Shao Qi, Wang Yihong, Xu Xuying, Wang Yaning, Pan Xiaochuan, Du Ying, Wang Rubin
Institute for Cognitive Neurodynamics, School of Mathematics, East China University of Science and Technology, 130 Meilong Road, Shanghai, 200237 China.
Center for Intelligent Computing, School of Mathematics, East China University of Science and Technology, 130 Meilong Road, Shanghai, 200237 China.
Cogn Neurodyn. 2025 Dec;19(1):99. doi: 10.1007/s11571-025-10282-6. Epub 2025 Jun 23.
Hippocampal place cells play a critical role in mammalian spatial navigation, episodic memory formation, and other relevant spatial cognitive functions. Experimental evidences suggest that when animals perform spatial navigation tasks in real or virtual environments, the number of place fields in the region adjacent to the target or reward location is significantly higher than in distal regions, a place cell representation phenomenon defined as "over-representation". The "over-representation" phenomenon shows dynamic changes in spatial representation: when the reward or target location moves, the location of maximum place field density shifts to the new reward position - a process termed "over-representation shift". Despite significant progress in understanding over-representation, current explanations predominantly focus on qualitative descriptions, lacking a comprehensive computational framework to systematically elucidate underlying neural mechanisms of over-representation. To address this question, we developed two distinct but related place cell sub-models based on the continuous attractor network framework: the Position-Integrated Model, which dynamically encodes spatial locations through place cell activity, and the Velocity-Driven Model, which incorporates speed cells to encode animal's movement speed. Both sub-models successfully achieved the path integration function observed in rodents. Building upon these foundational models, we implemented a reward-location-dependent dynamic gain mechanism to simulate goal-directed navigation in one-dimensional (1D) linear tracks and two-dimensional (2D) square environments. This mechanism dynamically modulates neural activity gains according to the Euclidean distance between reward locations and the animal's position. Our simulations revealed that place cells exhibit over-representation within 5-10 cm of reward zones, and the spatial distribution of place fields dynamically tracking reward location changes. This framework successfully reproduces over-representation and the dynamic shift of over-representation in place cells, revealing how reward locations shape spatial representations and trigger place field reorganization. These findings enhance our comprehension of hippocampal mechanisms in reward-based spatial navigation and establish a computational basis for studying experience-dependent neural remapping.
海马体位置细胞在哺乳动物的空间导航、情景记忆形成及其他相关空间认知功能中发挥着关键作用。实验证据表明，当动物在真实或虚拟环境中执行空间导航任务时，目标或奖励位置附近区域的位置野数量显著高于远端区域，这种位置细胞表征现象被定义为“过度表征”。“过度表征”现象在空间表征中呈现动态变化：当奖励或目标位置移动时，最大位置野密度的位置会转移到新的奖励位置——这一过程被称为“过度表征转移”。尽管在理解过度表征方面取得了显著进展，但目前的解释主要集中在定性描述上，缺乏一个全面的计算框架来系统地阐明过度表征的潜在神经机制。为了解决这个问题，我们基于连续吸引子网络框架开发了两个不同但相关的位置细胞子模型：位置整合模型，它通过位置细胞活动动态编码空间位置；速度驱动模型，它结合速度细胞来编码动物的运动速度。这两个子模型都成功实现了在啮齿动物中观察到的路径整合功能。基于这些基础模型，我们实现了一种依赖奖励位置的动态增益机制，以模拟在一维（1D）线性轨道和二维（2D）方形环境中的目标导向导航。该机制根据奖励位置与动物位置之间的欧几里得距离动态调节神经活动增益。我们的模拟结果显示，位置细胞在奖励区域5 - 10厘米范围内表现出过度表征，并且位置野的空间分布动态跟踪奖励位置的变化。这个框架成功地再现了位置细胞中的过度表征及其动态转移，揭示了奖励位置如何塑造空间表征并触发位置野重组。这些发现增强了我们对基于奖励的空间导航中海马体机制的理解，并为研究依赖经验的神经重映射建立了计算基础。
