STATE-DEPENDENT MODULATION OF SENSORY PROCESSING IN MOUSE NEOCORTEX

Perception and decision-making rely on distributed cortical circuits that transform sensory inputs into choices and actions. Such cortical processing varies with the internal state of the brain: the same inputs can be handled differently when animals are engaged or disengaged in the task. However, how internal state shapes sensory processing across distributed brain regions remains unclear. To address this question, we trained mice to perform a two-alternative multisensory choice task, where mice learn to discriminate two combinations of auditory tones with whisker texture cues. During task training, we simultaneously recorded the neuronal population activity from the primary somatosensory cortex (S1) and a higher association area, the posterior parietal cortex (PPC), using two-photon imaging of the fluorescent calcium indicator GCaMP6f. We applied a generalized-linear-model-coupled Hidden Markov Model (GLM-HMM) to model the choice of the mice using task inputs and identified three latent states, including one engaged state with high behavioral performance and two disengaged states with biased choice patterns. We found that these states are associated with different stimulus representations and inter-areal communication: during disengagement, stimulus representations degraded in both S1 and PPC; in parallel, the communication between S1 and PPC weakened, indicating a disruption of sensory transmission. We are further dissecting the computational mechanisms that produce these state contingencies. These results demonstrate that internal state governs not only how stimuli are encoded locally but also how task-relevant information is routed between cortical areas, providing a mechanism for flexible cognition.