Abstract The grid cell network in the Medial Entorhinal Cortex (MEC) is a cornerstone of the mammalian navigation system. While various theories explain the formation of these hexagonal firing patterns, the role of hippocampal input remains a subject of intense debate. This study re-analyzes landmark data from Bonnevie et al. (2013), where the hippocampus was inactivated while recording MEC grid cells. We investigated whether the firing associations of grid cells depend on hippocampal inputs by examining temporal and spatial correlations between simultaneously recorded cell pairs. Our analysis reveals that network coherence persists even in the absence of hippocampal input, both in cells that maintained their identity and those that transitioned to head-direction tuning. These findings suggest that the phase relations between grid cells are primarily driven by intrinsic MEC connectivity, supporting Continuous Attractor Network (CAN) models of grid cell formation.
Introduction
Grid cells are specialized neurons in the MEC that fire at regular, hexagonal intervals, providing a metric for spatial navigation. A fundamental question in systems neuroscience is whether these patterns are generated internally within the MEC or are the result of feedforward inputs from other regions, most notably the hippocampus.
Two primary theoretical frameworks compete to explain this phenomenon:
- Continuous Attractor Network (CAN) Models: These propose that grid patterns emerge from recurrent, internal wiring within the MEC. In this view, the hippocampus acts as a "spatial anchor," but the relative firing relationships between cells are hard-wired.
- Feedforward Models: These suggest that grid cells are formed by the summation of inputs from hippocampal place cells.
Previous experiments demonstrated that silencing the hippocampus causes the hexagonal grid pattern to collapse. While feedforward proponents argued this proved the hippocampus creates the grid, CAN theorists suggested the underlying network structure might remain intact even if the spatial mapping is lost. This study seeks to resolve this by analyzing the temporal synchrony of MEC cells during hippocampal inactivation.
Methods
This research involved a rigorous re-analysis of the dataset originally published by Bonnevie et al. (2013). The experimental setup included:
- Subjects: Eight male Long-Evans rats.
- Inactivation: The researchers administered muscimol, a $GABA_A$ agonist, to temporarily silence hippocampal activity.
- Electrophysiology: 301 well-separated neurons were recorded in the MEC and parasubiculum across 40 sessions, with 2–18 cells captured simultaneously per session.
- Analytical Approach: The team calculated temporal and spatial correlations between cell pairs across three phases: baseline (pre-inactivation), muscimol (during inactivation), and recovery (post-inactivation). Statistical significance was verified using shuffling methods, where spike times were randomly shifted 1,000 times to create a null distribution.
Key Results
Persistence of Temporal Synchrony
The most striking finding was that while the spatial "map" collapsed, the "network" remained functional. The mean grid score—a measure of hexagonal regularity—dropped from a highly organized 0.92 at baseline to -0.30 during inactivation. Despite this total loss of spatial structure, the temporal correlations between cell pairs remained remarkably stable, with a correlation coefficient of $r = 0.57$ ($P < 0.001$) between the pre-inactivation and inactivation phases.
Approximately 26% of cell pairs maintained statistically significant temporal correlations during inactivation. While this is a decrease from the 57% observed at baseline, it is significantly higher than what would be expected if the cells were firing independently or if the network had completely decoupled.
The Grid-to-Head-Direction Transition
Interestingly, a subset of grid cells shifted their functional identity during hippocampal inactivation, becoming tuned to the animal's head direction. Despite this change in what information the cells were encoding, they maintained their temporal synchrony with the rest of the grid cell population. This suggests that these cells belong to a unified functional module governed by the same internal dynamics, regardless of their specific sensory or spatial tuning at a given moment.
Spatial vs. Temporal Stability
The study found that spatial correlations were less persistent than temporal ones ($r = 0.34$). However, the authors identified a short temporal window (approximately one second) where some local spatial structure was preserved. This suggests that the MEC network attempts to maintain its internal logic and relative spatial relationships even as it loses its primary anchor to the external environment.
Discussion
The persistence of temporal synchrony in the absence of hippocampal input provides strong evidence for Continuous Attractor Network models. If grid cells were merely the product of feedforward hippocampal input, their firing relationships should have randomized once that input was removed. Instead, the cells continued to fire in a coordinated fashion, implying that the MEC possesses an intrinsic manifold—a pre-wired structure that dictates how neurons interact.
These results suggest a division of labor in the brain's navigation system: the MEC provides the internal metric and network structure, while the hippocampus provides the spatial context and environmental anchoring necessary to map that structure onto the physical world.
Conclusion
By looking beneath the surface of the collapsed spatial map, this re-analysis reveals a robust, internally-synchronized engine within the Medial Entorhinal Cortex. The finding that grid cells maintain their timing and synchrony without hippocampal support shifts the focus toward intrinsic MEC connectivity as the primary driver of the grid cell network. Future research should explore how other inputs, such as those from the retrosplenial cortex or parasubiculum, interact with this internal MEC engine to maintain spatial orientation.
Summary updated