Ensemble Spatial Coding in the Retrosplenial Cortex

Abstract

Spatial coding in the retrosplenial cortex (RSC) is critical for navigation and memory, yet how exactly it represents or encodes space remains to be fully understood. Additionally, the role of the medial entorhinal cortex (MEC) in driving RSC’s spatial representation is not well characterized. Using electrophysiological recordings from rats performing an open-field foraging task, we systematically investigated spatial coding at both the single-cell and population levels. At the single-cell level, we first validated the existence of border and head-direction cells in the RSC, and also explored the possibility of spatially-tuned “place” cells coding for spatial information using Bayesian decoding. At the population level, we employed Linear Discriminant Analysis (LDA) decoding, revealing high spatial decodability. To further explore the geometry of RSC spatial representations, we analyzed neural manifolds, uncovering a low-dimensional structured representation underlying the population activity. We additionally carried out high-dimensional population-vector analyses of spiking activity, which revealed spatially segregated subspaces and orthogonal representations of unique locations. Finally, we applied Variational Autoencoders (VAEs) and Supervised VAEs (S-VAEs) to extract latent representations of spatial coding, mitigating the limitations of LDA. Our results demonstrated that RSC codes space/position at the level of neuronal ensembles. These findings provide new insights into how space is represented in the RSC and the role the MEC plays in driving this representation.