the sample was large enough to reliably detect the observed effects (Fig. 5).

OFC Cortical Thickness and Power Analysis.
Results above indicated that ASPD + PP offenders had less cortical thickness in the left lateral orbitofrontal gyrus. The mean cortical thickness of the two groups and a range of variances were used to simulate the required sample size for a study with power = 0.8 and α = 0.05. These simulated required sample sizes are plotted against the actual sample size and variance of the cortical thickness measure. The plot shows that the sample size was sufficiently large enough to reliably detect the observed effects ( Fig. 6).

Discussion
This investigation is the first to demonstrate that MAO-A VNTR genetic polymorphisms show a relation with amygdala morphology in ASPD + PP. Our most noteworthy finding is that ASPD + PP offenders with the MAOA-L genotype exhibited decreased surface area of the right BLA nucleus and increased surface area of the right ACo nucleus. Another finding is that these alterations were associated with increased psychopathic traits. Previous studies of ASPD + PP reported amygdala surface area anomalies but did not compare results with additional biological measures 11,12 . By contrast, our investigation included genetic analyses, whose results suggest that the observed morphological abnormalities in ASPD + PP may relate to MAOA-L genotype effects. We believe that the results of the present investigation are significant, because they suggest a novel mechanism to help explain the morphological changes present in the amygdala of ASPD + PP. One of our two main results was a reduction in BLA nucleus surface area among MAOA-L carriers with ASPD + PP versus healthy subjects matched on genotype. The finding of decreased BLA nucleus surface area in ASPD + PP agrees with previous work that reported reduction effects in this region 11 , and it adds to a growing body of evidence linking abnormalities of the BLA nucleus to key cognitive impairments in psychopathy. Individuals with high trait psychopathy are insensitive to punishment and exhibit weakened autonomic responding to fearful or threatening stimuli 27 . Interestingly, acquisition of fear conditioning depends on intact microcircuitry within the BLA, such that the fear-conditioned startle response to conditioned stimuli is abolished in rodents with lesioned BLA 28,29 . Because serotonin is an important neuromodulator of BLA neurons 30 , genetic influences that regulate serotonin neurotransmission may affect the integrity of the BLA nucleus. We suggest that alteration of the BLA nucleus in ASPD + PP could relate to the effects of the MAOA-L genetic variant and its regulation of serotonergic neurotransmission in the developing amygdala. This position is strengthened by the observation of anomalous accumulation of serotonin in the amygdala of MAO-A knockout models 31 . Our model linking emotional deficits in psychopathy to altered monoamine functioning in the amygdala accords well with studies indicating preferential binding of MAO-A inhibitors to rodent BLA 32 and reduced fear conditioning in healthy subjects administered selective serotonin reuptake inhibitors 33 .
A second main finding is that the right ACo nucleus surface area was increased in ASPD + PP offenders carrying the MAOA-L genotype. A primary function of the ACo is to process smell directly from the main olfactory bulb 34 . Interestingly, the higher olfactory processes of identification and discrimination operate less efficiently among individuals with elevated psychopathic traits 35 . The notion that impaired sense of smell can be attributed to focal brain changes is congruent with clinical observations of faulty olfaction in other neurological disorders such as Parkinson's disease 36 . These neural alterations may also relate to the effect of specific MAO-A genotypes impacting sensory perception in vulnerable individuals. One consequence of reduced MAO-A activity that stems from the MAOA-L genotype may be increased dopaminergic neurotransmission. In the mouse main olfactory bulb, surges in dopamine impair detection of social odors localized to male urine 37 . Dopamine levels were additionally shown to have increased in the nucleus accumbens after rats engaged in physically aggressive confrontations with conspecifics 38 . Therefore, lowered MAO-A may link olfactory impairments in high psychopathy with physical aggression, perhaps through increased limbic dopamine transmission. Indeed, lower prefrontal MAO-A binding predicted greater anger/hostility in a PET study of healthy human subjects 39 .
