Introduction

Large-scale multi-photon imaging and holographic stimulation offers the ability to accurately track and record from the same neuronal populations over extended periods at high spatiotemporal resolution, record from axons, dendrites, and dendritic spines, and stimulate neurons with single-cell precision (Bando et al., 2021; Chaigneau et al., 2016; Dombeck et al., 2010; Dong et al., 2021; Emiliani et al., 2015; Fan et al., 2023; Hattori & Komiyama, 2022; Heer & Sheffield, 2024; Hernandez et al., 2016; Krishnan et al., 2022; Russell et al., 2022; Sheffield & Dombeck, 2015). These techniques facilitate the study of neural circuit dynamics, synaptic and plasticity mechanisms and enable the real-time manipulation of neural activity, all of which are crucial for understanding the mechanisms underlying memory formation and recall in the brain. However, head-fixation of animals, which is generally required to utilize these techniques, imposes constraints on the range of testable behaviors compared to freely moving conditions (Guo et al., 2014; Juczewski et al., 2020). This limitation is particularly evident in the Pavlovian contextual fear conditioning (CFC) task, a laboratory method for testing associative learning, fear memory formation, and recall (M. E. Bouton, 1993; Michael S. Fanselow, 1980, 1990; J. J. Kim & Fanselow, 1992; LeDoux, 2000; Maren, 2001; Maren et al., 2013).

In the traditional CFC paradigm for freely moving animals, an animal is placed in an environment (conditioned stimulus, CS) that is paired with an aversive stimulus, such as a mild foot shock (unconditioned stimulus, US). Animals are then removed from the environment. When reintroduced to the environment, animals exhibit a species-specific conditioned response, such as fearful freezing behavior in rodents, if they successfully associate the shock-paired environment with the aversive stimulus. The CFC paradigm is one of the most basic conditioning procedures for freely moving animals— it’s simple and robust (Michael S. Fanselow, 1980, 1990). Despite its longstanding use in studies involving freely behaving animals to explore questions on learning and memory (Anagnostaras et al., 2001; M. E. Bouton & King, 1983; Mark E. Bouton & Bolles, 1980; Debiec et al., 2002; M. S. Fanselow & Bolles, 1979; Fendt & Fanselow, 1999; Gewirtz et al., 2000; Holland & Bouton, 1999; Jeansok J. Kim & Jung, 2006; Maren et al., 2013; Stephen Maren & Holt, 2000; Pezze & Feldon, 2004), it has been noticeably absent in studies involving head-fixed animals, with some exceptions.

Two prior studies (Lovett-Barron et al., 2014; Rajasethupathy et al., 2015) reported a version of CFC in head-fixed mice with aversive air puffs as the US. These studies reported the conditioned response animals displayed as lick suppression, not freezing. Although lick suppression after fear conditioning has also been observed in freely moving animals (Leaf & Leaf, 1966; Leaf & Muller, 1965), it has been shown to result from increased freezing (Mark E. Bouton & Bolles, 1980). Therefore, relying solely on lick suppression in the head-fixed preparation makes it more challenging to compare with the extensive previous work in freely moving animals, where freezing is measured as the predominant conditioned fear response. Furthermore, the aforementioned head-fixed CFC studies also had the limitation of simultaneously having both a water reward and an aversive stimulus within the context, potentially confounding the results and complicating the interpretation. Hence, we aimed to develop a paradigm where freezing is the conditioned response and without the presence of a reward within the conditioned context, making it directly comparable to freely moving animals.

Here, we describe three different versions of a contextual fear-conditioning paradigm for head-fixed mice that resulted in freezing as the conditioned response. The context or CS was a virtual reality (VR)-based environment, which head-fixed mice explored by running on a cylindrical treadmill. In a previous study (Ratigan et al., 2023), we demonstrated the feasibility of this approach by combining it with imaging of thalamic axons innervating the hippocampus using a two-photon microscope. The present manuscript provides a comprehensive description and fundamentals of the behavioral framework, to enable the research community to reproduce and adapt our behavior paradigm. Our main aim is to show that VR-based CFC can elicit freezing responses similar to those seen in freely moving animals, while also examining how key parameters affect this response. Though we explored some important variables, this paper does not attempt to comprehensively investigate all parameters necessary for successful VR-based CFC. Additionally, we performed two-photon imaging of a large population of hippocampal CA1 cells during fear conditioning and tracked the same cells during memory recall across days. We analyzed the impact of our paradigm on place cells in the hippocampus and compared these findings to results from CFC in freely moving animals.

