Fig. 8: Grid cell orientation predicts homing behavior during navigation in darkness.
a–d, Examples of three grid cells with lever-anchored firing fields that predicted homing direction. a, Trials were divided based on whether the mouse headed left (orange) or right (green) relative to the median homing direction of the session. Lever-centered homing paths of the animal in three sessions are shown. The red circle indicates the center of the lever. b, Firing rate of three grid cells as a function of the direction of the mouse relative to the lever when the mouse was near the lever. The data are plotted separately for trials in which the animal homed left or right relative to the median homing direction. The MVL and peak rate are indicated. c, Trial matrix showing the firing rate of the three grid cells as a function of the direction of the mouse relative to the lever. Each row represents one dark trial. The rows were sorted based on the homing direction of the mouse (from right to left). Red ticks indicate the peak firing direction for each trial. A peak firing rate shifted in a positive direction means that the field is shifted anticlockwise compared to the median. d, Trial drift of the neuron against the homing direction of the mouse. Points represent individual dark trials. A regression line (red), the circular correlation coefficient (r) and the slope of the regression line (s) are shown. e–g, Examples of correlation between the decoded directional error and homing direction. e, Lever-centered homing paths of the animal in three sessions. Paths are color coded by the mouse’s median heading direction while homing (homing heading). The red circle indicates the center of the lever. f, Trial matrix of the decoded directional error on single trials during the search (left), when the mouse was at the lever (middle) or during homing (right). The trials were sorted by the homing heading of the mouse, as indicated by the color code on the left of each matrix. A white circle indicates the circular mean in the decoded directional error distribution (decoded trial drift). When the decoded directional error is shifted in a positive direction, this means that the decoded path deviates from the real path in an anticlockwise direction. g, Correlation of decoded trial drift and homing heading error. Data points represent individual dark trials. The decoded trial drift was calculated when the mouse was at the lever (middle) or during homing (right). A regression line (orange), the circular correlation coefficient (r) and the slope (s) are shown. h, Distribution of circular correlation coefficients between the decoded directional error and homing heading. Significant and nonsignificant values are represented in blue and orange, respectively; r values were considered significant when exceeding the 97.5th percentile. Distributions are shown during the search path (top, N = 24 sessions, two-sided one-sample Wilcoxon signed-rank test, H0: median equals 0, statistic = 19.0, P = 3.659 × 10−5), when the mouse was at the lever (middle, statistic = 0.0, P = 1.192 × 10−7) and during the homing path (bottom, statistic = 20.0, P = 4.422 × 10−5). ****P < 0.0001, ***P < 0.001, ** 0.001 < P < 0.01.
