Striatal dopamine encodes movement and value at distinct time points

Midbrain dopamine (DA) neurons are thought to be critical for reinforcement learning as well as motor control. One important target for DA neurons is the striatum where different subregions are innervated by different midbrain DA neurons. A wealth of evidence suggests that DA release in the ventral striatum acts as a reward prediction error (RPE) to support cue-outcome associations while DA manipulations in the dorsolateral striatum produce gross effects on movement. The dorsomedial striatum (DMS), located between the ventral and dorsolateral striatum, may represent an intermediate position along the continuum of value to action coding in the striatum and its associated DA innervation. To address how value and action are represented by DA in DMS, we trained rats on a novel value-based decision-making task that includes reward-and motor-related components at distinct points in time that facilitates relating DA to different aspects of behavior. Fiber photometry measurement of DA release in DMS reveals phasic DA responses at multiple task events, only some of which are accompanied by movement. The amplitude of these movement DA signals predicts the vigor of the upcoming contralateral movement. In the absence of movement, phasic DA release signals RPE by conveying reward magnitude and probability. Bilateral muscimol or DREADDs inhibition of DMS impaired rats’ ability to adapt their choices depending on the current reward context. To address which aspect of value processing is impaired, we simulated behavioral data with a Bayesian-inference-based model with softmax policy. We find that the qualitative features of inactivation data are captured by a high softmax parameter, suggesting that DMS inactivation renders policy more explorative. These data suggest that the heterogeneous DA signal in DMS supports value-based decision-making by promoting movement and learning at distinct timepoints for action policies.