An opponent striatal circuit for distributional reinforcement learning

We trained thirsty mice in a classical conditioning task in which random odors were paired with different probability distributions over water amounts. Behavior was monitored using an IR lick sensor, a face camera, and a body camera, and non-licking behavioral variables were extracted using Facemap. We recorded from neurons in the anterior striatum using acute Neuropixels probes and registered each cell to the Allen Common Coordinate Framework. In a subset of mice, we injected 6-hydroxydopamine unilaterally into the lateral nucleus accumbens shell (lAcbSh) before training to lesion dopamine axons projecting to this region, and then recorded after training. Separately, we recorded specifically from lAcbSh neurons expressing either the D1 or D2 dopamine receptor using two-photon calcium imaging in transgenic mice. Finally, we expressed excitatory (CoChR) or inhibitory (GtACR1) opsins in these cell types and measured changes in licking induced by optogenetic stimulation.