The fact that differences in the surface area compartments of the amygdala were not paralleled by changes in amygdala volume suggests that the observed morphological differences may have been due to alterations in dendritic arborization as opposed to alterations in actual number of neurons. Pre-clinical research has revealed that On the other hand, in a serotonin transporter (5-HTT) KO model, apical dendritic length and spine density in the BLA nucleus were increased in the mutant strain relative to wildtype 42 . Similarly, wild type mice subjected to a chronic stress paradigm produced hypotrophy of dendritic aborization and greater spine density in the BLA nucleus 43 . While the first two studies speak to possible genetic influences impacting dendritic arborization and the development of ASPD phenotype in males, the third study describes environmental exposures that may guide the assembly of dendritic morphology important to expression of ASPD. This last point is relevant, since individuals with ASPD are known to manifest low distress tolerance 44 . We were unable to locate any studies describing Figure 5. Plots of standard deviation versus required sample size for power = 0.8 and α = 0.05. At each vertex, the sample size was calculated that would be required to obtain 80% power given α = 0.05 and a variety of standard deviations. These sample sizes are plotted in blue as simulated data. In red, the actual standard deviation was plotted at each vertex versus the sample size of each group (8), allowing a comparison of our sample size versus the sample size required to measure the effect. This procedure was performed for four separate vertices. Plots (A) and (B) show data from vertices where ASPD subjects showed significantly decreased surface area; plots (C) and (D) show vertices where significantly increased surface area was observed in ASPD. Plots (A) and (C) correspond to "strong vertices, " where the effect passed 15% FDR easily, while (B) and (D) correspond to "weak vertices, " where the effect barely passed 15% FDR. The goal of this experiment was to examine the full breadth of observed effects. In (A), (B), and (C), the actual sample size of 8 is at least sufficient for observing the effect given the standard deviation of surface area measurements at the corresponding vertex. In (D), the actual sample size of 8 would not have been sufficient to achieve 80% power given the observed standard deviation. Overall, plots (A-D) give confidence that the sample size of the study is sufficient to observe the reported effects.
Scientific RepoRts | 7: 9607 | DOI:10.1038/s41598-017-08351-w an association between MAO-A VNTR genotype and amygdala dendritic arborization, nor were we able to find research that genetic or environmental influences affect arborization in the ACo nucleus.
It is likely that morphological changes of the amygdala in ASPD are related to other factors that may interact with genetic susceptibility. The serotonin system may represent one potential avenue. MAO-A KO mice exhibit increased serotonin levels in the amygdala and also display overt aggressive behavior 31 . Furthermore, pharmacological blockade of amygdala 5-HTT receptors has been shown to reduce aggression in MAO-A knockout models 45 . Other environmental factors may also shape the outcome of high serotonin levels that confer vulnerability to antisocial traits. For example, early childhood adversity is associated with altered serotonin systems in humans 46 that may be due to epigenetic modification of serotonin genes 47 and/or the reciprocal interaction that exists between the serotonergic system and the hypothalamopituitary axis 48,49 . Although ample evidence suggests that MAOA-L and early childhood adversity may interact to produce antisocial outcomes 50 , the literature is silent on whether the combination of these variables may alter amygdala morphology.
We did not detect any significant gene-diagnosis interactions in the OFC. We did ascertain, however, that ASPD + PP participants displayed smaller OFC volumes than the healthy controls, which accords well with findings in the literature 51,52 . Interestingly, a small voxel-based morphometry study comprising a non-clinical cohort of males with relatively high psychopathic traits did not report an interaction between MAOA-L and group in the OFC, although they did find grey matter reductions in the right superior temporal pole 53 . Given these findings, it may be that the OFC in early development is less sensitive to the effects of MAO-A VNTR genetic polymorphisms than the amygdala.
We note several limitations of the present study. First, the sample size was relatively small. However, as this study is the only neuroimaging investigation in an antisocial population that has, to the best of our knowledge, considered the effect of genotype on structural brain changes, we are emboldened by these novel results. We have further confidence in our findings, as our power analysis indicated that we were adequately powered to detect significant results given the effect sizes observed. Second, the imaging findings could reflect the influence of genetic variants other than MAO-A that are in linkage disequilibrium with MAO-A. However, the fact that we obtained signal from regions showing abundant MAO-A expression increases our confidence that we are observing activation of the MAOA system. Third, the observed effects could also relate to a range of other non-genetic influences. On the other hand, we measured and controlled for a host of clinical variables that could have impacted our results. Fourth, it is also possible that it was the effect of the MAOA-L genotype which exposed the individual to the influences that made him engage in antisocial behavior. However, not all MAOA-L males had ASPD and some of the typically developing group were also of this genotype. Moreover, the experimental group included some MAOA-H carriers with ASPD as well as typically developing individuals with MAOA-H or MAOA-L.
