Introduction
The global health burden of psychiatric illness (Whiteford et al.
2013) and its serious consequences for the individual and society (Organization
2014) have encouraged recent efforts to improve understanding of the etiology and pathophysiology of mental disorders. More precise phenotyping of disorders could in turn result in better diagnostics and therapeutics (Insel
2014). Despite the high heritability estimates for most major psychiatric disorders, including attention deficit hyperactivity disorder (ADHD), autism spectrum disorders (ASD), bipolar disorder and schizophrenia (Committee
2009), there are still few genetic markers that reliably associate with these disorders and limited significant genome-wide associations have been documented to date (Sullivan
2011). One possible reason for this is that psychiatric disorders are often heterogeneous at the symptom level and thus uncertainties about phenotype definition may have impeded the discovery of associated genetic risk factors (McLoughlin et al.
2013). A current goal for research in psychiatry is to close the gap in understanding between the symptoms and causes of psychopathology, and move beyond subjective and variable clinical diagnoses to classify disorders based on identifiable neural circuits, and further to link activity in these circuits to the cellular and genetic levels (Insel and Cuthbert
2009).
Within psychiatry, the concept of an endophenotype refers to a heritable quantitative trait, either cognitive or neurophysiological, that is more directly related to dysfunction in neural systems than diagnosis, and which therefore facilitates the identification of genetic variants associated with psychopathology (Gottesman and Gould
2003). Ideally an endophenotype should be heritable, associated with disorder diagnosis and be present in unaffected family members (Hawi et al.
2015). Brain dynamic measures known to relate to the pathophysiology of psychiatric illness can direct search for disorder endophenotypes. One strategy for identifying such measures is to examine cognitive and neural dysfunction closely related to core behavioural symptoms.
Cognitive or executive control, the capacity to flexibly direct and allocate resources to a goal by selecting and integrating relevant contextual information, is critical for higher mental abilities (Blasi et al.
2005). Cognitive control is among the most heritable of neurocognitive traits (Anokhin et al.
2004; Friedman et al.
2008; Macare et al.
2014) and deficits in cognitive control are consistently associated with multiple psychiatric illnesses (Laucht et al.
1997a; Millan et al.
2012; Nieoullon
2002; Royall et al.
2002). Thus there has been a focus on genetic approaches to uncover the biological underpinnings of these measures. Converging evidence from pharmacological, neuroimaging and animal studies indicates that dopamine is critical for the efficiency of cognitive control (for a review, see Cools
2008). Specifically, numerous studies suggest that too little or too much dopamine in the cortex, in particular in the prefrontal cortex (PFC), can be disruptive, impacting both brain dynamics and performance on tasks requiring cognitive control (Bruggemann et al.
2013; Laurens et al.
2010; Luu et al.
2004).
Recent evidence indicates that catechol-
O-methyltransferase (COMT), which has a crucial role in the regulation of dopamine in the PFC, accounts for some individual variability on tests of executive or cognitive control (Dickinson and Elvevåg
2009; Goldberg and Weinberger
2004). The common polymorphism (Val158Met) predicts activity of COMT; homozygosity for the COMT Met allele is associated with a 35–50% decrease in enzymatic activity and dopamine catabolism relative to Val homozygosity, ultimately resulting in increased availability of dopamine in PFC (Biederman et al.
2008; DeYoung et al.
2010; Dickinson and Elvevåg
2009; Lachman et al.
1996; O’Sullivan et al.
2009). Individuals with the homozygous Met allele (and concomitant PFC dopamine enhancement) show improved cognitive control, working memory, and general intelligence as indexed by both performance, and correlated event-related potential (ERP) and functional magnetic resonance imaging (fMRI) measures (Bertolino et al.
2004; Blasi et al.
2005; Cahoy et al.
2008; Diamond et al.
2014; Eroglu and Barres
2010; Freeman and Rowitch
2013; Joober et al.
2002; Wong and Van Tol
2003). However, other studies using cognitive performance, ERP and transcranial direct-current stimulation (tDCS) measures (Plewnia et al.
2013; Stefanis et al.
2004; Taerk et al.
2004; Tsai et al.
2003) have reported no association between COMT and cognitive control, or even that homozygous Met carriers exhibit abnormal cognitive control activity (Gallinat et al.
2003; Kramer et al.
2007).
Another polymorphism important for understanding genetic effects on executive control is the 7-repeat polymorphism of the DRD4 gene, which decreases dopamine transmission at the D4 receptor (Bellgrove et al.
2005; Johnson et al.
2008; Kiphardt
1974). D4 receptor function may play a role in the etiology of both personality traits and psychiatric illness (Ptacek et al.
2011) with a particular relation to ADHD—an association with the 7-repeat allele is the strongest and most consistently replicated molecular genetic finding in the disorder (Banaschewski et al.
2010; Faraone et al.
