Place cells on a maze encode routes rather than destinations
- University of Stirling, United Kingdom
- Edinburgh Medical School: Biomedical Sciences, University of Edinburgh, United Kingdom
Figures
Figure 1
Maze apparatus with two routes leading to the same goal.
(A) Top down view of the maze apparatus showing its layout including the start box, the three goal boxes, and the alleyways and choice points linking them. (B) The four trained routes through the maze. (C) Maze areas analysed for differential place cell firing. (D) Predictions of goal and route accounts of differential place cell firing. If differential firing of a place cells in the maze stem reflects the animal’s intended goal (Prediction 1 - left plots), then a given cell should fire when the animals takes either the left or right route to the same goal. If such firing reflects the animal’s route (Prediction 2 - right plots), firing should be seen on one route, but not the other. (E) Schematic of a representative daily session. Trials were blocked such that the same goal box was correct for at least 11 trials. The reward was then moved to a different goal box, and once it had been encountered by the rat, 11 further trials were run. In each session, all four routes were reinforced, although the order of these changed across sessions.
Figure 2
Acquisition of the win-stay task.
(A) The mean number of errors preceding the identification of the reinforced goal box in each block of trials (broken line) did not change significantly across training sessions. However, the number of errors following the identification of the reinforced goal box in each block of trials (solid line) decreased significantly across training. (B) The mean time taken to complete each trial preceding the identification of the reinforced goal box (broken line) did not change significantly across training sessions. However, the time taken following the identification of the reinforced goal box (solid line) decreased significantly. Error bars depict SEM. (C) Mean total number of errors summed across 10 training sessions on trials in which the centre goal box was rewarded (black) and on trials in which the Left and Right Goal Boxes were rewarded (white), after the rewarded goal box had been identified in a block of trials. Rats made significantly more errors on trials when the two routes to the Centre Goal Box were rewarded than on trials when the two routes leading to the Left and Right Goal Boxes were rewarded. (D) Mean total number of errors on each session for trials on which the Centre Goal Box (black) and the Left and Right Goal Boxes (white) were rewarded. (E) Mean number of errors summed across 10 training sessions broken down by the nature of the error. For example, the first bar shows the average number of times the rats incorrectly chose Route 1 when Route 2 was rewarded, plus the number of times they chose Route 2 when Route 1 was rewarded. The number of confusion errors between routes to different goal boxes (hollow bars) was similar, regardless of route combination. However, there were significantly more confusion errors for the two routes to Centre Goal Box (filled bar). Error bars depict SEM.
Figure 3 with 1 supplement
Differential firing throughout the maze.
(A) Heat map showing the position of every place field recorded on the maze apparatus, the corresponding colour axis is shown in the top left corner. Place cells over-represented the start of the maze as shown by the large number of place fields observed there. (B) There was a linear decrease in the percentage of active (firing above 1Hz) place cells as the distance from the start box increased. (C) Place cell firing rates tend to be highest in the central stem and first choice point. This pattern approximately follows the average running speeds of the rats in the maze, except in the goal boxes where a slight increase in firing rate is also observed. Marker colours follow the key given for panel A. Lines show fitted 3rd order polynomials. (D) Representative example cells which, clockwise from the top left, show differential firing in the start box, central stem, left and right stems of the maze. (E) Number of place cells (black bars) active in each area of the maze tested for differential firing, and the number of these which are showed differential firing (hollow bars). The percentage of active place cells that showed differential firing in each area is indicated by the number written within the hollow bars. (F) The average firing rate of place cells and differential cells in the start box of the maze when this firing was divided into four bins of equal duration. Higher firing was observed in bin 1, just after the rat is placed in the start box and in bin 4, just before the holding barrier was removed.
Figure 3—figure supplement 1
Diagram of parameters used in all ANCOVA analyses.
(A) All trajectories from a representative recording session. Paths are coloured according to the rat’s final destination, and this colour scheme is maintained in the panels below. (B) An enlarged view of each maze area analysed for differential activity. For this panel and the following ones, data for one maze area are shown per column. (C) The average x- y-coordinate recorded for each trajectory. The error bars signify the standard error of the mean in both the x and y-dimension for each trajectory. (D) The speed of the rat recorded for each trajectory. Speed was calculated as the total distance travelled in a maze area, such as the start box, divided by the total time spent in that area.