Results

A paradigm for contextual fear conditioning in head-fixed mice exploring virtual reality environments

To establish the contextual fear conditioning protocol (Figure 1A-C), head-fixed mice were water-restricted and trained to run on a treadmill and navigate VR environments for water rewards, as we have demonstrated previously (Dong et al., 2021; Krishnan et al., 2022; Krishnan & Sheffield, 2023; Ratigan et al., 2023). Forward movement on the treadmill translates to forward movement in VR, allowing mice to navigate the VR environment. The environments used were 2m-long linear tracks rich in visual cues (Figure 1). Reaching the track’s end triggered a water reward and defined a completed trial/lap. Mice were then virtually teleported back to the track’s start for the next lap. Unlike freely moving mice that tend to explore new environments, head-fixed mice placed on a treadmill do not automatically start exploring VR environments. However, they can be trained through reinforcement learning using water restriction and water rewards (Dombeck et al., 2010; Dong et al., 2021; Krishnan et al., 2022; Ratigan et al., 2023; Sheffield & Dombeck, 2015).

Figure 1.

Training to navigate VR environments on a treadmill (Stage 1), using water rewards, was achieved in approximately 10-14 days, after which well-trained mice could run several laps per minute. We set a criterion of >3 laps per minute (i.e., a lap velocity of 0.1 - 0.2 m/s) for mice to move on to the next stage (Stage 2). Once the mice reached this stage (Stage 2), the following day, we added a coat to their tails (Figure 1B). We developed a lightweight, wearable conductive apparatus for the mouse’s tail to ensure minimal discomfort for the mouse. We named this apparatus the “tail-coat” (Figure 1D). This tail-coat enabled us to administer mild electric shocks to the mouse’s tail (see Methods).

When the tail-coat was added, some mice reduced their running speed or refused to move. Some recovered the next day, but if their running speed did not reach at least three laps per minute (∼20% of mice failed to meet this criterion), they were not advanced to the next stage. Mice that maintained this running speed advanced to Stage 3. Data from mice that advanced with minimal changes to their lap running behavior after adding the tail-coat are shown in Supplementary Figure 1A.

The following day (Stage 3), mice were placed in the same VR environment with the tail-coat, but the water reward was removed (Figure 1B). Mice that continued to run at least three laps per minute with the tail-coat and without water reward advanced to the experiment stage (Stage 4). Around 40% of the mice dropped out at this stage. High levels of lap running during training were crucial to ensure mice were undeterred by both the tail-coat and the reward removal (Krishnan et al., 2022; Krishnan & Sheffield, 2023). Our criteria for consistent running behavior with minimal pauses throughout these three stages of training ensured that any baseline freezing behavior was minimal, similar to freely moving animals, and allowed us to attribute freezing after CFC to fear conditioning. In summary, our criteria for selecting mice for the experimental stage of contextual fear conditioning required them to run consistently (>3 laps per minute) with 1) rewards, then 2) with a tail-coat, and then 3) without rewards plus the tail-coat. This threshold resulted in about 40% of the mice proceeding to the experimental stage. Once mice qualified for the experimental stage, they were no longer excluded from analysis due to their behavior. We tested three versions of the head-fixed CFC paradigm. In Paradigm 1, the neutral environment was the training VR environment, and a new VR environment was paired with shocks. In Paradigms 2 and 3, both the neutral and shock-paired environments were new and distinct from the training environment to better control for environmental familiarity. In Paradigms 1 and 2, the tail-coat was removed on the days following CFC, whereas in Paradigm 3, it remained on the mouse on all post-CFC days. With these paradigms, we varied some parameters that may influence fear memory recall, contextual discrimination, fear extinction, and their associated behaviors in head-fixed VR-based CFC paradigms. We detail these variations in the following sections, starting with Paradigm 1.

When tested for contextual fear memory recall, mice displayed increased freezing behavior in the shocked-paired VR environment.

On Day 0 of the experimental stage in Paradigm 1 (CFC day), mice (n = 27, 25 males and 2 females) were exposed to the same VR environment we used for training, referred to as the Familiar VR from here onwards (Figure 1C), for 10 minutes. We then switched them to a new VR environment (CFC VR). Following baseline exploration of the CFC VR for 10 minutes, we repeatedly applied mild-electric shocks (4-12 shocks, 0.5-1mA in amplitude, 1s long) to the mouse’s tail. We call the period when shocks were delivered “during CFC”. Shocks were pseudo-randomly administered with respect to the animal’s position in the CFC VR, with an inter-shock interval (ISI) of 1 minute. To ensure that the shocks were associated with the context and not a specific position in the environment, no visual cues were explicitly paired with the shock. We quantified the animal’s response to the shocks and observed an immediate increase in running speed following each shock (Supplementary Figure 1B). This was consistent across all mice and shocks and was a reliable indicator that the apparatus was working correctly. Sixty seconds after the last shock was administered, the VR screens were switched off, the tail-coat was removed, and mice were returned to their home cages. The next day (Fear memory recall, Day 1), we tested for recall of the learned association between the VR environment and the fearful stimulus. Mice were exposed to both VRs (familiar VR and CFC VR) for five minutes each in a counterbalanced manner. Water reward remained absent on recall days; however, in this paradigm, the tail-coat was no longer present.

