BRAIN-WIDE NEURAL DYNAMICS UNDERLYING GOAL-DIRECTED SINGLE-SESSION LEARNING

Animals flexibly learn new associations by leveraging prior knowledge and using rewards as a teaching signal, a process affected by changes in behavioural strategies and internal states. Understanding how distributed neural circuits reorganize during learning requires measuring population dynamics continuously at the single-trial level, yet most studies compare static snapshots in naive versus expert animals or sample intermittently across days, missing how specific neuronal populations reorganize as learning unfolds. Here, we leveraged brain-wide Neuropixels recordings from 288 insertions across 80 mice yielding ∼100,000 neurons (∼40,000 well-isolated) spanning 60+ brain areas to investigate how neural activity reorganizes as mice rapidly acquire a novel sensorimotor mapping in a single session. Mice were first pretrained to lick for a reward following an auditory cue. After reaching expert performance, mice were then confronted with a novel whisker-reward contingency, with separate cohorts experiencing either Go or No-Go whisker stimuli, while maintaining auditory performance in interleaved trials. Both cohorts rapidly acquired the new sensorimotor mapping within tens of trials. Learning trajectories were highly individualized, with fast transitions between low- and high-performance, and correlated with ongoing movements and pupil area. Single-unit analyses revealed brain-wide encoding of performance states and licking, and a sparser encoding of sensory and choice variables. The activity of a small but distributed subset of neurons reflected rapid changes in performance and behavioural history. This brain-wide dataset recording from multiple brain areas simultaneously provides a foundation for dissecting the coordinated activity of distributed circuits during rapid learning.