Figure 4 with 2 supplements
Four representative place cells, two per row, which show differential firing in the start box of the maze.
For each cell, the firing rate map for the session is shown, but with data divided into trials in which the animals took each of the four possible routes. The area of this map in which differential activity was detected is highlighted, and this area is also shown, enlarged, below the firing rate map. The colour axis for the main and enlarged maps is scaled from 0Hz to the maximum firing rate in the map. The colour bar for these is given between the last two cells. The maximum firing rate in each map is denoted by a number found to its bottom right. The rat number, date of recording, electrode and cluster are given at the top left of the main firing rate map. The mean and SEM firing rate for the four trajectories is shown in a bar plot to the top right of the man firing rate map. These bars are coloured differently for each rat. Further examples can be found in Figure 4—figure supplement 1.
Figure 4—figure supplement 1
Representative differential cells included in the route/goal analysis.
Three example cells are given per row. For each cell, the firing rate map for the session is shown, but with data divided into the four possible trajectories. The area of this map in which differential activity was detected is highlighted, this area is also shown, enlarged, below the firing rate map. The colour axis for the main and enlarged maps is scaled from 0Hz to the maximum firing rate in the map, and the colour bar for these is shown to the right of the last example cell. The maximum firing rate in each map is denoted by a number found to its bottom right. The rat number, date of recording, electrode and cluster are given at the top left of the main firing rate map. The mean and SEM firing rate for the four trajectories is shown in a bar plot to the top right of the man firing rate map. These bars are coloured differently for each rat. One neuron (6202 130108 E6 C2) fires in a goal manner; similarly for routes 2 and 3.
Figure 4—figure supplement 2
Four examples of differential firing on all trials within a session.
The firing rate (black bars) and the velocity (black line) of each cell in the highlighted maze segment(s ) is plotted for each trajectory in a given session. Note that this firing is generally higher only for those trajectories leading to one goal, regardless of the order of these trajectories within the session.
Figure 5
Distribution of place cells with differential firing.
(A) Number of differential cells in the start box and central stem of the maze, sorted by preferred route. (B) Number of differential cells in left and right arms of the maze, again sorted by preferred route. See Figure 3—figure supplement 1 for a breakdown of the parameters used by the ANCOVA analyses to determine differential activity. See Supplementary file 1 (tables 1–4) for the results of the alternative differential activity statistical methods.
Figure 6
Population characteristics and signal quality measures and their impact on rANCOVA outcome.
(A) Density plot showing firing rate plotted against width of waveform for every cluster (n = 641) in our population. Our cut-offs for characterising units as pyramidal cells are shown as white dotted lines. We also used a spatial information content threshold of 0.5b/s to further differentiate pyramidal cells from place cells. (B) Distribution of isolation distance (left), Lratio (middle left), signal to noise ratio (middle right) and peak amplitude (right) values for place cells that did not show differential activity (black shaded area) and for place cells which fired differentially on the maze (grey shaded area). These measures of cluster and signal quality did not differ significantly between the two groups of cells (p>0.05 in all cases, Kolmogorov-Smirnov tests, test statistics are shown on each plot). (C) The relationship between isolation distance (left), Lratio (middle left), signal to noise ratio (middle right) and peak amplitude (right) values and rANCOVA F-statistics calculated for all four maze areas assessed for differential firing. Each point represents a place cell assessed for differential firing in a maze segment (thus there can be a maximum of four points for one cell); darker areas denote overlapping points and thus data density. None of these measures are significantly correlated with rANCOVA outcome and thus differential firing (p>0.05 in all cases, Spearman’s pairwise correlations, test statistics are shown on each plot).
Figure 7 with 1 supplement
Ensemble decoding of trajectories (routes).
