Social learning, defined as acquiring new knowledge through observing or directly interacting with others, is a fundamental behavior of humans and animals (Bandura, 1997; Meltzoff et al., 2009; Heyes and Galef, 1996). Behavioral studies have identified many potential mechanisms of how learning can occur through social observation or interaction. Besides the imitation learning that humans and some primates employ (Bandura, 1997; Meltzoff et al., 2009; Zentall, 2006), a common behavioral effect in many animal species is the so-called 'social facilitation' (Zentall, 2006; Zentall and Levine, 1972; Zajonc, 1965) or 'local enhancement' (Heyes and Galef, 1996): an animal’s understanding of an environment is facilitated by the presence of other social subjects in the same environment. Local enhancement can be caused possibly by enhanced sensory processing due to heightened attention, acquiring environmental attributes such as safety or food availability, or other unspecified means (Heyes and Galef, 1996; Zentall, 2006; Zajonc, 1965). According to this idea, the presence of social subjects in an environment impacts other animals’ neural processing of information related to the environment. In this study, we aim to explore the neural basis of such local enhancement effect of social observation.

In humans and rodents, spatial information of an environment is represented in spatial memory by hippocampal place cells (O'Keefe and Dostrovsky, 1971; Burgess and O'Keefe, 2003; Wilson and McNaughton, 1993), which fire at specific locations (place fields) of a given environment. For example, when a rat travels along a linear track from one end to the other, a subset of place cells display place fields on the track and they fire one after another in a unique sequence (Lee and Wilson, 2002). When the animal explores a different environment such as an open box, hippocampal place cells 'remap' (Leutgeb et al., 2005; Colgin et al., 2008; Alme et al., 2014), i.e., they either alter firing locations, stop firing, or become active. Given that spatial environments are represented by distinguished place cell activity patterns, we asked how the presence of a rat in an environment such as a track impacts an observer rat’s place cell activity patterns representing the environment. Specifically, we set out to test a hypothesis that an observer’s place cell sequence representing a track can be activated by another rat running on the track, even if the observer is physically in a different environment separated from the track. Through such activation of place cell activity patterns across different environments, here referred to as 'cross-activation', local enhancement may be achieved via strengthening spatial memory representation of an environment without actual self-exploration.

To test this hypothesis, we recorded neurons from the CA1 area of the hippocampus in rats as they stayed in a small box while a demonstrator rat running on a separate, nearby linear track and as they ran on the same track themselves. We found that place cell sequences during track running also appeared in the box, but only when a demonstrator was present on the track, supporting our hippocampal cross-activation hypothesis of local enhancement.

We recorded from hippocampal CA1 cells in 9 rats when they stayed in a small (25 × 25 cm) box and when they ran on a linear track. Prior to the recording, a group of 5 of these rats had gone through a training schedule that lasted 2–3 days, 15–30 min each day, where they were placed in the box while a well-trained demonstrator running back and forth for food reward on a nearby linear track (Figure 1A). On the first recording day, CA1 cells from these 5 rats were recorded during 3 sessions (Figure 1B). In two of these sessions, each of the rats stayed in the box performing the same task as in the training (Pre-box and Post-box sessions). We referred to this configuration of the recorded rat staying in the box while a well-trained demonstrator on the track as the Trained-demo box condition for a Pre- or Post-box session. In between these two sessions, the recorded rat ran the track itself for the first time (Track session). This recording procedure of 3 sessions with Pre- and Post-box under the Trained-demo condition was repeated for 1 or 2 more days. The same recording procedure continued for a total of 6–12 days, but with Pre- and Post-box under various other box conditions (Figure 1B), including removing the demonstrator from the track (Empty-track), removing both the track and demonstrator (No-track), replacing the well-trained demonstrator with a naïve demonstrator that had never been exposed to the track (Naïve-demo) or with a remotely controlled toy car running back and forth on the track (Toy-car), or blocking the recorded rat’s view in the box while a well-trained demonstrator running on the track (Blocked-view). The other group of 4 recorded rats were each trained for 2–3 days prior to the recording, similarly as in the first group, but without the presence of a demonstrator on the track. These rats were then recorded for 6-12 days with the same Pre-box, Track and Post-box schedule each day. However, in these rats, the first 2 recording days were under the Empty-track condition, followed by the Trained-demo and other conditions. A total of 3290 putative pyramidal CA1 cells (defined as mean firing rate < 6 Hz) were recorded from all animals on all recording days. Here we analyzed 2200 cells that were active (with mean rate > 0.5 Hz) in at least one of the Pre-box, Track, and Post-box sessions on a given recording day.

