Predictive Coding of Reward in the Hippocampus
A fundamental objective of the brain is to anticipate future outcomes1-7. This process requires learning the states of the world as well as the transitional relationships between those states. The hippocampal cognitive map is believed to be one such internal model8. However, evidence for predictive coding9-12 and reward sensitivity13-19 in the hippocampal neuronal representation suggests that its role extends beyond purely spatial representation. In fact, it raises the question of what kind of spatial representation is most useful for learning and maximizing future rewards? Here, we track the evolution of reward representation over weeks as mice learn to solve a cognitively demanding reward-based task. Our findings reveal a highly organized restructuring of hippocampal reward representations during the learning process. Specifically, we found multiple lines of evidence, both at the population and single-cell levels, that hippocampal representation becomes predictive of reward over weeks. Namely, both population-level information about reward and the percentage of reward-tuned neurons decrease over time. At the same time, the representation of the animals choice and reward approach period (the period between choice and reward) increased over time. By tracking individual reward cells across sessions, we found that neurons initially tuned for reward shifted their tuning towards choice and reward approach periods, indicating that reward cells backpropagate their tuning to anticipate reward with experience. These findings underscore the dynamic nature of hippocampal representations, highlighting their critical role in learning through the prediction of future outcomes.