2001; Li et al.
2006). Participants carrying the 7-repeat allele also have been reported to show poorer performance on tasks of attentional and executive functions and to have smaller attention-related ERP peaks in such tasks (Albrecht et al.
2014; Vogel et al.
2006). However, findings have again been inconsistent. Some studies have reported that the 7-repeat allele is associated with better performance on executive function tasks (Johnson et al.
2008; Swanson et al.
2000), while other studies have found no such differences (Barkley et al.
2006; Konrad et al.
2010).
Disparities between the results of studies investigating the role of COMT and DRD4 in cognitive control may relate to the use of psychiatric populations (Bertolino et al.
2004; Ehlis et al.
2007; Eroglu and Barres
2010; Stefanis et al.
2004; Taerk et al.
2004) or to the effects of other interacting genes (Wishart et al.
2011). Given that the level of dopamine in the cortex has specific effects on cognitive performance (either too much or too little can lead to impairments) (Bruggemann et al.
2013; Laurens et al.
2010; Luu et al.
2004) it is likely that the epistatic interaction between COMT and DRD4 has an impact on dopaminergic influences on cognitive control. Indeed, one recent study showed no impact of the single genes on an ERP-derived measure of response control (the “NoGo anteriorisation” or NGA) (Fallgatter et al.
1997), whereas there was a strong interaction between COMT and DRD4 on NGA amplitude and also on reaction time variability (Takahashi et al.
2011). This interaction effect may not extend to reward-related processing as measured by ERP markers to feedback on gains and losses (Cavanagh et al.
2012).
Another possible reason for not replicating COMT and DRD4 effects in cognitive control tasks is the relative inffectiveness of the measures used to characterise the underlying neurophysiology of the complex cognitive sub-processes (Cools
2008). Cognitive control has a multifactorial nature. Cognitive electrophysiology as well as functional imaging studies using fMRI have demonstrated that it is associated with activity in a network of brain regions, in particular the basal ganglia, prefrontal, and parietal attentional systems including medial frontal and dorsolateral prefrontal cortex, anterior cingulate cortex (ACC), and precuneus (Ogg et al.
2008; Yoon et al.
2008). Indeed there is evidence that the effects of COMT and DRD4 may contribute to anterior-posterior functional connectivity (Jaspar et al.
2014; Liu et al.
2010; Tian et al.
2013) and thus extend beyond the PFC to areas including the dorsal anterior cingulate (Blasi et al.
2005).
However, it should be noted that the associations previously found between COMT, DRD4, and cognitive ERP measures are not exclusive, and certainly do not suggest a one-to-one relation between dopamine-related gene functions and ERP component measures of cognitive control subprocesses at the scalp or source level.
One of the most studied tasks in relation to cognitive control in psychiatry is the continuous performance task (CPT; Dickinson et al.
2008). The cued version of the task (CPT-AX; Doehnert et al.
2010,
2008; McLoughlin et al.
2010,
2011; Valko et al.
2009) has the advantage of measuring two important components of cognitive control: attentional control (how to allocate attentional resources), via responses to ‘Go’ stimuli, and also response inhibition (monitoring performance in face of conflicts),via responses to “NoGo” stimuli. Deficits in both performance and ERP measures for the CPT-AX task, in particular in the amplitudes of the “Go N2/P3” complex and the corresponding “NoGo N2/P3” peaks aligned with response-initiating and response-inhibiting letter presentations, respectively, are consistently and strongly associated with psychiatric illness, including schizophrenia (Holmes et al.
2005; Lee and Park
2006; Salgado-Pineda et al.
2004), ASD (Lundervold et al.
2012; Tye et al.
2014; Wang et al.
2013) and ADHD (Albrecht et al.
2012; Banaschewski et al.
2004; Brandeis et al.
2006; Huang-Pollock et al.
2012; McLoughlin et al.
2010,
2011).
In the current study, we examined the relationship of two key components of cognitive control, namely the relationship of response execution and inhibition during the (CPT-AX) cued continuous performance task to the COMT val158met and DRD4 exon III VNTR polymorphisms. Previous publications based on this data set have already addressed the genetic modulation of motor postprocessing (Bender et al.
2012a), motor response variability (Bender et al.
2015), and visual postprocessing (Bender et al.
2012b) using statistical analysis at the sensor level of scalp ERPs. To more accurately characterise the complex cognitive sub-processes of cognitive control, we here increased the precision of these EEG measures by estimating brain activity at the source via decomposition of the data into maximally independent source processes using independent component analysis (ICA). ICA identifies component processes in the EEG data that are not only temporally near independent but typically also functionally independent in the sense that they exhibit more distinct response patterns to experimental events of interest than do the same measures applied to the raw channel data, consistent with the fact that scalp channel signals each sum potentials volume conducted to the scalp from a a wide variety of relevant and irrelevant brain and non-brain sources (McLoughlin et al.