Figure 1 with 1 supplement see all

Download asset Open asset

Firing sequences during rotation events in the box

In Pre- and Post-box under the conditions other than No-track and Blocked-view, a recorded rat was placed in the small box, which had opaque, high walls on its 3 sides with one side opening to the nearby linear track (Figure 1A). The small box and the track, both elevated from the floor, were separated by 25 cm to prevent direct physical contacts between the recorded rat and the demonstrator. The rat in the box exhibited two major types of behavior. The first was a relatively inactive behavior: facing the opening side of the box with little or small irregular movement. On average, the animals spent about 53% of the time facing the opening side of the box. However, this percentage did not differ among different box conditions (Figure 1C), except for the Blocked-view condition, where the opening side was blocked and rats predictably spent less percentage of time facing the opening side. This result is likely due to the fact that the opening side was the only way of exploring the outside environment regardless of the situation on the track. The second was an active, rotation behavior: the animal’s body rotating clockwise (CW) or counter-clockwise (CCW), mostly starting from the opening side of the box and ending at the same side (Figure 1D). We identified the active rotation events during all Pre- and Post-box sessions in all rats based on animals’ head directions. On average, 18.7 ± 1.1 rotation events were identified per session and there was no significant difference among different box conditions or between Pre- and Post-box (Figure 1E). The rotation events lasted for an average of 6.67 ± 0.05 s. Analysis of CA1 local field potentials (LFPs) shows that rotation events were accompanied by strong theta (6–10 Hz) oscillations (Buzsáki, 2002) without much high-frequency (100–250 Hz) sharp-wave ripples (Buzsáki et al., 1992) (Figure 1—figure supplement 1). Since theta is commonly associated with active behavior such as maze-running whereas ripples mostly occur during inactive behavior such as immobility or slow-wave sleep (Buzsáki, 2002; Buzsáki et al., 1992), the analysis indicates that the hippocampus was in an active state during rotation events. In this study, we focus on the CA1 activity within these active rotation events, because our primary goal is to test our hippocampal cross-activation hypothesis by comparing CA1 activity patterns in the box and those on the track in the same type of active, theta-associated behavior. In addition, since there were multiple rotation events per session, focusing on rotation events permitted the analysis of neural activity over multiple samples of the same behavior.

We identified those CA1 cells that were active in rotation events (14.7 ± 0.4 rotation-active cells per session) and examined their firing characteristics. As illustrated by firing rate maps of CA1 cells (firing rate versus the animal’s head position) in the small box (Figure 2A), we found that, despite the small size of the box, firing activities of many CA1 cells were not distributed across all the positions where an animal’s head was located, but highly restricted. Due to the circular nature of the rotation behavior, this location specificity led to the appearance of head-direction tuning, as illustrated by a circular firing rate curve (mean firing rate versus the animal’s head direction during rotation events) for each such cell in a box session (Figure 2B). We found that 40% of rotation-active CA1 cells (566 out of 1421) were significantly tuned to head direction during rotation events in the box (p<0.05, Rayleigh test). Consequently, simultaneously recorded CA1 cells in a box session displayed consistent firing sequences during these rotation events. When the animal rotated from different directions (CW versus CCW), the sequences appeared to be reversed to a large extent, but not always (Figure 2C and Figure 2—figure supplement 1).

Figure 2 with 1 supplement see all

Download asset Open asset

To quantify the consistency of each cell’s firing among rotation events of a box session, we computed a circular correlation between its firing rate curves of any two rotation events. The mean correlation among all different combinations of events in a session was compared to a distribution of correlation values obtained by random, independent shuffling of the cell’s rate curve in every event (Figure 2D) and z-score transformed. We refer to this z-scored mean cross-event correlation as the rotation-consistency of a cell and defined cells with z-score > 1.645 (p<0.05, Z-test) as rotation-consistent. We found that 74% of rotation-active cells were rotation-consistent. On average, we identified 10.2 ± 0.5 location-consistent cells per box session. We then compared rotation-consistency of the rotation-active cells in different types of rotation events under various conditions. First, we found that the average rotation-consistency within CW or CCW events was significantly greater than that across CW and CCW rotation events of same sessions (Figure 2E), suggesting that the activities of these cells on average showed a dependence on rotation direction. Nevertheless, 42% of rotation-consistent cells (29% of all rotation-active cells) were also consistent between CW and CCW events. This quantification provides an explanation for many, but not always, similar firing sequences in reverse order between CW and CCW rotation events (Figure 2C and Figure 2—figure supplement 1). Second, the rotation-consistency was similar among different box conditions (Figure 2F). These results demonstrate that CA1 cells did not fire throughout a rotation event, but at specific locations of its rotation trajectory, and therefore formed consistent firing sequences during these events.

Common cells between the box and the track

We showed that many CA1 cells were active during rotation events in box sessions. We then aimed to understand how their activities in the box were related to their activities during self-running on the track. To this end, we first identified individual lap events when the animal ran on the track back and forth from one end to the other (two running trajectories). We then analyzed whether rotation-active cells were also active during lap-running events on a track trajectory. On average, there were 18.8 ± 0.9 laps per trajectory per track session and each lap lasted for 5.8 ± 0.2 s. Among the 2200 cells active either during rotation in the box or during lap-running on the track across all days, 601 cells were active only during rotation, 779 cells only during lap-running, and 820 cells (common cells) during both rotation and lap running on at least one track trajectory (Figure 3A). These numbers indicate that, overall, 58% of all rotation-active cells and 51% of all running-active cells were common cells. However, we found that, in the Trained-demo and Naïve-demo sessions combined, these proportions of common cells (69% of all rotation-active cells, 58% of all running-active cells) were significantly higher than those in other (Empty-track, No-track, Toy-car, Blocked-view) box sessions combined (48% of all rotation-active cells, p=3.4 × 10^–21, binomial test comparing with that of Trained- and Naïve-demo combined; 44% of all running-active cells, p=7.5 × 10^–6). This finding suggests that many active CA1 cells were 'cross-activated' between the box and the track in the presence of a demonstrator, either well-trained or naïve.

Figure 3

Download asset Open asset

We further quantified this phenomenon by computing the proportion of common cells expected from chance between each box session and a track trajectory, assuming that CA1 place cells in the box and on the trajectory were randomly and independently drawn from a common set of CA1 cells (Alme et al., 2014). We then compared the actual proportion with the chance proportion and defined a proportion difference index (PDI) to measure the strength of cross-activation. We found that the actual proportion was significantly higher than the chance proportion for the Trained-demo and Naïve-demo sessions, but not for others (Empty-track, No-track, Toy-car, Blocked-view) (Figure 3B). Similarly, the PDI was significantly greater in the Trained- and Naïve-demo sessions than other sessions (Figure 3C). This analysis indicates that, as long as and only when a demonstrator was present, was there cross-activation of CA1 cells significantly more than the chance level between the box and the track.

