Five hundred fifty-seven young (age range 18–36 years, mean 22.9 ± 3.1), healthy volunteers participated in the behavioral part of our study. All participants had at least obtained university entrance diploma or were enrolled at the Otto von Guericke University Magdeburg or at the Magdeburg-Stendal University of Applied Sciences. The university entrance diploma was selected as one of our criteria of inclusion in order to reach a comparable educational background within our sample. Furthermore, the participants were recruited from all departments of both universities like medicine, natural sciences, educational sciences, social and health sciences, or economic sciences which ensured a certain degree of variance in the sample. Moreover, students of psychology were only allowed to take part in the study within their first academic year, to make sure, that they were as naïve as the other participants to our questionnaires and tests. In addition to the analyses of our data, a post hoc analysis of AKAP5 Pro100Leu effects on Buss–Perry aggression questionnaire (BPAQ) scores was performed in a cohort of 604 participants from the University of Barcelona community (457 female, mean age 21.6 ± 3.1 years; see Table 3). This cohort had been described in detail previously (Krämer et al., 2007).

Seventy young, healthy German volunteers (35 Pro homozygotes and Leu carriers, respectively) were recruited from our initial cohort for participation in the fMRI study after exclusion of contraindications. Three of these participants had to be excluded due to technical problems, leaving 67 participants for data analysis (see Table 4 for detailed demographics). All participants gave written informed consent in accordance with the Declaration of Helsinki and received financial compensation for participation. The work was approved by the Ethics Committee of the University of Magdeburg, Faculty of Medicine.

Genomic DNA was extracted from blood leukocytes using the GenElute DNA extraction kit (Sigma Aldrich) according to the manufacturer’s protocol. Genotyping was performed using PCR followed by allele-specific restriction analysis. The DNA fragment on Chr 14q21–24 containing the AKAP5 Pro100Leu polymorphism was amplified using the primers AKAP5_100-f (5′-GCT TCT GAT CAG CCA GAG CCC AC-3′) and AKAP5_100-r (5′-GCT TCT TCC TGG ACT TTG ATG CTG CAG-3′) and standard Taq polymerase (Qiagen and Fermentas). PCR products were digested with AluI (Fermentas), yielding two fragments (174 + 98 bp) for the Leu allele or a single fragment (272 bp) for the Pro allele. DNA fragments were separated on an ethidium bromide-stained agarose gel and visualized under UV light (see Figure 1C for genotyping example). All subjects suspected to be homozygous for the less common Leu allele were genotyped twice. The rater was blinded to the results of the personality inventories. PCR products from two homozygous subjects (one P/P, one L/L) were verified via custom sequencing (SeqLab), confirming that the PCR reaction had provided the desired DNA fragment on Chr 14q21–24.

To avoid population stratification effects and to exclude asymmetric distribution of other genetic variations that might affect aggression or anger, participants were also genotyped for the BDNF Val66Met polymorphism (Chr 11p13), the COMT Val108/158Met polymorphism (Chr 22q11), the DRD2 TaqIA polymorphism (Chr 11q23), the DRD3 Ser9Gly polymorphism (Chr 3q13), the MAOA VNTR polymorphism (Chr Xp11), the serotonin transporter fragment length polymorphism (Chr 17q11–12), for the serotonin receptor 5HT-2a His452Tyr polymorphism (Chr 13q14–21), and for additional polymorphisms in pre- and post-synaptic adaptor proteins (detailed genotyping protocols are available upon request), and X² tests were used to compare the genotype distributions of these polymorphisms between AKAP5 Pro homozygotes and Leu carriers.

To assess a potential influence of AKAP5 Pro100Leu on human aggression and anger, we systematically investigated genotype-dependent group differences in well-established self-report questionnaires of aggressive behavior and anger expression in a cohort recruited from the interdisciplinary campus community of the University of Magdeburg, Germany. From 527 young, healthy participants (289 women, 238 men), complete genetic, and questionnaire data were available. In this cohort, we identified 370 Pro homozygotes, 150 heterozygotes, and 7 Leu homozygotes. With 15.6% frequency of the Leu allele and 28% heterozygosity, the distribution was at Hardy–Weinberg equilibrium (X² = 3.646; p = 0.162), and frequencies were similar to those observed previously in a Caucasian population (Frank et al., 2008). Detailed demographics are shown in Table 1. To avoid population stratification effects, participants were also genotyped for additional polymorphisms in unrelated genes (see Materials and Methods). The polymorphisms were observed at allele frequencies similar to those reported previously, and no imbalance was observed in the distribution of any of these polymorphisms between Pro homozygotes and Leu carriers (all p > 0.133). However, the distribution varied between male and female participants, with the Leu allele being significantly more prevalent in females (X² = 9.37; p = 0.009; Figure 1D; see supplemental information for discussion). Given the low number of Leu homozygous subjects (n = 7), all Leu carriers (Pro/Leu and Leu/Leu) were grouped together for all further analyzes.