2013) (Makeig et al.
1996,
1997).
ICA decomposition has been shown to yield event-related measures that share more genetic variance with behaviour than channel-based EEG measures (McLoughlin et al.
2014), to better characterize individual differences in schizophrenia and ADHD symptomology than scalp channel measures (Lenartowicz et al.
2014; McLoughlin et al.
2014; Rissling et al.
2014). Thus, in general ICA decomposition may enable a more precise characterization of spatiotemporally complex cortical network dynamics associated with cognitive control than equivalent measures applied directly to the recorded data channels.
Based on previous findings, we predicted that carriers of the COMT Met/Met allele will demonstrate improved neural function in the cortical source activities related to response execution and inhibition compared to other COMT alleles, and conversely that individuals with the DRD4 7-repeat will show impaired response execution and inhibition. We also consider the impact of the interaction effect between the two polymorphisms on the neural measures of cognitive control.
Discussion
In this targeted candidate gene analysis, we identified several associations between polymorphisms of dopamine system genes and ERP indices of cognitive control during the cued continuous performance task CPT-AX. Specifically, we found that the homozygous Met allele of the COMT genotype was associated with smaller mean P3 amplitude in the IC ERP time-locked to Nogo stimuli and further that individuals with the 7-repeat polymorphism of DRD4 had a smaller peak IC-P3 to Go stimuli. Furthermore, an interaction effect between COMT and DRD4 emerged, indicating a smaller IC-derived Go-P3 in those with the 7-repeat if they also had the homozygous Val/Val polymorphism or the homozygous Met/Met polymorphism, but not in those with the heterozygous Val/Met polymorphism. In contrast to these findings, there was no association between any of the selected polymorphisms and performance measures or the IC-N2 measures.
The overall pattern of task–related ERP activity in our study was consistent with the fronto-parieto midline prominent N2-P3 peak complex observed in tasks that require attentional and cognitive control (Castellanos and Proal
2012; Jaspar et al.
2014). Due to proposed site of action of the COMT enzyme, most of the research on COMT has focused on prefrontally-mediated cognition. Our study instead found that Met allele homozygotes exhibited a smaller ERP measure of response inhibition for a parietal/temporal source domain, but did not identify a significant relationship between COMT and activity in prefrontal cortex. While, COMT alters enzyme activity in prefrontal cortex, and has been strongly associated with activation in frontal brain regions during task performance, the expression of this enzyme is widespread and relatively uniform within the human brain (Hong et al.
2014). Recent findings indicate that the effect of the COMT val
108/158
met polymorphism extends beyond the PFC and has different effects on brain activity and structure in other regions (Hong et al.
2015), and may specifically include parietal activity during inhibitory control (Van Rooij et al.
2015). These findings indicate that COMT-related inhibitory brain activity is not confined to prefrontal regions.
That the COMT Met/Met allele was associated with an impairment in a neural index of response inhibition may be viewed as surprising, given that the Met allele has been associated with improved performance in executive function and cognitive tests (Cahoy et al.
2008; Freeman and Rowitch
2013; Kiang et al.
2009; Wynn et al.
2010). However, the role of dopamine in the brain and its relationship with the neural network underlying cognitive control is complex (Durstewitz and Seamans
2008; Goldman-Rakic et al.
2000). The dual state or U shaped theories of dopamine regulation (Durstewitz and Seamans
2008; Meyer-Lindenberg and Weinberger
2006) predict that homozygous Met carriers have high tonic dopamine levels in the prefrontal cortex (PFC) and thus optimal levels of dopamine for tasks that require cognitive stability or maintenance, including working memory and competing programs (Nolan et al.
2004), while Val carriers have high phasic dopamine levels in the PFC, and are thus more efficient when cognitive flexibility is required, such as during rapid updating or task switching (Colzato et al.
2010; Drabant et al.
2006; Nolan et al.
2004). Our finding is largely consistent with studies examining the role of COMT that have found that those with the Met/Met polymorphism have impaired response inhibition performance in comparison to Val carriers (Weiss et al.
2014). Our findings are in further agreement with studies that show that the homozygous Met allele is associated with behavioural measures of impulsivity (Soeiro-De-Souza et al.
2013; Stipursky et al.
2011), particularly with a failure to plan ahead (Soeiro-De-Souza et al.
2013) and poorer delayed discounting (Stipursky et al.
2011).
We confirmed our prediction that the DRD4 7-repeat allele was associated with abnormalities in neural indices of attentional control. Previous studies have implicated this polymorphism in attention-related problems in both typically developing children (Auerbach et al.
2001; Schmidt et al.
2002) and in clinical samples of children with ADHD (Faraone et al.