Before fear conditioning, mice exhibited similar behavior in both the Familiar and CFC VRs. We evaluated the time taken to complete a lap and the freezing behavior (defined as time points when the animal’s instantaneous velocity dropped was 0 cm/s). Neither metric revealed a significant difference between the two VRs before shock administration (Figure 2A, Supplementary Figure 1D-E). However, after CFC, when the animals were reintroduced to the VRs on recall Day 1, we observed a significant increase in freezing and time taken to complete a lap in the CFC VR compared to the Familiar VR (Figure 2A-D). As soon as mice transitioned to the CFC VR (either from dark or after the Familiar VR) on the recall day, we observed behaviors such as slowing down, freezing, moving backward, and hesitation to move forward (Supplementary Videos 1-3). This response was markedly different from their behavior in the Familiar VR. This difference in behavior can be observed clearly in most mice on the very first lap (Figure 2A-B). Since we removed the reward, we didn’t find licking behavior to be indicative of recall behavior (Supplementary Figure 1C). However, some mice displayed licking behavior in the Familiar VR on Day 0 before fear conditioning, which was absent on Day 1. This is likely because mice had previously received a water reward in the Familiar VR. We will address this caveat in Paradigms 2 and 3.

Figure 2.

Supplementary Figure 1.

Supplementary Figure 2.

To quantify the distinct conditioned responses observed (Figure 2A-B, Supplementary Videos 1-3), we used two metrics: freezing behavior, defined as periods when the animal’s instantaneous velocity was 0 cm/s, and time taken to complete a lap, which captured the avoidance-like behaviors (such as slowing down, moving backward, and hesitation to move forward) as well as freezing. When we quantified the freezing behavior in mice, we found it to be, on average, high throughout the five minutes that mice spent in the CFC VR on recall Day 1 (Figure 2C-D, average freezing (%), mean ± 95% Confidence Intervals (CI), Familiar VR: 17.52 ± 3.25, CFC VR: 23.54 ± 3.26, p=0.001, Paired t-test, n = 27). We also observed a significant increase in freezing during the first lap in the CFC VR compared to the familiar VR (Figure 2B-C, average first lap freezing (%), Familiar VR: 15.94 ± 4.03, CFC VR: 25.11 ± 3.41, p<0.001, Paired t-test). Mice also took longer to complete a lap in the CFC VR compared to the Familiar VR, both in the first lap (average time taken to complete first lap (s), Familiar VR: 14.81 ± 3.74, CFC VR: 41.16 ± 9.33, p<0.001, Paired t-test) and on average (Figure 2E-F, the average time taken to complete all laps (s), Familiar VR: 16.17 ± 3.05, CFC VR: 38.82 ± 11.06, p<0.001, Paired t-test). The increase in freezing and time taken to complete a lap in the CFC VR compared to the Familiar VR was significantly higher on Day 1 than Day 0, indicating that it results from fear conditioning (Figure 2C-F, p<0.05, Paired t-test).

Based on studies in freely moving animals (Michael S. Fanselow, 1980, 1990; Ramanathan et al., 2018; Russo & Parsons, 2021; Trott et al., 2022), we settled on an intensity of 1 mA and six shocks as a reasonable protocol, which we tested on a sizeable cohort of mice (n = 12). We found that freezing at this intensity and shock number was higher on average in the CFC VR; however, mice showed variability in their responses (Supplementary Figure 2A-B). The majority of mice (7/12) displayed appropriate memory recall (Supplementary Figure 2A-B), indicated by more freezing in the CFC VR compared to the Familiar VR (delta > 0), while 3/12 mice froze more in the Familiar VR (delta < 0), and 2/12 showed equal freezing levels in both VRs (delta ≈ 0). Such variability has been regularly documented in freely moving animals and may not be unique to the VR setup used here (Chu et al., 2024; Dos Santos Corrêa et al., 2019; Navarro-Sánchez et al., 2024; Poulos et al., 2016; Russo & Parsons, 2021; Totty et al., 2021).

