Dorso-lateral prefrontal γ-amino butyric acid in men predicts individual differences in rash impulsivity

Introduction

The ability to flexibly regulate one’s behaviour in a changing environment is key to adaptive social life. Failure in this regulation is salient in psychiatric disorders, where impulsivity is the second commonest symptom in the Diagnostic and Statistical Manual of mental disease (1, see 2). However, impulsivity is an umbrella term comprising several facets, which have been clarified in the UPPS model as: Urgency (or ‘rash impulsivity’), the tendency to act rashly in response to distress or other strong emotions and urges; Premeditation (lack of), the tendency to envisage the consequences of an act before engaging in it; Perseverance (lack of), the ability to remain focused on a difficult task; Sensation Seeking, the tendency to engage in exciting activities (3).

Of these four traits, urgency has been most strongly linked to problematic levels of alcohol consumption, smoking, gambling, binge eating, illicit drug use, aggression and risky sexual behaviour (4,5). High levels of urgency represent one extreme of wide natural variation in this trait, and the UPPS model is an example of how it is fruitful to understand human behaviour and its maladaptive variants along a continuum (6). Thus identifying the neurobiological bases of individual variation in urgency may be of key importance for understanding the pathophysiology of psychiatric disorders.

Building from previous focus on serotonin and dopamine (2,7,8), there is now increasing interest in the role of GABA-ergic neurotransmission in impulsivity. Variation in GABA receptor genes has been associated with problematic levels of alcohol use, obesity and drug use, as well as bipolar spectrum disorders, conduct problems and distress regulation (9-16). Notably, Dick et al. (16) found that the GABRA2 gene is involved in the predisposition to alcohol dependence through a generalized ‘externalizing’ pathway, a concept that captures a range of disinhibited behaviours that commonly co-occur, including aggression, aspects of impulsivity, antisocial behavior and drug abuse (17), forming a coherent spectrum of personality and psychopathology. Research connecting externalizing psychopathology and personality traits (18) has shown that externalizing liability is a normally distributed dimension of risk, with urgency having the largest loading on the general externalizing factor (5,19). Thus, given that genetic variation in GABA transmission has been linked to general externalizing, and urgency is the best personality trait marker of the general externalizing liability, we predict that variation in GABA levels would be related to the urgency facet of impulsivity.

Any relationship between GABA and urgency may be specific to particular brain regions. Several studies have related general impulsivity to activity or grey matter thickness in the lateral prefrontal cortex (20-23). Increased delay discounting of rewards may reflect a general vulnerability to externalizing (24) and TMS disruption of dlPFC leads to such enhanced discounting (25). Increased externalizing traits in undergraduate subjects are related to reduced amplitude of the error-related negativity (ERN, 26) brain response, a signal that is abolished following lesions of lateral PFC (27). Variation in urgency is linked to the ability to adaptively regulate emotions and cravings (5), and recent research links dlPFC to this aspect of control (e.g.,28). Moreover, individual variation in emotion regulation abilities has also been linked to dlPFC function (29). In addition, MRS and post-mortem studies show alterations in GABA in dlPFC for psychiatric disorders linked with heightened emotional dysregulation and poor self-control, including cocaine addiction, depression and schizophrenia (30-32). Hence, dlPFC would be a strong candidate area in which a relationship between urgency and GABA levels might be manifest.

There are also reasons to propose the inferior frontal gyrus (IFG) or the anterior cingulate cortex (ACC) as candidate areas for a GABA-urgency relationship. IFG, for example, has been strongly linked with response inhibition in the stop-signal task (33), and variation in response inhibition has been suggested to be an important mechanism underlying externalizing vulnerability (34). ACC, together with dlPFC, is thought to be a neural generator of the ERN (35). Further, Matsuo et al. (36) found that gray matter volume in rostral ACC was correlated with impulsivity (as measured by the Barrat Impulsiveness Scale 37).

Advances in MRS (38-41) allow us to investigate individual differences in GABA concentration in specific regions of the human brain. This approach has proved successful in discovering group differences in regional GABA concentration between psychiatric and control cohorts (e.g., 32,42), and more recently in relating specific behavioural and physiological measures in healthy participants to individual differences in regional GABA levels, even in the small cohort size of typical MR studies (43-48). In frontal cortex in particular, higher GABA concentration measured in the region of the frontal eye fields was correlated with lower eye movement distractibility (i.e. better control, 49), leading us to predict that the likely direction of any association between frontal GABA and impulsivity would be higher GABA concentration, lower impulsivity (better control).

In this study we take a two-step approach. First, we explored whether resting GABA levels in any of the three candidate brain regions (dlPFC, IFG and ACC) were related to urgency, or indeed any of the other three impulsivity facets. Second, having identified a correlation between urgency and GABA in dlPFC, we aimed to test whether this would replicate in a second cohort. We also include data from two control regions, the parietal cortex and supplementary motor area (SMA), which were measured in the same participants for the purpose of a different study (46).