Similar firing sequences of common cells between the box and the track

After identifying the common cells, we next asked whether firing sequences of multiple common cells were preserved between rotation events in the box and lap-running events on the track. Indeed, we found that many sequences during lap-running also occurred in rotation events, either with the same or reverse order (Figure 4A and Figure 4—figure supplement 1). To quantify this, for each recording day we constructed a template firing sequence consisting of at least 5 common cells, by ordering their peak firing locations on each trajectory of the track (at most 2 templates per day). We then obtained the sequence of these cells within each rotation event (rotation sequence) in Pre- and Post-box that contained at least 5 active common cells, by ordering their peak firing angles. We quantified the similarity between a rotation sequence and a template by a circular matching score, given the circular nature of rotation events. In our method, a high similarity score means a sequence similar to a template with the same or reverse order. We also randomly shuffled a rotation sequence 1000 times and computed the score between each shuffled sequence and the template. If a rotation sequence yielded a score with any of the templates on the same day greater than the 95 percentile of the shuffle-generated scores, we referred to it as a matching sequence.

Figure 4 with 2 supplements see all

Download asset Open asset

We found that 125 out of 639 (20%) rotation sequences under Trained-demo and 73 out of 336 (22%) rotation sequences under Naïve-demo were matching sequences. In contrast, only 73 out of 712 (10%), 13 out of 116 (11%), 42 out of 419 (10%), and 22 out of 314 (7%) of rotation sequences under Empty-track, No-track, Toy-car, and Blocked-view were matching sequences, respectively. To assess how these numbers of matching sequences differed from the chance level, we generated 200 sets of randomized templates under a box condition, each by randomly shuffling cell identities in every template, and identified the number of matching sequences for each set of randomized templates, using the exactly same procedure as described. We used the proportion of randomized template sets that generated a greater number of matching sequences than our actual templates under a box condition as a measure of significance (p value). Relative to the shuffle-generated distribution of number of matching sequences, we found that the actual number of matching sequences was significant under Trained-demo and Naïve-demo, but not under Empty-track, No-track, Toy-car, or Blocked-view (Figure 4B). Furthermore, we also quantified the significance of matching sequences within each individual animal. In this case, we combined the rotation events under Trained-demo and Naïve-demo, in order to obtain sufficient samples for the analysis. We found a significant number of matching sequences consistently in 8 out of the 8 rats that yielded sufficient cells for the analysis (Figure 4—figure supplement 1). Therefore, our analysis indicates that firing sequences of common cells were preserved between lap-running and rotation events under Trained-demo and Naïve-demo conditions. Further analysis found that the precise firing times of common cells were not preserved (Figure 4—figure supplement 2), probably due to non-constant running speed on the track and non-constant turning speed in the box. The results nevertheless indicate that firing sequences of multiple common cells during lap-running were cross-activated in a significant number of rotation events in the box, but only when a demonstrator was present.

Experience independence of cross-activation

Our results so far demonstrate that the CA1 activity during a rat’s track running behavior was cross-activated in the box with the presence of a demonstrator on the track. A critical question is whether the track-running CA1 activity already started to appear during rotation events in the box prior to the very first experience on the track. Therefore, we analyzed CA1 cells from 5 rats in Pre-box on the first recording day under the Trained-demo condition, before they ran the track themselves for the very first time. As a key control, we also analyzed CA1 cells from other 4 rats in Pre-box on the first recording day under the Empty-track condition, prior to their first self-running on the track. We found that the actual proportion of common cells was significantly greater than the chance proportion only under Trained-demo, not under Empty-track (Figure 5A). Second, the actual number of matching sequences was also significantly greater than the chance level only under Trained-demo, not under Empty-track (Figure 5B). These results suggest that cross-activation of the track-running CA1 activity was present in the box session prior to any self-running experience, but only when a demonstrator was present.

Figure 5

Download asset Open asset

This cross-activation of place cell sequences prior to self-experience on the track raises the question of whether it was due to some kind of general, shared pre-existing activity patterns in the CA1 cell population as previous studies proposed (Dragoi and Tonegawa, 2011, 2013). The fact that the cross-activation only occurred under Trained-demo or Naïve-demo, but not under other conditions, suggests that this was not the case. To further demonstrate that the cross-activation was specific to the box, we conducted additional recordings of CA1 cells from 4 rats during rest sessions before (Pre-rest) and after (Post-rest) they ran the track themselves. During these rest sessions, the animals stayed on top of a small flower pot (20 cm diameter), placed at the center of an enclosed box with 4 high walls. We detected rotation events when animals moved around on top of the flower pot with high theta activity and then identified common cells that were active both in these rotation events and in lap-running events on the track. We found that actual proportion of common cells was similar to the chance proportion (Figure 6A) and the actual number of matching sequences was not significantly greater than the chance level (Figure 6B). The analysis suggests that cross-activation was unique to the box that had an opening to the nearby track where a demonstrator was present.

Figure 6

Download asset Open asset

Finally, although we have shown that cross-activation did not require the rat’s prior self-running experience on the track, a question remains whether it did change with the self-experience. To answer this question, we examined CA1 cells on the first and second day of recording (Day1, Day2) under the same condition of either Trained-demo or Naïve-demo. We first analyzed the experience dependence at the time scale of sessions, by comparing the CA1 activities between Pre-box and Post-box of Day1 and Day2 combined. We found no significant difference in the proportion of common cells between Pre-box and Post-box (using PDI as a measure, Figure 7A). The actual number of matching sequences was significantly greater than the chance level in both Pre-box and Post-box (Figure 7B). The data thus suggest that the cross-activation in Pre- and Post-box was largely similar. We then analyzed the experience dependence at a longer time scale by comparing these same measures between Day1 and Day2 (with Pre-box and Post-box combined on each day). We found that the cross-activation was also largely similar between Day1 and Day2 (Figure 7C,D). These results suggest that cross-activation of CA1 activity was insensitive to animals’ self-experience on the track.

