Information flow between subthalamic nucleus and posterior fronto-medial cortex during performance monitoring

Introduction:

Learning new motor tasks requires the ability to monitor ongoing behavior, detect errors and modify the performance accordingly. Several brain areas are involved in these processes. According to the reinforcement learning theory, when action outcomes are worse than expected, midbrain dopaminergic neurons convey signals to the posterior fronto-medial cortex (pFMC). During an erroneous response, an error detection signal can be observed in the ongoing electroencephalogram (EEG), the so called error-related negativity (ERN), probably used by the pFMC for behavioral adjustments. Recently, some neurons of the subthalamic nucleus (STN) have been shown to respond only during error trials. The pFMC is connected to the STN via the indirect pathway and the hyperdirect pathway. However, the direction of information flow between cortex and basal ganglia during error processing is not known so far. The aim of our study is to investigate the direction of information flow between STN and pFMC in error detection.

Methods:

We recorded simultaneously local field potentials (LFPs) from the STN and scalp EEG from electrodes FCz and Cz in 14 patients undergoing deep brain stimulation to treat severe Parkinson's disease (PD). Patients performed a flanker paradigm that was used to elicit laterality mistakes in motor responses in about 15–20% of trials. First, we performed a standard event-related potential (ERP) analysis of LFP and EEG signals. ERPs were derived by averaging response-locked epochs during incongruent correct and incongruent error responses. Averaged data were baseline-corrected (from 100 to 300ms prior to button press). Second, we used partial directed coherence (PDC) to assess the direction of information flow of oscillatory activity between STN and pFMC during error detection.

Results:

As previously shown, a significant ERN occurred in the pFMC-EEG with a peak latency of 69ms post-response (p=0.045, univariate permutation test). In the STN-LFPs, stimulus-locked activity of erroneous and correct responses was significantly different from -50 to 100ms around the button press (p=0.0096) and had a maximal difference amplitude at 30ms. Directionality analysis showed a significant unidirectional flow with ipsi- und contralateral STN activity leading over pFMC during erroneous responses. However, during correct responses a bidirectional information flow between STN and cortical activity was found. In both conditions the information flow was mainly mediated by high gamma oscillations (60–80Hz).

Conclusion:

In this study, we could demonstrate that error-related activity is elicited in the STN (STN-ERN) of PD patients. The STN-ERN was elicited in a similar time window like the scalp ERN and could therefore also represent a neural correlate of error detection processes. By means of the PDC, we found a unidirectional influence of the STN error-related LFP activity on cortical activity suggesting that during error detection STN activity precedes or drives pFMC activity.