FUNCTIONAL INDEPENDENCE OF ENTORHINAL GRID CELL MODULES DURING HIPPOCAMPAL REMAPPING

Hippocampal place cells reorganize their collective activity patterns (‘remap’) in response to environmental changes, producing orthogonal representations across environments. The high dimensionality of the place code enables the network to store many patterns with minimal overlap, a fundamental requirement for episodic memory. This orthogonality is not present upstream in the medial entorhinal cortex (MEC), where the activity of grid and head direction cells is constrained to low-dimensional manifolds. It is still unclear how spatial maps are converted from a fixed set of states in MEC to a vast number of uncorrelated representations in the hippocampus. Here, we used Neuropixels probes to simultaneously record from multiple grid modules and hippocampal place cells. We provide the first experimental evidence that differential changes in the spatial phase of grid modules coincide with complete orthogonalization (‘global remapping’) of place cells across familiar environments. In contrast, partial remapping occurred when at least one pair of grid modules remained coherent. The extent of disparity among module phase changes was the strongest predictor of remapping strength. Grid modules rotated coherently under all conditions. A simple network model linking idealized grid modules to downstream place reproduced our experimental results. Finally, a preliminary analysis of functional connectivity between regions identifies convergent input from multiple grid modules to single place cells, pending validation through additional control analyses. Together, these results point to a potential organizing principle of cortical computation: the construction of high-capacity, adaptive representations from the combination of a limited set of structured, low-dimensional representations.