Figure 7

Download asset Open asset

Improved behavioral performance and place field development on the track

The finding that cross-activation already occurred in Pre-box prior to self-running on the track prompted us to explore how the cross-activation impacted the rats’ behavior or their place fields on the track. We analyzed behavioral and place field parameters in the Track session on the first two days (Day1, Day2). These parameters were compared between the 5 rats that were recorded under the Trained-demo box condition (referred to as Trained-demo rats hereafter) and the 4 under the Empty-track box condition (Empty-track rats). In addition, 5 other rats were not recorded, but went through the same Pre-box, Track and Post-box procedure under either the Trained-demo (4 rats) or the Empty-track (1 rat) condition. The behaviors of these unrecorded rats on the track were included in the behavioral analysis. We need to point out again that all these Trained-demo and Empty-track rats had gone through a training procedure in the small box for 2–3 days prior to Day1, with and without the presence of a trained demonstrator, respectively. The behavior on the track was quantified by mean running speed and number of running laps per trajectory. Both parameters in the Trained-demo rats were significantly greater than those in the Empty-track rats on both Day1 and Day2 (Figure 8A), suggesting an improvement in track running performance in the former group. For place field analysis, since rats were not exposed to the track before Day1, we focused on how the development of novel place fields on the track differed between the Trained-demo and Empty-track rats. Because place fields developed quickly during first a few laps (Frank et al., 2004) and place cells in the Trained-demo and Empty-track rats should be compared using the same number of laps, we quantified place cell properties within the first 4 laps of each trajectory on each day, by spatial information (an overall measurement of spatial specificity; Skaggs et al., 1993) and lap-consistency (a measure of a cell’s firing reliability from lap to lap, similar to rotation consistency) (Cheng and Ji, 2013). We found that place cell firing on the track in the Trained-demo rats appeared to be more dispersed than those in the Empty-track rats on both Day1 and Day2 (Figure 8B). Indeed, spatial information of cells in the Trained-demo rats were significantly greater than those in the Empty-track rats on both Day1 and Day2 (Figure 8C). This improvement in spatial information was small on Day1 (15% increase in median spatial information from Empty-track to Trained-demo) and became modest on Day2 (38%). Lap consistency of cells in the Trained-demo rats also appeared greater than those in the Empty-track rats, but did not reach statistical significance (Figure 8D). The result suggests a mild speedup in novel place field development in the Trained-demo rats. Taken together, our data suggest that cross-activation was accompanied by an improvement in track-running performance and place field development.

Figure 8

Download asset Open asset

Ripple-associated replay in the box

We have focused on the rotation events in the small box, when CA1 LFPs displayed prominent theta oscillations with little high-frequency ripple oscillations (Figure 1—figure supplement 1). However, ripples did occur in the small box (Figure 9A), mostly when animals were immobile. It is known that place cell patterns on the track are often reactivated (replayed) within short ripple events of 50–400 ms during pausing/resting on the track (Foster and Wilson, 2006; Diba and Buzsáki, 2007; Karlsson and Frank, 2009). Here we examined whether the place cell templates on the track were also replayed within ripple events in the box during Pre-box and Post-box under various box conditions. We first identified individual ripple events in each box session. We found that the number of ripple events per unit time was slightly, but significantly increased from Pre-box (0.28 ± 0.03 ripples/s) to Post-box (0.31 ± 0.02 ripples/s) sessions (t = 2.7, p=0.009, paired t-test; N = 72 sessions with all conditions combined). However, the number of ripples was not significantly different across different conditions in either Pre-box (F_(5,66)= 2.0, p=0.09, One-way ANOVA) or Post-box (F_(5,66) = 1.9, p=0.09).

Figure 9

Download asset Open asset

We then identified those ripple events when a template sequence on a track trajectory was replayed. Similarly to the identification of matching sequences within rotation events, we used a sequence matching method to determine whether the firing sequence within a candidate ripple event (defined with at least 5 active cells) matched a template sequence, but using a rank-correlation score instead of a circular matching score (see Materials and methods). Also similarly, we counted the number of replay events under a box condition and assessed its significance by comparing it to the distribution obtained by random shuffling of templates. First, with Pre-box and Post-box sessions combined, we found a significant number of replay events under all conditions (Figure 9B). The percentage of replay events among all candidate events was 7.7% under Trained-demo, 12.5% under Naïve-demo, 7.8% under Empty-track, 14.6% under No-track, 9.3% under Toy-car, and 11.2% under Blocked-view. The result suggests that ripple-associated replay occurred in the box, consistent with the previously found 'remote replay' (Karlsson and Frank, 2009), but the replay was comparable with or without a demonstrator. Second, we compared the replay between Pre-box and Post-box with all box conditions combined (Figure 9C). We found that the number of replay events was significant in Post-box (9.2%), but did not reach the statistical significance in Pre-box (6.7%), suggesting that direct track experience enhanced ripple-associated replay. Finally, we specifically analyzed the replay between the rats under Trained-demo and Empty-track on the first day’s Pre-box and Post-box sessions. Under both conditions, although there appeared an increase in replay from Pre-box (Trained-demo: 4.7%; Empty-track: 4.7%) to Post-box Trained-demo: 7.7%; Empty-track: 6.8%), none of the number of replay events reached the statistical significant level (Figure 9D), suggesting the absence of replay before the first track running experience and weak replay after. Our data are thus consistent with the idea that, in contrast to the matching sequences within rotation events, ripple-associated replay was experience-dependent and was not specific to the presence of a demonstrator.

