Insights into the neuronal dynamics that support memory
Representational drift
In our laboratory, we have been investigating the role of the hippocampus in mediating episodic memory, particularly through the development and recall of context-specific cognitive maps. While it has been traditionally held that the hippocampus maintains stable representations of environments, our past research has observed that hippocampal cells often alter their tuning properties over time, even within the same setting—a phenomenon known as neural or representational drift. This observation has led us to further explore the impact of these long-term dynamics on the hippocampal code for context. Through extended studies, we have imaged large populations of CA1 neurons in freely behaving mice engaged in a geometric morph paradigm over periods exceeding a month. Our findings suggest that despite the long-term alterations in neural activity, the hippocampal network retains a consistent representation of context, as the changes occur orthogonally to context encoding in network space. This suggests that contextual information can be reliably extracted from hippocampal activity over extended periods. Intriguingly, at the individual cell level, we discovered varied patterns of activity, with some cells showing a positive correlation between their contextual coding and stability over time. These insights, part of our ongoing work, affirm that the hippocampal spatial code undergoes long-term changes while preserving the integrity of contextual representations, thereby contributing to our understanding of the hippocampus's role in episodic memory across time.
Path Integration
Our laboratory's recent studies have focused intensively on the phenomenon of path integration, a navigational process whereby animals and humans calculate their current position by keeping track of the distances and directions they have traveled from a starting point. This innate ability, akin to an internal GPS, allows for the accurate return to a starting location without relying on visual cues. Central to this process are grid cells in the medial entorhinal cortex (MEC), which provide a coordinate system for spatial navigation and are thought to play a critical role in path integration by maintaining an internal representation of the trajectory based on self-movement cues. We employed a specific path integration task to assess spatial navigation performance, wherein subjects must navigate to a goal location after accounting for variable starting positions and distances. Our studies utilized this task to explore how spatial navigation and memory are affected in mouse models of Alzheimer's disease (AD), especially during the early stages of AD pathology. Our research revealed that both grid cells and path integration capabilities are disrupted at these initial stages in the J20 transgenic mouse model of AD. We found that the grid cells were spatially unstable, particularly toward the center of the arena, and their spatial firing patterns were qualitatively altered, aligning parallel to environmental borders. Moreover, these cells showed a diminished ability to integrate the distance traveled, evidenced by reduced theta phase precession—a neural measure of path integration efficiency. Interestingly, while grid cells exhibited such impairments, the spatial coding of other cell types within the MEC and place cells within the hippocampus remained unaffected. This discrepancy suggests a unique vulnerability of grid cells to early AD pathology. The correlation between grid cell dysfunction and poor performance on the path integration task emphasizes the potential of grid cell integrity and path integration performance as early indicators of Alzheimer's disease, providing a window for early detection and intervention strategies in both familial and sporadic AD cases.
Geometric manipulations and remapping
Our laboratory has been utilizing experimental manipulations of environmental geometry to advance our understanding of the hippocampal representation's structure. The hippocampus, along with neocortical regions, is believed to support a variety of learning and memory processes through cognitive maps. Studies indicate that these maps encompass metric spaces, facilitating flexible navigation and memory, which become impaired with hippocampal damage. To quantitatively compare various theoretical models and empirical observations, we've adopted a systematic approach, employing a geometric deformation paradigm to collect large datasets. This method allows us to robustly quantify changes in neural representation and directly compare them with model predictions. Specifically, we've recorded from a substantial number of neurons within the CA1 subregion across different geometric configurations, using a representational-similarity framework to discern between competing theories of cognitive mapping. These experiments have not only provided a quantitative benchmark for theoretical advances but have also confirmed the strong influence of geometric remapping on population similarity within the hippocampus. Our findings offer a substantial contribution to the field, allowing for a more refined analysis of the geometric determinants of neural coding and enhancing the validity of cognitive mapping research.
Hippocampal dynamics during memory tasks
In our laboratory, the synergy of touchscreen-based tasks or contextual fear conditioning tasks with one-photon miniscope recordings to understand hippocampal neuronal encoding over extended learning periods. We monitor large neuronal populations in dorsal CA1 and CA3 as mice engage in weeks-long learning of tasks like the paired-associate learning (PAL) and the trial-unique nonmatching-to-location (TUNL) task. Our findings reveal that hippocampal neurons gradually develop context-dependent representations that become more stable as mice transition from novice to expert task performers. During the TUNL task, which measures spatial working memory, we witness hippocampal place fields becoming increasingly associated with reward locations, with distinct neuronal populations such as cue, reward prediction, and reward cells emerging and evolving. The activity of these cells aligns with reinforcement learning principles, suggesting a reorganization of hippocampal coding to facilitate memory-guided behavior. Moreover, as proficiency in these tasks increases, we observe that hippocampal neurons also become tuned not just to space but to directional trajectories and are modulated by the mouse's behavior, indicating a layered encoding of both allocentric space and behavior within the hippocampus. This extended learning framework allows us to capture the nuanced coding dynamics of memory tasks, significantly contributing to our understanding of hippocampal function in complex cognitive processes.
