Extended Data Fig. 7. Convex corner cells are not sensitive to non-geometric changes or corner angles but respond to certain properties of discrete objects.
(a) Related to Fig. 4f: comparing the overlap of corner cells with the chance. Left: overlap of corner cells (both concave) between square and rectangle environments (red bar). Middle: overlap of corner cells (concave vs. convex) between square and convex-1 environments (red bar). Right: overlap of corner cells (concave vs. convex) between rectangle and convex-1 environments (red bar). The gray histogram illustrates the corresponding distribution of overlap expected by chance, with the black bar denotes the 95th percentile of each distribution. This distribution is generated by randomly selecting the same number of neurons, as indicated above for each environment, 1000 times in each mouse (n = 9 mice). Corner cells in the square and rectangle showed an overlap that is higher than chance (left), while the overlap between corner cells encoding concave or convex corners was minimal and below the chance level. (b) Schematic of the normal (convex-1) and the modified (convex-m1) convex environments. In convex-m1, one of the convex corners (in pink) was composed of walls of a different color and texture from the other three. Orange bars indicate the location of local visual cues. (c) Two representative corner cells encoding convex corners from two different mice. Each column is a neuron in which its activity was tracked across the two conditions indicated in (b). Raster plot (left) indicates extracted spikes (red dots) on top of the animal’s running trajectory (grey lines) and the spatial rate map (right) is color-coded for maximum (red) and minimum (blue) values. Pink circles delineate the location of the modified corner in the convex-m1 arena. (d) Corrected peak spike rates of corner cells (convex) at the location of the modified corner in the convex-1 vs. convex-m1 arenas (two-tailed Wilcoxon signed-rank test: p = 0.85; n = 10 mice). Corner cells were defined in each session. (e) Schematic of a convex environment containing 270° and 225° corners, as in (b). The second session was also rotated 90 degrees counterclockwise, but was combined with the first session for analysis. (f) Raster plots and the corresponding rate maps of two corner cells encoding convex corners from two different mice, as in (c). Each column is a neuron in which its activity was tracked across the two sessions. (g) Corrected peak spike rates of corner cells (convex) at 270° and 225° corners (two-tailed Wilcoxon signed-rank test: p = 0.73; n = 9 mice). (h) Schematic of the experiments, as in (b). Corner cells (convex) were identified in the convex-1 environment on day 1, then their activity was tested with inserted objects (a triangle and a cylinder) on day 2. (i) Raster plots and the corresponding rate maps of three corner cells encoding convex corners from three different mice. Each column is a neuron in which its activity was tracked across the two sessions. (j) Illustration showing vertex and face locations for the triangular object. (k) Differences between spike rates at the vertices and faces of the triangular object in corner (convex) and non-corner cells (two-tailed Wilcoxon signed-rank test against zero: corner cells: p = 0.0078; non-corner cells: p = 0.95; two-tailed Wilcoxon signed-rank test: corner cells vs. non-corner cells: p = 0.0078; n = 8 mice). (l) Differences between spike rates at the cylinder and the faces of the triangular object in corner (convex) and non-corner cells (two-tailed Wilcoxon signed-rank test against zero: corner cells: p = 0.016; non-corner cells: p = 0.74; two-tailed Wilcoxon signed-rank test: corner cells vs. non-corner cells: p = 0.016; n = 8 mice).