Lack of spatial specificity to demonstrators’ positions

Besides the rotation events, animals in the small box spent much of the time facing the track (Figure 1C). This raises an interesting question of whether the CA1 cells in the recorded rats in the box were responding to the demonstrator’s positions on the track. Raster plots of CA1 cells’ spikes versus the demonstrator’s track positions did not show sign of spatial tuning (Figure 10A). We then used spatial information and lap consistency as quantitative measurements of spatial tuning for each cell active in a box session to the demonstrator’s positions. These measures were z-scored relative to those obtained using randomly shuffled spikes of the same cell (Haggerty and Ji, 2015). The mean z-scored spatial information and lap-consistency of all active cells under either Trained-demo or Naïve-demo were not significantly different from the chance level of 0 (Figure 10B,C). We found that 0.6% of cells in the Trained-demo rats and 0.4% in the Naïve-demo rats had significant spatial information (Z > 1.645, p<0.05), and 0% of cells in the Trained-demo rats and 0% in the Naïve-demo rats had significant lap consistency. These numbers were far below the 5% expected from chance. Therefore, our data did not support the idea that CA1 cells in the box respond to the demonstrator’s positions on the track.

Figure 10

Download asset Open asset

To explore the neural basis of local enhancement in social learning, we have analyzed CA1 place cells in rats as they stayed in a small box while a demonstrator running on a nearby track and as they ran the track themselves. We found that CA1 cells formed consistent firing sequences during rotation events in the box. A group of these cells (common cells) were also active during lap-running events on the track with similar firing patterns, as shown by similar multi-cell firing sequences across rotation and lap-running events. This cross-activation was specific to the box and took place only when a demonstrator (either Trained-demo or Naive-demo) was present on the track. Importantly, the cross-activation appeared before the animals’ very first running experience on the track, as well as after the experience, and was accompanied by improved behavioral performance and faster place field development on the track. Our data thus show that place cell activities representing a track can be activated by staying in a separate, nearby box while another rat running on the track. This finding supports our hypothesis that presence of social subjects in an environment can enhance the activation of an observer’s hippocampal place cell activity patterns that encode the environment.

We found that CA1 cells form firing sequences during rotation events in the box. Given the small size of the box, this finding is somewhat unexpected. The box size is only slightly bigger than the body length of a typical adult rat, which restricts typical walking behavior but permits body rotation. This box size is also smaller than the width of typical place fields (20–50 cm) in regular environments (Wilson and McNaughton, 1993). If CA1 cells in the box just fire like regular place cells with normal place fields, they would fire mostly throughout rotation events without a clear pattern. However, we found that different CA1 cells fire at different specific locations of even such a small box and multiple cells fire with consistent sequences during rotation events. Then, what gives rise to these firing sequences? We used head direction tuning (circular rate curve) to quantify rotation-consistent cells during rotation events, due to the circular nature of the rotation. But these cells are unlikely true head direction cells, because they constitute a large percentage of recorded cells in our experiment, but true head direction cells, if any, are very rare in CA1 (5%) (Leutgeb et al., 2000). Given that rotation-consistent CA1 cells fire at consistent head positions of the rotation events and many of them fire at the same positions even when animals rotate from two opposite directions, it is possible that these cells are still place cells that are tuned to head positions, but with place fields 'squeezed' into a small box. This implies that, even if an animal is confined to a small space, place cells are tuned to specific locations within the space in active behavior and they fire with a sequence when the animal actively moves his head along a trajectory. This interpretation is consistent with the well-known phenomenon that place fields shrink as the animal’s space shrinks (Gothard et al., 1996; O'Keefe and Burgess, 1996). Alternatively, the firing sequences during rotation could be internally generated to represent internal information or time, similarly as those occur when rats are running a wheel or a treadmill before resuming maze-running (Pastalkova et al., 2008; Wang et al., 2015; MacDonald et al., 2011, 2013). Although we cannot rule out this possibility, the consistent head-position tuning suggests that these sequences in the box are at least strongly modulated by spatial cues.

We found that CA1 place cell activities were cross-activated between lap-running events on the track and rotation events in the box when a demonstrator (Trained-demo and Naïve-demo) was present on the same track. This cross-activation is demonstrated by a significant number of common cells between rotation and lap-running events and their matched multi-cell firing sequences between the two types of events. Under control conditions (Empty-track, No-track, Toy-car, or Blocked-view) or between the track and an enclosed rest box, the number of commons cells is similar to the chance level, suggesting independent, random remapping under these conditions, consistent with a previous report (Alme et al., 2014). Using the sequence matching method that utilizes high-order firing patterns of multiple common cells (Lee and Wilson, 2002; Ji and Wilson, 2007), we found a significant number of sequences during rotation events in the box that match the template sequences on the track. However, we need to point out that this cross-activation is partial, that is, only for a set of common cells, not all place cells active on the track or in the box. In fact, common cells comprise about 58% of place cells on the track under Trained-demo or Naive-demo condition and about 20% of the sequences of these cells during rotation events matched with the track templates. Since we simultaneously recorded only a small sample of CA1 cells, the actual percentage of matching sequences in CA1 is unknown, but likely higher. In any case, our data means that the overall spatial presentations of the track and the box remain relatively separated. However, a significant number of firing patterns are common in both presentations. This finding suggests the existence of common firing motifs consisting of a subset of place cells that can be activated in different environments or in different behavior (Ravassard et al., 2013). Importantly, our results reveal that the presence of social subjects can enhance the cross-activation of such common firing motifs of place cells across different environments.

