COMPLEMENTARY ENCODING OF SPACE AND LEARNING STAGE IN HIPPOCAMPUS AND MEDIAL PREFRONTAL CORTEX

The hippocampus (HPC) is known to form robust spatial representations during navigation that are updated through learning. Although the medial prefrontal cortex (mPFC) also exhibits abstract spatial coding in spatial tasks, less is known about the refinement of these representations during learning per se. To address this, we simultaneously recorded neural activity from the HPC using 24 independently movable tetrodes and from the mPFC using a 384-channel Neuropixels probe. Rats were trained on a spatial memory task using an eight-arm radial maze, in which they retrieved rewards from three fixed arms per trial. Each daily session consisted of eight trials followed by two hours of sleep. Animals typically required at least three days to reach near-optimal performance. After learning, we continued recording for an additional week while animals performed the same task.

To examine how spatial coding evolved across trials, we applied uniform manifold approximation and projection (UMAP) to ensemble neural activity. In the mPFC, low-dimensional representations showed clear trial-by-trial progression early in learning, which progressively diminished as animals repeatedly performed the learned task. The UMAP representations also reflected linearized positions along maze arms, but only weakly distinguished arm identity. In contrast, HPC activity strongly differentiated spatial location and arm identity and showed little trial-by-trial progression. Together, these results reveal a functional dissociation, with the HPC represenations detailed spatial structures and the mPFC tracking learning-related task progression along with abstract spatial representation.