Next, in a small group of mice (n = 15), we investigated whether varying the number of shocks or the shock amplitude would affect the conditioned fear response (Supplementary Figure 2A-B). Though our parameter exploration was not exhaustive, we assessed shock amplitude more extensively than the number of shocks. (Supplementary Figure 2C-E). Overall, most shock amplitudes and numbers of shocks increased freezing behavior in the CFC VR on recall day (Supplementary Figure 2A-B). The observed variability was within the same range as that of the larger cohort of mice we tested using the 1 mA, 6-shock protocol (Supplementary Figure 2A-B). Notably, increasing the shock amplitude from 0.6 mA to 1 mA, while keeping the number of shocks consistent at 6 shocks, significantly enhanced both fear discrimination and freezing behavior (Supplementary Figure 2C-E). A more detailed analysis with a larger sample size may be necessary to understand further how variations in shock parameters affect the conditioned response, as has been done in studies with freely moving animals (Baldi et al., 2004; Cordero et al., 1998; Daviu et al., 2024; Poulos et al., 2016; Quinn et al., 2008).

Finally, we examined the evolution of this conditioned response across days (Supplementary Figure 2F-I). We placed the animals in both the Familiar and the CFC VR for five minutes over four days (Recall Days 1-4) in a counterbalanced manner and tested their recall behavior. We quantified both the degree of fear discrimination across days (Supplementary Figure 2 F-G) and the extinction of the fear response within the CFC VR compared to baseline (Supplementary Figure 2 H-I).

Fear discrimination was most prominent on Recall Day 1. Mice displayed the most substantial increase in freezing in the CFC VR compared to the Familiar VR on Recall Day 1 (Supplementary Figure 2F-G). Freezing on the first lap in the CFC VR was also most pronounced on Recall Day 1. Both average and first lap freezing in the CFC VR decreased on Recall Day 2 compared to Recall Day 1, becoming similar to the Familiar VR. While first lap freezing remained similar between the Familiar and the CFC VR on subsequent Recall Days 3 and 4, we observed some variability in the average freezing behavior. Average freezing increased significantly again on Recall Day 3 in the CFC VR but was similar in both VRs on Recall Day 4.

To determine the extinction of the fear response, we compared freezing within each VR to baseline freezing levels on Day 0 (Supplementary Figure 2H-I). We quantified this using a Linear Mixed Effects (LME) Model, treating recall days as fixed effects and mouse as a random effect, with baseline freezing on Day 0 as the reference (Tables 1-2). In the CFC VR, both first-lap and average freezing were most prominent on Recall Day 1 and returned to baseline levels by Recall Day 2. While we observed a small increase in average freezing on Recall Day 3, it remained lower than Recall Day 1 levels and returned to baseline by Recall Day 4. This pattern suggests that the pronounced first-lap freezing observed on Recall Day 1 extinguishes quickly, while average freezing, though more variable, also extinguishes by Recall Day 4.

Table 1:

Table 2:

In the Familiar VR, freezing behavior—both first lap and average—did not differ significantly from baseline on most days. However, we observed an unexpected increase in average freezing compared to baseline on Recall Day 4. This increase could reflect either fear generalization or decrease in motivation - a response to the prolonged absence of reward in a previously rewarded environment. To distinguish between these possibilities, we examined freezing behavior in control mice that underwent the same paradigm without shocks (no-shock controls, n = 7, black lines in Supplementary Figure 2H-I). Average freezing in the Familiar VR exceeded that of no-shock controls on both Recall Days 1 and 4. Since no-shock controls explored the environments under no-reward conditions, this suggests that the increase in freezing in the Familiar VR stems from fear conditioning and may indicate generalization rather than motivational changes. This aligns with findings from freely moving animals, where specific fear memories have been shown to generalize over time with repeated exposures to both neutral and conditioned contexts (Biedenkapp & Rudy, 2007; Wiltgen & Silva, 2007).

In summary, these results demonstrate that head-fixed mice can acquire fear associations in VR environments. Most mice displayed freezing as a conditioned response specifically to the shock-paired VR and showed robust fear discrimination—behavior that mirrors that of freely moving animals after fear conditioning. By the fourth recall day, freezing had largely extinguished. Notably, beyond freezing, other avoidance-like defensive behaviors were prominent in the shock-paired VR, demonstrating the range of behavioral responses our paradigm can capture.