An important question is, then, what drives the cross-activation of common firing motifs. The fact that cross-activation shows little experience-dependence suggests that it may be driven by sensory cues rather than experience. In our experimental setup, the small box had high walls on its 3 sides with the other side opening toward the track. This configuration provides common sensory input, especially distal visual cues in the room, when the animal is on the track and in the box. It is possible that this common external input biases common motifs to emerge in both environments. Interestingly, the rats’ rotation events consistently started and ended at the opening side, consistent with this idea that the initiation of common firing motifs could be triggered by common external visual cues. Furthermore, under the Blocked-view condition where there was a demonstrator on the track but external room cues and the demonstrator were invisible to the rat in the small box, cross-activation did not occur, supporting a crucial role of common visual cues in cross-activation. Presumably, cross-activation could occur during any active behavior when the rat is in the small box, as long as a demonstrator is present on the track and visible. However, given that no specific task is given to the rat in the box, the animal’s behavior is mostly irregular whether or not the animal is facing the opening side, which renders mostly irregular CA1 activity patterns. But the small size of the box produces one active behavior that is relatively consistent across animals and sessions: rotation behavior. According to our interpretation that the firing of CA1 cells during rotation mainly results from 'squeezed' place cell activity, the cross-activation of track templates during rotation events suggests a topological similarity between firing motifs representing the track-running trajectories and the head movement trajectories in the box. This is consistent with the proposal that place cell patterns may encode common topology in seemingly very different environments (Dabaghian et al., 2014).

Another important question is why cross-activation of common firing motifs occur in the box only when a demonstrator is visible on the track. One explanation is maybe related to the fact that the track is a separate environment from the box (both elevated from the floor). Without a demonstrator on the track, the track may appear as an object to the rat in the box and thus becomes part of the visual cues that drive the place cells in the box. When there is a demonstrator moving on the track, the rat in the box may 'realize' that the track is another space within the same room that can be explored. This sense of track as an explorable space might bias the drive to a subset of place cells in both the box and the track toward shared distal room cues, and lead to the emergence of common firing motifs and thus cross-activation. However, it remains unclear whether the actual moving of the demonstrator along the track is necessary for the rat in the small box to 'realize' that the track is explorable, or merely the presence of the demonstrator on the track, e.g. staying at one end of the track, may be sufficient. Interestingly, our data show that both a well-trained and a naive demonstrator are capable of triggering cross-activation, which suggests that cross-activation does not require precise behavioral performance of the demonstrator. Also, in our data we have not found any evidence that CA1 cells of the rat in the box responded to the demonstrator’s positions on the track. Therefore, the rat in the box may not necessarily understand the precise action of the demonstrator, but the presence of the demonstrator itself is sufficient to signal the spatial nature of the track (instead of merely a visual cue in the room), which leads to the cross-activation. In this way, animals can learn about an environment, without actual self-exploration. Indeed, it is known that rodents can learn a variety of tasks by observing a conspecific (Zentall and Levine, 1972; Leggio et al., 2000; Jeon et al., 2010), but how they do so is unknown. One theory is that, instead of imitation learning that requires detailed understanding of conspecifics’ actions, rodents rely on local enhancement or social facilitation as a general scheme of social learning (Heyes and Galef, 1996; Zentall, 2006; Zajonc, 1965). Our finding that cross-activation can be facilitated by both trained and naïve conspecifics is consistent with the local enhancement theory and may provide a specific neural mechanism of local enhancement in social learning.

Our data also suggest possible impacts of local enhancement on animals’ behavior and subsequent hippocampal activities. After 2–3 days’ experience in the box with the presence of a well-trained demonstrator on the track, the animals’ performance on the track is improved, compared to those animals staying in the box for 2–3 days without a demonstrator. More interestingly, their place fields during the first a few laps on the track seem already better tuned, suggesting that social observation facilitates the development of novel place fields. It is possible that this facilitation is related to the cross-activation in the box. However, a causal relationship between the two and how they are related to behavioral performance require further studies. Although ripple-associated replay may or may not take place prior to experience in other tasks (Dragoi and Tonegawa, 2011, 2013; Silva et al., 2015), ripple replay does not occur in the box before the first self-running on the track and are similar across various box conditions in our experiment. Our data thus suggest that ripple replay in our task requires direct experience and is not impacted by social observation.

In a broader sense, this study contributes to our understanding of how hippocampal place cells are generally impacted by social behavior. Previous studies find that introducing a rat to an open box induces no change or a small modification in CA1 place cell activities, but causes remapping of cells in the neighboring CA2 area (von Heimendahl et al., 2012; Alexander et al., 2016). Whether the demonstrator causes CA1 place cells to remap in the small box was not directly investigated in our experiment. However, the fact that consistent place cell patterns occur within rotation events in the box under all box conditions suggests that the presence of a demonstrator per se would not dramatically alter place cell activities in the box, consistent with these previous studies. Instead, we found that cross-activation of common firing motifs emerges with the presence of a demonstrator on the track. This effect could be due to our task design of using a small box with one opening side, which boosts the impact of distal visual cues and the demonstrator on place cells in the box, an effort not made in previous studies. The unique influence of the demonstrator on cross-activation may come from the modification of upstream CA2 activities by social subjects. Future studies will reveal what factors and pathways are important to social influence of hippocampal place cell patterns and how they contributes to social behavior.

