A systems model of alternating theta sweeps via firing rate adaptation

Zilong Ji, Tianhao Chu, Si Wu, Neil Burgess

TL;DR

This paper presents a computational model that explains how alternating left-right "theta sweeps" in the brain's navigation system are generated. By integrating head-direction and grid cell activity with firing rate adaptation and theta rhythm modulation, the model successfully reproduces experimental findings and offers testable predictions about neural dynamics during spatial navigation.

What Problem Does This Paper Address?

The brain's ability to navigate and form spatial memories relies on complex neural circuits, particularly within the hippocampus and entorhinal cortex. Specific neurons, like place cells and grid cells, fire in sequential patterns known as "theta sweeps" during movement. While these sweeps were initially understood as forward-progressing sequences along a linear path, recent research in open-field environments revealed a more complex pattern: alternating left-right sweeps. The problem this paper addresses is understanding the neural mechanisms underlying these alternating left-right theta sweeps in more complex, unconstrained environments, and how they relate to both directional and positional information.

Methods

The authors developed a computational model based on continuous attractor networks (CANs). This model comprises three interconnected layers:

  1. Theta-modulated Head-Direction (HD) cells: Modeled as a ring attractor, these cells encode an "internal direction" that can differ from the animal's actual head direction.
  2. Conjunctive Grid × Direction (Conj-grid) cells: These act as an intermediate layer, linking the internal direction from HD cells to grid cells.
  3. Pure Grid cells: Modeled as a 2D attractor network, these cells represent the animal's location.

Key mechanisms within the model include:

  • Firing Rate Adaptation: Neurons reduce their firing rate over time when presented with a constant input. This intrinsic property is crucial for generating the "sweeping" dynamics.
  • Medial-Septal Theta Modulation: The model incorporates the rhythmic theta input from the medial septum, which drives the frequency of the sweeps.
  • Attractor Dynamics: The networks maintain a stable "bump" of activity representing the current internal direction or location, which can then be moved by inputs.

The model integrates these components to show how an internal directional signal can drive a corresponding positional sweep in grid cells.

Datasets

The study utilized two existing datasets for validation and prediction testing:

  • Grid cell dataset: Collected by Gardner et al. from freely foraging rats in an open-field arena.
  • Head direction cell dataset: Collected by Lomi et al. from rats foraging in an arena, specifically focusing on anteroventral thalamic nucleus recordings.

No new datasets were introduced.

Key Insights and Ideas

  • Firing Rate Adaptation as a Sweep Generator: The most striking insight is that simple firing rate adaptation within the HD cell network, combined with theta modulation, is sufficient to generate alternating left-right sweeps of internal direction. This is a parsimonious explanation for a complex observed phenomenon.
  • Internal Direction Drives Location Sweeps: The model demonstrates a clear causal link: the sweeping internal direction signal from HD cells, relayed through conjunctive cells, directly drives the left-right sweeps of internal location in the grid cell network.
  • Theta Skipping and Phase Precession as Network Dynamics: The model unifies theta skipping (HD cells firing in alternating theta cycles) and theta phase precession (firing occurring progressively earlier in the theta cycle) as emergent properties of the same underlying network dynamics, dependent on head rotation and theta input.
  • Dorsal-Ventral Gradient in Adaptation: The model predicts that variations in firing rate adaptation strength along the dorsal-ventral axis of the medial entorhinal cortex (MEC), as observed experimentally, lead to systematic changes in sweep angle and length, potentially enhancing spatial coverage.

Key Results

  • The model successfully reproduces left-right alternating theta sweeps in both the internal direction (HD cells) and internal location (grid cells) in simulated open-field navigation.
  • It explains the observed alignment between internal direction and location sweeps.
  • It confirms the linear relationship between sweep length and grid spacing.
  • The model shows that running speed increases the alternation score and decreases the variance of sweep angle, consistent with experimental data.
  • It makes several testable predictions, including:
  • The relationship between theta phase precession during turning and theta skipping in HD cells.
  • How animals' behaviors (speed, turns) modulate sweep features.
  • The effect of medial-septal theta rhythm magnitude on theta skipping and phase precession.
  • The dorsal-ventral gradient in adaptation strength influencing sweep properties.

Limitations and Future Work

The authors acknowledge several areas for future investigation:

  • In vivo vs. in vitro adaptation: The adaptation strength was measured in vitro, and its role in vivo needs further confirmation.
  • Running speed and theta frequency: The model assumes running speed increases theta amplitude, but future work could explore how speed affects both theta amplitude and frequency.
  • Formation of attractor dynamics: The model assumes hardwired attractor networks; how these dynamics are formed during development remains an open question.
  • Parameter sensitivity: While the model is robust, further investigation into the precise values of parameters like the conj-grid offset is warranted.
  • Cholinergic component of theta: The authors note that rotational theta may have a cholinergic component, which was not explicitly modeled.

Future work could involve more detailed in vivo recordings to validate adaptation parameters and explore the impact of combined speed-dependent theta frequency and amplitude modulation.

Summary updated

Abstract

Place and grid cells provide a neural system for self-location and tend to fire in sequences within each cycle of the hippocampal theta rhythm when rodents run on a linear track. These sequences correspond to the decoded location of the animal sweeping forward from its current location ("theta sweeps"). However, recent findings in open-field environments show alternating left-right theta sweeps and propose a circuit for their generation. Here, we present a computational model of this circuit, comprising theta-modulated head-direction cells, conjunctive grid × direction cells, and pure grid cells, based on continuous attractor dynamics, firing rate adaptation, and modulation by the medial-septal theta rhythm. Due to firing rate adaptation, the head-direction ring attractor exhibits left-right sweeps coding for internal direction, providing an input to the grid cell attractor network shifted along the internal direction, via an intermediate layer of conjunctive grid × direction cells, producing left-right sweeps of position by grid cells. Our model explains the empirical findings, including the alignment of internal position and direction sweeps and the dependence of sweep length on grid spacing. It makes predictions for theta-modulated head-direction cells, including relationships between theta phase precession during turning, theta skipping, anticipatory firing, and directional tuning width, several of which we verify in experimental data from anteroventral thalamus. The model also predicts relationships between position and direction sweeps, running speed, and dorsal-ventral location within the entorhinal cortex. Overall, a simple intrinsic mechanism explains the complex theta dynamics of an internal direction signal within the hippocampal formation, with testable predictions.
BibTeX
@inproceedings{ji2025systems,
  author = {Zilong Ji and Tianhao Chu and Si Wu and Neil Burgess},
  title = {{A systems model of alternating theta sweeps via firing rate adaptation}},
  booktitle = {Current},
  year = {2025},
  doi = {10.1016/j.cub.2024.08.059},
}
Figure