It is widely accepted that the basal ganglia (BG) play a key role in action selection and reinforcement learning. However, despite considerable number of studies, the BG architecture and function are not completely understood. Action selection and reinforcement learning are facilitated by the activity of dopaminergic neurons, which encode reward prediction errors when reward outcomes are higher or lower than expected. The BG are thought to select proper motor responses by gating appropriate actions, and suppressing inappropriate ones. The direct striato-nigral (GO) and the indirect striato-pallidal (NOGO) pathways have been suggested to provide the functions of BG in the two-pathway concept. Previous models confirmed the idea that these two pathways can mediate the behavioral choice, but only for a relatively small number of potential behaviors. Recent studies have provided new evidence of BG involvement in motor adaptation tasks, in which adaptation occurs in a non-error-based manner. In such tasks, there is a continuum of possible actions, each represented by a complex neuronal activity pattern. We extended the classical concept of the two-pathway BG by creating a model of BG interacting with a movement execution system, which allows for an arbitrary number of possible actions. The model includes sensory and premotor cortices, BG, a spinal cord network, and a virtual mechanical arm performing 2D reaching movements. The arm is composed of 2 joints (shoulder and elbow) controlled by 6 muscles (4 mono-articular and 2 bi-articular). The spinal cord network contains motoneurons, controlling the muscles, and sensory interneurons that receive afferent feedback and mediate basic reflexes. Given a specific goal-oriented motor task, the BG network through reinforcement learning constructs a behavior from an arbitrary number of basic actions represented by cortical activity patterns. Our study confirms that, with slight modifications, the classical two-pathway BG concept is consistent with results of previous studies, including non-error based motor adaptation experiments, pharmacological manipulations with BG nuclei, and functional deficits observed in BG-related motor disorders.

No abstract

In this study, we explore the functional role of striatal cholinergic interneurons, hereinafter referred to as tonically active neurons (TANs), via computational modeling; specifically, we investigate the mechanistic relationship between TAN activity and dopamine variations and how changes in this relationship affect reinforcement learning in the striatum. TANs pause their tonic firing activity after excitatory stimuli from thalamic and cortical neurons in response to a sensory event or reward information. During the pause striatal dopamine concentration excursions are observed. However, functional interactions between the TAN pause and striatal dopamine release are poorly understood. Here we propose a TAN activity-dopamine relationship model and demonstrate that the TAN pause is likely a time window to gate phasic dopamine release and dopamine variations reciprocally modulate the TAN pause duration. Furthermore, this model is integrated into our previously published model of reward-based motor adaptation to demonstrate how phasic dopamine release is gated by the TAN pause to deliver reward information for reinforcement learning in a timely manner. We also show how TAN-dopamine interactions are affected by striatal dopamine deficiency to produce poor performance of motor adaptation.

Motor adaptation to perturbations is provided by learning mechanisms operating in the cerebellum and basal ganglia. The cerebellum normally performs motor adaptation through supervised learning using information about movement error provided by visual feedback. However, if visual feedback is critically distorted, the system may disengage cerebellar error-based learning and switch to reinforcement learning mechanisms mediated by basal ganglia. Yet, the exact conditions and mechanisms of cerebellum and basal ganglia involvement in motor adaptation remain unknown. We use mathematical modeling to simulate control of planar reaching movements that relies on both error-based and non-error-based learning mechanisms. We show that for learning to be efficient only one of these mechanisms should be active at a time. We suggest that switching between the mechanisms is provided by a special circuit that effectively suppresses the learning process in one structure and enables it in the other. To do so, this circuit modulates learning rate in the cerebellum and dopamine release in basal ganglia depending on error-based learning efficiency. We use the model to explain and interpret experimental data on error- and non-error-based motor adaptation under different conditions.