Development and reorganization of hippocampal representations during contextual spatial learning

The hippocampus supports flexible navigation by creating context-dependent spatial maps. How these maps emerge during learning and change upon manipulation of the context is not well understood. To elucidate these dynamics, we studied CA1 pyramidal cell (PC) representations in a contextual go/no-go task.We trained water-restricted, head-fixed Thy1-GCaMP6s mice to run on a treadmill in virtual environments, and distinguish two visually distinct corridors by selectively licking in a reward zone in one corridor. We monitored CA1PC activity by two-photon Ca2+ imaging and assessed behavior by measuring running speed and lick rate. To investigate the population-level representation of environmental variables, we performed Bayesian decoding on the neuronal data.In low-performance sessions, spatially tuned cells showed similar activity patterns in both corridors. As the animals learned the task, the fraction of spatially tuned and corridor-selective CA1PCs increased and decorrelated representations of the two corridors emerged, along with evolution of an accurate representation of corridor identity. Switching to a new wall pattern in the rewarded corridor or swapping the reward contingencies resulted in decreased fractions of both spatially and corridor-selective cells and decorrelation between representations before vs. after the context changes in both corridors. While population activity reorganized immediately upon switch, it was more gradual after the swap.Our results suggest that during learning, CA1PCs may initially generalize between environments but gradually reorganize their activity according to behavioral relevance of corridor identity. Upon abrupt context change, CA1PC representations rapidly reorganize even in an unchanged corridor, suggesting that context encoding represents global task structure.