Hippocampal place cell remapping occurs with memory storage of aversive experiences

  1. Garrett J Blair  Is a corresponding author
  2. Changliang Guo
  3. Shiyun Wang
  4. Michael S Fanselow
  5. Peyman Golshani
  6. Daniel Aharoni
  7. Hugh T Blair  Is a corresponding author
  1. Department of Psychology, UCLA, United States
  2. David Geffen School of Medicine, University of California Los Angeles, United States
  3. Department of Neurology, David Geffen School of Medicine, University of California, Los Angeles, United States
  4. Integrative Center for Learning and Memory, University of California, Los Angeles, United States

Abstract

Aversive stimuli can cause hippocampal place cells to remap their firing fields, but it is not known whether remapping plays a role in storing memories of aversive experiences. Here, we addressed this question by performing in vivo calcium imaging of CA1 place cells in freely behaving rats (n = 14). Rats were first trained to prefer a short path over a long path for obtaining food reward, then trained to avoid the short path by delivering a mild footshock. Remapping was assessed by comparing place cell population vector similarity before acquisition versus after extinction of avoidance. Some rats received shock after systemic injections of the amnestic drug scopolamine at a dose (1 mg/kg) that impaired avoidance learning but spared spatial tuning and shock-evoked responses of CA1 neurons. Place cells remapped significantly more following remembered than forgotten shocks (drug-free versus scopolamine conditions); shock-induced remapping did not cause place fields to migrate toward or away from the shocked location and was similarly prevalent in cells that were responsive versus non-responsive to shocks. When rats were exposed to a neutral barrier rather than aversive shock, place cells remapped significantly less in response to the barrier. We conclude that place cell remapping occurs in response to events that are remembered rather than merely perceived and forgotten, suggesting that reorganization of hippocampal population codes may play a role in storing memories for aversive events.

Editor's evaluation

This paper describes important results obtained from multi-cellular imaging of CA1 cells using large-field-of-view miniscopes in rats performing a shock avoidance task. By exploiting behavioral (barriers) and pharmacological (scopolamine) manipulations the authors provide convincing results on cell remapping dynamics during aversive learning. This work will be of interest for the neuroscience community by setting new methodological standards and providing data for across species comparisons.

https://doi.org/10.7554/eLife.80661.sa0

eLife digest

The human brain is able to remember experiences that occurred at specific places and times, such as a birthday party held at a particular restaurant. A part of the brain known as the hippocampus helps to store these episodic memories, but how exactly is not fully understood.

Within the hippocampus are specialized neurons known as place cells which ‘label’ locations with unique patterns of brain activity. When we revisit a place, such as the restaurant, place cells recall the stored pattern of brain activity allowing us to recognize the familiar location.

It has been shown that a new negative experience at a familiar place – for example, if we went back to the restaurant and had a terrible meal – triggers place cells to update the brain activity label associated with the location. However, it remains uncertain whether this re-labelling assists in storing the memory of the unpleasant experience.

To investigate, Blair et al. used a technique known as calcium imaging to monitor place cells in the hippocampus of freely moving rats. The rats were given a new experience – a mild foot shock – at a previously explored location. Tiny cameras attached to their heads were then used to record the activity of hundreds of place cells before and after the shock.

Initially, the rats remembered the aversive experience and avoided the location where they had been shocked. Over time, the rats began to return to the location; however, their place cells displayed different patterns of activity compared to their previous visits before the shock.

To test whether this change in place cell activity corresponded with new memories, another group of rats were administered a mild amnesia-inducing drug before the shock, causing them to forget the experience. These rats did not avoid the shock site or show any changes in place cell activity when they revisited it.

These findings imply that new events cause place cells to alter their ‘label’ for a location only if the event is remembered, not if it is forgotten. This indicates that alterations in place cell activity patterns may play a role in storing memories of unpleasant experiences. Having a better understanding of how episodic memories are stored could lead to better treatments for diseases that impair memory, such as Alzheimer’s disease and age-related dementia.

Introduction

The rodent hippocampus contains place cells that fire selectively during visits to specific locations in space (O’Keefe and Dostrovsky, 1971). When place cells were first discovered, they were hypothesized to encode ‘cognitive maps’ that store stable long-lasting memories for the spatial geometry of familiar environments (O’Keefe and Nadel, 1978). Supporting this view, electrophysiology studies showed that place cells can retain stable positional tuning across repeated visits to the same environment (Lever et al., 2002). However, methodological limitations of single-unit electrophysiology made it challenging to consistently record the same population of place cells across long time spans, and thus to investigate the duration of time over which place cell tuning properties remained stable.

Recent advances in in vivo calcium imaging have made it possible to perform longitudinal studies of place cell tuning over days or weeks of experience. Such studies suggest that place cells in mice (Ziv et al., 2013; Cai et al., 2016; Kinsky et al., 2018; Keinath et al., 2022) and to a lesser extent in rats (Wirtshafter and Disterhoft, 2022) can change their spatial tuning properties (i.e., ‘remap’) with the mere passage time (see also Mankin et al., 2012). This phenomenon suggests that place cells may not simply encode spatial maps of environmental geometry, but also encode mnemonic information about accumulated past experiences that have occurred within an environment (Leutgeb et al., 2005; Colgin et al., 2008; Mankin et al., 2012; Ziv et al., 2013; Cai et al., 2016; Sanders et al., 2020). If so, then the rate at which place cells remap over time within a given environment may be proportional to the rate at which an animal stores new episodic memories of events that are linked to that environment.

If the rate of place remapping reflects the rate of hippocampal memory storage, then motivationally significant events (such as encounters with appetitive or aversive stimuli) that are preferentially stored to episodic memory might be expected to accelerate the rate of place cell remapping within environments where such events occur. Consistent with this view, studies have shown that place cells remap in response to behaviorally significant events (Breese et al., 1989; Bostock et al., 1991; Hollup et al., 2001; Colgin et al., 2008; Dupret et al., 2010; Alvernhe et al., 2011; Mamad et al., 2017). Numerous studies have shown that when an aversive stimulus is encountered in an environment, place cells respond by remapping their firing fields in that environment (Moita et al., 2003; Moita et al., 2004; Wang et al., 2012; Wang et al., 2015; Kim et al., 2015; Mamad et al., 2019; Schuette et al., 2020; Ormond et al., 2023), but it is not well understood whether such remapping plays a role in storing memories of aversive experiences.

Here, we investigated the role of place cell remapping in memory by using the MiniLFOV (Guo et al., 2023), a recently developed large-field-of-view version of the UCLA Miniscope (Cai et al., 2016; Aharoni and Hoogland, 2019), to image population activity of CA1 place cells in freely behaving rats as they acquired and extinguished a spatial avoidance learning task. The task was performed on a rectangular maze where rats ran stereotyped linear trajectories, allowing the speed and direction of running behavior to be closely matched during sessions conducted before acquisition versus after extinction of avoidance learning. In agreement with prior studies, we observed that avoidance learning induced place cell remapping in an environment where footshock was encountered compared with control conditions where (1) rats encountered a neutral barrier instead of an aversive footshock or (2) rats received footshock after injections of the muscarinic ACh receptor (mAChR) antagonist scopolamine at a dose (1 mg/kg) that disrupted long-term retention of avoidance without acutely impairing immediate avoidance or spatial tuning of place cells. Remapping after avoidance learning was similarly prevalent in subpopulations of shock-responsive and non-responsive place cells, and in contrast with some prior findings (Mamad et al., 2019), remapping did not alter the spatial distribution of place fields on the maze (i.e., there was no net migration of place fields toward or away from the shocked location). These findings add to a large body of evidence showing that place cell remapping occurs in response to aversive events, and also provide new evidence that remapping occurs preferentially under conditions where aversive encounters are remembered rather than forgotten, consistent with the hypothesis that place cell remapping may play a role in storing memories of aversive encounters.

Results

Male (n = 7) and female (n = 8) Long–Evans rats underwent survival surgery to inject 1.2 uL of AAV9-Syn-GCamp7s (AddGene) in the hippocampal CA1 layer; ~2 wk later, a 1.8-mm-diameter GRIN lens stack (~0.5 pitch) was unilaterally implanted to image CA1 neurons. A baseplate was attached to the skull for mounting the miniscope, and then rats began a regimen of running one 15 min session per 48 hr on a 250 × 125 cm rectangular maze (Figure 1A). Rats earned 20 mg chocolate pellets by alternating between two rewarded corners. On each trial, rats were free to choose a direct short path (250 cm) or an indirect long path (500 cm) to reach the next reward. During early sessions, rats learned to prefer the short over the long path (Figure 1C and D). Upon reaching criterion for short path preference (>2 rewards/min, and twice as many short paths to long paths), rats began receiving drug-free shock, scopolamine shock, or barrier training sessions (Figure 1B).

Figure 1 with 1 supplement see all
Disruption of aversive avoidance acquisition by scopolamine.

(A) Overhead view of the rectangular maze. (B) Order of treatment for five subgroups of rats in the study. The number of sessions in ACQ and EXT boxes varied per rat, depending on how many sessions it took individual subjects to reach behavioral criteria. (C) Example trajectories from sessions 1, 3, and 5 during initial acquisition of short path preference. (D) Acquisition curves for short path preference; criterion was reached on session 0. Black lines show mean +/- 1 standard deviation for the session (E) Example maze trajectories from training sessions. (F) Left: rewards/min earned on the short path during the first 10 min (pre10) versus last 5 min (post5) of training sessions; symbols indicate whether each rat received its first (shk1) or second (shk2) shock during a session. Right: boxplots show post-shock reduction in short path rewards/min (post5/pre10). (G) Example maze trajectories 48 hr after training. (H) Left: rewards/min earned on the short path during drug-free sessions given 48 hr before (–48 hr) vs. after (+48 hr) training. Right: boxplots show 48 hr retention of shock avoidance (+48 hr/ –48 hr). (I) Example maze trajectories 144 hr after training. (J) Left: extinction curves over the first three post-extinction sessions; number of rewards earned on the short path is measured as a percentage of baseline (48 hr before training; thin lines: individual rats, thick lines: per session median). Right: cumulative distributions show percentage of rats in each training condition meeting extinction criterion (≥2 rewards/min on the short path, and at least twice as many short path rewards as long path rewards) at 1, 2, 3, or 4 d after training. *p<0.05, **p<0.01, ****p<0.0001.

Acquisition and extinction of avoidance learning

During shock training sessions (drug-free or scopolamine), rats performed the standard alternation task for 10 min. Then a 1.0 mA scrambled current was switched on to electrify grid bars in a 50 cm segment of the short path’s center (Figure 1A). Identical grid bars spanned the full length of the long and short paths, so there were no local cues to indicate the shock zone’s location. After encountering the shock zone 1–2 times, rats subsequently avoided the short path during the final 5 min of the session (Figure 1E). In the scopolamine shock condition, lens-implanted rats received 1.0 mg/kg scopolamine via intraperitoneal injection 30 min prior to the start of the shock session. In the drug-free condition, lens-implanted rats received no injection prior to the start of the shock session. To control for effects of the injection procedure, a separate cohort of rats without lens implants was given injections of saline or scopolamine prior to shock training; in animals without lens implants, scopolamine-injected rats showed impaired avoidance learning compared with saline-injected rats that was similar to the impairment seen between scopolamine-injected and drug-free (non-injected) rats with lens implants (Figure 1—figure supplement 1B and C), indicating that the injection procedure itself was not responsible for effects induced by scopolamine. In the barrier control condition, the maze was not electrified; instead, a clear plexiglass barrier was placed in the center of the short path, forcing the rat to take the long path during the final 5 min of the session.

During the first 10 m of training on scopolamine (before shock delivery), 6/10 rats (three males, three females) fell below the criterion for preferring the short path (compared with 0/14 drug-free rats). These rats earned a median of five short path rewards/min during the drug-free session given 48 hr before training, but only 1.9 short path rewards/min during the first 10 min (before shock) of their training session on scopolamine (Figure 1F, ‘scopolamine shock pre10’; Wilcoxon test, p=0.0098). Thus, prior to shock delivery, scopolamine impaired expression of the rats’ learned preference for taking the short path (Figure 1E, middle).

During the training session, rats earned fewer rewards/min on the short path after than before grid electrification (Figure 1F) in both the drug-free (n = 14, signed-rank test, p=1.2e-4) and scopolamine (n = 10, signed-rank test, p=0.004) conditions. Rats in the barrier condition (n = 6) were likewise forced to stop taking the short path after it was blocked by the barrier (Figure 1F, left panel). Rats in the drug-free shock condition showed long-term retention of avoidance (Figure 1G, left), earning fewer rewards/min on the short path 48 hr after than 48 hr before training (Figure 1H, Wilcoxon test, p=3.7e-4). By contrast, most rats trained on scopolamine showed similar short path reward rates during drug-free sessions given 48 hr before and after training (Figure 1H, Wilcoxon test, p=0.125), and thus failed to exhibit long-term retention of avoidance (Figure 1G, middle). When tested 48 hr after training, rats shocked on scopolamine showed less avoidance of the short path than rats shocked drug-free (rank-sum test: p=0.0235; Figure 1H). Hence, pre-training scopolamine injections impaired aversive learning, consistent with prior studies (Decker et al., 1990; Anagnostaras et al., 1999; Green et al., 2005).

Between- versus within-session place cell stability

A large-field-of-view version of the UCLA Miniscope (‘MiniLFOV’; Guo et al., 2023) was used to image neurons in the hippocampal CA1 layer through the implanted GRIN lens during maze sessions (Figure 2A). Although the GRIN lens caused some flattening and compression of stratum oriens (Figure 2—figure supplement 1), this did not appear to cause significant disruptions of CA1 activity since spatial tuning properties of place cells were stable and intact (see below).

Figure 2 with 8 supplements see all
Between- versus within-session population coding.

(A) Upper left: illustration of rat wearing MiniLFOV. Upper right: cell contours identified during pre-training (green) and training (red) sessions in one rat; regions of overlap between contours that recurred in both sessions appear yellow. Lower left: target position of GRIN lens over the CA1 layer. Lower right: fluorescence image of lens position from the example rat. (B) Top: rastergram shows normalized calcium fluorescence traces of place cells (one per row, sorted by preferred firing location and direction) during several traversals of the short path during an example session. Bottom: rat’s position (black) with running epochs colored by direction of travel (‘LR,’ left to right, in blue; ‘RL,’ right to left, in red). (C) Individual rat data (lines and symbols) and session means (bars) for the number of subsampled beeline trials (left) and median beeline running speed (right) during pre-training (pre) and training (trn) sessions; subsampled beeline trial counts were lower for trn than pre sessions because only trials from the first 10 min of trn (prior to shock delivery) were included; running speeds were lower for trn than pre sessions because scopolamine (SCP) reduced running speeds, resulting in preferential subsampling of slower beeline trials from drug-free (DF) sessions by the algorithm that minimized running speed differences between training conditions (see ‘Methods’). (D) Bar/line graphs show number of place cells imaged per rat (left) and percentage of all imaged cells classified as place cells (right) during pre and trn sessions. (E) Pie graphs show percentages of place cells imaged per condition (‘n’ gives total number summed over rats) that were spatially tuned in the LR only, RL only, or both LR and RL running directions. (F) Tuning curve properties of place cells imaged during pre and trn sessions. (G) Place cell recurrence ratios (RR) between pre and trn sessions. (H) Top: diagram shows timeline for pre and trn sessions given 48 hr apart. Bottom: tuning curve heatmaps for recurring place cells (from all rats combined, co-sorted by peak locations from the trn1 session) that were spatially tuned in the LR (top) or RL (bottom) running directions; separate heatmaps are shown for pre, trn part 1 (trn1), and trn part 2 (trn2) sessions. (I) Between- (B) and within- (W) session population vector correlation matrices are shown for DF and SCP shock training conditions; middle bar graph shows median place tuning stability (S) for each rat (lines and symbols) and mean over rats (bars) for B and W heatmap pairs. Asterisks in (C) and (D) denote significance for main effect of DF vs. SCP or uncorrected t-test comparing pre vs. trn sessions; asterisks in (E) and (G) denote significance for uncorrected t-tests. †p<0.1; *p<0.05; **p<0.01; ***p<0.001.