Materials and Methods

Results

The mean scores and their standard deviations in each subscale are given in Table 1. These were similar across the two cohorts (all p>.07), and comparable to those reported in similar populations of university staff and students.

The main result for the first cohort was a correlation between GABA concentration in dlPFC and urgency, such that participants with higher GABA concentration scored lower on this factor (Pearson’s r=−0.70, Spearman’s ρ=−0.77; Figure 2-left). There was also a correlation of dlPFC GABA with perseverance (r=−0.58). However, with our sample size, these correlations are not significant after correcting for multiple comparisons (5 areas □ 4 impulsivity facets). Therefore we tested for replication in a second independent cohort, using the results of the first cohort to specify planned comparisons. The correlation between dlPFC GABA and urgency was replicated (Pearson’s r=−0.67, p=.013, 1-tailed, corrected for 2 comparisons, Spearman’s ρ=−0.78, p=.0015; Figure 2-right). We also performed bootstrapping to check whether just one or two points drove the correlations. This involved taking N random samples from within our data, with replacement (where N is the number of participants), computing the correlation coefficient r, and repeating 10000 times to find the 95^th percentile value of r (i.e. the value for which 95% of the values produced in bootstrapping were greater or equal to this value). If a correlation is driven by just one or two points, since in many of the iterations these would be left out, near zero r values would occur more than 5% of the time. Thus if the 95^th percentile r-value were near zero or crossed it, it would indicate the correlation is not robust. The Bootstrap 95^th percentile values for R for the Pearson’s r for cohorts 1 and 2 were −0.31 and −0.33, respectively, indicating they are robust to subsampling.

This relationship did not generalize to the Glx peak in the spectrum (glutamate+glutamine, Figure 2-left insert; 1^st cohort: r=−0.15, p>.05, Figure 2-right insert; 2^nd cohort: r=−0.3, p>.05). Neither did Glx correlate with GABA (1^st cohort: r=0.39, p>.20; 2^nd cohort: r=0.37, p > .23) or with any impulsivity subscale (1^st cohort: all ∣r∣<0.39, all p > .2; 2^nd cohort: all ∣r∣<0.22, all p>.45). However, note that the Glx peak is not a pure measure of glutamate, because glutamate cannot be unambiguously separated from glutamine using this technique at a 3T field strength, therefore this absence of correlation cannot be taken to mean that individual differences in glutamate play no role.

GABA concentration in the dlPFC did not correlate with Premeditation or Sensation-seeking, and the correlation with Perseverance found at first did not replicate in the second cohort (Figure 3-a,b, top rows). This stresses the importance of an independent replication when evaluating correlations obtained with small samples and multiple possible comparisons. In terms of regional specificity, we found no evidence that any of the impulsivity traits correlated with GABA measures taken from voxels in the other two candidate regions (IFG and ACC) or with the other two regions for which we obtained measurements with these cohorts (SMA and parietal).

We tested whether the correlation between urgency and GABA in the dlPFC region might arise due to other individual factors such as age or grey matter fraction in the voxel. It did not; there were no significant correlations of either the urgency scores or dlPFC GABA with these factors (Figure 4-a,b), and when controlling for them, correlations between dlPFC GABA and urgency remained strong or even improved (Controlling for the Gray matter fraction: 1^st cohort: r_partial=−0.77, p<.005; 2^nd cohort: r_partial=−0.69, p<.012; Controlling for age: 1^st cohort: r_partial=−0.70, p<.016; 2^nd cohort: r_partial=−0.68, p<.014). We also tested whether the correlation might be related to irregular sleep patterns, tobacco, alcohol, drug use, or weight in our participants (for the 15 participants who provided this information). There was no relationship with any of these factors. None of the participants was taking any medication at the time of the study.

Because externalizing vulnerability has been linked to impaired response inhibition (34), we also tested whether results of the Stop-signal task, an experimental measure of response inhibition, correlated with dlPFC GABA (and with Urgency). We did not find any consistent relationship between GABA and either stopping ability (Figure 4-c), or overall response speed (1^st & 2^nd cohorts: all ∣r∣<0.08, p>.8). It is worth noting that performance in the stop task itself has previously been more strongly related to IFG and pre-SMA than to DLPFC (64-67), but we found no significant correlation between stopping and IFG GABA (1^st cohort: r=0.24, p>.4) or between stopping and GABA in our voxel centered on the SMA, which contained part of the pre-SMA (1st cohort: r =0.21, p>.5). We also did not find a relationship between stopping ability and any of the components of impulsivity.

