Hippocampal Place Cells Encode Local Surface-Texture Boundaries

Abstract

The cognitive map is often assumed to be a Euclidean map that isometrically represents the real world (i.e., the Euclidean distance between any two locations in the physical world should be preserved on the cognitive map). However, accumulating evidence suggests that environmental boundaries can distort the mental representations of physical space. For example, the distance between two locations can be remembered as longer than the true physical distance if the locations are separated by a boundary. While this overestimation is observed under different experimental conditions, even when the boundary is formed by flat surface cues, its physiological basis is not well understood. We examined the neural representation of flat surface cue boundaries, and of the space segregated by these boundaries, by recording place cell activity from CA1 and CA3 while rats foraged on a circular track or square platforms with inhomogeneous surface textures. About 40% of the place field edges concentrated near the boundaries on the circular track (significantly above the chance level 33%). Similarly, place field edges were more prevalent near boundaries on the platforms than expected by chance. In both one- and two-dimensional environments, the population vectors of place cell activity changed more abruptly with distance between locations that crossed cue boundaries than between locations within a bounded region. These results show that the locations of surface boundaries were evident as enhanced decorrelations of the neural representations of locations to either side of the boundaries. This enhancement might underlie the cognitive phenomenon of overestimation of distances across boundaries.

Keywords: CA1; CA3; boundaries; hippocampus; place fields; single units; spatial cognition; spatial segmentation.

Figure 1.. Local-Cue Boundaries Modulated the Locations of Place Field Edges

(A) (i) Top-down schematics of the double-rotation experiment sessions. Local textures on the circular track are denoted by the different patterns of the inner ring. Global cues are denoted by shapes on the black outer ring representing the black curtains surrounding the track. The rotation directions of the local and global cues in the mismatch (MIS) sessions are indicated by the dotted and solid lines, respectively. In this example, 180° (session 2) and 45° (session 4) MIS sessions were interleaved with 3 standard (STD) sessions. (ii) Photograph of the textured, double-rotation track. (B) Sorted firing-rate maps of all the place fields included in the analyses. The abscissa of the map is the track angle, and each row of the map is the firing-rate map of a unit. The locations of the local-cue boundaries are denoted by the green lines. The rate maps were normalized by the peak firing rates of each unit and were sorted by the centers of mass of the fields. About 20%–30% of the place cells had multiple place fields (CA1-STD: 171 out of 669 [26%] place cells with multiple place fields; CA1-MIS: 283 out of 890 [32%]; CA3-STD: 55 out of 293 [19%]; CA3-MIS: 95 out of 429 [22%]). For these cells, the same rate maps were included in the figure multiple times, with each subfield aligned by its center of mass. Gray-scale bar indicates normalized firing rate. (C) Examples of place fields observed in CA1 (cells 1–4) and CA3 (cells 5–8). Some fields confined within a texture quadrant (cells 1 and 5) or crossing a local-cue boundary (cells 2 and 6) had no edges close to any of the local-cue boundaries. Other fields had one edge near a boundary (cells 3 and 7) or had both of their edges near the boundaries (cells 4 and 8). Fields that had one or more edges near a boundary could be contained within a single texture quadrant or could span across multiple quadrants. For each cell, the trajectory-spike plot (left) and the linearized firing-rate map (right) are presented. The blue-shaded areas represent the ranges of the place fields, and all rate maps are duplicated and concatenated in order to show fields crossing 0°. The local-cue boundaries are labeled by green lines in both plots. (D) The cross-correlograms of the PVs. The narrow pinch points of the diagonal band near the local cue-boundaries (dashed lines) show that the PVs changed more rapidly across the boundaries than across similar distances within a texture quadrant. The firing-rate maps were normalized by the peak firing rates before constructing the cross-correlograms. Color bar indicates the correlation coefficients. See also Figures S1 and S7.

Figure 2.. Place Field Edges Coincided with Local-Cue Boundaries

Left: the distributions of place field edges peaked near the local-cue boundaries (denoted by the black lines). The abscissa of the map is the track angle and the ordinate is the number of field edges observed within the corresponding spatial bin. Middle, right: the random shuffling control distributions (middle column) and the bootstrapped distributions (right column) of the proportion of field edges observed within the local-cue windows. The experimentally observed values are denoted by the thick black lines, the 95% confidence intervals of the simulated distributions by the dotted lines, and the chance level (0.33) by the dashed lines. *significant at α = 0.05, Bonferroni corrected for 4 comparisons. See also Figures S2, S3, and S4A.

Figure 3.. Place Field Edges Modulated by Surface Boundaries on the Complex Board

(A) Photo of the complex board. (B) Examples of CA1 place fields that show the clearest modulation by the surface boundaries. For each cell, the trajectory-spike plot (left), the smoothed firing-rate map (middle), and the smoothed firing-rate map with superimposed cue boundaries (right) are presented. The cue boundaries appeared to modulate the edges of the place fields: cells 1–6 occupied one or multiple geometric shapes defined by the cue boundaries, and they developed triangular, rectangular, or complex-shaped fields. Cells 7 and 8 fired along one or more cue boundaries and had elongated, stripe-like fields. The cue boundaries also appeared to affect the number and locations of the place fields. For example, cell 9 developed 3 fields at the vertices of the brown triangle and cell 10 fired at the corresponding corners of the black triangles. Cells 11 and 12 developed complicated firing patterns. The field of cell 11 filled in the black area, with some “bleeding” into the lower left of the textured area, and the field of cell 12 occupied the rectangular area near the bottom but also extended along the diagonal boundary. Gray-scale bar indicates firing rate (Hz).

