Circuit-specific dopamine dynamics for distinct operations of decision computation in the primate striatum

Dopaminergic projections to the striatum are central to the reinforcement learning theory of decision-making, in which they are proposed to convey reward prediction errors. Extending this theory, we demonstrate structurally and functionally segregated dopamine dynamics that correspond to distinct cognitive operations during value-based decision-making. Using a genetically encoded dopamine sensor in macaque monkeys performing a decision-making task, we found that dopamine signals in the putamen primarily encoded reward prediction errors. In contrast, dopamine signals in the caudate exhibited temporally structured dynamics, sequentially representing option valuation, action selection, and counterfactual information about unchosen alternatives. These evolving signals paralleled the internal computational stages of decision-making. Furthermore, using multicolor retrograde viral vector tracers, we identified differential prefrontal inputs to each striatal recording site. Our findings uncover a previously unappreciated spatiotemporal organization of dopamine signaling, in which region-specific dynamics track the internal stages of decision computation and mediate anatomically distinct cortico-striatal circuit operations.