Long‐term dynamics of the entorhinal grid code

The hippocampus and entorhinal cortex form an “information loop”: The superficial layers of the entorhinal cortex provide the majority of excitatory input to the dorsal hippocampus, whereas the deep layers of the entorhinal cortex receive most of their input from the hippocampus. The medial entorhinal cortex (MEC) is crucial for the formation of spatial memory, and its inactivation impairs both recent and remote memory retrieval during spatial navigation tasks- indicating a key role for entorhinal neural codes in long term memory. Recent studies have found that hippocampal place codes gradually change over days and weeks when the animal repeatedly visits a fixed familiar environment. However, the long-term dynamics of entorhinal spatial codes have remained unexplored. Here, we developed a novel imaging preparation that enables simultaneous imaging from > 1,000 MEC cells in freely behaving mice. Unsupervised learning extracted hundreds of grid cells from up to four different modules within a single field of view. By tracking the same neurons over weeks, we longitudinally analyzed the long-term stability of grid cells, and compared it to that of hippocampal CA1 place cells. We found that over days entorhinal grid cells gradually changed their tuning while maintaining similar activity rates. Conversely, hippocampal place cells displayed gradual changes in cell activity rates, but their tuning changed to a lesser extent. We found similar differences between MEC grid cells and CA1 place cells when examining drift on shorter timescales (within a session). Overall, we found a double dissociation between MEC grid cells and CA1 place cells with respect to properties of neural code stability, suggesting that different mechanisms may govern representational drift in these circuits. Moreover, our results suggest that stable grid fields are not a prerequisite for stable place fields.