2005). The attention-related EEG activity differences for the parietal/temporal domain in this study are in line with multiple functional imaging studies that showed that attentional control is associated with the activity of a network of brain regions, including the parietal cortex (Buschman and Miller
2007), which has specific involvement in sustained (Coull et al.
1996) and orienting attention (Yantis et al.
2002).
Moreover, we found that the relationship between DRD4 and attentional control (IC-go-P3) varied depending on COMT genotype. This pattern of an epistatic interaction between DRD4 and COMT for attentional control is somewhat consistent with the proposed inverted U-shaped curve for the optimal range of dopamine availability for task performance, according to which too little or too much dopamine is disruptive and impairs functioning of the system (Bruggemann et al.
2013; Congdon et al.
2009; Durstewitz and Seamans
2008; Takahashi et al.
2011). Carriers of the DRD4 7-repeat allele with either of the homozygous COMT polymorphisms presumably had too high levels of dopamine availability for optimal attentional control and exhibited an attenuated pattern of activation to Go stimuli in the CPT-AX. Whereas individuals without the 7-repeat and either of the homozygous COMT polymorphisms may have had optimal levels of dopamine availability. In contrast, there was no influence of the DRD4 7-repeat allele on individuals with the heterozygous Val/Met allele. Dopamine is proposed to modulate the response of neural networks by suppressing spontaneous background firing and thus increasing signal-to-noise ratio and enhancing the task-specific response (Winterer et al.
2006). Our study suggests that both high and low dopaminergic states (Met/Met and Val/Val, respectively) may be compensated by increased D4 function and thus enhancing neural tuning in attentional networks.
A limitation of our study is the candidate gene approach. The selected polymorphisms will not cover all genetic variation within the examined neural circuits. We chose alleles with known functional effects and/or previously reported effects on relevant phenotypes instead of examining a larger set of SNPs tagging the major haplotypes within dopaminergic genes. Therefore, we covered only a small portion of the total dopamine signaling pathway. Already there are many other genes known to be involved in this signaling pathway (Beaulieu and Gainetdinov
2011) and a candidate gene approach will miss most of these signals. Replication is required and the relationship between source localisation measures in EEG and advanced genetic measures, including polygenic scores should be explored.
Despite evidence that psychiatric disorders have genetic etiology (Kendler), there has been a lack of identification of genetic risk factors that reliably associate with psychopathology ({Casey et al.
2013). Psychiatric disorders are likely to involve multiple brain systems and patients may differ in the extent to which processing in these systems is affected so that the biological roots within a clinical diagnosis may vary substantially (Miller
2010). Targeted analysis using functional neuroimaging measures within genetic designs could aid the identification of neural circuits affected in DSM diagnoses to more precisely diagnose and treat psychopathology (Insel and Cuthbert
2015).
No relationship between cognitive performance measures and the genetic markers was found. It may be that the infrequency of required response and inhibition in the CPT-AX (respectively, 10% of stimuli) does not provide enough data points to accurately describe these processes for use in genetic investigations as genetic effects may have relatively small effect sizes compared to, for example, group differences. Genetic effects may be more apparent in information-rich time series data at the neural level than detection via infrequent button presses at the behavioural level. This is reflected in the small effect sizes for all of the cognitive performance measures. Research focusing on overt behavioral correlates of genetic effects on attention and inhibition thus typically uses tasks that require more frequent responding, which may provide greater behavioral resolution of the cognitive process.
Although polymorphisms of COMT and DRD4 had been linked to EEG phenotypes (Loo et al.
2010; Solís-Ortiz et al.
2015), these potential links had not yet been investigated within tasks examining the neurophysiological indices of frontostriatal and parietal attentional networks. Our study aimed to investigate the relationship between COMT and DRD4 polymorphisms and ICA source-resolved brain activity during a cognitive control task, and our findings suggest higher genetic penetrance for IC-derived EEG measures of cognitive control than traditional channel based measures. The associations between these dopamine system gene function and ERP measures to be stronger at the source than at the scalp level, particularly for the COMT x DRD4 interaction; however our results remain correlational. They do not exclude mediating mechanisms, and should not be construed to suggest one-to-one relation between these gene functions and ERP source component measures.
The improved functional and anatomic separation of the cortical signal sources of EEG data produced by ICA decomposition may give measures that may be more informative for genetic studies of brain function. This may be because scalp data channels are each source signal admixtures, so that the effective signal-to-noise ratio of the ICA-separated cortical source activities is under favorable circumstances much higher than for the scalp channel signals (45). This result is consistent with the understanding that for biophysical reasons each scalp channel recording sums activities generated in many places in cortex, whereas many independent component processes separated from the data by ICA decomposition are compatible with an origin in just one cortical area or patch. Thus, neuroimaging methods, including source-based EEG measures, may be more powerful for unravelling gene-brain behavioural relationships than scalp channel-based EEG measures (McLoughlin et al.
2013; Weinberger et al.
2001).