Striatal D2-neurons gate D1-mediated associative encoding under uncertainty
Goal-directed actions are established through the encoding of predictive relationships between behaviour and its consequences, a process involving dopamine-dependent plasticity in the dorsal striatum. Early phases of learning entail complex associative landscapes from which useful action-outcome contingencies must be identified. Although dopamine processes are central to associative learning, the cellular mechanisms through which behaviourally relevant plasticity is selected from competing signalling events remain unknown. Here we show that D2-spiny projection neurons (D2-SPNs) are transiently recruited during the earliest stages of instrumental learning, when critical action-outcome contingencies emerge. We identify a dominant cell-to-cell transmodulation process by which D2-SPNs regulate dopamine-dependent signalling by constraining protein kinase A activity in D1-SPNs. This mechanism overrides extracellular dopamine signals and can be reproduced in vivo through calibrated optogenetic activation of cAMP signalling. Enhancing D2[->]D1-SPN transmodulation during uncertain reinforcement schedules durably improves action-outcome encoding, revealing a mechanism through which striatal networks promote relevant contingencies under associative ambiguity.