The colour bar for plots (A–C) is given to the right of (C), the colour bar for (E) and (F) is given to the right of (F). (B) Trajectory population vectors (PVs) were compared to session PVs and matched according to highest cosine similarity score. Tiles here show the percentage number of each trajectory matched to each of the session PVs only for the start box of the maze. For example the tile highlighted in blue shows the percentage number of Route 1 PVs correctly matched to the session PV for Route 1. (A) Matches were also made using shuffled data, where the session PVs were randomly shuffled for each neuron. The tiles here show the same as (B) except that this data is for one shuffle (10000 were conducted in total). (C) Shows how the contents of the lower left tiles in (A) and (B) can be used to calculate a probability or p-value for the number of matches made here. The grey shaded curve shows the distribution of percentage correct matches for Route 1’s PV to its session PV for all 10000 shuffles. The red line indicates where the value from the shown shuffled tile would fall. The blue line shows where the value from B (the actual data) would fall. A kernel smoothed cumulative density estimate (Epanechnikov) was then used to calculate the percentile value or probability of the real data value, given the shuffled distribution. (E) Shows the results of this analysis, with the probability of the percent correct matches made in (B), given the distributions in (A). Note that trajectories are only significantly matched to their corresponding session PVs. Furthermore, Routes 2 and 3 are not matched to each other significantly more than would be expected by chance. (C) and (F) show the same as (B) and (E) respectively, but for the central stem of the maze. See Figure 7—figure supplement 1 for ensemble analysis at the single trial level.
Figure 7—figure supplement 1
Analysis of ensemble decoding within blocks of trials.
When the firing rates of specific trajectories within blocks of trials were compared to their goal population vector, no significant increase or decrease in the number of correct matches (left graph) or cosine similarity value (right graph) over time were observed. This suggests that ensemble firing did not change significantly over the course of a block. The average number of trials completed per block was 12.5, and thus values after this point are not shown. Rats were well trained pre-surgery and made few errors during recording. Furthermore, the few errors that were made during recording were omitted from this ensemble analysis.
Figure 8
Place cell coding of the Centre Goal Box.
(A) A representative cell which fires similarly in the Centre Goal Box regardless of which route the animal took to get there. The firing rate map for a whole maze session is shown. This is divided into the four possible trajectories but is plotted using one colour axis scaled from 0–20 Hz. Surrounding this, activity in each of the four goal boxes has been enlarged. This pattern of firing suggests that the animal was aware this goal box occupied a single spatial location and was thus one box at the end of two distinct trajectories. (B) Left, the result of Spearman’s correlations comparing the firing of all place cells, from all rats and sessions in each pair of goal boxes, depending on the route taken to get there (i.e. firing in the Left Goal Box at the end of Route 1, the Centre Goal Box when at the end of Route 2, the Centre Goal Box at the end of Route 3 and finally the Right Goal Box at the end of Route 4). Correlations comparing the firing in each box to itself were not calculated and have thus been coloured white. Correlations are generally high, however, correlations comparing firing in the Centre Goal Box when it was accessed using the two different routes are the highest (i.e. Route 2 vs. Route 3 and vice versa), confirming that cells fire similarly in this box regardless of the route taken to get there. This suggests that rats and cells considered the Centre Goal Box to be one coherent spatial location. (B) Right, the results of a shuffling procedure to determine if the correlations are higher than would be expected by chance. Only those correlations between firing in the Centre Goal Box for Routes 2 and 3 are statistically significant (p<0.05). This test confirms that firing in the Centre Goal Box is more similar than would be expected by chance and that this firing is more similar than that between other pairs of goal boxes.
Figure 9
Histological confirmation of electrode placement.
(A) Coronal section of hippocampus with electrode track (inset: higher magnification view). (B) Schematic of individual electrode tracks towards the CA1 cell layer of the hippocampus. Arrows represent the angle and depth of implantation with the arrow tip showing the point at which the electrode passed through the CA1 cell layer. No electrodes contacted lower cell layers. Each arrow is labelled with a rat number and an estimated anterior-posterior (AP) coordinate (the schematic shows a slice at an AP -3.48 mm from bregma - the intended coordinate).
Videos
Video 1
Example of a place cell that fires differentially in the start box of the maze.
https://doi.org/10.7554/eLife.15986.007Additional files
-
Supplementary file 1
Analysis of differential firing using three different statistical methods.
- https://doi.org/10.7554/eLife.15986.017
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Place cells on a maze encode routes rather than destinations
eLife 5:e15986.
https://doi.org/10.7554/eLife.15986