1. 1. Alexander GM
   2. Farris S
   3. Pirone JR
   4. Zheng C
   5. Colgin LL
   6. Dudek SM

   (2016) Social and novel contexts modify hippocampal CA2 representations of space

   Nature Communications 7:10300.

   https://doi.org/10.1038/ncomms10300

   • PubMed
   • Google Scholar
2. 1. Alme CB
   2. Miao C
   3. Jezek K
   4. Treves A
   5. Moser EI
   6. Moser M-B

   (2014) Place cells in the hippocampus: Eleven maps for eleven rooms

   PNAS 111:18428–18435.

   https://doi.org/10.1073/pnas.1421056111

   • Google Scholar
3. Book

   1. Bandura A

   (1997)

   Social Learning Theory

   New York: General Learning Press.

   • Google Scholar
4. 1. Burgess N
   2. O'Keefe J

   (2003) Neural representations in human spatial memory

   Trends in Cognitive Sciences 7:517–519.

   https://doi.org/10.1016/j.tics.2003.10.014

   • PubMed
   • Google Scholar
5. 1. Buzsáki G
   2. Horváth Z
   3. Urioste R
   4. Hetke J
   5. Wise K

   (1992) High-frequency network oscillation in the hippocampus

   Science 256:1025–1027.

   https://doi.org/10.1126/science.1589772

   • PubMed
   • Google Scholar
6. 1. Buzsáki G

   (2002) Theta oscillations in the hippocampus

   Neuron 33:325–340.

   https://doi.org/10.1016/S0896-6273(02)00586-X

   • PubMed
   • Google Scholar
7. 1. Cheng J
   2. Ji D

   (2013) Rigid firing sequences undermine spatial memory codes in a neurodegenerative mouse model

   eLife 2:e00647.

   https://doi.org/10.7554/eLife.00647

   • PubMed
   • Google Scholar
8. 1. Colgin LL
   2. Moser EI
   3. Moser MB

   (2008) Understanding memory through hippocampal remapping

   Trends in Neurosciences 31:469–477.

   https://doi.org/10.1016/j.tins.2008.06.008

   • PubMed
   • Google Scholar
9. 1. Dabaghian Y
   2. Brandt VL
   3. Frank LM

   (2014) Reconceiving the hippocampal map as a topological template

   eLife 3:e03476.

   https://doi.org/10.7554/eLife.03476

   • PubMed
   • Google Scholar
10. 1. Diba K
    2. Buzsáki G

    (2007) Forward and reverse hippocampal place-cell sequences during ripples

    Nature Neuroscience 10:1241–1242.

    https://doi.org/10.1038/nn1961

    • PubMed
    • Google Scholar
11. 1. Dragoi G
    2. Tonegawa S

    (2011) Preplay of future place cell sequences by hippocampal cellular assemblies

    Nature 469:397–401.

    https://doi.org/10.1038/nature09633

    • PubMed
    • Google Scholar
12. 1. Dragoi G
    2. Tonegawa S

    (2013) Distinct preplay of multiple novel spatial experiences in the rat

    PNAS 110:9100–9105.

    https://doi.org/10.1073/pnas.1306031110

    • Google Scholar
13. 1. Foster DJ
    2. Wilson MA

    (2006) Reverse replay of behavioural sequences in hippocampal place cells during the awake state

    Nature 440:680–683.

    https://doi.org/10.1038/nature04587

    • PubMed
    • Google Scholar
14. 1. Frank LM
    2. Stanley GB
    3. Brown EN

    (2004) Hippocampal plasticity across multiple days of exposure to novel environments

    Journal of Neuroscience 24:7681–7689.

    https://doi.org/10.1523/JNEUROSCI.1958-04.2004

    • PubMed
    • Google Scholar
15. 1. Gothard KM
    2. Skaggs WE
    3. McNaughton BL

    (1996)

    Dynamics of mismatch correction in the hippocampal ensemble code for space: interaction between path integration and environmental cues

    Journal of Neuroscience 16:8027–8040.

    • PubMed
    • Google Scholar
16. 1. Haggerty DC
    2. Ji D

    (2015) Activities of visual cortical and hippocampal neurons co-fluctuate in freely moving rats during spatial behavior

    eLife 4:08902.

    https://doi.org/10.7554/eLife.08902

    • PubMed
    • Google Scholar
17. Book

    1. Heyes CM
    2. Galef JBG

    (1996)

    Social Learning in Animals: The Root of Culture

    New York: Academic Press.

    • Google Scholar
18. 1. Jeon D
    2. Kim S
    3. Chetana M
    4. Jo D
    5. Ruley HE
    6. Lin SY
    7. Rabah D
    8. Kinet JP
    9. Shin HS

    (2010) Observational fear learning involves affective pain system and Cav1.2 Ca2+ channels in ACC

    Nature Neuroscience 13:482–488.

    https://doi.org/10.1038/nn.2504

    • PubMed
    • Google Scholar
19. 1. Ji D
    2. Wilson MA

    (2007) Coordinated memory replay in the visual cortex and hippocampus during sleep

    Nature Neuroscience 10:100–107.

    https://doi.org/10.1038/nn1825

    • PubMed
    • Google Scholar
20. 1. Karlsson MP
    2. Frank LM

    (2009) Awake replay of remote experiences in the hippocampus

    Nature Neuroscience 12:913–918.

    https://doi.org/10.1038/nn.2344

    • PubMed
    • Google Scholar
21. 1. Lee AK
    2. Wilson MA

    (2002) Memory of sequential experience in the hippocampus during slow wave sleep

    Neuron 36:1183–1194.

    https://doi.org/10.1016/S0896-6273(02)01096-6

    • PubMed
    • Google Scholar
22. 1. Leggio MG
    2. Molinari M
    3. Neri P
    4. Graziano A
    5. Mandolesi L
    6. Petrosini L

    (2000) Representation of actions in rats: The role of cerebellum in learning spatial performances by observation

    PNAS 97:2320–2325.