Calcium activity was analyzed only during traversals of the short path since the long path was undersampled after rats had learned to prefer the short path. Analysis was further restricted to beeline trials during which the rat ran directly from one reward zone to the other on the short path, without any pause or change in direction (Figure 2B, Video 1); this assured that imaging data came from periods of active locomotion when place cells exhibit reliable spatial tuning, and that the rats’ behavior during imaging was similar across all experimental sessions and conditions. To further control for possible confounding effects of behavior differences between sessions (see Figure 2—figure supplement 2), beeline trials were randomly subsampled using an algorithm (see ‘Methods’) that minimized trial count and running speed differences between the drug-free and scopolamine conditions (Figure 2C). Spatial tuning was then analyzed and compared during these behaviorally homogeneous trials. To analyze spatial tuning, the short path was subdivided into 23 spatial bins (each 10.8 cm wide). Two spatial tuning curves (one per running direction) were derived for each neuron. Cell activity was measured in units of active frames per second (Af/s), defined as the mean number of imaging frames per second during which a neuron generated at least one inferred spike. A neuron was classified as a place cell if its LR or RL tuning curve (or both) met defined criteria for minimum Af/s, spatial selectivity, and spatial stability within a session (see ‘Methods’). About 1/3 of place cell tuning curves met the criteria for spatial selectivity only in the LR direction, 1/3 only in the RL direction, and 1/3 in both directions (Figure 2E). It should be noted that since we used a ubiquitous synapsin promoter to drive expression of GcaMP7s, some neurons classified as place cells might have been CA1 interneurons rather than pyramidal cells.

Video 1
Video demonstrating behavior and calcium imaging recorded from one rat performing linear alternation along the short path of the maze.

(A) Position of the rat (red dot). (B) Motion-corrected video of calcium fluorescence. (C) Processed demixed/denoised CaImAn output (spatial contours modulated by denoised calcium trace). (D) Right half shows calcium activity of 290 identified neurons in the video, sorted by their preferred firing direction (teal/pink) and location along the track. Current time highlighted in the center gray region.

To analyze acute effects of scopolamine on CA1 place cell activity, data from drug-free or scopolamine shock sessions was compared against data from pre-training baseline sessions (always drug-free) given 48 hr earlier. The drug-free condition only included rats (n = 9) that retained avoidance of the short path 48 hr aftershock training, whereas the scopolamine condition only included rats (n = 7) that failed to avoid the short path 48 hr aftershock training (indicating that the drug had blocked avoidance learning). A 2 × 2 ANOVA on the number of place cells detected per session (Figure 2D, left) found no effect of training condition (F1,14 = 0.03, p=0.87) or session (F1,14 = 0.014, p=0.92), and no interaction (F1,14 = 2.1, p=0.17). A similar ANOVA on the percentage of imaged cells per session that were classified as place cells (Figure 2D, right) found no effect of training condition (F1,14 = 0.07, p=0.79), but place cell percentages were lower during training than pre-training sessions (F1,14 = 5.1, p=0.04); this reduction was greater in the scopolamine than drug-free condition, resulting in a marginal interaction effect (F1,14 = 4.4, p=0.0539). An analysis of how spatial tuning varied with the size and location of cell contours revealed that contours of non-place cells tended to be slightly but significantly smaller (p<0.01) and nearer to the lens center (p<0.05) than contours of place cells (Figure 2—figure supplement 3A).

For tuning curves that met the criteria for spatial selectivity in each session, four tuning properties were analyzed: peak Af/s rate, out-of-field Af/s rate, tuning width, and spatial information content (see ‘Methods’). For each rat and each session, median values of these tuning properties were taken over all LR and RL tuning curves that met spatial selectivity criteria (Figure 2—figure supplement 4); 2 × 2 mixed ANOVAs were then performed on the resulting rate medians for each of the four tuning properties (Figure 2F). There were large main effects of session (pre-training vs. training) for all four tuning properties, which were driven by correspondingly large differences in behavior sampling and running speed (Figure 2C). However, there were no main effects of training condition (drug-free vs. scopolamine) or interactions between training condition and session (Figure 2F), indicating that none of the four tuning properties differed during training sessions given drug-free versus on scopolamine.

The place field recurrence ratio was defined as RR=NR/NT , where NR is the number of place fields that recurred during the pre-training and training sessions, and NT is the total number of recurring and non-recurring place fields detected during both sessions combined (see ‘Methods’). Mean RR values were lower (independent t14=2.15, p=0.0497) during scopolamine (40.7 ± 8.8%) than drug-free (50.7 ± 3.9%) training sessions, consistent with the findings reported above that the proportion of neurons classified as place cells during the training session was lower on scopolamine than drug-free (Figure 2D, right). It was observed that contours of recurring place cells tended to be slightly but significantly smaller (p<0.01) and nearer to the lens center (p<0.01) than contours of non-recurring place cells (see Figure 2—figure supplement 3B and C).

To investigate how scopolamine affected population coding by recurring place cells, two heatmaps (one per running direction, denoted HLRpre and HRLpre) were created from pre-training tuning curves of recurring place cells (Figure 2H, Figure 2—figure supplement 5). Additionally, two heatmaps per running direction were created from data collected during the first (HLRtrn1 , HRLtrn1) and second (HLRtrn2 , HRLtrn2) half of each training session’s subsampled beeline trials, all of which occurred prior to shock delivery. Between-session population vector correlation matrices (one per running direction) were derived for each rat (Figure 2—figure supplement 5) with matrix entries given by ρi,jbtwn=R(Hipre,Hjtrn1), where Hipre and Hjtrn1 denote population vectors in columns i and j of the pre-training and trn1 heatmaps, respectively, and R is the Pearson correlation coefficient. Within-session population vector correlation matrices were derived in a similar manner with matrix entries given by ρi,jwthn=R(Hitrn1,Hjtrn2), where Hitrn1 and Hjtrn2 denote population vectors in columns i and j of the trn1 and trn2 heatmaps, respectively. When population vector correlation matrices were averaged together across running directions and rats (Figure 2I), high mean ρwthn values were observed along diagonals of within-session correlation matrices for both the drug-free and scopolamine conditions. By contrast, lower mean ρbtwn values were observed along diagonals of between-session correlation matrices, in accordance with prior studies showing that place cell population vectors become decorrelated over time (Mankin et al., 2012; Ziv et al., 2013; Cai et al., 2016; Keinath et al., 2022). To quantify this decorrelation, a place tuning stability score (S) was computed as S = median(ρ1LR-ρoffLR , ρ2LR-ρoffLR ,..., ρKLR-ρoffLR , ρ1RL-ρoffRL , ρ2RL-ρoffRL ,..., ρKRL-ρoffRL), where ρoffLR and ρoffRL are mean off-diagonal ρ values and ρiLR and ρiRL are individual peri-diagonal ρ values for bins in the rat’s LR and RL correlation matrices (see ‘Methods’). A 2 × 2 mixed ANOVA on S scores with training condition (drug-free vs. scopolamine) as an independent factor and time interval (between vs. within session) as a repeated factor found no main effect of training condition (F1,14 = 1.73, p=0.21), but there was a large main effect of time interval (F1,14 = 71.5, p=7.1e-7) and a significant interaction (F1,14 = 8.1, p=0.0132). Uncorrected post hoc comparisons confirmed that this was because scopolamine injections reduced between- (t14 = 3.1, SDF < SSCP p=0.0084) but not within- (t14 = 0.83, SDF < SSCP p=0.42) session population vector correlations (Figure 2I, center). Hence, even though scopolamine did not acutely disrupt spatial tuning of CA1 neurons (Figure 2F), the drug reduced the percentage of recurring place cells and degraded the between-session (but not within-session) similarity of recurring place cell population vectors (Figure 2H and I). Similar disruption of between- but not within-session place cell stability has been previously reported in aged rats (Barnes et al., 1997).

Disruption of between-session place cell stability by scopolamine was similar in male and female rats (Figure 2—figure supplement 6). Cross-session stability of population vectors was completely abolished in control analyses (see ‘Methods’) where inferred spikes were circularly shifted against position tracking data during each individual beeline trial of the pre-training reference session (Figure 2—figure supplement 7). When analyses of Figure 2 were repeated on scopolamine rats (n = 2) that retained avoidance despite the injection, population vector similarity was not degraded between pre-training and training sessions, suggesting that impairment of between-session place tuning stability was related to impairment of avoidance learning (Figure 2—figure supplement 8). When the analyses were repeated on drug-free rats (n = 2) that failed to retain avoidance despite being trained drug free, between- and within-session population vector similarity were indistinguishable from drug rats that retained avoidance (Figure 2—figure supplement 8), implying that failure to acquire avoidance in these rats was not attributable to any noticeable deficiency in the fidelity of the hippocampal place code during training.

Induction of place cell remapping by aversive learning

Training-induced place cell remapping was analyzed in the same subset of rats shown in Figure 2. To measure remapping, post-extinction CA1 population vectors were compared against the 48 hr pre-acquisition baseline session. For the drug-free shock condition, post-extinction population vectors were sampled from the first session aftershock training during which the rat met criterion (>2 rewards/min, and twice as many short paths as long paths) for extinction of short path avoidance (Figure 3A, left); this was the earliest post-training session with adequate sampling of the short path. For the scopolamine shock condition, post-extinction population vectors were sampled from the session given 8 d (i.e., four sessions including the training session) after the pre-training session (Figure 3A, center); this matched the modal intersession interval for the drug-free condition, so that confounds would not arise from measuring remapping across different time intervals for the drug-free versus scopolamine conditions. Likewise, for the barrier control condition, post-extinction population vectors were also sampled 8 d after the pre-training session; one barrier training rat became startled due to external factors and stopped running early in the 8 d post-training session, so the 6 d session was used instead (Figure 3A, right).

Figure 3 with 5 supplements see all
Remapping of place cells is induced by avoidance learning.

(A) Top: bar graphs show distributions of intersession intervals between pre-training (pre) and post-extinction (post) session pairs included in the analysis. Bottom: diagram shows timeline for pre and post sessions. (B) Left: number of subsampled beeline trials did not differ significantly by training condition or session. Right: running speeds did not differ significantly by training condition; there was a marginal difference between pre vs. post running speeds in the barrier (but not drug-free or scopolamine shock) condition. (C) Left: total number of imaged place cells (recurring and non-recurring) did not differ significantly by training condition; there was a marginal difference between pre vs. post place cell counts in the barrier (but not drug-free or scopolamine shock) condition. Middle: percentage of all imaged cells (recurring and non-recurring) per session that were classified as place cells. Right: between-session place cell recurrence ratios did not differ by training condition. (D) Heatmap pairs show pre and post tuning curves for recurring place cells that were spatially tuned in the LR (top row) and RL (bottom row) running direction; both heatmaps in each pair are co-sorted by peak locations from the pre session. Pie charts show the proportion of recurring place cells (total number given at top) for which tuning curves were included in LR and RL heatmap pairs. (E) Kolmogorov–Smirnov tests show that distributions of place field peaks locations (LR and RL combined) were unchanged between pre and post sessions. (F) Mean population vector correlation matrices for the three training conditions. (G) Decoding accuracy index (DAI) at each position on the track for each of the three training conditions; ‘*’ indicates locations where decoding was significantly (p<0.05) less accurate for the drug-free condition than the other two conditions. (H) Analysis of place tuning stability scores (S) along the full short path are shown in the top row; results for L, C, and R track zones are shown in the bottom 3 rows, respectively. Left column: bar graphs show means and standard errors of S; Middle columns: boxplots show median and range of template-selected population vector correlation bins in each rat. Right column: templates used to select peri-diagonal correlation values for analysis from different track zones. †p<0.1; *p<0.05; **p<0.01; ***p<0.001.

To control for potential confounds from between-session behavior differences, beeline trials were subsampled using an algorithm (see ‘Methods’) that equalized the number of trials from pre- and post-extinction sessions in each rat (Figure 3B, left), while also minimizing running speed differences between sessions and training conditions (Figure 3B). Spatial tuning was analyzed during subsampled trials, and a 3 × 2 mixed ANOVA on the number of place cells detected during each session (Figure 3C, left) yielded no effect of training condition (F2,19 = 0.43, p=0.66) or session (F1,19 = 0.01, p=0.92), and no interaction effect (F2,19 = 1.57, p=0.24). A similar ANOVA on the percentage of imaged cells per session classified as place cells also yielded no significant effects (training condition: F2,19 = 0.14, p=0.87; session: F1,19 = 0.56, p=0.46; interaction: F2,19 = 1.28, p=0.3). The place field recurrence ratio (RR) between pre- and post-extinction sessions also did not differ among training conditions (Figure 3C, right; one-way ANOVA: F2,19=.65, p=0.53).

To investigate whether shock-induced remapping occurred, a pair of heatmaps (pre and post) was created from tuning curves of recurring place cells in each running direction (LR and RL) for each training condition (Figure 3D). As in prior calcium imaging studies (Ziv et al., 2013), place fields tended to be more densely concentrated near the ends of the short path than in the middle (Figure 3E). Heatmap pairs were used to derive two population vector correlation matrices per rat, PLR and PRL (Figure 3—figure supplement 1). When PLR and PRL were averaged across rats, a gap of low correlation appeared in the center of the diagonal for the drug-free shock condition (Figure 3F, left), which was not seen in the scopolamine shock (Figure 3F, middle) or barrier (Figure 3F, right) conditions, suggesting that drug-free shock training induced more remapping of the short path than scopolamine shock or barrier training, in accordance with prior studies (Moita et al., 2003; Moita et al., 2004; Wang et al., 2012; Mamad et al., 2019; Schuette et al., 2020; Ormond et al., 2023).

To statistically analyze this remapping effect, a place tuning stability score S was derived for each rat by the methods described above. Boxplots in Figure 3H (top row) show each rat’s median and range of S values from all peridiagonal bins in PLR and PRL combined. Bar graphs (Figure 3H, top left) for mean stability scores, S¯, for each training condition differed significantly between drug-free, scopolamine, and barrier conditions (one-way ANOVA: F2,19=9.7, p=0.0012); uncorrected post hoc t-tests confirmed that this was because S¯ was lower in the drug-free shock than scopolamine shock (t14=4.3,p=0.0007) or barrier (t13=2.2,p=0.0498) conditions, which differed only marginally from one another (t11=1.98, p=0.0834). Hence, drug-free avoidance training and extinction induced more place cell remapping than the scopolamine shock or barrier training conditions; this remapping occurred at all locations on the short path but a 3 × 3 repeated-measures ANOVA with training condition (drug-free shock, scopolamine shock, barrier) and track zone (left safe, center shock, right safe; bottom 3 rows of Figure 3H) as factors suggested that remapping was most pronounced in the center zone where shocks occurred than at L and R ends of the track where no shocks occurred (Figure 3H, bottom; F2,40=3.0, p=0.0596). Induction of remapping by avoidance training and extinction did not differ in male versus female rats, nor did the impairment of remapping by scopolamine (Figure 3—figure supplement 2). When the analyses of Figure 3 were repeated on scopolamine rats (n = 2) that successfully retained avoidance despite the injection, the drug was less effective at inducing place cell remapping (Figure 3—figure supplement 3), suggesting (but not proving) that impairment of remapping was causally related to impairment of avoidance learning. When the analyses were repeated on drug-free rats (n = 2) that failed to retain avoidance despite being trained drug free (Figure 3—figure supplement 2), one rat showed less and the other showed more place tuning stability across the entire length of the track when compared against rats that retained avoidance.

Kolmogorov–Smirnov tests found that place field density distributions (accumulated over all rats) did not change significantly between the pre versus post sessions in any training condition, including the drug-free shock condition (Figure 3E, left). Hence, remapping did not cause place fields to migrate toward or away from the shock zone (Figure 3—figure supplement 4), in contrast with prior reports that remapping can cause place fields to migrate toward locations where aversive stimuli are encountered (Mamad et al., 2019). All place cell remapping effects and differences among training conditions were abolished when remapping analyses were performed on shuffled data (Figure 3—figure supplement 5).