Buss–Perry aggression questionnaire and STAXI results are displayed in Figure 2 (see also Table 2), separated by AKAP5 genotype. Because of the unequal distribution of AKAP5 Pro100Leu across gender and because of a small age difference between genotypes (Table 1), we computed a MANOVA with gender and AKAP5 genotype as independent factors and age as covariate. Dependent variables were the four dimensions of the BPAQ and all STAXI domains except for state anger. The MANOVA yielded main effects of AKAP5 genotype and gender, but no significant interaction (Table 2, bottom). There was also a significant main effect of age (see below). Between-subjects univariate ANOVAs yielded significant effects of AKAP5 genotype on the physical aggression scale of the BPAQ (F_1,522 = 8.087; p = 0.005; η² = 0.015) and on the anger control scale of the STAXI (F_1,522 = 6.917; p = 0.009; η² = 0.013). Post hoc one-tailed two-sample t-tests showed that AKAP5 Leu carriers had significantly lower scores in physical aggression (t₅₂₅ = 2.874; p = 0.004, Bonferroni-corrected) and significantly higher scores in anger control (t₅₃₀ = 2.354; p = 0.019, Bonferroni-corrected).

To test, which of the dependent variables distinguished best between AKAP5 Pro homozygotes and Leu carriers, the four domains of interest from the BPAQ (physical aggression, verbal aggression, anger, hostility) and the STAXI scales for trait anger, anger-in, anger–out, and anger control were entered into a stepwise LDA. The LDA showed that physical aggression (Wilk’s lambda = 0.985; F_1,1,525 = 8.262; p = 0.004) and anger control (Wilk’s lambda = 0.975; F_2,1,524 = 6.687; p = 0.001) separated the two genotype groups best, with no other subscores of either questionnaire entering the model.

In order to validate our results obtained in the German cohort we performed post hoc genotyping for AKAP5 Pro100Leu in a previously described cohort of 604 participants from the University of Barcelona community (457 female, mean age 21.6 ± 3.1 years; see Krämer et al., 2007, for detailed demographics). The participants had completed the Spanish version of the BPAQ, but no STAXI scores data were available from that cohort. Table 3 displays demographic data and BPAQ scores. Carriers of the Leu allele showed trends for lower total BPAQ scores (t = 1.515; p = 0.065, one-tailed) as well as for the domains physical aggression (t = 1.478; p = 0.073, one-tailed), hostility (t = 1.666; p = 0.048, one-tailed), and, to a lesser degree, anger (t = 1.128, p = 0.130, one-tailed). These trends could be observed across the whole group and in female participants alone.

While age and gender did affect BPAQ and STAXI scores, they exerted their influence on domains distinct from those affected by AKAP5 Pro100Leu genotype. In our MANOVA model, a significant gender effect was observed for the anger-out dimension of the STAXI (F_1,522 = 4.292; p = 0.039; η² = 0.008), but not for either physical aggression or anger control (all p > 0.523). At the level of between-subject comparisons, we also observed a significant genotype by gender interaction for anger control (F_1,522 = 6.740; p = 0.010; η² = 0.013). Age exerted a significant effect on the trait anger dimension of the STAXI (F_1,529 = 10.051; p = 0.002; η² = 0.019), but there were no age effects on either physical or verbal aggression or on anger control (all p > 0.131). Post hoc Pearson correlations showed a weak but significant negative correlation between age and trait anger (r = −0.121; p = 0.005) and a trend for a negative correlation between age and the anger-out scale (r = −0.084; p = 0.053). No further significant correlations between age and BPAQ or STAXI subscores were observed (all p > 0.203).