Discussion

The neurochemical balance in different brain regions has long been thought instrumental in the regulation of personality traits and psychiatric disorders, but it is not straightforward to link genetic or pharmacological data implicating particular neurotransmitter systems with particular brain regions (e.g.,8). Employing magnetic resonance spectroscopy (MRS) offers a way to fill this gap, enabling us to report here that individual differences in GABA concentration in a specific region (dlPFC) were reliably related to a specific personality trait (urgency). It is worth emphasizing that the correlation did not arise from general GABA concentration over the whole brain – there was no correlation with the other frontal and parietal areas we measured, although low signal-to-noise in edited MRS techniques may have contributed to absence of detection of correlations in these regions. Previous reports have found no evidence for correlation of GABA concentration between brain regions (46,49,68). Thus on current data, it seems that having high GABA concentration in dlPFC does not necessarily mean that a person will have high GABA concentration in other brain regions, underlining the importance of targeting specific brain regions in clinical GABA MRS studies (32,69,70).

Recently, a series of genetic studies have linked variation in GABAergic neurotransmission with a variety of co-occurring disorders characterized by emotional dysregulation and poor self-control (9-11,14-16), thought to reflect a coherent externalizing spectrum (18). Our results support the utility of a dimensional approach to personality-psychopathology relations (e.g.,6). The results also link GABAergic neurotransmission with the role of dlPFC in self-control (e.g.,28,29,71). The exact mechanism by which differences in dlPFC GABA level would influence self-control ability is unknown. Our findings are consistent with the view that dlPFC neurons adapt their properties to robustly maintain information that is relevant to current goals (72). In the context of such ‘adaptive coding’ in the control of working memory, GABAergic inhibition has been posited as the mechanism for adjusting the focus of dlPFC neurons to the task at hand, leading to increased robustness of goal-representations in the face of distraction (e.g. from rewards or stressors, 73). Similarly, Sawaguchi et al. (74) found that GABA continuously and preferentially suppresses neuronal activity via GABA-A receptors to limit the population of prefrontal neurons activated by a particular behaviour, consistent with adaptive selectivity. This process may have parallels with the role of GABAergic inhibition in tuning the response properties of visual cortex neurons, sharpening orientation selectivity, for example (56,75). Thus it is plausible that higher GABA concentration can lead to more robust adaptive coding of goals in dlPFC, providing more effective integration of salient motivational information with current task goals, thus promoting adaptive behaviour (76).

The details of the relationship between GABAergic neurotransmission in dlPFC and functional inhibition of behaviour are likely to be complex. The dlPFC is part of a network involving loops through striatum and thalamus as well as numerous connections to other cortical and subcortical areas relevant for controlling behaviour. Similarly, the roles of dopamine and serotonin are yet to be clarified (7,8,23), and are likely to interact with that of GABA (77,78). Moreover, the reasons why people may differ in MRS measures of regional GABA concentration, and how this relates to variation in receptor genes, remain unknown. Also unknown is the contribution that macromolecules make to the GABA measurement: They are likely to make a significant contribution to the absolute measures, but are currently thought unlikely to drive the differences between individuals, which correlate with neurophysiological and behavioural differences that are readily explained by GABA variation, but would be difficult to explain by macromolecule variation (43-45,47,48,56,79). Most likely, differences in our measure of GABA concentration will be related to the number of GABA interneurons or the number and type of synapses, and we do not yet know what implications this could have for the wiring of GABAergic and glutamatergic excitation-inhibition circuits across the different cell layers of the neocortex (80,81). Importantly, however, such regional individual differences should not be expected to mimic the effects of agonist/antagonist pharmacological agents, which disturb neurotransmission over larger parts of the brain, and probably do not change the number of interneurons present. Even when regionally specific, pharmacological agents may not mimic natural variation: for example, GABA_A receptor agonists such as muscimol can be used to temporarily ‘deactivate’ regions of cortex (e.g.,82), but deactivating dlPFC would be expected to reduce impulse control (25) and thus be associated with higher urgency. Our finding that higher GABA predicts lower urgency is opposite to this prediction, and shows that natural variation in GABA should not simply be considered analogous to more or less cortical deactivation.

Our findings also imply that neither GABA nor impulsivity have a simple relationship with response inhibition as measured by the Stop-Signal reaction time alone. We found no correlation between either GABA or impulsivity and the stop-task in our samples. Although correlation between impulsivity and the stop-task has been reported previously (83,84), other studies have not replicated this (e.g.,85). Young et al.’s findings (34) are informative in this respect. They found that the relationship between externalizing disinhibitory traits and response inhibition is better captured at the level of a latent response inhibition variable, measured across several ‘inhibition’ tasks, than by any single task. Overall, these results suggest that dlPFC GABA is more related to a higher-level trait indexed by urgency, than specific to the performance in the Stop-signal task, although such conclusion based on absent correlations must be only tentative.

In sum, we have found that that one facet of impulsivity, urgency, a quantitative trait linked at high levels with several clinically problematic externalizing, disinhibitory conditions, correlates with the concentration of GABA measured in dlPFC, a region implicated in self-control. We hope that these results, as well as future research using regionally specific MRS in humans, will help to bridge the gap between the regional specificity provided by imaging studies and the biochemical data from genomic and pharmacological investigations.