Figure 6. Firing of RSC border cells provides local boundary information and is correlated with the animal’s future motion.

(A) Spike trajectory plots and spatial rate maps of typical border cells recorded in RSC and MEC. (B) Proportion of border cells in RSC and MEC that had significant directional tuning to allocentric head-direction or egocentric boundary-direction. (C) A decoder using a support vector machine (SVM) classifier estimated the animal's distance to the wall based on population spiking activity. Local distance information was present in both regions but extended further into the center of the arena only in MEC. (D) Self-motion maps were computed based on short trajectories of the animal, giving lateral and frontal displacements (Δx and Δy, respectively) and distance traveled, Dist, in 100 ms time bins relative to the animal’s forward head-direction, giving a self-centered moving direction, θ_m, at each timepoint. (E) Example motion map of an RSC border cell with spike times shifted in time relative to the animal’s motion data. A firing field emerged on left turns when spikes were shifted −200 to −500 ms before motion. (F) RSC border cells fired prospective to motion, where the amount of information present in motion maps is maximal when spike timings were shifted −50 to −300 ms earlier. (G) MEC border cells by contrast did not show any prospective or retrospective activity. (H) Spike-triggered average of changes in direction, calculated as the difference of moving directions in 250 ms bins, where positive values indicate right turns. RSC spikes preceded turning behavior of the animal by ~200 ms, with border cells in opposing hemispheres firing prospectively to ipsilateral turns. (I) MEC spikes by contrast were not locked to any change in the animal’s behavior. *p<0.05, t-test.

Figure 6—figure supplement 1. Additional population analysis on RSC and MEC border cells.

(A) Population vector correlation of firings rates, binned according to the wall distance for border cells in RSC (left) and MEC (right). (B) Population vector correlations decay from the diagonal to distal bins at a similar rate for MEC and RSC border cells in the small wall-distance range of 0–20 cm. In the larger distance range, decay is stronger for MEC, which suggests more heterogeneity in firing across the population, allowing for discrimination of the wall distance to a large extent. (C) Coefficient of variation (CV) between average firing rates alongside each wall for RSC and MEC border cells. (D) Border cells in MEC exhibited higher peak firing rates compared to RSC. (E) Distribution of peak distance tuning for MEC border cells. (F) Simulated spiking data using real behavioral position data. Spikes were generated based on a non-uniform Poisson distribution and selected from time points where the animal was both located between 5 and 20 cm distance of a boundary (randomly selected for each cell), and had a specific orientation toward the wall (width = 0.5*π, shifted by 90° for each neighboring wall to maintain consistent wall orientations). Shown are examples of trajectory spike plots and their associated spatial rate maps of two simulated cells. (G) Spike-triggered average of changes in moving direction using simulated spiking data. Data included all behavioral sessions used for Figure 6H, and an identical number of artificial cells as those recorded in each session were generated. Simulated cells did not show prospective activity before a change in direction, as the original peak at +200 ms disappeared. **p<0.01, Wilcoxon ranksum test.