Event-related brain potentials (ERPs) reveal neural substrates of cognitive flexibility

This user research summary is based on an article published as Rustamov, N., Rodriguez-Raecke, R., Timm, L., Agrawal, D., Dressler, D., Schrader, C., … Kopp, B. (in press). “Attention shifting in Parkinson’s disease: An analysis of behavioral and cortical responses”. Neuropsychology.

Cognitive inflexibility is a major symptom of dysexecutive syndromes that follow many types of brain lesions. For example, patients in early stages of Parkinson’s disease (PD) exhibit substantial inflexibility in their behavioural and cortical responding (Rustamov et al., 2013). Early-stage PD patients are primarily characterized by the degeneration of the nigro-striatal dopamine (DA) system (Robbins & Cools, 2014). Thus, our clinical study was designed as an empirical test of the hypothesis that a phasic release of nigro-striatal DA is necessary for cognitive flexibility to occur (Hikosaka & Isoda, 2010). This hypothesis conjectures that the nigro-striatal DA system serves as the eye of the needle for cognitive flexibility because it transiently enables thalamo-cortical gating (Fig. 1).

Figure 1: Adapted from Frank et al. (2004). A cartoon of cortico-striato-thalamo-cortical loops, including direct and indirect pathways of the basal ganglia. Left panel: Tonic nigrostriatal DA release supports cognitive stability via balanced excitation of the direct “Go” pathway and inhibition of the indirect “NoGo” pathway. Right panel: Additional phasic nigrostriatal DA release is associated with stronger excitation of the direct “Go” pathway and stronger inhibition of the indirect “NoGo” pathway. The net effect of phasic nigrostriatal DA release lies in thalamo-cortical disinhibition, thereby transiently enabling cognitive flexibility via thalamo-cortical gating.
SNc: Substantia nigra, pars compacta. SNr: Substantia nigra, pars reticulata. GPe: External globus pallidus. GPi: Internal globus pallidus. D1, D2: D1-, D2-DA receptor respectively.

The present investigation examined the N2 wave of the event-related brain potential (ERP) as a neural substrate of cognitive flexibility (Rustamov et al., 2013). Notice that there exist links between variations in fronto-centrally distributed negative ERP waveform deflections and signals conveyed by midbrain DA systems (Holroyd & Coles, 2002; Ullsperger, Danielmeier, & Jocham, 2014).

We hypothesized that thalamo-cortical gating underlies the capability to covertly shift spatial attention back and forth. We expected to observe a shift-related augmentation of N2 amplitudes as neural substrate of demands to shift attention through space. If cognitive inflexibility in early-stage PD patients (Rustamov et al., 2013) is attributable to suppressed or attenuated phasic nigrostriatal DA responses (Monchi et al., 2007), no phasic thalamo-cortical gating should occur in these patients. Thus, we hypothesized that the shift-related N2 amplitude augmentation would not occur in early-stage PD patients.

Participants

Forty individuals (20 patients with PD, 20 healthy controls (HCs) matched for age, sex, and education) participated in the experiment. All patients received a clinical diagnosis of idiopathic PD by experienced clinical neurologists. HCs had no known neurological or psychiatric disorders and had not been prescribed any psychiatric or neurological medications.

Design

To put the thalamo-cortical gating hypothesis to test in patients with PD, we created a novel task that basically combined three elements: (1) it consisted of a classical congruency task, (2) it corresponded to a shifting task, and (3) it invited the recording of event-related brain potentials (ERPs). On the flanker task, a (central) target stimulus is flanked, in immediate vicinity, either by response-congruent or by response-incongruent distractor stimuli. The flanker task is associated with strong and reliable congruency effects (Kopp et al., 1996). The flanker task was extended to a shifting task by requesting subjects to adjust the spatial spotlight of attention in one of the following ways: Figure 2a shows that the attention spotlight can be formed either like a “Mexican hat” (i.e., an “on-center, off-surround” attention field) or like a doughnut (i.e., an “off-center, on-surround” attention field), depending on whether central or peripheral information is attended or ignored. If one asks subjects to shift back and forth between these two attentional sets, there exist two potential shifting operations: (1) “zooming in” to a central, “Mexican hat”-like attention field, (2) “zooming out” to a peripheral, doughnut-like attention field.

Figure 2b shows how potential shifts between these two types of attentional sets can be used to create a shifting paradigm in which either central or peripheral stimuli of a flanker task must be attended to across runs of trials of unpredictable length. Termination of those trial runs was signaled by “incorrect response”-feedback stimuli, which indicated on incongruent trials that the wrong sensory information had been gated toward the response systems, and consequently that an attention shift was necessary on the upcoming trial.

We recorded event-related potentials (ERPs) and focused our analysis on the fronto-centrally distributed N2 wave which is usually evoked by stimuli that occur on incongruent trials of the flanker task (Kopp et al., 1996; Rustamov et al., 2013). The N2 has its peak around 250 to 300 ms after the onset of the eliciting stimulus (Folstein & Van Petten, 2008). Inspection of the ERP waveforms revealed that the parietally distributed P3 might in addition to the frontally distributed N2 be of relevance for an analysis of cortical substrates of cognitive flexibility.