To analyze the extent of remapping caused by aversive learning, the rat’s position on the track was decoded from place cell calcium activity (see ‘Methods’). The decoder was trained on data from one session (pre- or post-extinction) and tested on data from the other session. The decoding error in a given position bin was measured as the median of the absolute difference (in cm) between the rat’s actual versus predicted position over all visits to that bin. Wilcoxon signed-rank tests compared decoding errors obtained with shuffled versus non-shuffled population vectors; a decoding accuracy index (DAI) was then derived for each bin, with DAI = 0 corresponding to p=0.01 for more accurate decoding from unshuffled than shuffled population vectors (see ‘Methods’). When mean DAI values were compared among training conditions, significant differences were detected only in two bins at the center of the track (in the shock zone) where decoding in the drug-free shock condition was significantly less accurate (p<0.05) than in scopolamine shock and barrier training conditions combined (asterisks in Figure 3D, bottom). In accordance with results from population vector correlation analysis, these decoding results support the conclusion that drug-free shocks induced more remapping of the center zone than shocks while on scopolamine or from barrier training.

Shock-evoked responses of place cells

Shock-induced place cell remapping has been observed in prior studies (Moita et al., 2003; Moita et al., 2004; Wang et al., 2012; Schuette et al., 2020), but the cellular and synaptic mechanisms of such remapping are not well understood. Individual CA1 neurons can form new place fields at locations where they are artificially stimulated or inhibited (Bittner et al., 2015; Geiller et al., 2022), and since some CA1 place cells are naturally responsive to shocks (Moita et al., 2003; Moita et al., 2004), it is possible that individual CA1 neurons might gain or lose place fields at locations where they respond to shocks. If so, then this might partly account for remapping at the population level. If shock-evoked responses of CA1 neurons are involved in remapping, then blockade of remapping by scopolamine could be related to effects of the drug upon these shock-evoked responses. To investigate this, each place cell imaged during a training session was classified as shock-responsive if its Af/s rate was significantly higher (p<0.01) during entries to the electrified than non-electrified shock zone; otherwise, it was classified as non-responsive to shocks (see ‘Methods).

The mean percentage of shock-responsive cells per rat (Figure 4A, left column) was 35.5 ± 2.3% for drug-free training and 29.6 ± 3.3% for scopolamine training; these percentages did not differ significantly (independent t14= 1.58; p=0.13). Pre-shock baseline Af/s rates (Figure 4A, middle column) were higher for shock-responsive than non-responsive place cells in both the drug-free (t1931= 9.3; p<0.0001) and scopolamine (t1034= 4.34; p<0.0001) conditions, which was attributable to the fact that in both conditions, place field peaks for shock-responsive and non-responsive cells tended to be segregated on the ‘approach’ and ‘departure’ sides of the shock zone, respectively (Figure 4A, right column). Thus, as rats approached the shock zone in either running direction, shock-responsive cells were more likely than non-responsive cells to be active in their place fields just before shock zone entry.

Shock-evoked responses of place cells.

(A) Left column: frequency distribution of shock-responsive place cell percentages in each rat. Middle column: population-averaged Af/s rates (from all rats combined) for shock-responsive versus non shock-responsive place cells. Right column: spatial distribution of place field peak locations during the training session, prior to grid electrification; x-axis shows distance traveled between the start (left) and end (right) of the short path in the current running direction (LR or RL). (B) Heatmaps (LR and RL tuning curves combined) are plotted for shock-responsive place cells in the drug-free (top) versus scopolamine (bottom) conditions during pre- (left) and post- (right) training sessions (rows are co-sorted on peaks from the pre-training session, columns are sorted so that the rat’s running direction goes from left to right for both LR and RL tuning curves); P matrices show pre- vs. post-extinction population vector correlations for shocks given drug-free (top) or on scopolamine (bottom); line graphs show that signed p-values (y-axis, log scale) fall below .01 at center track positions where drug-free shocks induced more remapping than shocks on scopolamine. (C) Same as (B) but for shock non-responsive place cells. ****p<0.0001.

To compare remapping by shock-responsive versus non-responsive CA1 neurons, analyses were performed on place cells that recurred during pre- and post-extinction sessions and were also active during the training session (so that each cell could be classified as responsive or non-responsive to shocks). Pre- and post-extinction heatmaps were derived from tuning curves of shock-responsive (Figure 4B) and non-responsive (Figure 4C) place cells. LR and RL tuning curves were combined together in each heatmap, with RL (but not LR) tuning curves reflected on the horizontal so that the rat’s running direction was oriented from left to right in all rows; hence, columns to the left or right of center in each heatmap denoted locations where rats were approaching or departing the shock zone, respectively. From these heatmaps, shock-responsive and non-responsive population vector correlation matrices were derived for the drug-free (Pshk+DF , Pshk-DF) and scopolamine (Pshk+SCOP , Pshk-SCOP) conditions. A gap of low population vector correlation values in the center of Pshk+DF but not Pshk+SCOP (Figure 4B) and of Pshk-DF but not Pshk-SCOP (Figure 4C) indicated that drug-free shocks induced remapping of the center zone in both shock-responsive and non-responsive place cell populations, whereas scopolamine shocks failed to induce remapping in either population. Confirming this, a series of independent t-tests was performed on clusters of peri-diagonal bins arrayed along diagonals of Pshk+DF versus Pshk+SCOP , and Pshk-DF versus Pshk-SCOP (see ‘Methods’); resulting significance values (line graphs in Figure 4B and C) dipped well below p<0.01 in the center region of the track (but not at the ends), indicating that drug-free shocks caused shock-responsive and non-responsive place cells to remap the center zone more than scopolamine shocks.

Discussion

Place cell remapping has been shown to occur at locations where animals experience aversive events such as shocks, predator odors, or attacks (Moita et al., 2003; Moita et al., 2004; Wang et al., 2012; Wang et al., 2015; Kim et al., 2015; Mamad et al., 2019; Schuette et al., 2020; Ormond et al., 2023), but it is not well understood whether such remapping plays a functional role in storing memories for aversive encounters. Here we investigated this question by using a large-field-of-view version of the UCLA Miniscope (Guo et al., 2023) to image population activity of CA1 place cells in rats as they acquired and extinguished a spatial avoidance learning task. We found that significantly more place cell remapping was induced under experimental conditions where aversive encounters were remembered (drug-free shocks) than under control conditions where they were forgotten (shock training on scopolamine), or where a neutral rather than aversive path-blocking stimulus was encountered (barrier training). While these results do not definitively prove that place cell remapping is causally necessary for storing memories of aversive encounters, they provide correlational evidence that remapping occurs selectively under conditions where a stimulus occurs and is subsequently remembered rather than forgotten.

Spatial distribution of place fields

Hippocampal place fields are sometimes observed to be preferentially concentrated at locations where rodents spend large amounts of time, such as the two rewarded locations on our rectangular maze (Figure 3E). This makes sense from an information theoretic perspective since it is efficient for frequently occurring stimuli to be overrepresented (and conversely for rarely occurring stimuli to be underrepresented) among preferred stimuli of tuned neurons in a population code. In accordance with this principle, when animals learn to obtain reward at specific locations (and thus to spend more time at those locations), a form of remapping occurs whereby place fields migrate toward rewarded locations (Breese et al., 1989; Hollup et al., 2001; Dupret et al., 2010; Mamad et al., 2017).

Less is known about how aversive learning affects the spatial distribution of place fields, but based purely upon information theoretic considerations, one might expect that place fields should migrate away from (rather than toward) aversively reinforced locations as animals shift their behavior to spend less time there. Contradicting this intuition, there is evidence to suggest that place fields migrate toward (rather than away from) aversively reinforced locations (Mamad et al., 2019). Possibly, overrepresentation of aversively reinforced locations increases the likelihood that place cells encoding such locations will become active during ‘replay events’ that occur during sleep (Girardeau et al., 2017), or during waking as the animal deliberates over whether to approach a remote location where danger has been encountered (Wu et al., 2017; Ormond et al., 2023). This may help animals to avoid dangerous places by anticipating in advance the threats that await them there.

We analyzed place cell remapping after rats first learned to avoid a shocked location, then extinguished avoidance and resumed spending time at the formerly shocked location. In agreement with prior findings (Moita et al., 2003; Moita et al., 2004; Wang et al., 2012; Schuette et al., 2020), we observed that place cell population vectors were less correlated (and thus showed more remapping) in rats that underwent acquisition and extinction of avoidance than in control rats that encountered a neutral barrier or failed to retain avoidance after being trained under the influence of scopolamine (Figure 3F–H). In contrast with Mamad et al., 2019, place cells did not shift their center of mass toward the site of the aversive stimulus; instead, the spatial distribution of place fields remained unchanged by remapping (Figure 3E), indicating that place fields shifted toward and away from the shocked location in equal proportions.

The discrepancy in place field migration findings might be related to our use of calcium imaging rather than electrophysiology to record neural activity; mChARs can regulate calcium entry into cells via a wide range of mechanisms (Brown, 2018) and thus scopolamine may have different effects on spike detection from calcium traces versus electrophysiological recordings. The discrepancy might also be a consequence of using different aversive stimuli (shock versus noxious odor), different degrees of fear acquisition and extinction during the task, or differences in the spatial layout of the maze. Rats in our experiment only received one or two self-initiated shocks during the shock sessions, thus a stronger conditioning paradigm might be necessary to induce place field migration. Another possible source of discrepancy could be different methods for normalizing speed and occupancy across experimental conditions. Aversive events cause lasting behavioral changes that alter the manner in which an animal samples its spatial environment (including reduced movement speed and avoidance of locations where aversive stimuli have been encountered). To control for confounding effects of these learning-induced behavior changes upon measures of place cell remapping, studies rely upon subsampling methods to normalize speed and occupancy distributions from experimental sessions conducted before versus after encounters with aversive events (Moita et al., 2003; Moita et al., 2004; Wang et al., 2012; Schuette et al., 2020). Here, we minimized confounds of behavioral sampling by restricting data analysis to ‘beeline trials’ during which rats ran highly stereotyped trajectories on a linear track, so that behavior was matched as closely as possible between sessions conducted before acquisition versus after extinction of avoidance learning. Using this approach, we replicated prior findings that aversive learning induces place cell remapping (Moita et al., 2003; Moita et al., 2004; Wang et al., 2012; Wang et al., 2015; Kim et al., 2015; Mamad et al., 2019; Schuette et al., 2020; Ormond et al., 2023) but found no change in the spatial distribution of place fields.

Acute effects of scopolamine

Significantly more place cell remapping occurred when aversive encounters were remembered (as assessed by 48 hr avoidance expression) after drug-free training than when they were forgotten after training on scopolamine. This suggests that place cell remapping might play a functional role in storing memories of aversive encounters, but potential confounds must be considered. A trivial explanation for our finding that scopolamine impaired both avoidance learning and place cell remapping could be that, at the dose given, the drug impaired rats’ awareness of their surroundings and their ability to perceive shock stimulus during the training session. This interpretation is not well supported by the data for several reasons. First, scopolamine did not acutely affect median running speeds during beeline trials (Figure 2C), consistent with previous demonstrations that 1 mg/kg scopolamine does not impair locomotion in rodents (Sun et al., 2021; Anagnostaras et al., 1999). Second, scopolamine did not acutely impair immediate avoidance of the shocked location during training (Figure 1F), consistent with prior evidence that mAChRs are not necessary for immediate shock avoidance (Decker et al., 1990). Third, under the influence of scopolamine, place cells retained stable within-session spatial tuning (Figure 2F–I) and exhibited shock-evoked responses similar to those observed in the drug-free condition (Figure 4A). By contrast, a recent calcium imaging study in mice (Sun et al., 2021) reported that tuning properties of place cells on a linear track were acutely disrupted by systemic scopolamine injections at the same dosage used here (1 mg/kg) (but see also Brazhnik et al., 2004). More consistent with our current findings, electrophysiology studies in rats have reported that systemic ACh antagonists alter temporal firing properties of place cells, such as their synchronization with hippocampal theta rhythm, while largely preserving their spatial firing properties (Douchamps et al., 2013; Newman et al., 2017). Conversely, cholinergic activation of the hippocampus acutely alters EEG states by promoting theta oscillations and suppressing sharp-wave ripples (Zylla et al., 2013; Vandecasteele et al., 2014; Hunt et al., 2018; Ma et al., 2020; Jarzebowski et al., 2021). Therefore, it is possible that theta oscillations and sharp waves were acutely altered by scopolamine during our experiments, but these signals were not detectable using our calcium imaging methods, so it was not possible to analyze these phenomena in the current study.

Although rats in our study were still able to locomote through the environment and keep track of their own location while on scopolamine, some acute effects of the drug upon behavior and place cell activity were observed. For example, while on scopolamine, rats showed a significant reduction of their previously learned preference for the short path (Figure 1F) prior to any delivery of shocks. This suggests that scopolamine may have affected the animals’ motivation to seek reward on the maze or impaired recall of previously learned strategies for finding the reward, consistent with prior evidence that scopolamine can impair navigational learning and decision-making (Nieto-Escámez et al., 2002; Vales and Stuchlik, 2005; Huang et al., 2011; Svoboda et al., 2017; Rashid and Ahmed, 2019; Hales et al., 2020). Although place cells continued to exhibit spatial tuning while rats were on scopolamine, between-session place cell recurrence probabilities were reduced (Figure 2G) and population vectors showed degraded between-session similarity with drug-free maze visits on prior days (Figure 2H and I); interestingly similar impairment of between-session place cell stability has been reported in aged rats (Barnes et al., 1997).

By selectively degrading between-session (but not within-session) stability of place cells, scopolamine may have prevented CA1 from retrieving accurate memory representations of the familiar maze environment during the training session. If so, then this could help to explain why rats acutely stopped preferring the short path while on scopolamine; expressing the learned preference for the short path may depend upon recalling stored maze representations that were previously active when the preference was learned. It may also help to explain why scopolamine prevented the aversive shock experience from inducing long-term remapping of place cells since failure to recall previously stored maze representations during training may have shielded those representations from modification by the aversive experience, and thus from remapping.

Effects of scopolamine on memory storage and retrieval

Many studies have shown that ACh receptor antagonist drugs can impair hippocampal memory processing in animal models (Decker et al., 1990; Anagnostaras et al., 1999; Atri et al., 2004; Hasselmo and McGaughy, 2004; Green et al., 2005; Hasselmo, 2006; Huang et al., 2011; Dannenberg et al., 2017; Solari and Hangya, 2018). Conversely, drugs that upregulate ACh neurotransmission can facilitate learning and memory (Barten and Albright, 2008; Digby et al., 2010; Ragozzino et al., 2012) and are commonly prescribed to treat aging-related amnesia and dementia in human patients (Lombardo and Maskos, 2015). Here, rats that received avoidance training on scopolamine were impaired when tested for retention 48 hr later (Figure 1H), consistent with prior studies showing that mAChRs are necessary for acquisition of aversive learning (Decker et al., 1990; Anagnostaras et al., 1995; Anagnostaras et al., 1999; Wallenstein and Vago, 2001). In conjunction with blocking avoidance learning, scopolamine also blocked place cell remapping between pre-training and post-extinction sessions (Figure 3D–H). Prior studies have similarly shown that scopolamine interferes with the ability of place cells to form and update cognitive maps across repeated visits to the same environment (Douchamps et al., 2013), but to our knowledge, the present study is the first to show that impairment of memory for an episodic event (in this case, an aversive shock) by scopolamine is accompanied by disruption of place cell remapping that normally occurs in response to such an event.

Since rats both acquired and extinguished avoidance behavior during the timespan over which remapping was measured, it was not possible to fully dissociate whether place cell remapping in drug-free rats was driven mainly by acquisition, extinction, or both. Rats trained on scopolamine did not undergo extinction (since they failed to exhibit long-term retention of avoidance behavior in the first place), so if remapping in drug-free rats was driven mainly by extinction, then scopolamine may have impaired remapping not by disrupting learning during the training session but instead by depriving rats of the extinction experience that would otherwise have caused remapping to occur after training. Prior evidence indicates that acquisition and extinction of conditioned fear can both contribute to place cell remapping (Moita et al., 2004; Wang et al., 2012; Wang et al., 2015), so it is likely that impairment of remapping by pre-training injections of scopolamine was at least partly caused by disruption of experience-dependent changes to the hippocampal population code during the training session.

