Learning-related oscillatory dynamics in the human cortico-basal ganglia network during motor sequence initiation in Parkinson’s Disease

Learning fine motor sequences is crucial to quality of life and can be altered in Parkinson’s Disease (PD). It may be partially driven by the optimization of motor preparation and initiation, but neither pre-movement neural activity nor its optimization is well understood. One theory of general motor initiation posits that network beta (β) desynchronization releases motor cortical excitability, reflected as delta (δ) phase response, which in turn facilitates sequence-specific motor cortical ensemble activity to produce voluntary movement. With motor sequence learning, increases in cortico-basal ganglia δ synchrony may facilitate enhanced recruitment of both cortical and basal ganglia neural ensembles, increasing the reliability of their firing patterns. As gamma (γ) activity can correlate with spiking population dynamics, this model predicts learning-dependent pre-movement sequence-specific γ activity, alongside β→δ→γ interactions and cortico-basal ganglia δ phase synchrony—any of which could differ in PD. We test this model in four subjects with PD, on dopaminergic medication without deep brain stimulation, by leveraging invasive cortico-basal ganglia field potential recordings during a multi-day, multi-sequence typing task. Possibly indicative of PD-related neuropathophysiology, there was no consistent evidence for a full cascade of β→δ→γ gating of general motor initiation. However, subjects who improved with practice did demonstrate 1) practice-driven increases in the discriminability of sequence-specific γ activity and 2) cortico-basal ganglia δ synchrony that appeared to reflect a framework for learning-dependent δ-β and δ-γ coupling. In subjects who did not improve, discriminability of sequence-specific γ activity rather decreased with practice, and basal ganglia was functionally disconnected from cortex, instead dominated by ongoing subcortical δ oscillations. These findings suggest a model in which phase-responsiveness of basal ganglia δ to cortex is a predictor of motor learning.