A second paradigm where head-fixed mice discriminate between two novel VRs

As previously mentioned, the incomplete extinction of licking behavior in the familiar VR before fear conditioning or the gradual increases in freezing behavior in the familiar VR during recall tests may indicate that mice still associate the VR with water rewards, which could be a caveat. Furthermore, in traditional CFC paradigms with freely moving mice, fear behavior has been assessed mainly by comparing two novel contexts, one associated with shocks and the other acting as a control. To address this, we tested another paradigm (Paradigm 2), using a novel VR instead of the familiar VR to act as the control (Figure 3A). One novel VR was assigned as the Control VR, and a second novel VR was assigned as the CFC VR, which would later become paired with the shocks. In this paradigm, we also added an extra habituation day to increase the animal’s pre-exposure time to the novel VRs prior to conditioning (Day -1). Pre-exposure to the context before fear conditioning can enhance contextual fear, as seen in freely moving animals (Biedenkapp & Rudy, 2007; Michael S. Fanselow, 1980, 1990; Maren et al., 2013; Rudy et al., 2002; Wiltgen & Silva, 2007).

Figure 3.

The training was done similarly to the previous paradigm. After mice showed consistent running behavior in the Familiar VR with the tail-coat and without reward, they were advanced to the experimental stage. On Day -1, mice were exposed to the two novel VRs for ten minutes each. On Day 0, they were re-exposed to the same two VRs for ten minutes before receiving six one-second-long mild electric shocks (at 1 mA and 1 minute ISI) in one of them - the CFC VR. On Day 1, mice were exposed to the two VRs for five minutes to evaluate memory recall (Figure 3A-B).

Compared to Paradigm 1, in the novel VRs, mice showed no differences in licking behavior across all experimental sessions (Supplementary Figure 3A). There was also no significant difference in freezing behavior and time taken to complete a lap between the two VRs on Day -1 and Day 0 before fear conditioning (Supplementary Figure 3B-C).

Head-fixed mice show heightened fear discrimination during the first lap when discriminating between two novel VRs.

We compared the freezing behavior between Paradigm 1 and Paradigm 2 on Recall Day 1 to the same intensity and number of shocks (1 mA, 1 s long, 60 s ISI, six shocks, n=12 mice each paradigm). While first-lap freezing in the CFC VR was markedly higher in Paradigm 2, significant differences between the Control VR and CFC VR were observed only for first-lap freezing, not for average freezing or lap completion time across all laps (Figure 3B-D, 4A-E). The effect sizes for the difference in first-lap freezing between the two VRs were nearly identical between the paradigms (Cohen’s d: First-lap freezing: Paradigm 1: 1.038, Paradigm 2: 1.040), but the effect size for average freezing was notably smaller in Paradigm 2 compared to Paradigm 1 (Cohen’s d: Average Freezing: Paradigm 1: 0.797, Paradigm 2: 0.093).

When we quantified freezing by lap on Recall Day 1 in Paradigm 2, we found that after the first lap, mice showed similar freezing in both the Control VR and the CFC VR across all subsequent laps (Supplementary Figure 4). In contrast, in Paradigm 1, mice consistently froze more in the CFC VR compared to the Familiar VR throughout Recall Day 1 (Supplementary Figure 4).

Indeed, when we quantified freezing behavior in the Control VR and CFC VR in Paradigm 2, we found that the most significant increase in freezing and other avoidance-like defensive behaviors in the CFC VR was observed in the first lap (freezing in first lap %: Control VR: 11.04 ± 1.8 CFC VR: 16.05 ± 3.94, p = 0.016, time taken to complete first lap (s): Control VR: 14.27 ± 2.85, CFC VR: 25.68 ± 5.55, p = 0.001, Paired t-test, Figure 3B-E, Figure 4A-E). Neither average freezing across all laps (Figure 4C, average freezing in all laps %: Control VR: 20.54 ± 4.13, CFC VR: 21.15 ± 4.21, p=0.738, Paired t-test) nor time taken to complete all laps (Figure 4E, the average time taken to complete all laps: Control VR: 21.73 ± 8.11, CFC VR: 24.27 ± 6.36, p=0.558, Paired t-test) showed significant differences between the two VRs.

Figure 4.

Supplementary Figure 3.

Supplementary Figure 4.

Supplementary Figure 5.

When mice were returned to the two VRs the following day for a second day of recall (Recall Day 2), they continued to exhibit modest levels of fear discrimination, as indicated by a small but statistically significant difference in freezing in the first lap and on average between the CFC VR and Control VR (Supplementary Figure 5 A-B). However, when assessing fear extinction by comparing freezing in the CFC VR across recall days to baseline levels (Supplementary Figure 5 C-D, Tables 3-4), we found no significant differences beyond the very first lap on Recall Day 1. This suggests that while mice maintained their ability to discriminate between environments by freezing more in the CFC VR, their overall fear response had extinguished. Thus, in Paradigm 2, fear extinction occurs rapidly within the first recall session after the initial lap.