Manipulations that impair ACh neurotransmission have been found to selectively impair storage but not retrieval in certain memory tasks, such as remembering semantic information in humans (Atri et al., 2004; Hasselmo and McGaughy, 2004) and fear learning in rodents (Decker et al., 1990; Huang et al., 2011). Hippocampal synaptic plasticity is likewise thought to be more important for storage than retrieval of hippocampal-dependent memory tasks (Martin et al., 2000). Hippocampal plasticity is sensitive to mAChR antagonists (Drever et al., 2011) and has been implicated in place cell remapping (Dragoi et al., 2003), so it is possible that in our study scopolamine impaired avoidance learning and place cell remapping by interfering with synaptic plasticity in the hippocampus, and thus preventing the storage of memories for the aversive shock experience. Some CA1 place cells are excited by shocks (Moita et al., 2003; Moita et al., 2004), and it has been shown that individual CA1 neurons can form new place fields at locations where they are artificially excited or inhibited (Bittner et al., 2015; Schoenenberger et al., 2016; Geiller et al., 2022; McKenzie et al., 2021). This raises the possibility that during acquisition of fear learning, shock-evoked responses of CA1 neurons might trigger synaptic plasticity that leads to place cell remapping. If scopolamine disrupted shock-evoked hippocampal plasticity, it likely did so by impairing mechanisms downstream from neural responses to shock since we found that scopolamine had no effect on shock-evoked responses of CA1 place cells and shock-induced remapping was similarly prevalent in subpopulations of shock-responsive and non-responsive place cells.

Since the drug was given systemically, scopolamine may have interfered with plasticity not only in the hippocampus but also in other brain regions that contribute to the storage of fear memories, such as the amygdala, where associations between predictive cues and aversive events are thought to be stored by synaptic plasticity that is both triggered by noxious stimuli (Blair et al., 2001; Grewe et al., 2017; d’Aquin et al., 2022) and susceptible to cholinergic modulation (Izquierdo et al., 1992; Jiang et al., 2016). Regardless of whether scopolamine disrupted place cell remapping via its effects upon synaptic plasticity or other mechanisms (such as retrieval of previously stored maze respresentations; see above), or whether the drug acted upon pharmacological targets inside versus outside of the hippocampus (or both), our findings indicate that the drug disrupted shock-induced place cell remapping and thus support the idea that reorganization of the hippocampal place code occurs in conjunction with hippocampal storage of episodic memories for aversive events. Where and how scopolamine acts in the brain to impair remapping is a worthwhile question for further research that could help to elucidate the broader role of acetylcholine in memory storage and retrieval.

Implications for brain disease and aging

If remapping is a mechanism for memory, then disruption of remapping by systemic cholinergic antagonists may serve as a useful animal model for memory loss accompanied by degeneration of cholinergic signaling in dementia (Wenk et al., 1980; Arendt et al., 1983; Hardy, 2006; Schliebs and Arendt, 2011). Degradation of cholinergic signaling also occurs with alcohol abuse disorder (Arendt, 1994; Miller and Kamens, 2020). By further investigating how the cholinergic neuromodulatory system interacts with the hippocampus, we may develop a deeper understanding of how memory systems are impacted by a variety of neurodegenerative insults and make progress toward improved therapies for memory diseases.

Like humans, rodents show memory deficits with aging (Barnes, 1979; Barnes et al., 1980). Interestingly, prior studies have shown that place cell population dynamics in aged rats are altered in ways similar to those observed here under scopolamine. For example, we found that scopolamine degraded between- but not within-session stability of the CA1 population code, which has also been observed in aged rats (Barnes et al., 1997). This is in contrast to other brain diseases, such as epilepsy, which degrades both within- and between-session place coding (Shuman et al., 2020). Additionally, we found that in rats that showed impaired avoidance learning after pre-training injections of scopolamine, the drug concurrently prevented CA1 place cells from remapping during learning and extinction (Figures 3 and 4). Place cells in aged rodents have likewise been reported to show deficits in experience-dependent modification of their tuning properties (Oler and Markus, 2000; Wilson et al., 2003; Wilson et al., 2004; Tanila et al., 2018). These parallels between the effects of aging and scopolamine upon place cells suggest that mAChR blockade and aging may produce similar memory deficits by causing similar impairments to hippocampal population stability. Further study is warranted to investigate whether such impairments of hippocampal population coding can be reversed using anti-amnestic drugs, such as those that upregulate ACh transmission or other novel compounds. By identifying interventions that can improve the long-term stability of hippocampal population coding in animal models, it may be possible to make progress toward improved therapies for memory deficits associated with aging and neurodegenerative disease.

Methods

All experimental procedures were approved by the Chancellor’s Animal Research Committee of the University of California, Los Angeles, in accordance with the US National Institutes of Health (NIH) guidelines, protocol #2017-038. Raw data files and Matlab analysis code are openly available at https://github.com/tadblair/tadblair or https://doi.org/10.5068/D1ZT2S.

Subjects and apparatus

15 Long–Evans rats (eight females, seven males) acquired from Charles River at 3 months of age were used in this experiment. Subjects were singly housed within a temperature and humidity controlled vivarium on a 12 hr reverse light-cycle. Beginning at 4 months of age, rats were reduced 85% of their ad lib weight (450–550 g) through limited daily feeding, and then underwent virus injection and GRIN lens implantation surgery (see below). After 5 d recovery from surgery, rats began a regimen of behavior and recording sessions on the rectangular maze. The maze was constructed from a 4″ wide track composed from metal grid bars along the full length of the long and short paths. The track was elevated 20″ from the floor. There were extramaze cues in the surrounding environment (doors, shelving, checkerboard poster on the wall) but no local cues on the maze to indicate the shock zone’s location. During shock training sessions (drug-free or scopolamine), a 1.0 mA current from a grid scrambler (Med Associates Corporation, Fairfax, VT) was switched on to electrify grid bars in a 50 cm segment of the short path’s center (Figure 1A). The exact duration of the rat’s exposure to shock varied with the length of time it spent in contact with the electrified bars as it ran across or retreated from the shock zone, but was typically very brief (on the order of 1 s or less).

Figure 1B shows the order in which experimental manipulations were given to different groups of rats. Six rats (three males, three females) received drug-free shock training as their first manipulation after reaching criterion for preferring the short path. Of these, five rats (three males, two females) then received extinction sessions and ended the experiment. The remaining female rat was mistakenly given shock at the start of the drug-free training session (rather than after 10 min) and was thus dropped from imaging analysis of the drug-free shock condition, but after extinction, this rat subsequently received scopolamine shock training and was thus included in analysis for that condition (see below). Four rats (two males, two females) were given their drug-free shock sessions after previously receiving barrier training as their first manipulation (so these rats were still naive to shock during drug-free shock avoidance training). Three rats (one male, two females) were given their drug-free shock session after previously receiving a scopolamine shock session (followed by extinction) as their first manipulation, and two rats (one male, one female) were given their drug-free shock session after previously receiving both barrier training and a scopolamine shock session (followed by extinction). Among rats that were given scopolamine shock sessions, five (two males, three females) had not previously experienced shock at the time of their scopolamine shock avoidance training session, whereas five rats (three males, two females) had previously received drug-free shock avoidance training followed by extinction. To compare imaging results across training conditions, rats from the drug-free condition were only included if they ran <60% as many trials during their first post-extinction session as during their final pre-training session (indicating that they had successfully acquired short path avoidance), whereas rats from the scopolamine shock condition were only included if they ran >60% as many trials during their first post-extinction session as during their final pre-training session (indicating that scopolamine had successfully blocked avoidance learning).

Surgical procedures

Subjects were given two survival surgeries prior to behavior training in order to record fluorescent calcium activity from hippocampal CA1 cells. During the first surgery, rats were anesthetized with 5% isoflurane at 2.5 L/min of oxygen, then maintained at 2–2.5% isoflurane while a craniotomy was made above the dorsal hippocampus. Next, 1.2 uL of AAV9-Syn-GCamp7s (AddGene) was injected at 0.12 uL/min just below the dorsal CA1 pyramidal layer (–3.6 AP, 2.5 ML, 2.6 DV) via a 10 uL Nanofil syringe (World Precision Instruments) mounted in a Quintessential Stereotaxic Injector (Stoelting) controlled by a Motorized Lab Standard Stereotax (Harvard Apparatus). Left or right hemisphere was balanced across all animals. One week later, the rat was again anesthetized and four skull screws were implanted to provide stable hold for the subsequent implant. The viral craniotomy was reopened to a diameter of 1.8 mm, and cortical tissue and corpus callosal fibers above the hippocampus were aspirated away using a 27- and 30-gauge blunt needle. Following this aspiration, and assuring no bleeding persisted in the craniotomy, a 1.8-mm-diameter Gradient Refractive INdex lens (‘GRIN lens,’ Edmund Optics) was implanted over the hippocampus and cemented in place with methacrylate bone cement (Simplex-P, Stryker Orthopaedics). The dorsal surface of the skull and the bone screws were cemented with the GRIN lens to ensure stability of the implant, while the dorsal surface of the implanted lens was left exposed.

Two to three weeks later, rats were again placed under anesthesia in order to cement a 3D-printed baseplate above the lens. First, a second GRIN lens was optically glued (Norland Optical Adhesive 68, Edmund Optics) to the surface of the implanted lens and cured with UV light. The pitch of each GRIN lens is 0.25, so implanting 2 together provides a 0.5 pitch. This half pitch provides translation of the image at the bottom surface of the lenses to the top while maintaining the focal point below the lens. This relay implant enables access to tissue deep below the skull surface. Magnification of the image is performed by the miniscope. The miniscope was placed securely in the baseplate and then mounted to the stereotax to visualize the calcium fluorescence and tissue. The baseplate was then cemented in place above the relay lenses at the proper focal plane and allowed to cure. Once rats had been baseplated, they were placed on food restriction to reach a goal weight of 85% ad lib weight and then began behavioral training. Time between the beginning of the surgical procedures and the start of behavior training was typically 6–8 wk.

Scopolamine Injections

On some training days rats were injected intraperitoneally with vehicle (0.9% sterile saline) or scopolamine (1 mg/kg dissolved at a concentration of 1 mg/mL) prior to shock sessions. After the injection, rats were returned to their home cage for 30 min (located on a cart in the experiment room) for 30 min prior to starting the session.

Histology

At the end of the experiment, rats were anesthetized with isoflurane, intraperitoneally injected with 1 mL of pentobarbital, then transcardially perfused with 100 mL of 0.01 M PBS followed by 200 mL of 4% paraformaldehyde in 0.01 M PBS to fix the brain tissue. Brains were sectioned at 40 µm thickness on a cryostat (Leica), mounted on gelatin prepared slides, then imaged on a confocal microscope (Zeiss) to confirm GFP expression and GRIN lens placement (see examples in Figure 2—figure supplement 1).

MiniLFOV calcium imaging system

To record calcium activity during unrestrained behavior, we utilized a large-field-of-view version of the Miniscope imaging system, ‘MiniLFOV’ (Guo et al., 2023). This open-source epifluorescence imaging camera weighs 13.9 g with a 3.6 × 2.3 mm field of view while maintaining a 2.5 um resolution at the center of view and 4.4 um at the periphery. The MiniLFOV can record at 22 Hz, has a 5 MP CMOS imaging sensor, an electrowetting lens for digitally setting the focal plane, and a modular lens configuration to enable a longer working distance (either 1.8 mm, used here, or 3.5 mm). The system is 20× more sensitive than previous v3 Miniscopes and twice as sensitive as current v4 Miniscopes. Power, communication, and image data are packed into a flexible 50 Ω coaxial cable (CW2040-3650SR, Cooner Wire) using power-over-coax filtering and a serializer/deserializer pair for bidirectional control communication with I2C protocol and unidirectional high bandwidth data streaming. The MiniLFOV device interfaces with UCLA open-source Miniscope DAQ Software to stream, visualize, and record neural dynamics and head orientation data. This DAQ platform and software allow for excitation intensity adjustment, focus change, image sensor gain selection, and frame rate setting. See the MiniLFOV website (github.com/Aharoni-Lab/Miniscope-LFOV) and methods paper (Guo et al., 2023) for further details and printable part files.

Trace extraction and spike inference

To accelerate image processing, each session’s video stack was cropped to a rectangle that excluded edge regions containing no neural activity in any session from that animal. The stack was also temporally downsampled to every other frame, yielding an effective sample rate of ~11 Hz. Non-rigid motion correction was applied to remove residual non-uniform tissue motion artifacts (Pnevmatikakis and Giovannucci, 2017). Source extraction was then performed using the Python implementation of the Calcium Imaging Analysis package, ‘CaImAN’ (Giovannucci et al., 2019), yielding individual spatial contours and demixed temporal fluorescence traces for each detected neuron; parameters used for source extraction can be found in scripts provided at https://github.com/tadblair/tadblair/tree/Blair_et_al. Deconvolved spikes were derived from denoised fluorescence traces by CaImAN ‘s ‘deconvolveCa’ function using a second-order autoregressive model with automated estimation of baseline and convolution kernel parameters. To match cell contours across sessions, spatial contour weights from CaImAn were thresholded to 50% of their peak value to generate binary pixel masks that were then analyzed using CellReg (Sheintuch et al., 2017) for matching contours across all sessions included in the analysis for each rat.

Behavior tracking

A webcam mounted in the behavior room tracked a red LED located on the top of the miniscope, and this video was saved alongside the calcium imaging via the miniscope software with synchronized frame timestamps. These behavior video files were initially processed by custom Python code, where all the session videos were concatenated together into one tiff stack, downsampled to 15 frames per second, the median of the stack was subtracted from each image, and finally they were all rescaled to the original 8-bit range to yield the same maximum and minimum values before subtraction. Background subtracted behavior videos were then processed in MATLAB. The rat’s position in each frame was determined using the location of the red LED on the camera. Extracted positions were then rescaled using an inverted model of the empirically measured camera distortion and convert the pixel position to centimeters according to the maze size. Positional information was then interpolated to the timestamps of the calcium imaging video.

Behavioral performance measures and criteria

During the initial phase of training where rats learn to prefer the short path over the long path on the maze, rats were required to reach a criterion of earning at least two rewards per minute by taking the short path and earning at least twice as many rewards on the short versus long path during a single 15 min session. After reaching this criterion, rats were given avoidance training in the next session as explained in ‘Results.’ After a single-avoidance training session, rats received standard (non-shock) sessions until avoidance behavior was extinguished. Rats were considered to have extinguished avoidance behavior on the first post-training day that they recovered to the original criterion (defined above) for preferring the short path.

48 hr retention of short path avoidance was quantified by the ratio A=N+48/N-48 , where N+48 and N-48 are the number of rewards earned on the short path during the first session after and before the training session, respectively. Hence, lower values of A indicated better 48 hr retention of short path avoidance. In the drug-free shock condition, an individual rat was considered to show failed 48 hr retention of short path avoidance if A > 0.6, which was the cutoff for the top quartile of A values in all rats (n = 11) from which imaging was successfully obtained during the drug-free shock condition (of the 14 rats that received drug-free shock training, one was omitted from imaging analysis because shocks were mistakenly delivered at the start of the session, and two were omitted because they failed to ever reach criterion for extinction of avoidance behavior after multiple post-training sessions). In the scopolamine shock condition, an individual rat was considered to show successful 48 hr retention of short path avoidance if A < 0.31, which was the cutoff for the bottom quartile of A values in all rats (n = 9) from which imaging was successfully obtained during the scopolamine shock condition (of the 10 rats that received scopolamine shocks, 1 was omitted from imaging analysis because the GRIN lens shifted and imaging was lost prior to the avoidance training session).

Behavioral downsampling

Since behavior on the maze was altered by training and drug administration (Figure 1), imaging analyses only included data from beeline trials during which rats ran from one end of the track to the other with no change in direction (a beeline trial was defined as a continuous traversal of the short path throughout which the rat’s 1D position along the track changed monotonically in one direction). This assured that rats were behaving as similarly as possible during analyzed data segments from all sessions. Beeline trials from each session were then further downsampled prior to imaging analyses to control for potential effects of behavior or sampling differences on the outcome of imaging analyses.

