Spatial Memory Engram in the Mouse Retrosplenial Cortex

Abstract

Memory relies on lasting adaptations of neuronal properties elicited by stimulus-driven plastic changes [1]. The strengthening (and weakening) of synapses results in the establishment of functional ensembles. It is presumed that such ensembles (or engrams) are activated during memory acquisition and re-activated upon memory retrieval. The retrosplenial cortex (RSC) has emerged as a key brain area supporting memory [2], including episodic and topographical memory in humans [3-5], as well as spatial memory in rodents [6, 7]. Dysgranular RSC is densely connected with dorsal stream visual areas [8] and contains place-like and head-direction cells, making it a prime candidate for integrating navigational information [9]. While previous reports [6, 10] describe the recruitment of RSC ensembles during navigational tasks, such ensembles have never been tracked long enough to provide evidence of stable engrams and have not been related to the retention of long-term memory. Here, we used in vivo 2-photon imaging to analyze patterns of activity of over 6,000 neurons within dysgranular RSC. Eight mice were trained on a spatial memory task. Learning was accompanied by the gradual emergence of a context-specific pattern of neuronal activity over a 3-week period, which was re-instated upon retrieval more than 3 weeks later. The stability of this memory engram was predictive of the degree of forgetting; more stable engrams were associated with better performance. This provides direct evidence for the interdependence of spatial memory consolidation and RSC engram formation. Our results demonstrate the participation of RSC in spatial memory storage at the level of neuronal ensembles.

Keywords: 2-photon imaging; c-Fos; engram; immediate-early gene; mouse; radial-arm maze; reference memory; retrosplenial cortex.

Figure 1

In Vivo Immediate-Early Gene Imaging (A–C) Regions of interest (ROIs) selected for 2-photon imaging. (A) Representation of the top view of the mouse brain based on the Allen Institute’s 3D mouse brain atlas with the positions of the ROIs. The dorsal (dysgranular) retrosplenial cortex (RSC) is highlighted in yellow. (B) A zoomed-in image showing the RSC. The images in (C) are brightfield photographs of the craniotomy areas in each animal. The numbering of cases and associated color-code is carried forward to subsequent figures. (D) Task-evoked change in fluorescence. The white outlines denote the positions of cells detected across the whole study, while the coloring indicates the level of fluorescence observed at four timepoints following exposure to the maze on a single session. The intervals are 1.5–2 hr (T1), 2–2.5 hr (T2), 2.5–3 hr (T3), and 3–3.5 hr (T4) from the beginning of the task, respectively. See also Figure S1.

Figure 2

The Pattern of Cell Activity Depended on Experimental Conditions (A) Reference memory errors in the radial-arm maze. An error was defined as entry into a non-baited arm (see inset above graph). Animals trained for 19 days, followed by negative control sessions on days 22 (C1) and 40 (C2), retention tests on days 25 (R1) and 43 (R2), and exposure to a novel environment on day 28 (N). Black symbols represent the mean error score, while colored symbols, the performance of individual animals. H, habituation; C1, C2, negative control sessions; R1, R2, retention sessions; N, novelty condition. (B) The black symbols show the mean percentage of active cells throughout the study, while colored symbols show data from individual animals. On the right-hand side of the graph, days 1–19 were collapsed into a single category of “training” (T), negative control sessions C1 and C2 into a single category of “control” (C), and sessions R1 and R2 into a single category “retention” (R). See also Figure S2.

Figure 3

The Pattern of Cell Activity Depended on Experimental Conditions (A) The similarity index J was calculated as the ratio of elements in common C and the sum of all available unique elements A + B − C. (B) A representative example of a population of cells tracked throughout the study. The outlines denote detected cells, while the coloring denotes the signal difference between timepoint 3 and timepoint 1, with red denoting an increase and blue denoting a decrease in fluorescence. The arrows point to putative “engram” cells. (C) Similarity index heat maps generated for individual animals. The arrangement of sessions is the same as in (D). Each square of the heat map displays the similarity index between two experimental sessions. Warm colors highlight high similarity index values, and cool colors highlight low similarity index values. (D) Mean similarity index heat map and corresponding dendrogram. The squares below the diagonal of the heat map represent mean similarity indices, while values above the diagonal represent corresponding (log10) p values with those below (Bonferroni-adjusted) α = 0.05 highlighted in orange. The dendrogram to the left reveals the clustering of conditions based on the mean similarity index values. Four clusters were detected: habituation (H) and days 1 and 4 of RAM training in light green; RAM training days 7–19 with the two retention sessions R1 and R2 in dark green; negative control sessions (C1, C2) in blue; and novelty on its own (N) in orange.

Figure 4

Engram Stability Predicted the Degree of Memory Retention (A) Relationship between mean similarity index and mean difference in error score, between training session 19 and remaining RAM sessions (green symbols); error bars are SEM. As animals become better at the RAM task (error score difference approaching 0), the similarity index increases, whereas the two memory retention sessions (red symbols) show increasing error score difference with decreasing similarity index. (B and C) (B) Relationship between error score difference and normalized similarity index (see STAR Methods) for individual animals plotted for sessions R2 and 19 and, in (C), for sessions R2 and R1. See also Figure S3.