Table 3:

Table 4:

A shorter ISI and leaving the tail-coat on during fear memory recall improved fear discrimination when using two novel VRs.

Freezing behavior in the CFC VR in Paradigm 2 was weaker than in Paradigm 1, with rapid fear extinction after the first recall lap. In Paradigm 3, two parameters were changed relative to Paradigm 2 to test whether freezing behavior and fear discrimination could be enhanced. First, the interval between consecutive shocks was reduced to see if a shorter ISI could produce a more robust fear response (McNally & Westbrook, 2006; Williams, 1994). Second, the tail-coat was kept on during memory recall, based on evidence that removing contextual cues can reduce conditioned freezing (González et al., 2003). We hypothesized that the tail-coat might act as a contextual cue and that its presence during recall might enhance freezing behavior. To further test these effects, we used a lower shock intensity of 0.6 mA—which we previously showed as producing only modest fear responses in Paradigm 1 (Supplementary Figure 2 C-E)—to determine whether a shorter ISI and the retention of the tail-coat during recall could enhance conditioned fear responses produced by the lower shock intensity.

In Paradigm 3, mice followed the same experimental protocol as Paradigm 2, with two novel VRs: one assigned as the Control VR and the other as the CFC VR. A habituation day in the two new VRs (Day -1) was followed by the CFC day (Day 0), where mice received six shocks (0.6 mA, 20-26 s ISI) in the CFC VR. This was followed by recall days (Days 1-3), where mice explored the two VRs presented in a counter-balanced manner. Unlike the other two paradigms, in Paradigm 3, the tail-coat remained on during the recall days (Figure 5A).

Figure 5.

We found that mice (n = 20 male mice) showed better fear discrimination between the Control and CFC VRs in Paradigm 3 compared to Paradigm 2 (Figure 5B-D). Mice not only showed significant freezing in the first lap (freezing in first lap %: Control VR: 13.61 ± 2.24 CFC VR: 19.73 ± 5.17, p = 0.021, Paired t-test) but average freezing across all laps was also higher in the CFC VR compared to the Control VR (Figure 5C-D, Figure 6A-D, average freezing in all laps %: Control VR: 15.28 ± 2.00, CFC VR: 21.92 ± 2.95, p < 0.001, Paired t-test).

Figure 6.

Supplementary Figure 6.

While first-lap freezing showed a weaker effect size in Paradigm 3 compared to the other paradigms, there was a strong effect of average freezing across the session (Cohen’s d: First lap freezing: Paradigm 1: 1.038, Paradigm 2: 1.040, Paradigm 3: 0.719. Average Freezing: Paradigm 1: 0.797, Paradigm 2: 0.093, Paradigm 3: 1.232). On average, we found that Paradigm 1 still produced the highest absolute freezing behavior in the CFC VR (average freezing in all laps %: CFC VR: Paradigm 1: 23.54 ± 3.26, Paradigm 2: 21.51 ± 4.21, Paradigm 3: 21.92 ± 2.95). The difference in freezing behavior between the CFC VR and the neutral VR was comparable between Paradigms 1 and 3 (Δ average freezing % (CFC VR – Familiar/Control VR): Paradigm 1: 6.74, Paradigm 2: 0.61, Paradigm 3: 6.63). Thus, unlike Paradigm 2, where the largest differences in freezing between the CFC VR and the Control VR were restricted to the first lap, the shorter ISIs and the addition of the tail-coat during the recall day in Paradigm 3 improved contextual fear discrimination between the two novel VRs. When tested for memory recall in Paradigm 3, mice froze more in the CFC VR than in the Control VR, and this effect persisted beyond the first lap (Figure 5C-D, Supplementary Figure 4). Mice also maintained their ability to discriminate between the two VRs for the first two recall days, freezing significantly more in the CFC VR than in the Control VR on both Recall Days 1 and 2 (Supplementary Figure 6A-B). However, by Recall Day 3, this discrimination diminished, as freezing levels no longer differed significantly between the two VRs.

Unlike the rapid fear extinction observed in Paradigm 2, mice in Paradigm 3 showed delayed extinction (Table 5-6, Supplementary Figure 6 C-D). Freezing levels in the CFC VR remained elevated above baseline during the first lap and across all laps on Recall Days 1 and 2, with average freezing still elevated above baseline on Recall Day 3. While freezing levels showed a consistent downward trend across days, these results suggest that fear had not completely extinguished by Recall Day 3. As in Paradigm 1, the most pronounced increases in first lap freezing in the CFC VR occurred in early recall days, while elevated average freezing persisted longer. These findings suggest that the shorter ISI and presence of the tail-coat during memory recall may enhance conditioned fear and delay fear extinction in head-fixed mice discriminating between two novel VRs, though the individual effects of these parameters require further investigation.