For comparisons of pre-training versus training sessions (Figure 2C–H) in the drug-free versus scopolamine shock conditions, it was necessary to correct for two imbalances in behavior between the drug-free versus scopolamine conditions (Figure 2—figure supplement 2A): (1) rats ran more slowly (and thus also ran fewer beeline trials) during training sessions on scopolamine than drug-free, and (2) rats ran more pre-training trials per session in the scopolamine than the drug-free condition. To correct for imbalance #1, we computed that the mean number of beeline trials (in the LR and RL directions combined) during training in the scopolamine condition (23.85 trials) was 64% of mean beeline trials in the drug-free condition (37.22 trials); since running speed was also slower for scopolamine training, we sorted each rat’s drug-free training beeline trials in order of running speed and omitted the fastest 36% of trials from each running direction (rounded to the nearest integer). This simultaneously equalized beeline trial counts and running speeds during training sessions between the drug-free versus scopolamine conditions. To correct for imbalance #2, we computed that the mean number of beeline trials during pre-training (in the LR and RL directions combined) in the drug-free condition (53.78 trials) was 71% of mean beeline trials in the drug-free condition (76.0 trials); we omitted the last 29% of trials from each running direction in the pre-training session. This equalized trial counts across training conditions without significantly affecting mean running speeds. Also, since omitted pre-training trials were removed from the end of the session, the effective duration of pre-training sessions was reduced from 15 min down to a shorter duration that was more similar to the 10 min pre-shock period during which beeline trials were sampled from the training session. After these behavioral downsampling measures had been taken, rats in both conditions still ran fewer beeline trials during the analyzed portion of the training session (the first 10 min prior to shock delivery) than during the included portion of the pre-training session as a consequence of the fact that scopolamine rats ran fewer trials in the training session and this low trial count had to be matched in the drug-free group. But the difference in trial counts between pre-training and training sessions, though large, was similar for the drug-free and scopolamine training conditions, and was therefore assumed not to be a confound in analyses of the training condition effects (Figure 2C–H).

For comparisons of pre-training versus post-extinction sessions (Figure 3), it was necessary to correct for two imbalances in behavior between the drug-free versus scopolamine conditions (Figure 2—figure supplement 2): (1) rats in the drug-free shock condition tended to run fewer laps at lower speeds during the post than pre session (presumably because they still had some residual fear of the short path even after reaching the extinction criterion), whereas rats in the barrier condition tended to run more laps at higher speeds during the post than the pre session (presumably because they continued to acquire a preference for the short path without being interrupted by avoidance learning), and (2) during both the pre and post sessions, rats in the scopolamine shock condition ran more short path beeline trials than rats in the other two conditions. Imbalance #1 was corrected in two steps. The first step was to take the difference between the number of pre and post session trials for each rat in the drug-free condition: NdiffDF=NpreDF-NpostDF . Trials from the pre session were then sorted by speed, and the fastest NdiffDF trials were removed from the pre session, which equalized the number of pre and post session beeline trials while also making their mean running speeds much more similar, so that there was no longer a significant running speed difference (Figure 3B, right). The second step was to take the difference between the number of pre and post session trials for each rat in the barrier condition: NdiffBAR=NpostBAR-NpreBAR . Trials from the post session were then sorted by speed, and the fastest NdiffBAR trials were removed from the post session, which equalized the number of pre and post session beeline trials while also making their mean running speeds much more similar, so that there was no longer a significant running speed difference (Figure 3B, right). Then an additional 20% of the fastest trials were removed from both sessions to render the mean number of beeline trials per session similar across the drug-free and barrier conditions. To correct for imbalance #2, the absolute difference was taken between the number of pre and post session trials for each rat in the drug-free condition: NdiffSCOP=NpreSCOP-NpostSCOP ; trials from the scopolamine shock condition were then sorted by running speed, and the number of pre and post trials was equalized by removing fastest NdiffSCOP trials from whichever session had fewer trials in a given rat. Then an additional 25% of the fastest trials were removed from both sessions to render the mean number of beeline trials per session similar across the drug-free and scopolamine shock conditions.

Spatial tuning properties

To analyze acute effects of scopolamine on spatial tuning properties, we measured each peak Af/s rate, out-of-field Af/s rate, tuning width, and spatial information content from place cell tuning curves. Peak Af/s rate was defined as the maximum Af/s rate observed for any of the 23 bins in the tuning curve. Out-of-field Af/s rate was defined as the mean rate in the subset of spatial bins that had rates <50% of the peak Af/s rate. Tuning width was computed as W×N50 , where W=10.8 cm is the width if each bin and N50 is the number of bins that had rates >50% of the peak Af/s rate. Spatial mutual information content was measured in units of bits per active frame (bits/Af), which was computed as (Skaggs et al., 1993)

rB_log2(rB_)+i=1Bripilog2(ri)

where B is the set of spatial bins with non-zero mean Af/s rates, ri is the mean Af/s rate in spatial bin i, pi is the probability that the rat was occupying bin i, and rB_ is the mean Af/s rate over all bins in the set B.

Selection of place cells

Stable place cells were identified in each imaging session by randomly splitting the session’s beeline trials (separately for each running direction) into two halves, and then computing two spatial tuning curves, one for each randomly chosen half of the beeline trials. A Pearson correlation coefficient (R) between these two tuning curves was then computed, and its p-value was recorded. A cell was classified as a place cell in a given session if the median p-value for R was <0.01 over 200 iterations of random beeline splits in either of the two running directions. To be classified as a place cell, a neuron also had to generate an average of at least one inferred spike per beeline trial and was not permitted to fire at >70% of its peak inferred spike rate over more than 50% of the track (this was done to exclude likely interneurons and thereby restrict analysis to putative pyramidal neurons).

Recurring place fields

A recurring place field was defined as an LR or RL tuning curve that (1) met the spatial selectivity criteria described above during at least one of the two sessions and (2) contained active frames (i.e., generated at least one inferred spike) during subsampled beeline trials in the specified running direction from both sessions. Place cell tuning curves were excluded from the between-session heatmap pair if they contained non-recurring place fields, defined as those that (1) met spatial selectivity criteria in during one of the two sessions, but (2) generated no inferred spikes (and thus went completely undetected) during subsampled beeline trials in the specified running direction from the other session.

Shock responses

To analyze shock-evoked responses of CA1 place cells, we identified each frame during which the rat made an entry into the electrified shock zone and noted in which direction the rat was running when shock onset began. For each entry into the shock zone, a significant shock-evoked response occurred when the count of active frames in a 2 s window after entry into the electrified shock zone exceeded the confidence limit (p<0.01) for a Poisson fit to the baseline distribution of active frame counts from prior visits (during the first 10 min of the session) to unelectrified shock zone in the same running direction. Most rats experienced the shock in both running directions before they began to avoid the short path, but a few rats only experienced the shock in one direction, and in a small number of rats, more than two shocks were received before the rat began avoiding the short path. To equalize statistical power for detecting shock responses across all rats (regardless of the shocks they received), a neuron was classified as shock-responsive only if it exhibited a significant shock response during the rat’s first or second visit (or both) to the electrified shock zone; later visits to the shock zone were not considered, so that place cells would not have a higher likelihood of being classified as shock-responsive in rats that made more visits to the shock zone.

To compute population averaged responses of place cells to shocks, only the first significant shock-evoked response of each shock-responsive place cell (from either the first or second shock encounter, whichever yielded the first significant response) was included in the average, so that each cell’s contribution to the average always came from a single significant shock response trial, and therefore did not vary depending upon the number of shocks different animals received. For shock-responsive place cells, the population average included only the first shock encounter (which by definition failed to elicit a significant shock response), so that again, each cell’s contribution to the average came from a single trial.

Distributions of tuning curve peak locations (Figure 4A, right) for shock-responsive place cells included only tuning curve peaks for the running direction in which the rat received its first significant shock-evoked response (from either the first or second shock encounter, whichever yielded the first significant response). Distributions of tuning curve peak locations for shock-responsive place cells included tuning curve peaks from both running directions if the rat received its first and second shock in different directions, but if the first and second shock were both received in the same running direction, then only the tuning curve peak for that direction was included in the distribution of peak locations.

To assess whether a place cell’s tuning curve peak was on the approaching versus departing side of the shock zone, the 23-bins of the tuning curve were reflected across the middle (bin 13) for RL tuning curves, but not for LR tuning curves, so that for both running directions, tuning curve peaks in bins 1–12 were on the approaching side of the shock zone, and peaks in bins 14–23 were on the departing side of the shock zone. For tuning curves with peaks in the center (bin 13), an additional step was taken to measure the percentage of the area under the tuning curve on either side of center. If the area in bins 1–12 was greater than in bins 14–23, then the tuning curve peak was considered to be on the approaching side; otherwise, it was considered to be on the departing side.

Position decoding

For decoding analysis, frame positions in one imaging session were predicted using place cell heatmaps derived from a different session. For each analyzed pair of pre- and post-sessions, decoding was performed using memoryless Bayesian inference (Davidson et al., 2009; Zhang et al., 1998) with place maps constructed from each running direction. Post-training positions were predicted from pre-training heatmaps (pre→post), and vice versa (post→pre), taking the maximally likely position bin at each frame, and results from both directions were averaged together to obtain the results in Figure 3G. A subsampling procedure was used to equalize the number of place cells used for each running direction (LR and RL, 100 subsample iterations), so that any observed differences in decoding performance were not attributable to differences in the sizes of the population vectors (i.e., the number of recurring place cells) that were used for pre-training versus cross-training decoding.

The decoding error (DE) in each frame was computed as the absolute error (in cm) between the rat’s actual versus predicted position in that frame, DE = abs(xact-xpred). The shuffled decoding error (DEshuffle) was derived in the same manner, except that place cells in the population vector were randomly permuted against their place maps before deriving the predicted position, xpred. Shuffled decoding error was computed multiple times per frame using different random place cell permutations (n = 2000 shuffles). The decoding error in a given position bin was measured as the median of the absolute difference (in cm) between the rat’s actual versus predicted position over all visits to that bin. The non-shuffled decoding error was subtracted from the median DEshuffle to obtain the accuracy gain in each frame, G = median(DEshuffle)-DE, which measured by how many centimeter the decoding error was reduced when position was predicted from non-shuffled rather than shuffled population vectors. The median value of G was derived for all frames in each actual position bin, x, yielding an accuracy gain measurement at each position on the track, G(x) = median(Gx1, Gx2, … Gxn). Accuracy gain was computed separately at each position in each running direction (LR and RL) and each decoding direction (pre→post and post→pre) to obtain four different G(x) values per session pair: GprepostLRx , GprepostRLx , GpostpreLRx , and GpostpreRLx . For statistical analyses, each of these four values was treated as an independent observation at each position. Accuracy gains tended to be greater near the ends of the track than in the middle, but this was mainly a consequence of bias in the G(x) accuracy measure, rather than better decoding at the ends of the track, because greater reductions of G(x) were attainable near the ends of the track, where xact-xpred could span the entire track length, than in the middle, where xact-xpred could only span half the track length at most. To eliminate this positional bias, we used Wilcoxon signed-rank tests (rather than subtraction) to test whether the median error difference, DEshuffle-DE, differed significantly from zero (see ‘Methods’). p-Values from the Wilcoxon tests provided a measure of decoding performance that was unbiased by track position, but these p-values were non-normally distributed, and thus could not be analyzed to test for interaction effects using parametric statistics. We therefore took the negative log of the Wilcoxon p-value and multiplied by the sign of G to derive a decoding accuracy index, DAI = –log10 (P) × sign(G) – 2, which was both normally distributed and unbiased by track position. Subtracting 2 at the end of the expression was done so that DAI = 0 would correspond to p=0.01 for more significant decoding from unshuffled than shuffled population vectors since log10 0.01 = 2.

Data availability

Source data and code for reproducing the figures are available at: https://github.com/tadblair/tadblair or https://doi.org/10.5068/D1ZT2S.

The following data sets were generated
    1. Blair GJ
    2. Guo C
    3. Wang S
    4. Fanselow MS
    5. Golshani P
    6. Aharoni D
    7. Blair HT
    (2023) Dryad Digital Repository
    Hippocampal place cell remapping occurs with memory storage of aversive experiences.
    https://doi.org/10.5068/D1ZT2S

References

    1. Barnes CA
    2. Nadel L
    3. Honig WK
    (1980) Spatial memory deficit in Senescent rats
    Canadian Journal of Psychology / Revue Canadienne de Psychologie 34:29–39.
    https://doi.org/10.1037/h0081022
  1. Book
    1. O’Keefe J
    2. Nadel L
    (1978)
    The Hippocampus as a Cognitive Map
    Oxford University Press.
  2. Conference
    1. Skaggs WE
    2. McNaughton BL
    3. Gothard KM
    4. Markus EJ
    (1993)
    An information-theoretic approach for Deciphering the hippocampal code
    Advances in Neural Information Processing Systems.

Decision letter

  1. Liset M de la Prida
    Reviewing Editor; Instituto Cajal, Spain
  2. Laura L Colgin
    Senior Editor; University of Texas at Austin, United States
  3. Liset M de la Prida
    Reviewer; Instituto Cajal, Spain
  4. Alexandra T Keinath
    Reviewer; University of Illinois at Chicago, United States

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Disruption of place cell remapping by scopolamine during aversive learning" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Liset M de la Prida as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Laura Colgin as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Alexandra T Keinath (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

This paper describes results obtained from multi-cellular imaging of CA1 cells using large-field-of-view miniscopes in rats performing a shock avoidance task. By exploiting behavioral (barriers) and pharmacological (scopolamine) manipulations, the authors explore cell remapping dynamics during aversive learning. We very much value using large-field-of-view miniscope imaging to provide place cell data from rats, thus favoring inter-species comparisons. We also appreciate the inclusion of an appropriate set of control analyses and experiments to exclude potential confounds, especially when it comes to comparison between groups. While all reviewers agreed that the work will be of interest, there were also some concerns requiring substantial revision.

A point of discussion was whether the results of scopolamine are novel and specific enough to support claiming a direct effect on memory processes. There were also concerns regarding additional methodological controls which are required. We feel they can be addressed with additional analyses and experiments. In a consultation session, we agreed the following revisions are required before a definitive recommendation can be made:

1) Control experiments for scopolamine are required. These include:

a) A control is needed to show that the injection procedures do not impair general behaviour (e.g. motivation, attention, level of stress) as well as other forms of learning;

b) Consider effects in the absence of shock learning. To be able to claim that the effects of scopolamine are specific to aversive learning, a control with either no learning or appetitive learning (e.g. of a reward location, which also involves place field reorganization in some cases) would be useful. We feel that acknowledging the systemic/brain-wide/unspecific effect of scopolamine is not enough and you should provide some additional evidence of specificity.

c) Consider behavioral side effects of a systemic approach. Similarly, consider that functional effects of scopolamine (power and coordination of theta oscillations brain-wide) can additionally provide an explanatory value of the effects. Some of these are discussed but some others need support.

2) Additional control analyses are required, including:

a) Population vector analysis should be tested against shuffled distributions.

b) Population-level analysis in terms of the percentage of cells meeting place-cell criteria across manipulations; the equivalency of place cell measures when including all principal cells, not just those meeting (same-session) pre-selection criteria.

c) There are also some issues with the different pre-/post-training timing (8 days vs 6 days after pre-training) used in remapping effects with/without scopolamine that require clarification.

3) Clarification regarding LFOV:

a) Using large GRIN lenses may bring undesired histological damage and we wonder whether flattening the dorsal hippocampus may have consequences on functional readouts.

b) Given improvements in LFOV imaging with your approach, we wonder whether center/periphery ROI imaging may be or may not be affected. Given individual variability of imaging ROIs per rat, we would like to see this more carefully addressed.

4) Regarding impact and novelty, you may consider the following suggestions:

a) We discussed whether the above revisions may offer a chance to reinforce the novelty of your observations, by presenting pharmacological and behavioral (barriers) experiments as manipulations to better focus on the study of place cell remapping dynamics during aversive learning in rats. Actually, the abstract poorly integrates barrier results, which we feel should be mentioned. The authors may consider whether by highlighting/refocusing around some of these angles their manuscript would gain additional impact and broader interest.

b) Consider also tuning the title accordingly if required. Please, note that the title and/or the abstract should provide an indication of the biological system under study.

c) We also discussed whether using LFOV imaging in rats is an important addition to your paper and we would ask you to consider expanding the value of the approach when it comes to imaging from larger brains such as in rats versus mice. We feel in addressing questions regarding GRIN lenses and image aberration you may find the opportunity to highlight this better.