EEG Recording and Data Analyses

Continuous electroencephalogram (EEG) was recorded by means of a 32-channel BrainAmp amplifier (Brain Products, Gilching, Germany), using active electrodes (Brain Products, Gilching, Germany) which were mounted on an actiCAP (Brain Products, Gilching, Germany) in an International 10–20 System montage. ERPs were prepared for statistical analysis according to all standards in the field. N2 amplitude was measured as mean amplitudes in a -40 to +40 ms time interval around individual peak N2 latencies, which were defined as the maximum negative amplitude in a 200 to 350 ms post-stimulus interval at electrode Fz. Likewise, P3 amplitude was measured as mean amplitudes in a -40 to +40 ms time interval around individual peak P3 latencies, which were defined as the maximum positive amplitude in a 400 to 600 ms post-stimulus interval at electrode Pz. Response times (RTs), error rates, N2 and P3 amplitudes were analysed via analyses of variance (ANOVAs) with the between-subjects factor Group and the within-subjects factor Stage within task runs (with the levels 1 = shift trial, 2 = trials 2 and 3, 3 = trials 4 to 6, and 4 = trials 7 and 8).

Patients with PD responded more slowly than HCs. Importantly, the ANOVA revealed that the interaction Group by Stage was significant, F (3, 114) = 3.84, p < .05, indicating an overall difference between groups with regard to RTs as a function of Stage. Specifically, a Group by Stage Helmert contrast compared RTs on shift trials with those on all remaining trials in HCs and in patients with PD, F (1, 38) = 9.78, p < .05, indicating that RT shift costs differed significantly between groups (HCs > patients with PD). With regard to error rates, the interaction Group by Stage proved significant, F (3, 114) = 13.73, p < .05, indicating that patients with PD were more error prone across the early stages of task runs. The first Group by Stage Helmert contrast (shift trials vs. all other trials) revealed that shift costs on error rates did not differ significantly between groups, F (1, 38) = 4.02, p = .052. However, the second, stage (2) versus the remaining trials, F (1, 38) = 27.93, p < .05, and third, stage (3) versus stage (4), F (1, 38) = 17.11, p < .05, comparisons showed that error rates gradually declined across stages (2) to (4) in patients with PD, whereas HCs achieved a stable level of erroneous responses as early as stage (2). Thus, patients with PD, but not HCs, committed a multitude of perserveration errors throughout multiple stages of task runs.

Figure 3 shows grand-average stimulus-locked ERP activity at midline electrodes for incongruent trials, separately for groups across the four stages within task runs. Inspection of these waveforms revealed an enhanced negativity (N2) on shift trials (labeled N2_s), which achieved its maximum around 300 ms post-stimulus at electrode Fz in HCs. However, a similar shift-related N2_s could not be detected in patients with PD. The ANOVA on frontally distributed N2 amplitudes revealed that the interaction Group by Stage proved significant, F (3, 114) = 7.70, p < .05. A Group by Stage Helmert contrast, comparing shift trials with all remaining trials in HCs and patients with PD, revealed that N2_s amplitudes were associated with group, F (1, 38) = 9.89, p < .05. Thus, an N2 amplification (N2_s) was triggered on shift trials in HCs but not in patients with PD, whereas HCs and patients with PD did not differ with regard to N2 amplitudes on non-shift trials. Further, The ANOVA on parietally distributed P3 amplitudes revealed that the Group by Stage interaction proved significant, F (3, 114) = 8.79, p < .05. A Group by Stage Helmert contrast, comparing shift trials with all remaining trials in HCs and patients with PD, revealed that P3_s amplitudes were associated with group, F (1, 38) = 6.85, p < .05. Thus, a P3 amplification (P3_s) was triggered on shift trials in HCs but not in patients with PD, whereas HCs and patients with PD did not differ with regard to P3 amplitudes on non-shift trials.

Figure 3: Grand-average stimulus-locked event-related potential (ERP) activity at midline electrodes in controls (HC; left panels) and in patients (PD; right panels), as a function of stage across task runs. The N2 and P3 components of the ERP waveform are of particular importance. Note that HCs show enhanced N2 and P3 amplitudes on shift trials; these enhancements are labeled N2s and P3s, respectively. No comparable amplitude modulations were observed in patients with PD. The scalp maps show N2 (260–340 ms; medio-frontal maximum) and P3 (460–540 ms; parietal maximum) scalp topographies.

HCs, but not patients in early stages of PD, showed signs of shift costs in our novel attention shifting task. Specifically, RTs on shift trials were prolonged and N2 as well as P3 amplitudes were enhanced in HCs, but all these shift costs were completely absent in patients with PD. Given that patients with PD committed more perseveration errors than HCs, the lack of behavioural and cortical shift costs should be interpreted as signs of cognitive inflexibility in these patients. These data show that PD is associated with a loss in the capability to change attentional sets in accordance with contextual requirements. Thus, PD should no longer be conceptualized as a movement disorder, but rather as a disorder that disrupts the capability to change motor as well as non-motor sets in accordance with contextual requirements. Our ERP-based approach opens a new window onto an understanding of cognitive flexibility that seems to be associated with thalamo-cortical gating which, in turn, might be enabled via phasic release of nigro-striatal DA. Direct tests of the nigro-striatal DA hypothesis of cognitive flexibility will be possible through an analysis of local field potentials that can be obtained from patients with PD or dystonia who are treated with deep brain stimulation and whose stimulation electrodes are located in the GPi.