Table 5:

The tail-coat’s presence during recall days likely contributed to some transfer of fear to the Control VR, evidenced by elevated freezing in the Control VR compared to baseline. Nevertheless, freezing levels in the Control VR remained consistently lower than in the CFC VR across all days (Supplementary Figure 6, Tables 5-6).

Table 6:

In summary, our results show that mice exhibited greater average freezing behavior in the CFC VR in Paradigms 1 and 3 compared to Paradigm 2. In Paradigm 2, the most noticeable increase in freezing occurred during the very first lap in the CFC VR compared to the Control VR. Fear extinction was rapid in Paradigm 2 and was back to baseline levels within the first day of recall. We found that fear discrimination could be improved, and fear extinction could be delayed with a shorter ISI when presenting tail shocks and when the tail-coat was present during memory recall, as shown in Paradigm 3. Thus, we provide three different variations for performing CFC in head-fixed mice that can be used to understand the neural underpinnings of fear memory.

Place cells in the CA1 subregion of the hippocampus displayed remapping and decreased field widths following contextual fear conditioning

The hippocampus plays a critical role in the encoding, consolidation, and recall of memories and is necessary for the expression of contextual fear memory (Anagnostaras et al., 2001; Ji & Maren, 2007; Liu et al., 2012; Maren et al., 2013; Stephen Maren & Holt, 2000; Phillips & LeDoux, 1992; Ramirez et al., 2013; Wiltgen & Silva, 2007). Context is thought to be represented in the hippocampus by place cells that fire at specific spatial locations (place fields) in an environment (O’Keefe & Dostrovsky, 1971). These place cells have been thought to help discriminate between feared and neutral contexts. Previous studies on freely moving animals have shown that fear conditioning can cause place cells to remap (Blair et al., 2023; E. J. Kim et al., 2015; Kinsky et al., 2023; Mamad et al., 2019; Moita et al., 2004; Ormond et al., 2023; Schuette et al., 2020; Wang et al., 2012, 2015; Wu et al., 2017). Remapping is defined by place cells shifting their firing fields, indicative of place cells adapting to incorporate new information into memory (Colgin et al., 2008; Leutgeb et al., 2005; Muller & Kubie, 1987). Additionally, a recent study (Schuette et al., 2020) that performed calcium imaging of hippocampal cells in freely moving mice found that fear conditioning narrows the width of place fields in the feared environment. This suggests that place cells encode the same environment on a finer scale following fear conditioning, potentially enhancing context discrimination and threat avoidance (Schuette et al., 2020).

We replicated these findings in our head-fixed version of CFC in VR. To investigate the effect of fear conditioning on place cell activity in our task, we expressed a calcium indicator, GCaMP6f (Chen et al., 2013), in hippocampal pyramidal cells and, using two-photon microscopy, imaged from the same hippocampal neurons across days during fear conditioning and fear memory recall in Paradigm 1 (Figure 7A-B, n=8 mice). We found that a subset of place cells formed before CFC remained stable, while others remapped or changed their preferred firing locations after CFC (see examples in Figure 7C, 7D-G). This is similar to previously reported observations (Blair et al., 2023; Moita et al., 2004; Wang et al., 2012, 2015).

Figure 7.

We found remapping to occur in both the Familiar and the CFC VR to similar extents, as indicated by the lower correlation across sessions/days compared to the within-session correlation (Figure 7F-G). This contradicts previous findings by Moita et al. (2004), who reported more stability in the control environment compared to the CFC environment after fear conditioning. This discrepancy may be explained by the different reward protocols used. Moita et al.’s rats foraged continuously for food pellets during the experimental sessions, while our mice received no rewards. We removed water rewards from the previously rewarded Familiar VR, and as we have shown in earlier work, removing rewards from a previously rewarded environment causes the place map to become more unstable and likely to remap across days (Krishnan et al., 2022; Krishnan & Sheffield, 2023). Additionally, Moita et al. measured place cell stability within a single day, with just one hour between conditioning and retrieval sessions, while our recordings were separated by a day, including intervening sleep periods. This longer interval may have contributed to more pronounced remapping compared to the shorter interval used by Moita et al (Cai et al., 2009; Maboudi et al., 2018).

