Grid cells show field-to-field variability and this explains the\n aperiodic response of inhibitory interneurons

Benjamin Dunn, Daniel Wennberg, Ziwei Huang, Yasser Roudi

Abstract

Research on network mechanisms and coding properties of grid cells assume that the firing rate of a grid cell in each of its fields is the same. Furthermore, proposed network models predict spatial regularities in the firing of inhibitory interneurons that are inconsistent with experimental data. In this paper, by analyzing the response of grid cells recorded from rats during free navigation, we first show that there are strong variations in the mean firing rate of the fields of individual grid cells and thus show that the data is inconsistent with the theoretical models that predict similar peak magnitudes. We then build a two population excitatory-inhibitory network model in which sparse spatially selective input to the excitatory cells, presumed to arise from e.g. salient external stimuli, hippocampus or a combination of both, leads to the variability in the firing field amplitudes of grid cells. We show that, when combined with appropriate connectivity between the excitatory and inhibitory neurons, the variability in the firing field amplitudes of grid cells results in inhibitory neurons that do not exhibit regular spatial firing, consistent with experimental data. Finally, we show that even if the spatial positions of the fields are maintained, variations in the firing rates of the fields of grid cells are enough to cause remapping of hippocampal cells.

Introduction

The medial entorhinal cortex (MEC) is home to grid cells, which exhibit strikingly regular hexagonal firing patterns as an animal navigates its environment. Traditional Continuous Attractor Network (CAN) models have long assumed that these firing "fields" are identical in magnitude. However, a significant discrepancy exists between these models and biological reality: CAN models predict that inhibitory interneurons within the same network should also display periodic, grid-like firing. In contrast, experimental recordings show that these interneurons typically exhibit aperiodic, spatially irregular activity. This study investigates whether the observed "messiness"—specifically, the variability in firing rates across different fields of a single grid cell—is the key to reconciling network models with experimental data.

Methods

The researchers performed a secondary analysis of 373 grid cells recorded from five rats navigating open-field environments, originally reported by Stensola et al. (2012). To distinguish genuine biological variability from random noise, they compared experimental spike trains against synthetic ones with identical spatial fields.

Statistical Quantification

Field-to-field variability was quantified using a coefficient of variation ($c_W$) and F-statistics. This allowed the team to identify stable, inherent variations in firing rates that could not be attributed to time-dependent effects or differences in spatial occupancy.

Distribution of the coefficient of variation ($c_W$) across the recorded population of grid cells.

Computational Modeling

The team developed two-population recurrent network models consisting of excitatory and inhibitory neurons. They utilized Non-negative Matrix Factorization (NMF) to reverse-engineer connectivity matrices. Two primary architectures were compared: a "fully factorized" model with highly structured connectivity and a "partially factorized" model characterized by sparse, random excitatory-to-inhibitory projections.

Results

Field-to-Field Variability is Widespread

Analysis revealed that approximately 35% of grid cells exhibit statistically significant amplitude variations across their fields. These variations are not merely noise but represent a stable feature of the cell's spatial representation.

Grid cell firing rate maps (a, c) and their corresponding normalized field firing rates (b, d) illustrating high and low field-to-field variability.

Sparse Connectivity Explains Aperiodic Inhibition

The "partially factorized" model successfully resolved the conflict between grid cell periodicity and interneuron aperiodicity. By implementing sparse, random connectivity from grid cells to interneurons, the periodic signals from individual grid cells effectively "wash out" at the population level. This results in the spatially irregular firing patterns observed in biological interneurons.

Implications for Hippocampal Remapping

The study further demonstrated that variations in grid field amplitudes alone—without any change in field position—are sufficient to drive "global remapping" in hippocampal place cells. This suggests that amplitude modulation provides a mechanism for encoding contextual information within a stable spatial framework.

Discussion

While the sparse model accurately reflects biological interneuron activity, it introduces a trade-off in path integration stability. The "partially factorized" model is more susceptible to "drift" over time compared to highly structured models. This suggests that the brain likely relies on external sensory cues to periodically recalibrate the internal spatial map. Future work should explore how synaptic plasticity might refine these sparse networks to balance biological realism with navigational accuracy.

Summary updated

Abstract

Research on network mechanisms and coding properties of grid cells assume\nthat the firing rate of a grid cell in each of its fields is the same.\nFurthermore, proposed network models predict spatial regularities in the firing\nof inhibitory interneurons that are inconsistent with experimental data. In\nthis paper, by analyzing the response of grid cells recorded from rats during\nfree navigation, we first show that there are strong variations in the mean\nfiring rate of the fields of individual grid cells and thus show that the data\nis inconsistent with the theoretical models that predict similar peak\nmagnitudes. We then build a two population excitatory-inhibitory network model\nin which sparse spatially selective input to the excitatory cells, presumed to\narise from e.g. salient external stimuli, hippocampus or a combination of both,\nleads to the variability in the firing field amplitudes of grid cells. We show\nthat, when combined with appropriate connectivity between the excitatory and\ninhibitory neurons, the variability in the firing field amplitudes of grid\ncells results in inhibitory neurons that do not exhibit regular spatial firing,\nconsistent with experimental data. Finally, we show that even if the spatial\npositions of the fields are maintained, variations in the firing rates of the\nfields of grid cells are enough to cause remapping of hippocampal cells.\n
BibTeX
@misc{dunn2017grid,
  author = {Benjamin Dunn and Daniel Wennberg and Ziwei Huang and Yasser Roudi},
  title = {{Grid cells show field-to-field variability and this explains the\textbackslash\{\}n aperiodic response of inhibitory interneurons}},
  year = {2017},
  eprint = {1701.04893},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1701.04893},
}
Figure