A Generalized Linear Model of a Navigation Network.

Ehud Vinepinsky,Ronen Segev,Shay Perchik

doi:10.3389/fncir.2020.00056

Ehud Vinepinsky, Ronen Segev + Show 1 more

Open Access

https://doi.org/10.3389/fncir.2020.00056

Copy DOI

Abstract

Navigation by mammals is believed to rely on a network of neurons in the hippocampal formation, which includes the hippocampus, the medial entorhinal cortex (MEC), and additional nearby regions. Neurons in these regions represent spatial information by tuning to the position, orientation, and speed of the animal in the form of head direction cells, speed cells, grid cells, border cells, and unclassified spatially modulated cells. While the properties of single cells are well studied, little is known about the functional structure of the network in the MEC. Here, we use a generalized linear model to study the network of spatially modulated cells in the MEC. We found connectivity patterns between all spatially encoding cells and not only grid cells. In addition, the neurons’ past activity contributed to the overall activity patterns. Finally, position-modulated cells and head direction cells differed in the dependence of the activity on the history. Our results indicate that MEC neurons form a local interacting network to support spatial information representations and suggest an explanation for their complex temporal properties.

Highlights

The way mammals navigate is considered to rely on networks of neurons in the hippocampal formation, which includes the hippocampus, the dentate gyrus, the subiculum, and the medial entorhinal cortex (MEC)
The generalized linear model (GLM) approach is based on the well-known linear nonlinear Poisson (LNP) model where the input of each cell is described by a set of linear filters (Figure 1)
The network within the MEC contains grid cells, which encode navigational information in the form of firing fields that are organized in a regular periodic structure, and additional cells that encode space with non-periodic firing fields

Summary

Introduction

The way mammals navigate is considered to rely on networks of neurons in the hippocampal formation, which includes the hippocampus, the dentate gyrus, the subiculum, and the medial entorhinal cortex (MEC). In other words, correlating the cellular response with spatial variables (e.g., position, head direction, or speed) can indicate how the cells’ firing rates are related to these variables This approach has led to the discovery of many types of neurons that encode navigational variables and are located in the MEC, including grid cells (Hafting et al, 2005), border cells (Solstad et al, 2008), head direction cells (Taube et al, 1990), speed cells (Kropff et al, 2015), spatial-modulated cells (Fyhn, 2004), and cells with a conjunctive representation of these spatial variables (Sargolini, 2006; Hardcastle et al, 2017). In addition to pure cell types, a study using an unbiased statistical approach recently reported a high degree of mixed selectivity to navigational variables and heterogeneity in the responses of MEC neurons (Hardcastle et al, 2017)

Methods

Results

Conclusion