Dynamics among overlapping memory representations in the hippocampus at long timescales

The hippocampus plays a central role in episodic memory and spatial navigation. The activity of individual neurons and ensembles of cells encodes location within an environment, the spatial context, and non-spatial behavioral task demands, creating unique codes for these features. The hippocampus plays a role both during the initial encoding of memory representations, and also at extended intervals in tasks which require flexible retrieval or self-localization. While behavior and memory can be stable for long periods of time, many studies have shown that their neural basis is more dynamic than expected. In the studies presented here, I used single-photon calcium imaging in freely behaving mice to track hippocampal single-unit activity over many recording sessions to test how circuit instability interacts with ongoing behavioral demands. In the first study, I asked whether the representation of multiple task demands remained stable alongside an animal’s behavior. Previous work has indicated that hippocampal activity will change as an animal’s performance in a task improves. Additionally, drift, the inactivation and replacement of neuron membership within the active population, may affect neurons that code for different aspects of a task at different rates. I tested this hypothesis by recording hippocampal activity in an alternation task which animals performed stably for multiple weeks. I found that the population code separating each task dimension was highly stable in spite of cell turn over, but that the distribution of task dimensions encoded by single neurons changed as a function of time. This change in distribution of task dimensions encoded by single neurons was not driven by different levels of stability among the different coding populations, as indicated by previous reports, but instead was driven by changing rates with which newly active neurons encoded task dimensions. In the second study, I looked at how new learning affected a previously-encoded task representation. Mice performed two different tasks in a plus-shaped maze in the sequence A-B-A over a nine-day sequence. One group performed the entire sequence on a single maze, while another group performed the second rule on a second maze. This allowed me to test the hypothesis that new learning in a single environment would cause greater change in the hippocampal representation for that environment than can be accounted for alone by time between recordings. This hypothesis is confirmed by multiple measures of single unit activity, and in the population code. Together, these results demonstrate that the instability observed in long term patterns of neuronal activity does not impair behavior, and that it may have a role in the ongoing refinement of the organization of hippocampal memory representations.