CNS*2025 Florence

Coordinated Cortical Dynamics During Sleep Stabilize Low-Dimensional Motor Representations

Jaekyung Kim^*1,2,3, Linmeng He^2,3, Karunesh Ganguly<sup>2,3

1</sup> Department of Biological Sciences, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea
² Neurology and Rehabilitation Service, San Francisco Veterans Affairs Medical Center, San Francisco, California, USA
³ Department of Neurology, University of California-San Francisco, San Francisco, California, USA

^*Email: jaekyung.kim@kaist.ac.kr



Introduction
Sleep is thought to play a critical role in stabilizing neural population dynamics that underlie long-term memory [1]. However, the specific cortical interactions during sleep that support the consolidation of low-dimensional neural manifolds remain poorly understood. Here, we investigated how slow oscillation (SO) coupling between the prefrontal cortex (PFC) and primary motor cortex (M1) during NREM sleep contributes to motor memory consolidation and population-level neural representation stability.
Methods
Rats were trained on a reach-to-grasp motor task over 20 days, during which motor behavior and neural activity were recorded daily. To assess the causal contribution of sleep-dependent PFC-M1 coordination, we applied closed-loop optogenetic disruption targeting SO coupling during NREM sleep. Neural population activity in M1 during task performance was analyzed using Gaussian Process Factor Analysis (GPFA) to extract latent trajectories and assess manifold dynamics [2].
Results
In control animals, GPFA revealed a transition from early-phase manifold exploration to late-phase consolidation, reflected by increasing trajectory fidelity and reduced variability over time. In contrast, animals with disrupted PFC-M1 coupling exhibited persistent instability in neural trajectories and impaired consolidation of low-dimensional representations. Furthermore, the degree of neural trajectory stabilization during the learning phase correlated strongly with SO coupling strength during preceding sleep, indicating a mechanistic link between cortical synchronization during sleep and waking neural stability.
Discussion
These findings provide causal evidence that NREM sleep, via PFC-M1 slow oscillation coupling, supports the computational structuring of motor-related neural manifolds. Our results bridge behavioral consolidation with population-level neural analysis, highlighting a key role for sleep-dependent cortical communication in shaping long-term memory representations through manifold refinement.







Research was supported by awards from the Korea Basic Science Institute (NREC) (RS-2024-00401876 to J.K.); the National Research Foundation of Korea (NRF) grant (RS-2024-00345236 to J.K.); the NIH, NINDS (5K02NS093014 to K.G.); the NIH RF1 (NS132913 to K.G.).

[1] Buzsáki, G. (1989). Two-stage model of memory trace formation: a role for "noisy" brain states. Neuroscience, 31(3), 551-570. https://doi.org/10.1016/0306-4522(89)90423-5
[2] Kim, J., Joshi, A., Frank, L., & Ganguly, K. (2023). Cortical-hippocampal coupling during manifold exploration in motor cortex. Nature, 613(7942), 103-110. https://www.nature.com/articles/s41586-022-05533-z