SPATIAL PATTERN OF DOPAMINE RELEASE IN THE FRONTAL CORTEX ENCODES REWARD PREDICTION

Dopamine released in the striatum plays a critical role in guiding reinforcement learning. However, the roles that dopamine signals in the frontal cortex might play during learning remain unclear. It is also unknown whether frontal dopamine release is spatially uniform, consistent with volume transmission, or organized into local spatial domains. To address these questions, we used two-photon imaging of a genetically encoded dopamine sensor (GRAB-DA) in the dorsal frontal cortex of head-fixed mice as they progressed from naive to expert in an auditory classical conditioning task. We partitioned each field-of-view into an 8x8 grid and quantified dopamine signals of sectors to reward-predictive and non-predictive cues throughout learning. In expert animals, cue-evoked dopamine transients formed reproducible spatial patterns that distinguished the cues. Cross-validated logistic regression decoded trial-by-trial predicted reward more accurately from dopamine release spatial patterns than from the field-mean signal. This decoder also outperformed spatially-uniform synthetic control data, preserving the field-mean signal while removing spatial structure, indicating that cue information is embedded in the spatial pattern of dopamine release. During learning, spatial variability and the local autocorrelations of dopamine release sectors increased together with behavioral discrimination. To control for potential artifacts, we imaged a spectrally-matched static fluorophore in a separate cohort of mice during the same task, demonstrating that these signals did not reproduce the cue-locked patterns of dopamine release. Our results indicate that reproducible spatial patterns of dopamine release in the frontal cortex emerge with learning and carry reward-predictive information.