In summary, we demonstrated that ASPD + PP with MAOA-L display decreased surface area in the right BLA and increased surface area in the right ACo. We take the position that these alterations in the amygdala were influenced by the effect of MAOA-L. Future investigations examining the neurobiological substrate of ASPD + PP should pay heed to the genetic make-up of this group to better determine whether other genotypes implicated in violent behavior also impact brain structure and function in ASPD.

Methods
Participants. Thirty-eight males completed the study protocol: 18 participants with ASPD + PP and 20 healthy controls. Subjects provided written consent to take part in the investigation after study procedures were explained to them. All components of the study were approved by the Research Ethics Board for Human Subjects at the Centre for Addiction and Mental Health, Toronto, Ontario. All methods were performed in accordance with the relevant guidelines and regulations. ASPD Participants. Subjects with ASPD + PP were recruited from the community and from half-way houses operated by correctional services. One forensic psychiatrist (NJK) clinically assessed and diagnosed participants with ASPD using the Structured Clinical Interview for DSM IV Axis II Disorders (SCID-II) 54 and also used the Structured Clinical Interview for DSM IV-TR Axis I Disorders (SCID-I) 55 to assess for other psychiatric illness. Each participant had a history of violent offending that included charges of assault, sexual assault, robbery, uttering threats, and manslaughter. A history of major depressive disorder, bipolar disorder, or a schizophrenia spectrum disorder was exclusionary. Current use of illicit substances, cigarette smoking 56 , and use of psychotropic medications were also exclusionary. To verify the non-smoking status of subjects, breathalyzer testing for carbon monoxide was conducted (MicroSmokerlyzer; Bedfont Scientific Ltd., Kent, United Kingdom). Subjects additionally provided negative urine toxicology screens on all scanning and assessment days.

Control Participants.
Healthy participants responded to recruitment ads seeking study participants. These ads were placed in the hospital, community, and on websites. Control subjects were also recruited and administered the SCID-I and SCID-II by NJK. Control subjects had no lifetime history of psychiatric disorder. All were non-smokers, as determined by self-report and breathalyzer testing, and urine toxicology screens were uniformly negative throughout the study. Criminal record checks from federal government agencies determined that none of the control participants had criminal charges.
Additional Inclusion and Exclusion Criteria. All ASPD and control participants were right-handed.
None had a history of neurological illness, and those with a history of head trauma resulting in loss of consciousness were excluded from participation.

PCL-R.
All subjects were administered the Psychopathy Checklist -Revised (PCL-R) 57 by a trained forensic psychiatrist (NJK). The PCL-R is a clinical instrument that operationalizes psychopathy based on 20 personality and behavioral items. Each item is rated from 0 to 2 based on whether the trait is present (0 = absence of trait; 1 = some indicators of the trait; 2 = definite presence of the trait). Total PCL-R scores range from 0 to 40. Four facet scores can also be generated from the PCL-R that reflect different aspects of prototypical psychopathy. Facet 1 describes interpersonal features; facet 2 indexes affective components of psychopathy; facet 3 assesses lifestyle items; and facet 4 relates to antisocial behavior. Intelligence. The Wechsler Test of Adult Reading 58 was administered to all participants to provide an estimate of full-scale IQ.

ASPD and Conduct
Genetics. Standard PCR procedures using primers as previously described were used to amplify the MAO-A VNTR locus 59 . We implemented minor changes, including labeling the forward primer with 5′ HEX modifier that allowed for electrophoresis and visualization on a capillary sequencer. In brief, 125 ng of total genomic DNA was combined with the following elements: 1X PCR x Amplification Buffer, 1.5 mM MgSO 4 , and 1X PCR x Enhancer Solution that accompanied the Invitrogen TM PCR x Enhancer Kit, 0.2 mM of each dNTP, 0.0975 ug of each primer, and 0.5 U Taq polymerase. A total reaction volume of 20 uL was produced. The cycling conditions were the same as previously described 59 , save for a further denaturation step of 5 min at 95 °C. The ABI 3130 Genetic Analyzer system and GeneMapper software (ThermoFisher Scientific, Waltham, MA) were used to electrophorese and visualize 1 µL of the amplified product. Subjects with 2, 3, or 5 copies of the MAO-A VNTR were assigned the MAOA-L genotype, while individuals with 3.5 or 4 copies were designated as MAOA-H carriers.
