Canonical reinforcement learning models postulate that dopamine neurons encode reward prediction error (RPE) and provide a teaching signal to striatal spiny projection neurons (SPNs) in the form of dopamine (DA) release. DA is thought to guide learning via dynamic modulation of protein kinase A (PKA) in SPNs. However, this fundamental assumption remains untested in behaving animals. Here we utilized multi-channel fiber photometry and fluorescence lifetime photometry (FLiP) to monitor the

more »

... y of DA neurons, extracellular DA levels, and net PKA activity in SPNs in the nucleus accumbens during learning. We found dynamic encoding of RPE in the activity of DA neurons, which is both necessary and sufficient to explain striatal DA levels and SPN PKA activity. The modulation of PKA in SPNs that express type-1 (D1R-SPNs) and type-2 (D2R-SPNs) DA receptors was dichotomous such that in each cell class it is selectively sensitive to increases and decreases in DA, respectively, and occur at and support different phases of learning. Thus, PKA-dependent pathways in D1R- and D2R-SPNs are asynchronously engaged by RPE-encoding DA signals to promote different aspects of reinforcement learning: the former responsible for the initial association between action and outcome and the latter responsible for refining the learned association.