Our investigation suggests that although SERT genotype influences aggression and brain function, this effect appears to be independent from that of MA. We found no evidence for higher SERT risk allele loads in the MA-dependent group, and no interaction between MA abuse and genetic risk in predicting aggression (AQ scores). The imaging findings suggest that genotype and MA abuse act via distinct neural substrates, overlapping only in right ventrolateral PFC. However, it should be noted that the present sample was relatively small, and negative findings should be interpreted with caution.
Our finding of comparable SERT risk allele loads between MA-dependent and control groups in the present sample is consistent with a previous study,55
and is the first report of the relative distribution of STin2 alleles in these groups. The findings argue against the possibility that a disproportionate load of SERT risk alleles in MA-dependent populations drives observed differences in aggression and neurocircuitry. However, only three SERT polymorphisms were queried, and SERT is among many polymorphic genes influencing aggression,65
suggesting that other genetic loci may play a role. It is also possible that owing to differences in early environmental experiences,66
differential gene × environment interactions led to distinct outcomes despite similar SERT allele distributions. Given the relatively small sample size, the possibility that the analysis was not adequately powered to detect differences in genotype distribution cannot be excluded. Replication in a sufficiently large sample (n
>7866, given post-hoc
power analyses with present effect sizes) would be necessary to conclusively interpret this negative finding.
The aggression self-report findings, showing effects of both MA abuse and genetic risk, are consistent with previous reports of heightened hostility and aggression in MA-abusing samples,2, 3, 4, 5, 6, 7, 8
as well as evidence for LPR short allele and STin2 12-repeat allele involvement in social–cognitive and emotional phenotypes associated with aggression.31, 32, 33, 34, 35
However, there was no interaction between these factors, suggesting that MA abuse and SERT genotype influence aggression via distinct mechanisms.
One such distinction may involve the amygdala, as its activation showed an effect of genetic risk, but not MA abuse. Our finding of SERT-related variation in amygdala activation is consistent with previous reports;42, 43, 44
however, the direction of the effect was unexpected. In previous studies, LPR short-allele carriers have exhibited amygdala hyperactivation, whereas in the present sample, signal change was lower in this group. It is possible that the finding reflects tonic hyperactivity of the amygdala in the high-risk group, creating a ceiling effect, while individuals in the low-risk group can accommodate a larger change in activation. Several lines of evidence support this view, including a ‘tonic model',67
suggesting that carriers of the LPR short allele have high amygdala activity at rest and during neutral and undefined conditions.68, 69, 70
In addition, amygdala activation in the present MA-dependent sample correlated inversely with MA use measures (amount and frequency), so that that the lowest signal change occurred in those participants who used MA most heavily. Heavy MA use has been associated with high glucose metabolism in the amygdala,21
suggesting that in the present sample, the lowest activation reflected the highest glucose metabolism (tonic activity). The direction of the effect may therefore reflect the relative nature of functional magnetic resonance imaging, rather than an actual reversal of the effect.
Our finding that amygdala activation did not differ between MA-dependent and healthy participants suggests that heightened MA-related aggression may occur via mechanisms independent of the amygdala (for example, higher cognitive and executive processes, life and social experiences or personality traits). Again, we cannot exclude the possibility that our sample size did not permit detection of differences. Replication in a sufficiently large sample (n>275, given post-hoc power analyses with present effect sizes) would be necessary to conclusively interpret this negative finding.
Finally, both MA abuse and genetic risk modulated activation in cortical regions linked to emotion processing, but the regions showed little overlap. Genetic risk influenced the core system for visual analysis of faces,71
and PFC regions involved in recognition and interpretation of facial affect, calculation of stimulus contingencies and regulation of emotional responses,11, 72
consistent with the idea that SERT-related aggression is linked to social–cognitive neurocircuitry. Effects of MA abuse were less manifest, revealing only small activation differences in PFC and occipital cortex. While also part of emotion processing circuitry, the roles of these regions are less clear. Among the regions identified, the only one that showed overlap between genotype and MA effects was the right ventrolateral PFC, a region implicated in social cognition and inhibitory control.13, 14
The overlap suggests that this region modulates aggression regardless of sources of variation (genotype or MA abuse), suggesting it may have utility as a biomarker for risk or intervention. Taken together, the data suggest that the effects of MA abuse add to—but mostly do not interact with—genetic risk in modulating aggression and associated neurocircuitry.
Several limitations of the study should be noted. First, statistical power was limited owing to sample size. We maximized power by combining groups, using a small number of outcome measures and including an ROI approach in functional magnetic resonance imaging analyses. Second, not all withdrawal symptoms resolved by the time of testing, possibly confounding results; however, symptoms and outcome measures were not correlated, suggesting that outcomes were independent of withdrawal. In addition, data were obtained at only one point following a brief abstinence period; differences in brain activation may vary with time. Third, our method for combining genotypes was based on one prior study in a prison population,51
and generalization to other populations is uncertain; however, our results using both combined and individual variants (see Supplementary Materials
) support using this strategy more broadly. Finally, comparable amygdala activation between negative and neutral faces calls into question the psychological significance of the probe task. However, although broad, the measured psychological operation (‘detection of socially salient environmental cues') appears highly relevant to aggression, and as such, meaningful to investigate. Potential reasons for comparable amygdala activation between stimulus types are small sample size/low statistical power, attribution of negative valence to neutral faces73, 74
or general role of the amygdala in detecting socially relevant stimuli rather than merely threat.75
In summary, this study adds to our understanding of genotype- and MA-related modulation of aggression and social–cognitive neurocircuitry, suggesting that differences associated with MA abuse occur independently of, and in addition to, the predisposing influence of SERT genotype; it is therefore important to recognize that in a certain proportion of MA-dependent individuals, mere abstinence from MA or attention to MA-use effects will not be sufficient in decreasing propensity for aggression and violence (and, thereby, minimizing risk for stress-induced relapse). At the same time, the findings point to right ventrolateral PFC as a potential biomarker for both MA- and genotype-related aggression, suggesting that modulation of its function may allay this behavior regardless of its source. In light of these findings, continued investigation of socio-emotional function and its neurobiological underpinnings will be critical in addressing problematic behaviors and developing appropriate strategies for intervention.