    https://doi.org/10.1073/pnas.040554297

    • Google Scholar
23. 1. Leutgeb S
    2. Leutgeb JK
    3. Barnes CA
    4. Moser EI
    5. McNaughton BL
    6. Moser MB

    (2005) Independent codes for spatial and episodic memory in hippocampal neuronal ensembles

    Science 309:619–623.

    https://doi.org/10.1126/science.1114037

    • PubMed
    • Google Scholar
24. 1. Leutgeb S
    2. Ragozzino KE
    3. Mizumori SJ

    (2000) Convergence of head direction and place information in the CA1 region of hippocampus

    Neuroscience 100:11–19.

    https://doi.org/10.1016/S0306-4522(00)00258-X

    • PubMed
    • Google Scholar
25. 1. MacDonald CJ
    2. Carrow S
    3. Place R
    4. Eichenbaum H

    (2013) Distinct hippocampal time cell sequences represent odor memories in immobilized rats

    Journal of Neuroscience 33:14607–14616.

    https://doi.org/10.1523/JNEUROSCI.1537-13.2013

    • PubMed
    • Google Scholar
26. 1. MacDonald CJ
    2. Lepage KQ
    3. Eden UT
    4. Eichenbaum H

    (2011) Hippocampal "time cells" bridge the gap in memory for discontiguous events

    Neuron 71:737–749.

    https://doi.org/10.1016/j.neuron.2011.07.012

    • PubMed
    • Google Scholar
27. 1. Meltzoff AN
    2. Kuhl PK
    3. Movellan J
    4. Sejnowski TJ

    (2009) Foundations for a new science of learning

    Science 325:284–288.

    https://doi.org/10.1126/science.1175626

    • PubMed
    • Google Scholar
28. 1. Mou X
    2. Ji D

    (2017) Social Observation Task in a Linear Maze for Rats

    BIO-PROTOCOL 7:e2361.

    https://doi.org/10.21769/BioProtoc.2361

    • Google Scholar
29. 1. O'Keefe J
    2. Burgess N

    (1996) Geometric determinants of the place fields of hippocampal neurons

    Nature 381:425–428.

    https://doi.org/10.1038/381425a0

    • PubMed
    • Google Scholar
30. 1. O'Keefe J
    2. Dostrovsky J

    (1971) The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat

    Brain Research 34:171–175.

    https://doi.org/10.1016/0006-8993(71)90358-1

    • PubMed
    • Google Scholar
31. 1. Pastalkova E
    2. Itskov V
    3. Amarasingham A
    4. Buzsáki G

    (2008) Internally generated cell assembly sequences in the rat hippocampus

    Science 321:1322–1327.

    https://doi.org/10.1126/science.1159775

    • PubMed
    • Google Scholar
32. 1. Ravassard P
    2. Kees A
    3. Willers B
    4. Ho D
    5. Aharoni D
    6. Cushman J
    7. Aghajan ZM
    8. Mehta MR

    (2013) Multisensory control of hippocampal spatiotemporal selectivity

    Science 340:1342–1346.

    https://doi.org/10.1126/science.1232655

    • PubMed
    • Google Scholar
33. 1. Silva D
    2. Feng T
    3. Foster DJ

    (2015) Trajectory events across hippocampal place cells require previous experience

    Nature Neuroscience 18:1772–1779.

    https://doi.org/10.1038/nn.4151

    • PubMed
    • Google Scholar
34. Book

    1. Skaggs WE
    2. McNaughton BL
    3. Gothard KM
    4. Markus EJ

    (1993)

    An information-theoretic approach to deciphering the hippocampal code

    In: Hanson SJ, Cowan JD, Giles CJ, editors. Advances in Neural Information Processing Systems. San Mateo: Morgan Kaufmann. pp. 1030–1037.

    • Google Scholar
35. 1. von Heimendahl M
    2. Rao RP
    3. Brecht M

    (2012) Weak and nondiscriminative responses to conspecifics in the rat hippocampus

    Journal of Neuroscience 32:2129–2141.

    https://doi.org/10.1523/JNEUROSCI.3812-11.2012

    • PubMed
    • Google Scholar
36. 1. Wang Y
    2. Romani S
    3. Lustig B
    4. Leonardo A
    5. Pastalkova E

    (2015) Theta sequences are essential for internally generated hippocampal firing fields

    Nature Neuroscience 18:282–288.

    https://doi.org/10.1038/nn.3904

    • PubMed
    • Google Scholar
37. 1. Wilson MA
    2. McNaughton BL

    (1993) Dynamics of the hippocampal ensemble code for space

    Science 261:1055–1058.

    https://doi.org/10.1126/science.8351520

    • PubMed
    • Google Scholar
38. 1. Zajonc RB

    (1965) Social facilitation

    Science 149:269–274.

    https://doi.org/10.1126/science.149.3681.269

    • Google Scholar
39. 1. Zentall TR
    2. Levine JM

    (1972) Observational learning and social facilitation in the rat

    Science 178:1220–1221.

    https://doi.org/10.1126/science.178.4066.1220

    • PubMed
    • Google Scholar
40. 1. Zentall TR

    (2006) Imitation: definitions, evidence, and mechanisms

    Animal Cognition 9:335–353.

    https://doi.org/10.1007/s10071-006-0039-2

    • PubMed
    • Google Scholar

Author details

1. Xiang Mou

   Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, United States

   Contribution

   XM, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0002-8579-7316

2. Daoyun Ji

   1. Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, United States
   2. Department of Neuroscience, Baylor College of Medicine, Houston, United States

   Contribution

   DJ, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   dji@bcm.edu

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0003-4115-5888

• 1,962

  views

• 441

  downloads

• 22

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.