Ultraslow entorhinal oscillations shape spatial memory through grid cell drifting

This study proposes a computational model demonstrating that ultraslow entorhinal oscillations shape spatial memory by inducing grid cell drifting, which generates new grid-place cell associations to enable flexible access to different spatial memories during navigation.

The Gist

The Big Idea: The Brain's "Drifting Compass"

Imagine your brain has a built-in GPS system that helps you know where you are, even when you can't see anything (like walking in the dark). This system relies on two main types of "cells":

1. Grid Cells: Think of these as the grid lines on a map. They fire in a perfect, repeating hexagonal pattern to measure distance and direction.
2. Place Cells: Think of these as landmarks (like "the big oak tree" or "the coffee shop"). They tell you, "You are here."

For years, scientists knew these cells existed, but they noticed something weird in the medial entorhinal cortex (the brain's navigation hub): even when mice were just sitting still in the dark, these grid cells were humming with a very slow, rhythmic pulse. It was so slow it happened less than once every minute.

The Mystery: What is this slow pulse doing? Is it helping the mouse navigate? Or is it just background noise?

The Discovery: This paper uses a computer simulation to show that this slow pulse isn't helping the mouse navigate right now. Instead, it acts like a slow-motion shuffler that changes the map entirely, allowing the brain to create new memories of the same place.

---

The Analogy: The Shifting Mural

Imagine the brain's navigation system is a giant, magical mural painted on a wall.

• The Grid Cells are the grid lines drawn on the mural.
• The Place Cells are the paintings of specific spots (like a house or a tree) drawn on top of those grid lines.

1. The "Drift" (The Ultraslow Oscillation)

In the experiment, the researchers introduced a tiny, invisible force that slowly pushed the entire grid of the mural. It wasn't a fast shake; it was a slow, creeping drift, like a glacier moving.

• What happened? As the grid lines slowly drifted, the paintings of the houses and trees (the place cells) had to move with them.
• The Result: After the drift stopped, the "House" painting wasn't in the same spot on the grid anymore. It had shifted.

2. The Cost: Getting Lost (Path Integration Error)

When the mouse was actually moving and trying to navigate (Path Integration), this slow drift was a problem.

• Analogy: Imagine trying to walk in a straight line while standing on a moving walkway that is slowly sliding sideways. You think you are walking straight, but you are actually drifting off course.
• The Finding: The paper shows that when these slow oscillations are active, the mouse's internal GPS gets less accurate. It accumulates errors faster because the "map" is constantly sliding underneath its feet.

3. The Benefit: New Memories (The "Shuffle")

Here is the twist: While the drift made navigation worse in the short term, it created something amazing in the long term.

• The Analogy: Imagine you have a photo album of your house. Usually, the photos are in the same order. But if you slowly rotate the entire album, the photos end up in new positions relative to the pages.
• The Finding: Because the grid lines shifted, the brain created new associations. The same physical location in the room was now encoded by a different combination of neurons.
• Why is this good? It means the brain can store multiple versions of the same memory. It's like having a "Day Version" of your house and a "Night Version" of your house, even though the house hasn't changed. This allows the brain to be flexible, accessing different "maps" of the same space depending on the context.

Key Takeaways in Plain English

• The "Hum" isn't for navigation: That slow, minute-long rhythm in the brain doesn't help the animal find its way right now. In fact, it messes up the calculation of where the animal is.
• It breaks the map to rebuild it: The oscillation causes the brain's internal map to "drift." This drift is permanent; even after the oscillation stops, the map stays shifted.
• One room, many maps: Because the map shifts, the brain can create a brand new memory of the same room. This is like having a "reset button" for your mental map without actually leaving the room.
• Flexibility over Accuracy: The brain seems to sacrifice short-term accuracy (getting lost a little bit) to gain long-term flexibility (the ability to store and access different versions of the same memory).

The Bottom Line

Think of these ultraslow oscillations as the brain's way of shuffling the deck. It's not trying to deal a perfect hand for the current game (navigation); it's shuffling the cards so that when the game changes, the brain has a fresh, new set of memories ready to play with. It turns a static map into a dynamic, ever-changing library of spatial experiences.

Technical Summary

1. Problem Statement

Grid cells in the medial entorhinal cortex (MEC) are known to support path integration (PI), allowing animals to estimate their position by integrating self-motion cues. Recent experimental evidence (Gonzalo-Cogno et al.) revealed that grid cells in head-fixed mice exhibit ultraslow oscillations (USO) (<0.01 Hz) during movement in darkness. However, the functional role of these oscillations remains unclear:

• There is no established link between USO frequency and the time scale of path integration.
• It is unknown whether these oscillations aid navigation, degrade it, or serve a different function entirely (e.g., memory formation).
• Existing models struggle to explain how intrinsic network dynamics can generate such slow, persistent oscillations without external sensory input.

The authors hypothesize that USO does not support path integration directly but instead plays a critical role in linking spatiotemporal memories by inducing drift in grid cell dynamics, which subsequently reshapes hippocampal place cell representations.

2. Methodology

The authors developed a computational model combining a Continuous Attractor Network (CAN) for grid cells and a Hebbian learning network for place cells.