Amygdala Volume and Morphometry. These parameters were assessed using the MAGeT Brain pipeline (http://cobralab.ca/software/MAGeTbrain/) 63 based on previous neuroanatomical definitions of the amygdala derived by our group 64 . In brief, five high-resolution atlases with amygdala segmentation were used as inputs 65 . MAGeT Brain bootstraps the segmentation by using an intermediate template library of 21 subjects sampled from the entire population under study and chosen to represent that variability with respect to age and sample distributions (e.g., case vs. control status). Each template receives five segmentations via model-based segmentation with each of the atlases. Finally, at the subject-level, each subject undergoes model-based segmentation with each of the templates, thereby growing the number of possible candidate labels to 105. These labels are then fused via majority-vote to generate the final segmentation. Surface-based representations of the amygdala were generated using a similar methodology as previously described 66,67 , where coordinates for developing the final surface were generated based on the median location of the 105 possible surfaces. Vertex-wise measures of surface area were generated by averaging the individual areas of the three polygons at the intersection of each vertex and were blurred using a 5 mm surface-based blurring kernel 68 .
To gain anatomical specificity, significant vertex-wise statistics were described as occurring in a particular amygdala subregion. We used an anatomical reference atlas 69 to determine the amygdala subregion in which any morphological changes occurred. Specifically, we assessed vertex-wise results in terms of their placement along both the anterior-posterior and medial-lateral axes of the amygdala, as well as their location relative to neighboring structures (e.g. hippocampus). We then matched these results to a corresponding region in the reference atlas to determine the subregion where morphological changes had occurred.
In an exploratory analysis, we also tested whether amygdala surface area differences were related to PCL-R scores among subjects with the MAOA-L genotype. These results as well as the above analyses were corrected using FDR thresholding.
Cortical Thickness. Cortical thickness was estimated using the CIVET processing pipeline (version 1.1.12; Montreal Neurological Institute), in an application similar to previous studies 70,71,72 . Briefly, T1-weighted images were registered to the ICBM 152 average template with a nine parameter transformation 60 , followed by intensity inhomogeneity correction 73 and classification into grey matter (GM), white matter (WM) and CSF 74 . Hemispheres were then modeled as GM and WM surfaces using a deformable model strategy. This approach generates four separate surfaces, each defined by 40,962 vertices 68 . Cortical thickness was determined in native space through non-linear surface-based normalization that uses a mid-surface between pial and WM surfaces. Images were then smoothed with a 20-mm surface-based diffusion kernel and non-linearly registered to a minimally biased surface-based template 75 . Native-space thicknesses were used in all analyses.
The middle and lateral orbitofrontal cortex were defined as ROIs using the cortical parcellations available in the LPBA40 atlas 76 . The intersection of atlas labels and the cortical surface allow computation of average cortical thickness measures per ROI as opposed to per vertex. The middle and lateral orbitofrontal cortex were selected a priori as regions to investigate.
Statistics. Volumetric analyses of the amygdala and OFC were performed in the R statistical environment 77 using a two-way (group × genotype) ANOVA to investigate the effect of the MAO-A genotype on amygdala volume in ASPD + PP offenders and healthy controls. Age, IQ, substance use, and BeAST derived total brain volumes were also included in the model. Subsequent post hoc analyses were conducted using the Tukey's HSD method.
Vertex-wise surface area measures were analyzed with the RMINC statistical package (https://github.com/ Mouse-Imaging-Centre/RMINC) 78 using a two-way (group × genotype) ANOVA to investigate the effect of the MAO-A genotype on amygdala morphology in ASPD + PP offenders. Age, substance abuse history, and IQ were included in the model. In the event that the group × genotype interaction was significant, the interaction was analyzed by subsetting the data into MAOA-L and MAOA-H groups and running a one-way (group) vertex-wise ANOVA in each subset. All vertex-wise surface area values were divided by the total amygdala surface area to account for global structure-wise differences in total size. Vertex-wise measures were corrected for multiple comparisons using FDR 79 as in our previous work 80 . This approach computes an FDR threshold at a given percentage for the F statistic output from the ANOVA. Therefore, among the vertices above the FDR threshold, only a certain percentage can be expected to be false positives.
Power Analysis. Post-hoc power analyses were conducted in order to assess the statistical power of the given sample in relation to the observed means and variances. For each result, the required sample size was plotted to obtain power = 0.80 (α = 0.05) using the observed means and a range of variances. Data Availability. All data generated or analyzed during this study are included in this published article.