Please see the comments from each reviewer for more details.

Reviewer #1 (Recommendations for the authors):

Here is a list of comments for the authors to improve their work:

1. The body of the analysis is centered on the evaluation of population vector correlation matrices between behavioral phases and conditions. To build them, the authors rely on tuning curve heatmaps calculated from place cells in a reference condition (pre). One major problem with population tuning curves is that their typical 'diagonal' shape is very sensitive to noise so all curves and derived analysis should be tested against shuffled distribution as well. This affects data in Figure 2H,I; Figure 3D-F; Figure 4B,C.

2. The authors nicely control for their main figure analysis using the subset of scopolamine-treated rats that retained avoidance, but they do not provide data on the similarly important group of drug-free rats that DO NOT retain avoidance. They may help to provide additional controls for all main figure analyses as well. In addition, the paper would benefit from examining whether remapping indices can explain the behavioral performance of all rats, without resorting to categorizing them by performance level (i.e. using correlation analysis and/or predictive model analysis).

3. Using large GRIN lenses for large-field-of-view is a major strength of the paper when applied to imaging from the rat hippocampus. However, this can yield undesired histological damage. For instance, pushing the lens against the alveus for image flattening results in a reduction of the stratum oriens and atrophy of basal dendrites. This is inevitable and should not be considered by the authors as a criticism, but worth mentioning/controlling for effects. Are any of the behavioral and/or functional readouts correlated with some histological damage?

4. Similarly, there is natural variability in imaging from cells at the center/periphery of the GRIN, which differ in aberration effects. Since there is a different contribution across animals another useful control is to evaluate the consistency of conclusions for data obtained from center/periphery within and between animals. This would also help to set standards for the field, especially for LFOV imaging. Importantly, there is poor information on how the ROIs are defined, what methods are used for segmentation and how imaging from different sessions are treated (same ROIs, different ROIs, etc…), and what signals are evaluated dF/F with/without denoising, models of spike inference, etc…

Reviewer #2 (Recommendations for the authors):

Here are some detailed cpmmrnts.

1) Please add page and line numbers to ease review.

2) Given that several of the rats did the drug-free shock learning twice, how were these data analysed? Was only one of the sessions per rat chosen to contribute to the analysis, if yes, how?

3) Was there an influence of the order of sessions on the results (e.g. scopolamine during the first shock session, the second one, after a barrier session or not)?

4) Any analysis of the number of trials, if comparing across sessions of different durations, should be done on a number of trials/minute to allow for cross-sessions comparisons (e.g. Sup Figure 2 A).

5) Analysis of remapping: remapping is not the same as instability. It corresponds to a consistent activation of place fields in specific locations, that change (in a consistent manner) in a different condition, with the activity pattern, mostly reinstated when coming back to the original condition. Otherwise, changes in place cell firing that are not consistent or reproducible are considered drift or instability of the spatial firing. It is unclear if the current paradigm and analysis tests remapping and not simply instability.

6) In the 'Scopolamine does not disrupt place cell responses to shock' section: individual place cells are considered to 'respond' to the shock if their activity increases during entry into the shock zone. However, place cells can also respond by decreasing their activity, or even changing their place field shape or location, but this is not analysed.

7) Given that place cells can modulate their firing spontaneously with time (mentioned by the authors), any analysis of place cells between two different conditions where time has elapsed should have as a control a similar comparison, without the parameter of interest being changed, but with approximately the same amount of time passed or at least number of sessions (as the authors do in one of their analyses).

8) Nomenclature: In general, it would be good to avoid the use of abbreviations as much as possible, including in the figures, to help readability. Also, the different terms used (mostly condition or session names) can be confusing and there may be errors in figures or legends that make the interpretation difficult. The use of pretraining, training, and post in some instances, but pre10 post 5 in others (within the same session), is also a bit confusing. Perhaps the 10 min part of the shock training sessions that is pre-shock should be called the normal session and then the shock session would be called the shock session (or barrier session), without lumping both in the same session. Also, there are 2 forms of training: training to choose the shortest path, then training of the shock – and then extinction of the shock learning. Perhaps the authors could find a more intuitive and consistent way to name their sessions. What is 'post-extinction' (e.g. in 'post-extinction CA1 population vectors)? Is it referring to the activity during the extinction sessions? Or during the sessions that followed the extinction sessions, which according to Figure 2B will always be a shock session? It seems that this is sometimes called post-training (e.g. Figure 1J) and sometimes post-extinction (e.g. Figure 1J, legend). Is the place field recurrence rate (Figure 2G) the same as place field recurrence ratio (in results)? Do subsampled and down sampled mean the same here?

9) Related to this, in several instances the axis labels or legends of a figure are not self-sufficient and could be enhanced with more information about the measure that is shown (e.g.: Figure 1J: how is %baseline computed?).

10) The graph in Figure 1B seems to show that after each shock session, the rats have 1 extinction session. But in the main text, it is said that 'rats continued to run standard sessions on the maze every 48h until they reached the avoidance extinction criterion", which seems to assume that this would take several sessions. It seems that the graph should be adapted to reflect how many sessions were run. It is not clear what happens after the barrier sessions – do the rats move on to the shock learning sessions immediately? Does the '3s' mean that they do instead of the barrier session 3 times, or maybe extinction sessions?

11) It is a bit unclear what the -48/+48 measure means on some of the figure axes and in text. Does that mean that what is compared is data from 4 days apart? Or just 2 days apart? If it's the latter, shouldn't the measure be 0-48h or -24/+24?

12) Rejecting specific trials from the analysis does not change the behaviour of the rats during data collection; it only makes the analysis more balanced. The sentence "This assured that rats were behaving as similarly as possible during data collection in all sessions" should be rewritten to reflect this.

13) Figure 2E could show proportions out of the total number of putative pyramidal cells analysed, not just out of detected place cells, to see if the percentage of active place cells changes.

14) Tuning curve 'sequences' (showing the tuning curves of all cells ordered according to peak firing) are used in many of the plots, but it is unclear what information to take from them. Reorganizing the figures to emphasize the quantification of the results might be useful and perhaps removing some of the sequence plots unless they convey a specific message or can be used as an example.

15) Related to this, most of the figures are really small. Focusing on the ones that provide quantitative results (i.e. less 'tuning curves' plots) and making those bigger would probably help the reader.

16) Descriptions of some specific procedures could be added in the methods (listed in no specific order):

a) How were the beeline trials selected?

b) How were putative spikes extracted/reconstructed from the calcium imaging signal (deconvolution?)?

c) How was the injection procedure performed, and where were the rats kept after it? Were there any observable effects on the behaviour?

d) what was the weight/age of the rats at the start of the experiment?

e) what is the statistical significance threshold chosen?

f) how were the shocks delivered and calibrated? Did they not generate artefacts in the imaging?

17) The difference between past findings and the novelty of the present study could be explained better (for example, Sun et al. 2021 seems very relevant to the current study but is practically not discussed, also, Wu et al. (2017) [Hippocampal awake replay in fear memory retrieval. Nature Neuroscience 20, 571-580] and Girardeau et al. (2017) [Hippocampal awake replay in fear memory retrieval. Nature Neuroscience] seem like they should be discussed).

18) How were the A>0.6 threshold and A<0.31 threshold chosen?

19) The authors have selected for analyses of the 'control' (non-scopolamine) group only rats that learn to avoid the shock zone, and for analyses of the 'test' (scopolamine) group the rats that do not learn. Possibly related to this, it seems that the different groups have different behaviours in the pre-scopolamine session. I believe this is not the right approach; instead, if anything, the groups that are going to be compared should be equalized in as many parameters as possible at least prior to the manipulation of interest. If anything, the variability within each group should be used to assess the correlation between remapping/ instability of place cells and performance.

Reviewer #3 (Recommendations for the authors):

On the whole, I very much enjoyed reading this work, I found the results to be quite interesting and the presentation to be clear. I have only one question that can be addressed with further analysis. An essential claim of the manuscript is that the spatial codes measured during the drug-free and scopolamine shock sessions are equivalent in terms of within-session quality (and I appreciated all of the subsampling controls to match gross behavioral measures), yet I think this is only shown to be the case for 'place cells' which were preselected on the basis of various within-session measures. Can you confirm that: (1) the percentage of cells meeting place-cell criteria is equivalent across the two manipulations (not just the raw number of cells), and that (2) the equivalency of these measures holds when including all principal cells, not just those meeting (same-session) preselection criteria? I would be convinced about this claim if this were the case.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed. Please, check the reviewers' comments and correct/make changes to address the remaining comments. Regarding R2 comments (previous comment 6) asking for additional analysis, please feel free to make a decision on whether this can be addressed textually.

Reviewer #2 (Recommendations for the authors):

Hippocampal place cell remapping occurs with memory storage of aversive experiences.

Tracking no: 29-05-2022-RA-eLife-80661R1

The scientific approach and interpretation of this revised manuscript have been substantially updated (mostly in the introduction and discussion) and appear much more straightforward and supported by the results. The authors have answered most of my comments and added a behavioural-only control for the injection procedure, which is very useful (although the results for the control group might need some clarification, see comment below). Overall, I find the manuscript to be much improved and the results very interesting.

Comments related to previous comments:

Previous Comment 4: Normalizing trial counts: Currently, the data shown in Figure 2 – supplement 2 in the 'trn' group is taken from 10 minutes of behaviour, while data in the other groups are obtained from 15-minute sessions. Thus, the figure gives the impression of the trial number decreasing between the pre and trn sessions, but this is due (at least partially) to the shorter session duration for trn. This is why I was suggesting using trials/min in order to see if the rats still run fewer trials for an identical duration. If the authors prefer to keep the absolute number of trials as a measure, I suggest adding in the figure itself a clear mention of the duration of data included in each group to avoid misunderstandings.

Previous Comment 5 Analysis of remapping vs instability: I believe the authors have actually done an analysis that answers this, by computing the within-session correlation of rate maps (showing that the cells are still spatially stable) in Figure 2I, by showing that the spatial information is similar between the pre and trn sessions even in the scopolamine condition, together with computing the pre vs post correlations; thus no action is needed.

Previous Comment 6 Response to shock by increased firing vs decreased firing: My suggestion would be to detect all active place fields contacting the shock zone in the pre10 session part, and analyze their firing in the post5 or subsequent session part. Using an active place field should prevent the low baseline problem described by the authors.

Previous Comment 16.e Choice of significance threshold: Sorry if the previous comment was unclear – I meant that the methods should specify which is the threshold for statistical significance chosen to be used throughout the manuscript to decide when a test result that produces a p-value is deemed significant vs not, e.g. 'tests were considered significant at p<0.05'.

New Comment 1 (NC1): Thank you for adding a behavioural control with injection of saline. I have some comments related to this new figure (Figure 1 – Supplementary figure 1):

a) Is data in the 'Shock' session from the entire 15 min or only the first 10 min pre-shock? I recommend showing separately pre10 and post5 to get a better understanding of the effects.

b) If all the data were included: why is it that the saline rats do not seem to avoid the shock zone during the shock session (compare pre and shock data in B or C)? This should be discussed in the main text.

c) Separating the data in the graph would also allow us to replicate the effect seen in the implanted rats where scopolamine rats avoid the short path even before the shock session. Does this effect underlie an increase in impulsivity, or impairment of decision-making, possibly linked to a decrease in general attention? Are such side-effects of the scopolamine known? Perhaps these types of possible side effects of the scopolamine could be added to the relevant discussion (e.g. in around L 596).

d) Ideally, some measure of general activity (average speed or total distance covered for the whole session, not just included trials) should be shown in the same way as the existing figures to further assess the effects of the scopolamine, injection procedures and shock on activity.

e) Why is the 'shock' session not called 'trn' as in Figure 2 Supplement 2?

NC2: L 24, L 52 (and also L 503) "We conclude that place cell remapping occurs in response to motivationally significant (rather than neutral) stimuli that are remembered rather than merely perceived and forgotten"; "studies have shown that place cells remap in response to behaviorally significant events": this is partially correct, however, these sentences and the corresponding one in the discussion imply that only behaviourally significant events can cause place cell remapping. But place cells also remap in response to events of low/no behavioural significance, e.g. a change in the odour/colour of the environment (Anderson and Jeffery 2003, J Neurosci; Jeffery et al., 2003) or events of intermediary behavioural relevance like the introduction of barriers in the absence of a navigation task (Rivard et al., 2004, Journal of General Physiology, Muller and Kubie, 1987, J Neurosci). These studies generally observe partial remapping (of a sub-population of the place cells) while perhaps the remapping in the shock condition in the current study is more global (but to test this, additional analysis of place fields away from the shock zone would be necessary). For these reasons, I suggest reformulating/removing the references to "motivationally significant (rather than neutral) stimuli" to perhaps focus on the remembered vs non-remembered aspect. Perhaps the salience of an event influences the degree of remapping throughout the place cell population… but that hasn't really been demonstrated by the current study.

NC3: L508 +: "Hippocampal place fields tend to be preferentially concentrated at locations where rodents spend large amounts of time […]" : this is not always the case (see for example Hok et al., 2007, J Neurosci; Pfieffer, 2022, J Neurosci), and might depend on specific task demands and behavioural stereotypy (reviewed in Nyberg et al., 2022). I suggest adding a clarification that the goal overrepresentation phenomenon is not ubiquitous and might depend on the task and navigational strategy.

NC4: L 536 'indicating that place fields shifted towards and away from the shocked location in equal proportions': could the authors point to a result showing this, instead of fields, for example, changing their size or firing rate?

https://doi.org/10.7554/eLife.80661.sa1

Author response

Essential revisions:

A point of discussion was whether the results of scopolamine are novel and specific enough to support claiming a direct effect on memory processes. There were also concerns regarding additional methodological controls which are required. We feel they can be addressed with additional analyses and experiments. In a consultation session, we agreed the following revisions are required before a definitive recommendation can be made:

1) Control experiments for scopolamine are required. These include:

a) A control is needed to show that the injection procedures do not impair general behaviour (e.g. motivation, attention, level of stress) as well as other forms of learning;

We have run an additional saline-only control group (along with additional scopolamine groups) to demonstrate that scopolamine does impair avoidance learning compared against saline-injected controls. As shown in the Supplement to Figure 1, rats receiving saline alone show significantly greater post-training avoidance of the short path than rats receiving scopolamine.

b) Consider effects in the absence of shock learning. To be able to claim that the effects of scopolamine are specific to aversive learning, a control with either no learning or appetitive learning (e.g. of a reward location, which also involves place field reorganization in some cases) would be useful.

We believe that the barrier control condition, in which rats were exposed to a neutral barrier rather than an aversive shock, is sufficient to address this concern. However, we agree with this and other reviewers that insufficient attention was given to this control in the original manuscript, so it is now mentioned in the abstract and considered more prominently in the introduction and discussion.

We feel that acknowledging the systemic/brain-wide/unspecific effect of scopolamine is not enough and you should provide some additional evidence of specificity.

We have added a paragraph to the discussion (ll. 665-678) in which it is explained why the main conclusions of our study (namely, that place cell cell remapping is related to storage of memories for aversive events) do not depend upon whether or not scopolamine’s pharmacological actions were localized to the hippocampus (as almost certainly they were not).

c) Consider behavioral side effects of a systemic approach. Similarly, consider that functional effects of scopolamine (power and coordination of theta oscillations brain-wide) can additionally provide an explanatory value of the effects. Some of these are discussed but some others need support.

These considerations have been thoroughly integrated into the discussion. We acknowledge that scopolamine likely affected hippocampal and non-hippocampal circuits (ll. 665-678) and discuss the drug’s acute effects upon behavior and place cell activity (ll.561-612), including citation of prior work on how mAChRs modulate theta oscillations (ll.580-589).

2) Additional control analyses are required, including:

a) Population vector analysis should be tested against shuffled distributions.

New analyses show that all of the effects reported in the paper are abolished when the analyses are performed on shuffled data (Supplement 7 to Figure 2; Supplement 4 to Figure 3). Running this control analysis helped us to identify a minor flaw in the population vector correlation metric from the original manuscript (see reply to referee #1 comments below); this flaw has been corrected throughout the main text and figures by using a new population vector correlation metric referred to as place tuning stability (S). All previously reported effects survive the correction of converting to this slightly modified correlation measure.

b) Population-level analysis in terms of the percentage of cells meeting place-cell criteria across manipulations; the equivalency of place cell measures when including all principal cells, not just those meeting (same-session) pre-selection criteria.

