Striatal Acetylcholine Dip-Rebound Is Induced by Direct Pathway Neurons and Encodes Action-Outcome Contingency

Learning contingencies between actions and outcomes is pivotal in adaptive behavior and requires ongoing flexible sensory, motor, and reward information integration. The striatum is central to the integration. However, it remains unclear how the striatal neurons interact with each other to facilitate learning. Here we show that in the dorsomedial striatum, direct-pathway medium spiny neurons (dMSNs), but not dopamine (DA), inhibit cholinergic interneurons, inducing a characteristic acetylcholine (ACh) dip-rebound, encoding action-outcome (A-O) contingency. Using genetically encoded sensors and in vivo fiber photometry, we find that dMSN activation, ACh dip-rebound, and DA transients emerge only after mice acquire the contingency. dMSN activity and ACh dynamics persist even when rewards are probabilistic, adapt in time as the learned relationship between action and reward evolves, and vanish when the contingency is degraded. Ex vivo recordings and in vivo optogenetics further show that dMSN activity is both sufficient and required to generate the ACh dip-rebound through GABAergic inhibition of cholinergic interneurons. Disrupting the dip-rebound slows acquisition and accelerates extinction. Together, these findings reveal a previously unrecognized dMSN-ACh circuit mechanism that encodes contingency during instrumental learning, operating alongside DA signals that track reward outcomes.