To further explore the neural mechanisms that might underlie the observed associations of AKAP5 Pro100Leu with aggression and anger control, we conducted an fMRI study in 67 young, healthy participants (34 Pro homozygotes, 33 Leu carriers). Because of the observed effect of AKAP5 Pro100Leu on anger control and the known trait anger-related attentional bias for angry faces (Van Honk et al., 2001), we hypothesized that Pro homozygotes and Leu carriers might differ at neural level during the cognitive control of emotional interference. We addressed this hypothesis by using an Erikson flanker paradigm with task-irrelevant background pictures (Larson et al., 2006; Wiswede et al., 2009), which, in our study, were angry vs. neutral faces (Figure 3A). In the Eriksen flanker paradigm, processing of incongruent stimuli is typically associated with activations in medial frontal cortex (mFC) regions like the supplementary motor area (SMA, Brodmann area/BA 6) and the ACC (Ridderinkhof et al., 2004).

Previous studies had demonstrated AKAP5 expression in the human amygdala, hippocampus and striatum, but no information was thus far available on the expression of AKAP5 in frontolimbic cortical structures. In our experiments AKAP5 Pro100Leu affected cortical rather than subcortical limbic activity during emotional interference. Therefore, expression of AKAP5 in the human frontal limbic neocortex was taken into account. We used immunohistochemistry on post mortem human brain tissue to assess expression of AKAP5 in human anterior cingulate, medial frontal, and orbitofrontal regions, based on our fMRI results. Indeed, we found AKAP5 immunoreactivity widely distributed in the human cortex. In the ACC (BA 24, 32) and in the mFC (BA 6) AKAP5 was expressed in pyramidal cells and a subset of interneurons (Figure 4A, top row, middle row, left). A similar expression pattern was observed in the medial OFC (BA 11). To verify the validity of these findings, AKAP5 immunoreactivity was also assessed in the hippocampus and striatum, brain regions that have previously been shown to express AKAP5 (Sik et al., 2000; Ulfig et al., 2001; Figure 4A, bottom row). Western blotting using the same primary antibody revealed expression of the previously observed AKAP5-immunoreactive band at approximately 75 kDa (Gardner et al., 2006) in the detergent-insoluble pellet fractions of the human hippocampus and cingulate cortex (Figure 4B), but not in the cerebellum, compatible with previous observations for the AKAP5 ortholog in rodents (Ostroveanu et al., 2007), thus confirming the specificity of the antibody.

Previous studies on genetically mediated variability of aggression and anger have largely focused on genes directly related to neuromodulatory, particularly monoaminergic, transmitter systems (Buckholtz and Meyer-Lindenberg, 2008; Kang et al., 2008; Kulikova et al., 2008; Hess et al., 2009). Here we could demonstrate that a genetic variation in an adaptor molecule that is involved most likely in intracellular signaling pathways of several monoaminergic neurotransmitter systems is associated with human anger and aggression phenotypes in young, healthy, participants. Notably, carriers of the less common variant show lower aggression and stronger anger control when compared to homozygotes of the presumed wild type allele. The association was observed in young, healthy participants and is therefore likely to reflect variability in the normal population. It should be noted that effect sizes were small, but given the complexity of both the phenotypes and the multiple other genetic and environmental influences, no larger influence could have been reasonably expected. In our analysis, effects of AKAP5 Pro100Leu were primarily observed for the dimensions of physical aggression and anger control, while other BPAQ and STAXI subscales showed trends into the same direction that did not reach significance.

The AKAP5 gene product interacts with beta-adrenergic receptors (Gardner et al., 2006), which are highly expressed in the mammalian brain. As beta-adrenergic signaling has been associated with human aggression (Fava, 1997; Hess et al., 2009), modulation of CNS beta-adrenergic activity might be a promising explanation for our observations. AKAP5 has also been linked to dopaminergic signaling and its interaction with glutamatergic neurotransmission (Swayze et al., 2004). Among the neuromodulatory monoamines, serotonin has previously been particularly strongly linked to aggressive behavior (Buckholtz and Meyer-Lindenberg, 2008; Popova, 2008). The role of AKAP5 in serotonergic signaling is thus far unclear, however, an in vitro study suggests a molecular association of AKAP150 with different 5HT receptors (Sandoz et al., 2006). Investigating a potential role of AKAP5 in the serotonergic system is therefore warranted to help explain the present results.