Percentage of cells meeting place cell criteria are now reported in Figure 2D (right graph) and 3C (middle graph).

c) There are also some issues with the different pre-/post-training timing (8 days vs 6 days after pre-training) used in remapping effects with/without scopolamine that require clarification.

This appears to be a reference to an issue raised in a comment from referee #2. As explained in our reply to that comment below, we have controlled for this; the amount of time that elapses between sessions is identical for all analyses in Figure 2 (48 h) and very nearly so for all analyses in Figures 3 and 4, where all but 3 rats had identical intervals of 8d.

3) Clarification regarding LFOV:

a) Using large GRIN lenses may bring undesired histological damage and we wonder whether flattening the dorsal hippocampus may have consequences on functional readouts.

We agree that the issue of tissue compression is a concern for GRIN lens experiments, but no more so for our current study than for numerous other studies that have used similar methods in mice and rats. We do not convinced that histological damage can be quantified in a reliable way from the micrographs, but we have acknowledged this issue by adding a sentence to the Results section (ll. 151-154):

“Although the GRIN lens caused some flattening and compression of stratum oriens (Supplementary Figure 1), we do not believe this caused significant disruptions of CA1 activity since spatial tuning properties of place cells were quite similar to those reported in prior studies of rats without lens implants.”

b) Given improvements in LFOV imaging with your approach, we wonder whether center/periphery ROI imaging may be or may not be affected. Given individual variability of imaging ROIs per rat, we would like to see this more carefully addressed.

Supplement 3 to Figure 2 now shows how spatial tuning and between-session recurrence of place cells depends upon the size (area in pixels) and eccentricity (distance from lens center) of cell contours in the imaging plane. Although there are small statistically significant effects of contour size and eccentricity upon the percentage of cells classified as place cells (and also on the probability that a place cell will recur from one session to the next), we not believe that these small effects or are of much interest or that they are likely to have impacted the outcome of our other analysis, for reasons discussed in the caption.

4) Regarding impact and novelty, you may consider the following suggestions:

a) We discussed whether the above revisions may offer a chance to reinforce the novelty of your observations, by presenting pharmacological and behavioral (barriers) experiments as manipulations to better focus on the study of place cell remapping dynamics during aversive learning in rats. Actually, the abstract poorly integrates barrier results, which we feel should be mentioned. The authors may consider whether by highlighting/refocusing around some of these angles their manuscript would gain additional impact and broader interest.

b) Consider also tuning the title accordingly if required. Please, note that the title and/or the abstract should provide an indication of the biological system under study.

c) We also discussed whether using LFOV imaging in rats is an important addition to your paper and we would ask you to consider expanding the value of the approach when it comes to imaging from larger brains such as in rats versus mice. We feel in addressing questions regarding GRIN lenses and image aberration you may find the opportunity to highlight this better.

We are grateful for these suggestions and have made structural revisions to address them. The abstract (as well as the introduction and discussion) now address the barrier result and its importance for interpreting the findings. The title, introduction, and discussion have been thoroughly revised to shift emphasis away from implying that the scopolamine effects were end-goal of the study, and toward depiction of the scopolamine manipulation as a means to the more fundamental end of elucidating the role of remapping in storing memories of aversive experiences.

Please see the comments from each reviewer for more details.

Reviewer #1 (Recommendations for the authors):

Here is a list of comments for the authors to improve their work:

1. The body of the analysis is centered on the evaluation of population vector correlation matrices between behavioral phases and conditions. To build them, the authors rely on tuning curve heatmaps calculated from place cells in a reference condition (pre). One major problem with population tuning curves is that their typical 'diagonal' shape is very sensitive to noise so all curves and derived analysis should be tested against shuffled distribution as well. This affects data in Figure 2H,I; Figure 3D-F; Figure 4B,C.

The requested controls against shuffled distributions are now reported in Supplement 7 to Figure 2 and Supplement 4 to Figure 3, which show results after circularly shifting inferred spikes against position tracking data during each beeline trial. When deconvolved spikes were shuffled prior to the analysis, spatial tuning and remapping effects were abolished, as expected. It is worth noting that in the course of performing this control analysis, we became aware of a minor flaw in the population vector correlation measure that we used in the original manuscript; this flaw has been corrected in the revised paper without major impact upon the reported results. Briefly, the flaw was as follows. We previously measured population vector correlations by taking the median R value in bins near the diagonal of the population vector correlation matrix. However, when running our analyses on shuffled distributions, we realized that this measure can become large not only when place cells exhibit similar spatial tuning across sessions, but also when non-place cells fire at similar baseline rates across the entire track during both sessions. The proper control for this is to subtract away the mean off-diagonal r-values (which tend to be small for place cells and large for non-place cells) in each individual correlation matrix before taking the median. Making this correction to our metric produced almost no differences from the outcomes of our population vector correlation analyses, because the analyses were already being performed upon tuning curves of place cells that had been selected for their spatial firing properties, and thus for the fact that they already had near-zero mean r-values off the diagonal (hence, subtracting these near-zero means under the new measure had minimal effect upon analysis outcomes). It does become important to perform this correction in the requested analyses of population vector correlation matrices computed from shuffled spike trains, since these matrices can have large values on the diagonal in the absence of spatial tuning. We thank the referee for proposing the shuffle control and thereby helping us to improve our population vector correlation measure.

2. The authors nicely control for their main figure analysis using the subset of scopolamine-treated rats that retained avoidance, but they do not provide data on the similarly important group of drug-free rats that DO NOT retain avoidance. They may help to provide additional controls for all main figure analyses as well. In addition, the paper would benefit from examining whether remapping indices can explain the behavioral performance of all rats, without resorting to categorizing them by performance level (i.e. using correlation analysis and/or predictive model analysis).

In the original manuscript, the results for drug free rats that did not retain avoidance were included in Supplementary Figures 4-7 (in the revised manuscript, these results are shown in Supplement 8 to Figure 2 and Supplement 3 to Figure 3). However, the reviewer is correct to point out that these results were not previously mentioned in the main text. We have addressed this by adding sentences to the main text that describe results for drug free rats that did not retain avoidance along with results for the scopolamine rats that did retain avoidance.

3. Using large GRIN lenses for large-field-of-view is a major strength of the paper when applied to imaging from the rat hippocampus. However, this can yield undesired histological damage. For instance, pushing the lens against the alveus for image flattening results in a reduction of the stratum oriens and atrophy of basal dendrites. This is inevitable and should not be considered by the authors as a criticism, but worth mentioning/controlling for effects. Are any of the behavioral and/or functional readouts correlated with some histological damage?

We agree that the issue of tissue compression is a concern for GRIN lens experiments, but no more so for our current study than for numerous other studies that have used similar methods in mice and rats. We do not convinced that histological damage can be quantified in a reliable way from the micrographs, but we have acknowledged this issue by adding a sentence to the Results section (ll. 151-154):

“Although the GRIN lens caused some flattening and compression of stratum oriens (Supplementary Figure 1), we do not believe this caused significant disruptions of CA1 activity since spatial tuning properties of place cells were quite similar to those reported in prior studies of rats without lens implants.”

4. Similarly, there is natural variability in imaging from cells at the center/periphery of the GRIN, which differ in aberration effects. Since there is a different contribution across animals another useful control is to evaluate the consistency of conclusions for data obtained from center/periphery within and between animals. This would also help to set standards for the field, especially for LFOV imaging.

Supplement 3 to Figure 2 now shows how spatial tuning and between-session recurrence of place cells depends upon the size (area in pixels) and eccentricity (distance from lens center) of cell contours in the imaging plane. Although there are small statistically significant effects of contour size and eccentricity upon the percentage of cells classified as place cells (and also on the probability that a place cell will recur from one session to the next), we not believe that these small effects or are of much interest or that they are likely to have impacted the outcome of our other analysis, for reasons discussed in the caption.

Importantly, there is poor information on how the ROIs are defined, what methods are used for segmentation and how imaging from different sessions are treated (same ROIs, different ROIs, etc…), and what signals are evaluated dF/F with/without denoising, models of spike inference, etc…

We have addressed this by specifying the requested information more clearly in the methods section (ll. 808-840; as noted therein, exact parameters for the analysis may be found in the scripts provided on Github). We have also updated Figure 2A (upper right quadrant) to more clearly illustrate an example of how cell contours align across sessions.

Reviewer #2 (Recommendations for the authors):

Here are some detailed comments.

1) Please add page and line numbers to ease review.

The revision is paginated and line numbered.

2) Given that several of the rats did the drug-free shock learning twice, how were these data analysed? Was only one of the sessions per rat chosen to contribute to the analysis, if yes, how?

Only one rat (the female shown in the bottom row of Figure 2B) received two drugfree shocks, and the second drug-free shock session was not included in the analysis. To avoid confusion, the second drug-free shock session is no longer shown in Figure 2 B (since no data from this animal was used after the extinction session that is still plotted in the Figure). We thank the reviewer for catching this.

3) Was there an influence of the order of sessions on the results (e.g. scopolamine during the first shock session, the second one, after a barrier session or not)?

A ranksum test yielded no significant effect of training session order on results within either the drug free or scopolamine shock conditions; this is now indicated by ‘ns’ labels underneath the horizontal axis in Figure 1H, right panel.

4) Any analysis of the number of trials, if comparing across sessions of different durations, should be done on a number of trials/minute to allow for cross-sessions comparisons (e.g. Sup Figure 2 A).

It is not clear to us why the referee is recommending this. The goal of beeline trial filtering is to assure that behavior sampling of the short path (total time spent in each bin) is as similar as possible across analyzed sessions; this mandates equalizing the TOTAL NUMBER of beeline trials per session, not trials per unit time. Hence, the primary reason for showing panel A in Supplement 2 to Figure 2 (formerly Sup Figure 2A to which the referee refers) is to show what pre-subsampled trial counts look like when they are NOT normalized by time, and thereby demonstrate why it was necessary to equalize trial counts by subsampling beeline trials. If the graph were normalized by time, then the information the graph is intended to convey would be lost.

5) Analysis of remapping: remapping is not the same as instability. It corresponds to a consistent activation of place fields in specific locations, that change (in a consistent manner) in a different condition, with the activity pattern, mostly reinstated when coming back to the original condition. Otherwise, changes in place cell firing that are not consistent or reproducible are considered drift or instability of the spatial firing. It is unclear if the current paradigm and analysis tests remapping and not simply instability.

It is unclear what remedy, if any, the referee is requesting to address this concern.

6) In the 'Scopolamine does not disrupt place cell responses to shock' section: individual place cells are considered to 'respond' to the shock if their activity increases during entry into the shock zone. However, place cells can also respond by decreasing their activity, or even changing their place field shape or location, but this is not analysed.

Prior to submitting the original manuscript we did perform an analysis of inhibitory responses to shock. The problem with such an analysis is that statistical power for detecting excitatory versus inhibitory responses is highly asymmetric; the majority of place cells have a low baseline spike rate at the shocked location, so an immense number of trials would be required to detect a shock-evoked decrease of the spike rate from the already low baseline rate. Consequently, very few cell can be categorized as inhibited by shock, but whether this reflects an actual paucity of shock inhibition or a statistical power problem is impossible to determine.

7) Given that place cells can modulate their firing spontaneously with time (mentioned by the authors), any analysis of place cells between two different conditions where time has elapsed should have as a control a similar comparison, without the parameter of interest being changed, but with approximately the same amount of time passed or at least number of sessions (as the authors do in one of their analyses).

Yes, we have controlled for this. The amount of time that elapses between sessions is identical for all analyses in Figure 2 (48 h) and very nearly so for all analyses in Figures 3 and 4 (mode of 8d).

8) Nomenclature: In general, it would be good to avoid the use of abbreviations as much as possible, including in the figures, to help readability. Also, the different terms used (mostly condition or session names) can be confusing and there may be errors in figures or legends that make the interpretation difficult. The use of pretraining, training, and post in some instances, but pre10 post 5 in others (within the same session), is also a bit confusing. Perhaps the 10 min part of the shock training sessions that is pre-shock should be called the normal session and then the shock session would be called the shock session (or barrier session), without lumping both in the same session. Also, there are 2 forms of training: training to choose the shortest path, then training of the shock – and then extinction of the shock learning. Perhaps the authors could find a more intuitive and consistent way to name their sessions. What is 'post-extinction' (e.g. in 'post-extinction CA1 population vectors)? Is it referring to the activity during the extinction sessions? Or during the sessions that followed the extinction sessions, which according to Figure 2B will always be a shock session? It seems that this is sometimes called post-training (e.g. Figure 1J) and sometimes post-extinction (e.g. Figure 1J, legend). Is the place field recurrence rate (Figure 2G) the same as place field recurrence ratio (in results)? Do subsampled and down sampled mean the same here?

We sympathize with the referee’s confusion over these issues and have endeavored to clarify them throughout the revised manuscript.

9) Related to this, in several instances the axis labels or legends of a figure are not self-sufficient and could be enhanced with more information about the measure that is shown (e.g.: Figure 1J: how is %baseline computed?).

This has been clarified in the caption for panel J of Figure 1 which now reads:

“J) Left: Extinction curves over the first 3 post-extinction sessions; number of rewards earned on the short path is measured as a percentage of baseline (48 h before training; thin lines: individual rats, thick lines: per session median).”

10) The graph in Figure 1B seems to show that after each shock session, the rats have 1 extinction session. But in the main text, it is said that 'rats continued to run standard sessions on the maze every 48h until they reached the avoidance extinction criterion", which seems to assume that this would take several sessions. It seems that the graph should be adapted to reflect how many sessions were run. It is not clear what happens after the barrier sessions – do the rats move on to the shock learning sessions immediately? Does the '3s' mean that they do instead of the barrier session 3 times, or maybe extinction sessions?

This has been clarified by adding the following to the caption for panel B of Figure 1:

“The number of sessions in ACQ and EXT boxes varied per rat, depending on how many sessions it took individual subjects to reach behavioral criteria.”

11) It is a bit unclear what the -48/+48 measure means on some of the figure axes and in text. Does that mean that what is compared is data from 4 days apart? Or just 2 days apart? If it's the latter, shouldn't the measure be 0-48h or -24/+24?

4 days apart. To clarify this, the label “time from training session” has been added to the horizontal axis of Figure 1H.

12) Rejecting specific trials from the analysis does not change the behaviour of the rats during data collection; it only makes the analysis more balanced. The sentence "This assured that rats were behaving as similarly as possible during data collection in all sessions" should be rewritten to reflect this.

The statement has been rewritten as follows (ll. 884-887):

“This assured that rats were behaving as similarly as possibly during analyzed data segments from all sessions.”

13) Figure 2E could show proportions out of the total number of putative pyramidal cells analysed, not just out of detected place cells, to see if the percentage of active place cells changes.

Percentage of cells meeting place cell criteria are now reported in Figure 2D (right graph) and 3C (middle graph).

14) Tuning curve 'sequences' (showing the tuning curves of all cells ordered according to peak firing) are used in many of the plots, but it is unclear what information to take from them. Reorganizing the figures to emphasize the quantification of the results might be useful and perhaps removing some of the sequence plots unless they convey a specific message or can be used as an example.

We have retained the heatmap plots because they show al data from individual cells and therefore provide readers with important validation of the size and quality of the dataset that was analyzed in the other summary graphs.

15) Related to this, most of the figures are really small. Focusing on the ones that provide quantitative results (i.e. less 'tuning curves' plots) and making those bigger would probably help the reader.

We believe that the sizing of our figures conforms with current practice for publishing calcium imaging data in peer reviewed journals, but we will be happy to consider any figure reformatting guidance from the editorial staff if the paper is accepted for publication.

16) Descriptions of some specific procedures could be added in the methods (listed in no specific order): a) How were the beeline trials selected?

This is explained in great detail in the Methods section (ll. 879-944).

b) How were putative spikes extracted/reconstructed from the calcium imaging signal (deconvolution?)?

This is now explained in the Methods (ll. 834-840).

c) How was the injection procedure performed, and where were the rats kept after it? Were there any observable effects on the behaviour?

