CELL TYPE-SPECIFIC REPRESENTATION REWARD PREDICTION ERROR SIGNALS IN THE PRIMARY VISUAL CORTEX

Learning-related signals such as reward prediction errors are widely represented in the sensory cortex. However, how these signals are structured across defined neuronal populations and coordinated during changing task demands is not fully understood. Here, we investigated how major neuronal subtypes in the mouse primary visual cortex, vasoactive-intestinal polypeptide (VIP) and somatostatin-expressing (SOM) interneurons as well as pyramidal cells represent visual and reward-related signals during a visual discrimination task. Water-restricted mice were trained in a Pavlovian conditioning paradigm in which visual stimuli predicted reward delivery, with rewarded and unrewarded trials interleaved to dissociate sensory and reinforcement components within single trials. To monitor neuronal activity, we performed large-volume, three-dimensional calcium imaging using a custom-developed acousto-optical two-photon microscope. This approach enabled fast population- and dendrite-level recordings across hundreds of neurons within extended cortical volumes during task performance. Neuronal responses were analyzed during stimulus presentation and reward-related epochs across different stages of learning and following rule reversal. We characterized response dynamics, amplitude distributions, and spatial and temporal correlations within and between neuronal populations. These analyses allowed us to assess how sensory and reinforcement-related signals are represented and reorganized across cell types and cortical layers. In particular, we examined how response patterns evolve with learning and adapt to changes in task contingencies. Together, this study combines behavioral training with large-scale 3D functional imaging to investigate how visual and reward-related information is distributed across defined neuronal populations in the primary visual cortex.