Because our study cohort consisted mainly of students, one might argue that it was somewhat stratified. It should be noted, though, that students were from a large variety of fields of studies (including economics, educational and natural sciences, and medicine, among others), and that the distribution of numerous additional genetic polymorphisms, which all occurred with the previously reported allele frequencies, showed no significant linkage disequilibrium with AKAP5 Pro100Leu. Moreover, the observation of a behavioral phenotype related to a genetic variation in a socially rather homogeneous population, such as the one investigated here, would be expected to be rather subtle and thus, if detected, even more likely to result from genetic differences rather than a sampling bias. Despite a small difference in the distribution of AKAP5 Pro100Leu between men and women (see Results), multivariate statistics revealed independent main effects of gender and genotype that were observed for distinct subscales.

Further though limited support for the validity of the present results comes from post hoc analysis of AKAP5 Pro100Leu on BPAQ scores in a previously reported cohort from Barcelona (Krämer et al., 2007), where comparable trends for lower aggression in Leu carriers were observed. Although the investigation of the AKAP polymorphism and its link to human aggression was not the primary goal in the original design of the Barcelona cohort, it yielded comparable trends for lower aggression in Leu carriers, but the effects were generally less pronounced and did not reach statistical significance in two-tailed comparisons. However, it should be noted that the participants from the Spanish cohort had overall lower BPAQ scores those from the German cohort. The most likely explanation for this might be that the majority of participants in the Spanish study were female (75.6%), and women showed generally lower aggression. Furthermore, the lower overall BPAQ scores might also be related to linguistic or cultural aspects of the Spanish BPAQ version, as in an earlier validation study, García-León et al. (2002) had also reported lower scores in their Spanish subjects as compared Buss and Perry’s original validation of the questionnaire in an American sample, and a direct comparison of BPAQ scores with Dutch and Japanese samples also revealed lower BPAQ scores in a Spanish cohort (Ramírez et al., 2001). The sensitivity of the BPAQ might also have been reduced to some degree by the fact that the participants from Barcelona were largely native speakers of Catalán rather than Spanish, but the Spanish version of the questionnaire was used. Support for the notion that the BPAQ might show linguistically or culturally mediated differences in sensitivity comes from several studies on adaptations of the BPAQ in different languages and cultures (Meesters et al., 1996; Nakano, 2001; Fossati et al., 2003). Those studies were able to confirm the four-factor structure, but reported disagreements on item level and concerning the interpretation of the factors (Von Collani and Werner, 2005), which might be reflected here by hostility being the BPAQ factor most strongly associated with AKAP5 Pro100Leu in the Spanish, but not in the German cohort. Future attempts to replicate our results in different ethnic and cultural populations should therefore include additional instruments, particularly the STAXI, which had not been obtained in the Barcelona cohort.

In the present study, we had focused our investigations on aggression and anger traits. It should be noted, however, that we cannot exclude further influences of the polymorphism on individual differences in emotional processing beyond aggression and anger. Previous animal research has demonstrated a role for AKAP5 in anxiety and fear memory (Moita et al., 2002; Nijholt et al., 2008; Tunquist et al., 2008). Further research is required to assess potential effects of AKAP5 Pro100Leu on anxiety and possibly on the processing of other negative emotions.

Compatible with significantly higher anger control in Leu carriers, our fMRI study showed stronger activation of the left (ACC, BA 24) in Leu carriers relative to Pro homozygotes when an angry face was presented in the background of an incongruent trial, that is, when emotional interference occurred in the condition that required higher attentional control. Increased activation of brain regions involved in attention and executive functions has previously been linked to decreased processing efficiency (Egan et al., 2001; Blasi et al., 2005; Meyer-Lindenberg and Weinberger, 2006). In those studies, increased prefrontal or ACC activation occurred despite similar or even inferior behavioral performance. Here we observed a significant correlation between the ACC BOLD signal in the emotional incongruent condition and shorter RTs. Moreover, as error rates did not significantly covary with the ACC response and were not higher in Leu carriers relative to Pro homozygotes, the reduced RTs likely reflect an actual behavioral advantage rather than impulsivity or a behavioral tradeoff, which would be associated with a corresponding error rate increase (Caldu et al., 2007). Compatible with our results, a recent study investigating the role of the well-described catechol-O-methyl transferase (COMT) Val108/158Met polymorphism on inhibitory control of memory has shown a similar pattern of increased cortical activity linked to better task performance. In that study, COMT Met carriers showed increased prefrontal activation during selective retrieval associated with more efficient inhibition of irrelevant memories (Wimber et al., 2011). Also, the observation that genetically mediated differences in brain activity predicted RTs more accurately than a direct comparison of the different allele carriers (see Results) is compatible with previous studies in which, similarly, fMRI correlates of cognitive processes showed correlations with both genetic variations and behavior that were stronger than the direct relationship between the respective genotypes and behavioral phenotypes (Bertolino et al., 2006; Forbes et al., 2009). For the commonly investigated COMT Val108/158Met polymorphism, a meta-analysis further demonstrated that BOLD responses are more sensitive to genetic modifiers as compared to direct behavioral measures, and the authors suggested that the fMRI endophenotype might reflect a neural mediator of pleiotropic effects of genetic variations on complex human behavior (Mier et al., 2010).

