Robust conjunctive item-place coding by hippocampal neurons parallels learning what happens where

Abstract

Previous research indicates a critical role of the hippocampus in memory for events in the context in which they occur. However, studies to date have not provided compelling evidence that hippocampal neurons encode event-context conjunctions directly associated with this kind of learning. Here we report that, as animals learn different meanings for items in distinct contexts, individual hippocampal neurons develop responses to specific stimuli in the places where they have differential significance. Furthermore, this conjunctive coding evolves in the form of enhanced item-specific responses within a subset of the preexisting spatial representation. These findings support the view that conjunctive representations in the hippocampus underlie the acquisition of context-specific memories.

Figure 1.

a, Conditional discrimination task. The two contexts (represented by different shadings) differed in their flooring (wood vs black paper for the initial conditional discrimination and rubber vs sandpaper for the second discrimination problem) and in the walls (white vs black paper for the initial conditional discrimination and vertical vs horizontal stripes for the second). The stimulus objects (X or Y) differed in odor and in the medium that filled the pots (shown as blue and yellow). b, Lesion marks made at tips of tetrodes, one located in CA1 and two others in the CA3 region of hippocampus.

Figure 2.

Example cells recorded from a single animal. Rasters and perievent histograms are plotted for stimulus sampling events for each item (X or Y) at each position (1 or 2) within each context (A or B). Each line of the raster represents a single stimulus sample ordered from 0 to the maximum number of samples for each particular item–position combination for all trials, correct and incorrect. Time point 0 denotes the time when the animal's nose crosses the edge of the stimulus pot, and each bar of the histogram represents the average activity in hertz for a 250 ms time window. Cells 1 and 2 are examples of item–position cells recorded during an overtraining session, whereas cells 3 and 4 are examples of item–position cells during a learning session. Cells 5 and 6 are examples of position cells during the same learning session.

Figure 3.

Spatial distributions of firing rates for the same cells for which raster plots are shown in Figure 2. Each panel includes spike activity recorded from the onset of the first stimulus sampling event until 1 s after the onset of the last stimulus sampling event in the trial, during which the rat can sample each stimulus multiple times before digging for the reward. Stimulus identities are indicated just outside their locations within the environment. Color-coded firing rates are indicated in the legend to the right of the plots for each cell. Gray indicates visited areas associated with no neural activity.

Figure 4.

Example cells recorded during a learning session. Panels in each row show the average ± SE firing rate of a cell during stimulus sampling for early, criterion, and late trial blocks. Each panel shows the average firing rate during sampling of items X and Y at each of two positions within a context (represented with different shades of gray). Cells 7–10 were classified as item–position cells in the last 30 trial block but not in the first 30 trials. Cell 11 was classified as a position cell in all trial blocks.

Figure 5.

Changes in proportions of item–position and position cells in learning (a) versus overtraining (b) sessions. Changes in selectivity for items during learning and overtraining sessions (c, d) and positions for the same sessions (e, f). The selectivity index was calculated using firing rates for each cell averaged across all stimulus samples. Bars represent the average ± SE proportions of cells in each 30 trial block, whereas the line shows average performance over the same trial blocks. The dotted white lines indicate average values calculated from the shuffling analyses.

Figure 6.

Changes in item selectivity during learning for individual animals. Average item selectivity index and performance accuracy are plotted over sliding 30 trial windows.

Figure 7.

Average ± SE z-score firing for all item–position cells during item samples for the preferred item–position combination of each cell on correct and incorrect trials compared with averaged z-score firing to nonpreferred item samples at that same position on correct trials.

Figure 8.

Example cells recorded during an overtraining session. Format is the same as in Figure 4. Cells 12–15 were classified as item–position cells in all trial blocks, whereas cell 16 was classified as a position cell over the same three intervals. Cell 17 shows a significant interaction of stimulus identity with position in the last 30 trials of the session but fails to show this effect in the first 30 trials.

Figure 9.

Changes in firing rate to preferred and nonpreferred stimuli and positions. a, Average z-score firing rate (see Materials and Methods) for item–position cells increases selectively in response to preferred stimuli after learning but remains unchanged in response to nonpreferred stimuli. b, Average z-score firing rate for position cells remains unchanged in response to both preferred and nonpreferred positions.