Injection methods are now explained in the Methods (ll. 794-798). Behavior effects are detailed in Results (ll. 112-146) and Discussion (ll. 561-612).

d) what was the weight/age of the rats at the start of the experiment?

This is now stated in the Methods (ll. 724-727).

e) what is the statistical significance threshold chosen?

It is unclear to what analysis this question is referring.

f) how were the shocks delivered and calibrated? Did they not generate artefacts in the imaging?

Shock delivery apparatus and procedures are now explained in the Methods (ll. 731-737). Shocks did not generate artefacts in the imaging. Unlike electrophysiological recordings, calcium imaging is not susceptible to shock artefacts. The housing of the miniscope is made of non-conductive plastic and is therefore electrically insulated from the rat and the shock current.

17) The difference between past findings and the novelty of the present study could be explained better (for example, Sun et al. 2021 seems very relevant to the current study but is practically not discussed, also, Wu et al. (2017) [Hippocampal awake replay in fear memory retrieval. Nature Neuroscience 20, 571-580] and Girardeau et al. (2017) [Hippocampal awake replay in fear memory retrieval. Nature Neuroscience] seem like they should be discussed).

We have added references to the Wu et al. (2017) and Girardeau et al. (2017) studies in the context of our discussion about how shock-induced remapping affects the spatial distribution of place fields (l. 522-24). We have also added new references to recently published work by Wirtshafter and Disterhoft (2022) and Ormond et al. (2023) that has come out during the time we have been working on revisions.

18) How were the A>0.6 threshold and A<0.31 threshold chosen?

As explained in the Methods (ll. 873-877), these values were chosen because they were the cutoff thresholds for the top and bottom quartiles, respectively, of the percentage of trials on which rats chose the short path.

19) The authors have selected for analyses of the 'control' (non-scopolamine) group only rats that learn to avoid the shock zone, and for analyses of the 'test' (scopolamine) group the rats that do not learn. Possibly related to this, it seems that the different groups have different behaviours in the pre-scopolamine session. I believe this is not the right approach; instead, if anything, the groups that are going to be compared should be equalized in as many parameters as possible at least prior to the manipulation of interest. If anything, the variability within each group should be used to assess the correlation between remapping/ instability of place cells and performance.

We report an analysis of correlation between remapping measures and behavioral performance in Panel H of Supplement 8 to Figure 2.

Reviewer #3 (Recommendations for the authors):

On the whole, I very much enjoyed reading this work, I found the results to be quite interesting and the presentation to be clear. I have only one question that can be addressed with further analysis. An essential claim of the manuscript is that the spatial codes measured during the drug-free and scopolamine shock sessions are equivalent in terms of within-session quality (and I appreciated all of the subsampling controls to match gross behavioral measures), yet I think this is only shown to be the case for 'place cells' which were preselected on the basis of various within-session measures. Can you confirm that: (1) the percentage of cells meeting place-cell criteria is equivalent across the two manipulations (not just the raw number of cells),

Percentage of cells meeting place cell criteria are now reported in Figure 2D (right graph) and 3C (middle graph).

and that (2) the equivalency of these measures holds when including all principal cells, not just those meeting (same-session) preselection criteria? I would be convinced about this claim if this were the case.

The within-session quality measures to which the referee is referring (place field width, spatial info, peak in-field firing rate, mean out-of-field firing rate) are measures that can only be applied to cells with verified spatial tuning, so we are unsure what the referee means by requesting that these measures be performed on “all principle cells…not just those meeting preselection criteria.”

[Editors’ note: what follows is the authors’ response to the second round of review.]

Reviewer #2 (Recommendations for the authors):

Hippocampal place cell remapping occurs with memory storage of aversive experiences.

Tracking no: 29-05-2022-RA-eLife-80661R1

The scientific approach and interpretation of this revised manuscript have been substantially updated (mostly in the introduction and discussion) and appear much more straightforward and supported by the results. The authors have answered most of my comments and added a behavioural-only control for the injection procedure, which is very useful (although the results for the control group might need some clarification, see comment below). Overall, I find the manuscript to be much improved and the results very interesting.

Comments related to previous comments:

Previous Comment 4: Normalizing trial counts: Currently, the data shown in Figure 2 – supplement 2 in the 'trn' group is taken from 10 minutes of behaviour, while data in the other groups are obtained from 15-minute sessions. Thus, the figure gives the impression of the trial number decreasing between the pre and trn sessions, but this is due (at least partially) to the shorter session duration for trn. This is why I was suggesting using trials/min in order to see if the rats still run fewer trials for an identical duration. If the authors prefer to keep the absolute number of trials as a measure, I suggest adding in the figure itself a clear mention of the duration of data included in each group to avoid misunderstandings.

As explained in our reply to the referee’s original comment, retaining the nonnormalized scale is essential for conveying the information provided by the figure; we agree with the referee’s concerns about lack of clarity and have made the recommended changes by adding data durations to x-axis labels of the figure, along with a clarification of what these labels mean in the caption.

Previous Comment 5 Analysis of remapping vs instability: I believe the authors have actually done an analysis that answers this, by computing the within-session correlation of rate maps (showing that the cells are still spatially stable) in Figure 2I, by showing that the spatial information is similar between the pre and trn sessions even in the scopolamine condition, together with computing the pre vs post correlations; thus no action is needed.

Acknowledged.

Previous Comment 6 Response to shock by increased firing vs decreased firing: My suggestion would be to detect all active place fields contacting the shock zone in the pre10 session part, and analyze their firing in the post5 or subsequent session part. Using an active place field should prevent the low baseline problem described by the authors.

It would certainly be possible to skirt the problem of asymmetrical statistical power (noted in our prior reply) by performing such an analysis, but it is unclear to us that the results of such an analysis would be germane to the questions we are asking in the present study about whether shock responses are predictive of remapping responses. Analyzing inhibitory responses by the proposed method would only be possible for a subset of cells selected for the property of being active in the shock zone, a property that itself might be confounded with the dependent variable of interest: remapping. Moreover, some kind of inclusion threshold on statistical power would have to be formulated and justified to select cells by this criterion (the referee does not propose suggestions for how to do this). For these reasons, it is our view that the only question to which the proposed analysis could provide a meaningful answer would be “are there any place cells that show evidence of being inhibited by shock?” That question, although interesting, is not primary to the scope and focus of the present study.

Previous Comment 16.e Choice of significance threshold: Sorry if the previous comment was unclear – I meant that the methods should specify which is the threshold for statistical significance chosen to be used throughout the manuscript to decide when a test result that produces a p-value is deemed significant vs not, e.g. 'tests were considered significant at p<0.05'.

We have addressed this point by locating every statement about “significance” in the Results and Methods sections where threshold information was not provided in the text or figures; threshold information has been added on lines 215, 216, 235, 236, 460.

New Comment 1 (NC1): Thank you for adding a behavioural control with injection of saline. I have some comments related to this new figure (Figure 1 – Supplementary figure 1):

a) Is data in the 'Shock' session from the entire 15 min or only the first 10 min pre-shock? I recommend showing separately pre10 and post5 to get a better understanding of the effects.

We have clarified this by adding the duration of each session in minutes to the session labels above the plot in Figure 1 – supplementary figure 1B; the caption text has also been updated with the following text:

“shock session includes only data from the 10’ pre-shock period whereas all other days include data from the full 15’ session.”

b) If all the data were included: why is it that the saline rats do not seem to avoid the shock zone during the shock session (compare pre and shock data in B or C)? This should be discussed in the main text.

As noted above (see reply to ‘a’) shock session data only includes the pre-shock period, so there is no inconsistency to address here.

c) Separating the data in the graph would also allow us to replicate the effect seen in the implanted rats where scopolamine rats avoid the short path even before the shock session. Does this effect underlie an increase in impulsivity, or impairment of decision-making, possibly linked to a decrease in general attention? Are such side-effects of the scopolamine known? Perhaps these types of possible side effects of the scopolamine could be added to the relevant discussion (e.g. in around L 596).

We assume that when the referee says “scopolamine rats avoid the short path even before the shock session” they mean before the shock STIMULUS (not “session”). It is indeed the case that both implanted and non-implanted scopolamine rats show reduced preference for the short path prior to shock. “Separating the data in the graph” is not necessary to see this; in the existing version of Figure 1 – supplementary figure 1 (panels B and C) it can be seen that on the shock day, scopolamine rats show a greatly reduced preference for the short path compared with saline-injected rats. New statistics have been added to the figure caption to show that this effect is significant.

d) Ideally, some measure of general activity (average speed or total distance covered for the whole session, not just included trials) should be shown in the same way as the existing figures to further assess the effects of the scopolamine, injection procedures and shock on activity.

We have now included the measure of median running speed during beeline trials for this data as well in Supplement to Figure 1A (bottom) and the caption has been revised to reflect this (L 1076):

“A) Top: Mann-Whitney U tests found that 48 hour retention of avoidance behavior did not differ for males versus females in the drug-free or scopolamine shock conditions. Bottom: Median beeline running speeds per session.”

e) Why is the 'shock' session not called 'trn' as in Figure 2 Supplement 2?

We have changed the label from ‘shock’ to ‘trn’ for consistency with figures in the main text.

NC2: L 24, L 52 (and also L 503) "We conclude that place cell remapping occurs in response to motivationally significant (rather than neutral) stimuli that are remembered rather than merely perceived and forgotten"; "studies have shown that place cells remap in response to behaviorally significant events": this is partially correct, however, these sentences and the corresponding one in the discussion imply that only behaviourally significant events can cause place cell remapping. But place cells also remap in response to events of low/no behavioural significance, e.g. a change in the odour/colour of the environment (Anderson and Jeffery 2003, J Neurosci; Jeffery et al., 2003) or events of intermediary behavioural relevance like the introduction of barriers in the absence of a navigation task (Rivard et al., 2004, Journal of General Physiology, Muller and Kubie, 1987, J Neurosci). These studies generally observe partial remapping (of a sub-population of the place cells) while perhaps the remapping in the shock condition in the current study is more global (but to test this, additional analysis of place fields away from the shock zone would be necessary). For these reasons, I suggest reformulating/removing the references to "motivationally significant (rather than neutral) stimuli" to perhaps focus on the remembered vs non-remembered aspect. Perhaps the salience of an event influences the degree of remapping throughout the place cell population… but that hasn't really been demonstrated by the current study.

On lines 24 and 502, we have followed the referee’s suggestion by changing the wording to read:

“…remapping occurs selectively under conditions where a stimulus occurs and is subsequently remembered rather than forgotten…”

We have retained existing wording on line 52 in the citation of prior studies showing that “place cells remap in response to behaviorally significant events” because we believe this wording is accurate and it is clearly not intended as a statement about what is demonstrated in the current study. We agree with the referee’s point that “behavioral significance” may not be a core prerequisite for remapping, but the existing wording of the citation is fully compatible with this, especially since the Introduction explains that remapping can occur with the mere passage of time and that the rate of remapping may be proportional to the rate at which episodic memories are being stored.

NC3: L508 +: "Hippocampal place fields tend to be preferentially concentrated at locations where rodents spend large amounts of time […]" : this is not always the case (see for example Hok et al., 2007, J Neurosci; Pfieffer, 2022, J Neurosci), and might depend on specific task demands and behavioural stereotypy (reviewed in Nyberg et al., 2022). I suggest adding a clarification that the goal overrepresentation phenomenon is not ubiquitous and might depend on the task and navigational strategy.

“tend to be” has been replaced with “are sometimes observed to be”

NC4: L 536 'indicating that place fields shifted towards and away from the shocked location in equal proportions': could the authors point to a result showing this, instead of fields, for example, changing their size or firing rate?

To better support this conclusion, we have added another supplementary figure (Figure 3—figure supplement 4) to quantify the field shifts.

https://doi.org/10.7554/eLife.80661.sa2

Article and author information

Author details

  1. Garrett J Blair

    Department of Psychology, UCLA, Los Angeles, United States
    Present address
    Center for Neural Science, New York University, New York, United States
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    For correspondence
    gblair@ucla.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2724-8914
  2. Changliang Guo

    1. David Geffen School of Medicine, University of California Los Angeles, Los Angeles, United States
    2. Department of Neurology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    3. Integrative Center for Learning and Memory, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Conceptualization, Software, Investigation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Shiyun Wang

    Department of Psychology, UCLA, Los Angeles, United States
    Contribution
    Collection of supplementary behavioral data
    Competing interests
    No competing interests declared
  4. Michael S Fanselow

    1. Department of Psychology, UCLA, Los Angeles, United States
    2. Integrative Center for Learning and Memory, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Conceptualization, Resources, Funding acquisition, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3850-5966
  5. Peyman Golshani

    1. David Geffen School of Medicine, University of California Los Angeles, Los Angeles, United States
    2. Department of Neurology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    3. Integrative Center for Learning and Memory, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Supervision, Funding acquisition, Project administration, Writing – review and editing
    Competing interests
    No competing interests declared
  6. Daniel Aharoni

    1. David Geffen School of Medicine, University of California Los Angeles, Los Angeles, United States
    2. Department of Neurology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    3. Integrative Center for Learning and Memory, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Conceptualization, Resources, Software, Supervision, Funding acquisition, Validation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4931-8514
  7. Hugh T Blair

    1. Department of Psychology, UCLA, Los Angeles, United States
    2. Integrative Center for Learning and Memory, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Writing – original draft, Project administration
    For correspondence
    tadblair@ucla.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8256-5109

Funding

National Science Foundation (NeuroNex 1704708)

  • Hugh T Blair
  • Peyman Golshani
  • Daniel Aharoni

National Institute of Mental Health (RO1-MH062122)

  • Michael S Fanselow

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. Andrew Howe, Shiyun Wang, Ryan Grgurich, Umais Khan, and Lorelei Mae Cordes for helpful comments on the experimental design, data collection, and manuscript. This work was supported by NSF NeuroNex 1704708 (HTB, PG, DA) and RO1-MH062122 (MSF).

Ethics

All experimental procedures were approved by the Chancellor's Animal Research Committee of the University of California, Los Angeles, in accordance with the US National Institutes of Health (NIH) guidelines. protocol #2017-038.

Senior Editor

  1. Laura L Colgin, University of Texas at Austin, United States

Reviewing Editor

  1. Liset M de la Prida, Instituto Cajal, Spain

Reviewers

  1. Liset M de la Prida, Instituto Cajal, Spain
  2. Alexandra T Keinath, University of Illinois at Chicago, United States

Version history

  1. Preprint posted: May 29, 2022 (view preprint)
  2. Received: May 30, 2022
  3. Accepted: June 22, 2023
  4. Accepted Manuscript published: July 19, 2023 (version 1)
  5. Version of Record published: July 25, 2023 (version 2)

Copyright

© 2023, Blair et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,719
    Page views
  • 227
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

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)

  1. Garrett J Blair
  2. Changliang Guo
  3. Shiyun Wang
  4. Michael S Fanselow
  5. Peyman Golshani
  6. Daniel Aharoni
  7. Hugh T Blair
(2023)
Hippocampal place cell remapping occurs with memory storage of aversive experiences
eLife 12:e80661.
https://doi.org/10.7554/eLife.80661

Share this article

https://doi.org/10.7554/eLife.80661

Further reading

    1. Computational and Systems Biology
    2. Neuroscience
    Tony Zhang, Matthew Rosenberg ... Markus Meister
    Research Article

    An animal entering a new environment typically faces three challenges: explore the space for resources, memorize their locations, and navigate towards those targets as needed. Here we propose a neural algorithm that can solve all these problems and operates reliably in diverse and complex environments. At its core, the mechanism makes use of a behavioral module common to all motile animals, namely the ability to follow an odor to its source. We show how the brain can learn to generate internal “virtual odors” that guide the animal to any location of interest. This endotaxis algorithm can be implemented with a simple 3-layer neural circuit using only biologically realistic structures and learning rules. Several neural components of this scheme are found in brains from insects to humans. Nature may have evolved a general mechanism for search and navigation on the ancient backbone of chemotaxis.

    1. Neuroscience
    Frances Skinner
    Insight

    Automatic leveraging of information in a hippocampal neuron database to generate mathematical models should help foster interactions between experimental and computational neuroscientists.