Decision strategy adaptation is supported by shifts in dopaminergic RPE-mediated contingency representation

Aims: Decision making relies on cognitive maps of available options and expected outcomes. The dopaminergic (DA) system can encode such expectations, and discrepancies between expected and actual outcomes, the so-called Reward Prediction Error (RPE). Here we addressed the link between decision strategies, contingency representation and RPE.Methods: In a circular arena with three locations associated with rewards, mice are conditioned to perform sequences of binary choices. To promote shifts in decision strategy and the associated neural representation of actions, we used exactly the same setting but changed the reward delivery rule. First, a Deterministic rule rewards every trial (p=100%). Second, in a Probabilistic rule, each target has a specific reward delivery probability (p=100%, 50% or 25%). In a third, Complex rule, mice are rewarded for non-repetitive choice sequences. Analysis of choice behaviour was coupled with Grab-DA recording in the NAc to assess DA release.Results: First, NAc DA release shows trial-by-trial signaling of outcomes and errors through reward-induced peaks and omission-induced dips of fluorescence, whose amplitude is modulated by expectations, consistently with the RPE framework. Second, as mice use specific decision strategies in each task, we tested three possible configurations of contingency representations to explain choice behaviour from DA signal. Analysis of DA amplitude modulation by RPE suggests a trajectory-based action selection in Deterministic, a target-based action selection in Probabilistic, and a random-like strategy based on one-order representation in Complexity.Conclusion: Decision strategy adaptation in each rule can be explained by shifts in contingency representation observed in the DA-RPE signatures.