Method: The goal of this analysis is to investigate whether cells with cue-correlated activity patterns show consistently shifted responses to individual cues. Since cue cells were largely chosen based on the correlation of their activity patterns to a specific cue template (Figure 1B), this procedure could artificially select cells with activity patterns consistently shifted from individual cues and thus having high correlations to the template (comparability, high cue scores). Consequently, when these selected cue cells were ordered based on their spatial shifts, their responses were very likely to form consistent sequences at individual cues (as in Figures 4A and 5D). To avoid this artifact, here we classified cells with cue-correlated activity using a different approach in order to investigate whether having responses with consistent spatial shifts to individual cues is a true feature of cue cells. This analysis was performed on data collected in layers 2 and 3 of the MEC when mice navigated along an 18-meter track (Figure 5—figure supplement 4A and B). The track contained a large number of cues (10 cues), which allowed a more precise classification of cells with cue-correlated activity even when choosing only half the number of cues (method described below). Each cue template was split into two half-templates (templates 1 and 2), each containing half of the cues of the original template. The cues on the two templates were non-overlapping. Cue wells were first classified using one of the half-templates (i.e., template 1). The spatial shifts found from the correlation to template 1 was compared to that found for the other half-template (template 2), which was not used to classify the cells. The hypothesis was that if the response of a cue cell was shifted by the same distance from each cue, then the spatial shifts would be similar between these two half-templates. An example with two half-templates is shown in A. R1 and R2 are two half-templates with cues on the right side of the track. We calculated the percentage of cells that maintained similar spatial shifts across the two half-templates (the difference of the spatial shifts on R1 and R2 is less than 25 cm). We found that a large fraction of cue cells (76.9% and 80.3% for cells identified on R1 and R2, respectively) had very similar shifts on the two half-templates. A similar example for left-side cues is shown in (B). To further confirm that this high percentage of cells with consistent spatial shifts to cues was not found only by using a particular set of half-templates, we repeated this analysis for cells in both layers 2 and 3 using multiple sets of half-templates comprised of various combinations of cues from the original templates. This more strict analysis of spatial shifts of cue cells data together indicate that cue cells respond to individual cues with consistent spatial shifts.
Figure details: (A) Spatial shifts of cue cells classified using two half-templates of the right cue template (R template). The R template was split into two half-templates: R1 and R2. Spatial shifts of cue cells classified using R1 and R2 are shown in a1 and a2, respectively. a1: From top to bottom: (1) Classification of cue cells (R1 cue cells) using R1 template. (2) Ordered R1 cell responses based on their spatial shifts to R1 (R1 shifts). (3) Ordered R1 cell responses based on their spatial shifts to R2 (R2 shifts). Note that the activity patterns in both (2) and (3) consistently shift from individual cues, indicating that R1 cue cells generally had similar spatial shifts on R1 and R2. (4) Difference in spatial shifts of R1 cells on R1 and R2. Each dot is one cell. The fraction of cue cells (76.9%) with very similar spatial shifts on R1 and R1 (less than 25 cm absolute shift differences, marked by red parenthesis) a2: similar to a1 but for cue cells (R2 cue cells) classified using R2 template. (B) Similar to A but for left cue template (L template). (C) Summary of the percentages of cells in layers 2 and 3 with very similar spatial shifts (less than 25 cm absolute shift differences) on multiple pairs of half-templates (5 and 2 pairs for layers 2 and 3, respectively) comprised of different combinations of cues on the left and right cue templates (L1, L2, R1, R2). All analyses showed similar results. Red lines indicate the mean of each group.