Given the observed relationship between increased ACC activation and shorter RTs, our results are in line with previously reported anger-related attentional biases for angry faces (Van Honk et al., 2001). While the ACC activation was found to be higher in Leu carriers, we observed a reverse effect in the right medial OFC, where Pro homozygotes exhibited a stronger fMRI response in the emotional incongruent condition, that is, in the same experimental condition where Leu carriers show their increased ACC activity. Moreover, activation of the ACC was observed in Pro homozygotes during emotional interference in the congruent condition. This activation might indeed reflect inefficient recruitment of the ACC as it was not correlated with a behavioral advantage during task performance.

Notably, these activation patterns reflect the previously proposed differentiation of a ventral affective processing system (VAPS) and a dorsal executive control system (DECS; Dolcos and Mccarthy, 2006). It is supposed that emotionally salient stimuli are processed in VAPS regions like the amygdala and the OFC while executive functions recruits DECS regions like the dorsolateral prefrontal cortex (DLPFC), posterior parietal cortex, and ACC (Pessoa, 2008; Hart et al., 2010). Related to our findings we suggest that Leu carriers, who exhibit higher anger control and a stronger ACC activation in the emotional incongruent flanker condition, emphasize the requested executive control during task performance. Pro homozygotes, in contrast, might react more sensitively to the emotional (angry) background stimuli, as reflected by a lower anger control and stronger activation of the VAPS in the emotional incongruent flanker condition.

Additionally, in order to test whether the observed effects of AKAP5 on emotional interference processing in the ACC and the OFC might reflect a local phenomenon related to monoaminergic modulation of ACC and OFC activity or rather the result of large-scale network activity, we had investigated the expression of AKAP5 in post mortem human brain tissue. Since AKAP5 was detected in both ACC and OFC pyramidal cells and interneurons (Figure 4), a locally mediated effect of AKAP5 on cognitive control of anger and aggression is possible, however, a more indirect mechanism, for example by influencing amygdala-dependent processing of anger, cannot be excluded. Although the mean age of the donors in the post mortem investigations (52.0 years) was substantially higher than that of the participants in the questionnaire and fMRI studies (22.9 years), it should be noted that mRNA profiling in younger donors (24 and 39 years, respectively) has also provided evidence for AKAP5 expression in frontolimbic structures (Allen’s Brain Atlas¹).

An important direction for future research will be the investigation of molecular consequences of AKAP5 Pro100Leu. The polymorphism is located close to the second basic membrane localization domain and the GPCR-binding region (Figure 1A) and affects the length of an alpha-helical structure and the geometry of the membrane-attached N-terminus (Figure 1B), with a potential effect on the folding of the downstream domains (Protein folding was predicted using the PEP-FOLD prediction software²).

Given the localization of the polymorphism within the AKAP5 gene, we tentatively suggest that Pro100Leu might modulate interactions with GPCRs. The consequences of such an interaction are, however, be difficult to predict, particularly in relation to a complex behavioral phenotype such as aggression. AKAP5 Pro100Leu might, for example, affect the colocalization of GPCRs with their intracellular targets. Apart from a potential influence of AKAP5 interaction with GPCRs, it appears also plausible that the polymorphism might affect the interactions of AKAP5 with other post-synaptic adaptor molecules or with the post-synaptic membrane.