Figure 4.. Place Field Edges Modulated by Surface Boundaries on the Simple Boards

(A) Photos of the simple boards and schematics of the simple board foraging task protocol. (B) Examples of CA1 place fields modulated by the cue boundaries. The figure format is as described in Figure 3. Based on visual inspection, ~5% of the place fields occupied geometric shapes defined by the cue boundary and the edges of the experiment board (cells 1–4), and ~2% of the fields extended along the cue boundary (cells 5–7). Some of the neural correlates near the cue boundary were similar to those near walls or other traditionally defined boundaries reported in previous studies. Cells 8 and 9 resembled boundary cells in that they developed multiple fields at corresponding locations with respect to the cue boundary and the board edge. However, for cell 9 the two fields were different in size, which might reflect the reset of cell activity at the cue boundary or might imply that the cell has one large field that was split by the boundary. Similarly, cell 10 seemed to have two fields that were intersected by the boundary, as if its firing activity was inhibited by the cue boundary. No other cells demonstrated firing patterns similar to cells 8–10. Gray-scale bar indicates firing rate (Hz). (C) Schematics of the cross-correlogram construction process for the simple boards. (i) The spatial binning of the simple board. The firing-rate map was constructed based on binning (denoted by the yellow grid) aligned with the surface boundary. (ii and iii) The construction of the cross-correlogram. Assuming there were only three cells (denoted by the squares in ii), the firing rates of the third columns of the grids from each cell (the purple, cyan, and green stripes in ii) were stacked together and formed the third population vector matrix (PVM) (M₃ in iii), and the 8^th columns (the red, pink, and orange stripes in ii) formed M₈, etc. For each element of the cross-correlogram, the Pearson’s product-moment correlation coefficients were calculated between the corresponding columns of the selected PVMs, and the mean correlation was calculated across the columns. (D) The cross-correlograms of the PVs. Along the direction perpendicular to the surface boundary, the correlation dropped more abruptly near the surface boundaries (denoted by the white dashed lines) for the leather boards. The firing-rate maps were normalized by the peak firing rates before constructing the cross-correlograms. Color bar indicates the correlation coefficient. See also Figures S5 and S7.

Figure 5.. Place Field Edges Concentrated Near the Surface Boundaries

(A) The place-field-edge density maps of the simple boards (top) and the control plain board with zone markings from the corresponding simple boards (bottom). The rim of the platform and the cue boundary are labeled by thick lines, and the boundary and nonboundary zones used in the analyses are labeled by thin lines. Color bar indicates field-edge density (number of field edges per bin). (B) The bootstrapped distributions (top) and the permutation distributions (bottom) of the BPI differences. The figure formats are as described in Figure 2. *Significant at a = 0.05, Bonferroni corrected for 4 comparisons. See also Figure S4B.

Figure 6.. Adjacent Fields Extending along the Surface Boundaries

(A) Schematics of different hypotheses explaining why the BPI difference was larger on the experiment boards than on the plain board. Hypothesis 1 suggests that a higher proportion of place fields were observed within the boundary zone on the simple boards than on the plain board. Hypothesis 2 suggests that place fields tended to extend along the cue boundary on the simple boards, and thus the average length of field edges observed within the boundary zone was larger on the simple boards than on the plain board. The place fields are denoted by colored circles, and the red fields increase the field-edge density within the boundary zone. B) The permutation test of the field proportion difference (no boards pass significance test at α = 0.05, two-tailed, with Bonferroni correction). The denotations are as described in Figure 2. (C) Longer field edges were found near the surface boundaries on the leather boards than on the plain board. Top: the distributions of the field-edge lengths within the boundary zone. The distributions of the field-edge lengths collected from the simple boards are represented by black bars, with the median values denoted by the solid lines, and the distributions of the plain board control are represented by white bars, with the median values denoted by the dashed lines. The field-edge length of the leather boards were significantly larger than the plain board control (two-tailed Mann-Whitney U test, n is the number of fields, and m is the median of the contour lengths: leather-standard, n_texture = 59, n_plain = 54, m_texture = 21.53, m_plain = 14.27, U = 1099.0, p = 0.002*; leather-shift, n_texture = 33, n_plain = 51, m_texture = 21.93, m_plain = 12.06, U = 397.0, p < 0.001*; tape-standard, n_texture = 54, n_plain = 50, m_texture = 12.94, m_plain = 12.92, U = 1142.5, p = 0.089; tape-shift, n_texture = 21, n_plain = 44, m_texture = 15.22, m_plain = 13.07, U = 381.0, p = 0.129; α = 0.05 with Bonferroni correction). Bottom: the permutation test of the edge length difference. The denotations are as described in Figure 2.

Figure 7.. Population Activity Changes More Abruptly Across, Not Along, the Surface Boundary

(A) Visualizations of the directions along which the PVs changed most abruptly. Red represents the direction parallel to the surface boundary, and blue represents the direction perpendicular to the boundary. The denotations of the lines are as described in Figure 2. In contrast to the divergent color distributions in the nonboundary zones, the boundary zones of the leather boards were dominated by blue and green, which represent directions perpendicular to the cue boundaries. (B) The Rayleigh plots of the directions observed within the boundary zone (red) and the non-boundary zone (blue). The movement direction ranged from 0° to 180°, 0° when parallel with the cue boundary and 90° when perpendicular to the cue boundary. (C) The bootstrap distributions of the MVL differences. (D) The bootstrap distributions of the spatial bin proportion difference. The denotations of (C) and (D) are as described in Figure 2. See also Figures S6 and S7.