INTERROGATING CEREBELLAR CONTRIBUTIONS TO COMPUTATIONS IN THE BASAL GANGLIA DURING PROBABILISTIC DECISION MAKING

The cerebellum and basal ganglia (BG) are key subcortical brain regions for the execution of learned behaviours. Traditionally, the cerebellum is thought to use error-based learning for motor refinement, while the BG is thought to use reward-based learning to guide decision-making. However, recent studies suggest these regions are in direct communication with one another and may both use reward to learn. We aim to dissect the cerebellum’s contributions to reward-related computations in the BG by defining the shared dynamics in cerebellar modules and BG during the execution of value-guided behaviour. We have developed a head-fixed, forelimb reaching-based two-armed bandit task in which mice must reach towards one of two available reward ports, which deliver rewards with high and low probability in alternating blocks. To maximize reward, mice have to infer which port has a higher probability of reward payout. We used two-photon imaging of Purkinje cells (PC) to measure cerebellar activity in Lobule simplex (LS), which is known to encode forelimb movements, during task execution. We found that reaches towards the two ports were associated with distinct activity patterns in cerebellar modules. We then probed the functional connectivity between the cerebellum and basal ganglia using a combination of holographic two-photon photostimulation of PCs in LS and Neuropixel probe recordings in the dorsomedial striatum (DMS). These functional mapping experiments revealed that stimulating PCs can evoke diverging activity in neurons in the DMS. Going forward, we are investigating the relationship between LS and the DMS using dual-site recordings in behaving animals.