Grid cell disruption in a mouse model of early Alzheimer’s disease reflects reduced integration of self-motion cues
The McGill-Mouse-Miniscope platform: A standardized approach for high-throughput imaging of neuronal dynamics during behavior
Hippocampal Neural Circuits Respond to Optogenetic Pacing of Theta Frequencies by Generating Accelerated Oscillation Frequencies
The medial entorhinal cortex is necessary for temporal organization of hippocampal neuronal activity
New and Distinct Hippocampal Place Codes Are Generated in a New Environment during Septal Inactivation
Related publications
Representational drift
Path Integration
Geometric manipulations and remapping
Hippocampal dynamics during memory tasks
In our laboratory, we have been investigating the role of the hippocampus in mediating episodic memory, particularly through the development and recall of context-specific cognitive maps. While it has been traditionally held that the hippocampus maintains stable representations of environments, our past research has observed that hippocampal cells often alter their tuning properties over time, even within the same setting—a phenomenon known as neural or representational drift. This observation has led us to further explore the impact of these long-term dynamics on the hippocampal code for context. Through extended studies, we have imaged large populations of CA1 neurons in freely behaving mice engaged in a geometric morph paradigm over periods exceeding a month. Our findings suggest that despite the long-term alterations in neural activity, the hippocampal network retains a consistent representation of context, as the changes occur orthogonally to context encoding in network space. This suggests that contextual information can be reliably extracted from hippocampal activity over extended periods. Intriguingly, at the individual cell level, we discovered varied patterns of activity, with some cells showing a positive correlation between their contextual coding and stability over time. These insights, part of our ongoing work, affirm that the hippocampal spatial code undergoes long-term changes while preserving the integrity of contextual representations, thereby contributing to our understanding of the hippocampus's role in episodic memory across time.
Our laboratory's recent studies have focused intensively on the phenomenon of path integration, a navigational process whereby animals and humans calculate their current position by keeping track of the distances and directions they have traveled from a starting point. This innate ability, akin to an internal GPS, allows for the accurate return to a starting location without relying on visual cues. Central to this process are grid cells in the medial entorhinal cortex (MEC), which provide a coordinate system for spatial navigation and are thought to play a critical role in path integration by maintaining an internal representation of the trajectory based on self-movement cues. We employed a specific path integration task to assess spatial navigation performance, wherein subjects must navigate to a goal location after accounting for variable starting positions and distances. Our studies utilized this task to explore how spatial navigation and memory are affected in mouse models of Alzheimer's disease (AD), especially during the early stages of AD pathology. Our research revealed that both grid cells and path integration capabilities are disrupted at these initial stages in the J20 transgenic mouse model of AD. We found that the grid cells were spatially unstable, particularly toward the center of the arena, and their spatial firing patterns were qualitatively altered, aligning parallel to environmental borders. Moreover, these cells showed a diminished ability to integrate the distance traveled, evidenced by reduced theta phase precession—a neural measure of path integration efficiency. Interestingly, while grid cells exhibited such impairments, the spatial coding of other cell types within the MEC and place cells within the hippocampus remained unaffected. This discrepancy suggests a unique vulnerability of grid cells to early AD pathology. The correlation between grid cell dysfunction and poor performance on the path integration task emphasizes the potential of grid cell integrity and path integration performance as early indicators of Alzheimer's disease, providing a window for early detection and intervention strategies in both familial and sporadic AD cases.
Our laboratory has been utilizing experimental manipulations of environmental geometry to advance our understanding of the hippocampal representation's structure. The hippocampus, along with neocortical regions, is believed to support a variety of learning and memory processes through cognitive maps. Studies indicate that these maps encompass metric spaces, facilitating flexible navigation and memory, which become impaired with hippocampal damage. To quantitatively compare various theoretical models and empirical observations, we've adopted a systematic approach, employing a geometric deformation paradigm to collect large datasets. This method allows us to robustly quantify changes in neural representation and directly compare them with model predictions. Specifically, we've recorded from a substantial number of neurons within the CA1 subregion across different geometric configurations, using a representational-similarity framework to discern between competing theories of cognitive mapping. These experiments have not only provided a quantitative benchmark for theoretical advances but have also confirmed the strong influence of geometric remapping on population similarity within the hippocampus. Our findings offer a substantial contribution to the field, allowing for a more refined analysis of the geometric determinants of neural coding and enhancing the validity of cognitive mapping research.
In our laboratory, the synergy of touchscreen-based tasks or contextual fear conditioning tasks with one-photon miniscope recordings to understand hippocampal neuronal encoding over extended learning periods. We monitor large neuronal populations in dorsal CA1 and CA3 as mice engage in weeks-long learning of tasks like the paired-associate learning (PAL) and the trial-unique nonmatching-to-location (TUNL) task. Our findings reveal that hippocampal neurons gradually develop context-dependent representations that become more stable as mice transition from novice to expert task performers. During the TUNL task, which measures spatial working memory, we witness hippocampal place fields becoming increasingly associated with reward locations, with distinct neuronal populations such as cue, reward prediction, and reward cells emerging and evolving. The activity of these cells aligns with reinforcement learning principles, suggesting a reorganization of hippocampal coding to facilitate memory-guided behavior. Moreover, as proficiency in these tasks increases, we observe that hippocampal neurons also become tuned not just to space but to directional trajectories and are modulated by the mouse's behavior, indicating a layered encoding of both allocentric space and behavior within the hippocampus. This extended learning framework allows us to capture the nuanced coding dynamics of memory tasks, significantly contributing to our understanding of hippocampal function in complex cognitive processes.
