Visual cue-related activity of cells in the medial entorhinal cortex during navigation in virtual reality

Amina A. Kinkhabwala, Yi Gu, Dmitriy Aronov, David W. TankeLife 2020 · DOI

TL;DR

Researchers have discovered a new class of neurons in the medial entorhinal cortex (MEC) called "cue cells" that are specifically tuned to visual landmarks. These cue cells fire in sequences around these landmarks, providing a potential mechanism for correcting errors that accumulate during path integration, a fundamental process for spatial navigation.

What Problem Does This Paper Address?

Animals navigate by combining self-motion cues (path integration) with external environmental cues (landmarks). While path integration is crucial, it's prone to accumulating errors over time. How these errors are corrected, especially in complex, visually rich environments, remains a significant question. The medial entorhinal cortex (MEC) is known to be involved in path integration, with grid cells forming a spatial metric. However, the MEC's role in using external cues for error correction, particularly beyond simple boundaries, is less understood. This paper investigates whether there are specific cell types within the MEC that encode visual landmarks and could contribute to this error correction process.

Methods

The researchers used a combination of electrophysiology (tetrode recordings) and two-photon calcium imaging in mice navigating virtual reality (VR) environments.

  • Virtual Reality Setup: Mice were trained to navigate linear tracks in VR, allowing for precise control over visual cues and the environment.
  • Electrophysiology: Tetrodes were used to record the activity of individual neurons in the MEC while mice ran on virtual tracks.
  • Two-Photon Imaging: This technique allowed for cellular-resolution imaging of neural activity in specific layers of the MEC, particularly in response to side-specific visual cues.
  • Data Analysis:
    • Cue Score: A novel metric was developed to quantify how well a cell's firing pattern correlated with the presence and location of visual cues.
  • Spatial Firing Fields: Identified locations where cells showed increased firing activity.
  • Perturbation Experiments: Cues were removed from the VR environment to test the dependence of cue cell activity on these landmarks.
  • Real Arena Recordings: To assess the stability and properties of these cells in a more naturalistic setting, recordings were also made while mice foraged in a physical arena.
  • Cell Classification: Cells were classified based on their responses to cues, as well as compared to known MEC cell types like grid, border, and head direction cells.

Datasets

The study utilized data collected from mice navigating custom-designed virtual reality environments. No new benchmark datasets were introduced; rather, the researchers created controlled experimental paradigms within VR to elicit specific neural responses.

Key Insights and Ideas

  • Discovery of "Cue Cells": The most significant finding is the identification of a new class of neurons in the MEC that are specifically responsive to visual landmarks. This expands our understanding of the MEC's functional repertoire beyond just grid cells.
  • Landmark-Specific Firing: These cue cells don't just fire anywhere; they have distinct "firing fields" precisely located around visual cues.
  • Sequential Activity: When multiple cues are present, the population of cue cells fires in a sequential pattern, creating a "chain" of activity that mirrors the sequence of landmarks. This suggests a dynamic representation of the environment.
  • Cue Dependence: The activity of these cells is directly tied to the presence of visual cues. When cues were removed in VR, the corresponding cue cell activity disappeared, demonstrating their reliance on external visual input.
  • Side Preference: Interestingly, many cue cells showed a preference for cues on one side of the track (left or right), suggesting a potential for more refined spatial coding.
  • Cross-Environment Stability: Cue cells maintained their cue-related firing patterns across different VR environments, indicating they represent a stable functional property rather than being environment-specific.
  • Role in Error Correction: The authors propose that these cue cells could be crucial for correcting errors in path integration by providing reliable landmark information to the MEC's internal navigation circuits.

Key Results

  • Approximately 13% of recorded MEC cells were identified as "cue cells" based on their strong correlation with visual cues.
  • When cues were removed in VR, cue cell firing fields disappeared in those locations, and the population activity sequence was disrupted.
  • In a VR track with side-specific cues, a significant proportion of cue cells showed a preference for either left or right cues.
  • In a real arena, cue cells exhibited spatially stable activity and accounted for a substantial portion (32%) of previously unclassified, spatially stable MEC cells.
  • Cue cells showed distinct properties from grid, border, and head direction cells, though some overlap (conjunctive cells) was observed.

Limitations and Future Work

  • Virtual vs. Real World: While cue cells showed stable activity in VR, their representation in complex real-world environments is still being explored. The paper notes that forming a representation in real arenas might require triangulation from multiple cues at various angles, which is more complex than the linear VR tracks.
  • Circuit Mechanisms: The precise circuit mechanisms by which visual inputs drive cue cells and how their activity is integrated with path integration remain open questions.
  • Head Direction Tuning: The significant head direction tuning observed in some cue cells warrants further investigation into its functional role and circuit basis.
  • Future Directions: The authors suggest further experiments to probe other features of cue cells with more perturbations of the virtual environment. They also emphasize the need for further characterization of their precise coding of unique landmarks and the development of new models to determine how effectively they might be used for path integration and error correction. Investigating the role of mixed selectivity in cue cells and their integration with other sensory inputs is also highlighted as a future avenue.

Summary updated

Abstract

During spatial navigation, animals use self-motion to estimate positions through path integration. However, estimation errors accumulate over time and it is unclear how they are corrected. Here we report a new cell class ('cue cell') encoding visual cues that could be used to correct errors in path integration in mouse medial entorhinal cortex (MEC). During virtual navigation, individual cue cells exhibited firing fields only near visual cues and their population response formed sequences repeated at each cue. These cells consistently responded to cues across multiple environments. On a track with cues on left and right sides, most cue cells only responded to cues on one side. During navigation in a real arena, they showed spatially stable activity and accounted for 32% of unidentified, spatially stable MEC cells. These cue cell properties demonstrate that the MEC contains a code representing spatial landmarks, which could be important for error correction during path integration.
BibTeX
@inproceedings{kinkhabwala2020visual,
  author = {Amina A. Kinkhabwala and Yi Gu and Dmitriy Aronov and David W. Tank},
  title = {{Visual cue-related activity of cells in the medial entorhinal cortex during navigation in virtual reality}},
  booktitle = {eLife},
  year = {2020},
  doi = {10.7554/elife.43140},
}
Figure