Experience dependent changes in neural representations in the dentate gyrus and CA3

The hippocampus is essential for memory and spatial representation. However, how neurons encode distinct aspects of experience and how this coding changes with learning remain unclear. We recorded neural activity in the dentate gyrus (DG) and CA3 during a virtual reality task in which mice learned associations between odors, visual contexts, and reward outcomes. Using a Poisson generalized linear model, we quantified the contribution of sensory, motor, and cognitive variables to neuronal activity. In expert animals, CA3 neurons encoded more task variables than DG neurons. Moreover, the interaction between the odors and the rewarded context made a larger contribution to neural activity in CA3 than in DG. Context was decoded from both regions, more accurately in CA3 within the context and at earlier corridor positions, whereas odor and reward were decoded similarly. Contextual information emerged with learning in both regions. Reward information was evident from the first session, but odor information was decodable before learning only in DG. Learning increased the contribution of speed, position, and reward in DG, and of position and the odor–context association in CA3. Finally, LSTM-based decoding showed that position and speed could be decoded from CA3 before learning, but from DG only after learning. These results reveal that learning differentially reorganizes neural coding, with CA3 showing greater multiplexing and stronger associative coding than DG.