• Grid Cell Model (CAN):

  • Consisted of 5 grid modules, each with 400 neurons arranged on a toroidal manifold.
  • Modules had increasing spatial scales (mimicking dorsal-ventral MEC organization).
  • Path Integration: Implemented via an external velocity signal that moves an activity "bump" across the torus.
  • Ultraslow Oscillation (USO) Implementation: The authors introduced a directed force to the activity bump by applying small, structured perturbations to the synaptic weights. This created a gradient that induced a slow, directional drift of the activity bump, independent of the animal's movement.
  • Parameters: Drift strength ($\alpha$) and direction ($\theta$) were tuned to match experimental frequencies (~0.0066 Hz).

• Place Cell Model:

  • A population of place cells received input solely from the grid cells.
  • Synaptic weights between grid and place cells were trained using Hebbian plasticity with k-Winner-Take-All (k-WTA) competition to ensure sparse, stable firing fields.
  • Place cells acted as indicators of spatial memory and error correction.

• Experimental Conditions:

  • Simulations ran for 90 minutes across three stages: (1) Immobility/No USO, (2) Movement with/without USO, (3) Immobility/No USO.
  • Four movement scenarios were tested: 1D fixed (running wheel), 1D free, 2D constrained, and 2D unconstrained.
  • Analysis Metrics: Power Spectral Density (PSD), Oscillation Score, Euclidean position estimation error, Spatial Information, Spatial Entropy, Effective Spatial Coverage (ESC), and Population Vector Correlation (PVC) for memory recall.

3. Key Contributions

1. Mechanistic Origin of USO: Demonstrated that ultraslow oscillations can emerge naturally in a CAN model solely through synaptic weight perturbations, without requiring external oscillatory inputs or sensory cues.
2. Decoupling PI and USO: Showed that while USO and path integration coexist, they are functionally distinct. USO is effectively masked by path integration dynamics during movement but persists as a hidden force.
3. Drift-Induced Memory Remapping: Proposed that USO causes a persistent drift in grid and place fields. This drift is not merely an error but a mechanism that generates new spatial representations (memories) of the same environment.
4. Dynamic Memory Access: Provided evidence that the timing and phase of the USO determine which internal spatial representation is active, suggesting a mechanism for flexibly accessing different spatial memories.

4. Key Results

• Generation of Oscillations:

  • Tuning synaptic weights successfully generated population-wide synchronized oscillations at ~0.0066 Hz, matching experimental data.
  • The oscillation pattern was confirmed via PCA and autocorrelation, showing that neurons within a module fire in a cyclical, non-traveling wave pattern.

• Impact on Path Integration:

  • Suppression: During active movement, the power of the USO signal decreased significantly or vanished in spectrograms, indicating it is suppressed by the dominant path integration dynamics.
  • Error Accumulation: Despite being hidden, USO degraded position estimation. The drift induced by USO added to the intrinsic error of path integration, causing exponential growth in positional error.
  • Directionality: The estimation error was minimized when the animal's trajectory aligned with the USO drift direction, confirming the drift acts as a constant bias.

• Lasting Changes in Spatial Coding:

  • Grid Cell Drift: After the USO period ended, grid cell activity did not return to its original state. Instead, the activity bump settled into a new phase, resulting in a shifted grid pattern.
  • Place Cell Remapping: Consequently, place fields shifted to new locations. This resembled global remapping observed experimentally when animals enter new environments, even though the physical environment remained unchanged.
  • Spatial Metrics: During USO, spatial information decreased, and spatial entropy/coverage increased (grid fields became less specific and covered more of the arena). Post-USO, the new place fields maintained high spatial information but at different locations.

• Memory Recall and Dynamics:

  • New Memories: The shifted grid-place associations represented new spatial memories encoded in the same neural population.
  • Reproducibility: If the same USO parameters were used, the system consistently returned to the same new representation, regardless of the specific trajectory taken during the oscillation.
  • Phase Dependence: The specific memory accessed depended on the duration and phase of the oscillation. Different oscillation durations led to distinct internal representations (low PVC correlation between different durations).
  • Reversibility: Attempting to reverse the drift (using complementary angles) did not perfectly restore the original memory, suggesting the system settles into a new stable attractor state rather than a reversible cycle.

5. Significance and Discussion

• Reinterpretation of Grid Cell Function: The paper challenges the view that grid cells are solely for precise navigation. Instead, it suggests USO acts as a mechanism for cognitive map renewal. By inducing controlled drift, the brain can explore different internal representations of the same space, potentially to find a more suitable map when sensory cues are ambiguous (e.g., in darkness).
• Memory Flexibility: The findings suggest that the brain can dynamically switch between different "versions" of a spatial map. This provides a theoretical basis for how animals might update their cognitive maps or access episodic memories associated with specific spatial contexts without needing new sensory input.
• Theoretical Framework: The work bridges the gap between the "temporal-spatial scaffold" hypothesis and recent experimental findings, proposing that ultraslow dynamics are crucial for the adaptability of the cognitive map rather than the precision of path integration.

In conclusion, the study posits that ultraslow entorhinal oscillations are not a bug in the navigation system, but a feature for memory plasticity, allowing the brain to generate and access multiple spatial memories through controlled grid cell drifting.