Next, we looked at the place cells in each session. Interestingly, we found an increase in the number of place cells in the CFC VR during memory recall, a finding not previously reported (Figure 8A-B). We also found that place cells had narrower fields in the CFC VR during memory recall (Figure 8C), as previously shown (Schuette et al., 2020). We did not find a significant difference in other place cell parameters we examined, such as lap-by-lap reliability (Figure 8D) and out-of-field firing (Figure 8E). Thus, our observations of remapping and narrower place field widths replicate findings from freely moving animals, validating our paradigm’s use to explore further questions using the head-fixed preparation to study the neural underpinnings of contextual fear memories. We add to this field by showing an increase in the proportion of place cells following CFC, a possible neural substrate of fear memory recall.

Figure 8.

Discussion

We present three paradigms for performing CFC in head-fixed mice, using a conductive “tail-coat” to deliver aversive shocks as mice navigate VR environments on a treadmill. Our key findings are: 1) Head-fixed mice exhibited freezing behavior in response to an aversive stimulus in VR, consistent with conditioned responses seen in freely moving animals. Freezing was especially pronounced during the first lap in the fear-conditioned environment compared to the neutral environment across all paradigms tested. 2) With mice on a cylindrical treadmill, additional conditioned responses such as hesitation and backward movement were detected. 3) The choice of neutral environment influenced fear discrimination: mice froze more in the fear-conditioned VR when the neutral environment was familiar rather than novel. Freezing in the fear-conditioned VR could be enhanced by shorter ISIs and the presence of the tail-coat during recall when the neutral environment was novel. 4) Following fear conditioning, place cells in the hippocampal CA1 region showed remapping, increased numbers, and narrower fields, paralleling findings in freely moving animals and offering new insights.

Here, we experimented with three paradigms for fear conditioning in mice, with the main difference being the neutral environment where the mice didn’t receive any shock. In Paradigm 1, the neutral VR was familiar, where the mice had previously received rewards, and the CFC VR was novel. In Paradigms 2 and 3, both the neutral and CFC VRs were similarly novel. In Paradigm 3, we decreased the ISI and kept the tail-coat on the animal when testing for memory recall. Thus, we have tested three different versions for performing CFC in head-fixed mice. On average, mice showed more freezing behavior in the CFC VR compared to the neutral VR. A common feature across all paradigms was pronounced freezing during the very first lap in the CFC VR. While mice in Paradigm 2 extinguished fear rapidly after the first lap, freezing behavior persisted up to 3-4 recall days in Paradigms 1 and 3, to varying degrees. We also observed variability in conditioned responses across mice. Some displayed an opposite pattern, freezing more in the neutral VR or equally in both VRs. This variability in behavior, similar to that observed in freely moving animals (Chu et al., 2024; Dos Santos Corrêa et al., 2019; Navarro-Sánchez et al., 2024; Poulos et al., 2016; Russo & Parsons, 2021; Totty et al., 2021), can be insightful when combined with investigations of neural activity, helping to understand the neural dynamics that contribute to an animal’s ability to learn the association between a context and fearful stimuli. This can reveal the mechanisms by which such memories may generalize or become maladaptive, as seen in post-traumatic stress disorders.

We have shown here an example of using Paradigm 1 to study place cells in the hippocampus and their role in fear memory formation and recall. We recorded from a large group of the same hippocampal CA1 cells across days and demonstrated that place cells remap and narrow their place field widths with fear conditioning. This occurs in head-fixed mice navigating VR environments, similar to freely moving animals. Narrow place fields may suggest sharper spatial tuning to incorporate a salient emotional memory and may aid in context discrimination (Schuette et al., 2020). However, how place cells narrow their widths in response to fear conditioning remains an open question. One possible way to answer this is by imaging CA1 place cell dendrites to understand how they integrate inputs before and after fear conditioning. We can explore what dendritic firing and plasticity mechanisms lead to narrow place field widths. We can then perturb these firing patterns to attribute causality to the narrowing of place field widths in fear discrimination. Our head-fixed paradigm enables the recording of dendrites and the study of plasticity mechanisms (Sheffield et al., 2017; Sheffield & Dombeck, 2015), thus facilitating these investigations.

We recently used Paradigm 3 to understand the role of the thalamic nucleus reuniens-hippocampal pathway (Ratigan et al., 2023). Using a two-photon microscope, we identified neural activity in this pathway by imaging axons from the nucleus reuniens that project to the CA1 with single-axon resolution. We found this pathway necessary for suppressing fear memory retrieval, context discrimination, and fear extinction. The robust freezing response in Paradigm 3 in the fear-conditioned VR, compared to a neutral novel VR, allowed us to align our findings with the existing literature on the role of the nucleus reuniens-hippocampal circuit in fear memory in freely moving animals (Ramanathan & Maren, 2019; Ramanathan et al., 2018; Totty et al., 2023; Troyner et al., 2018; Vetere et al., 2017; Xu & Südhof, 2013). Furthermore, the subcellular resolution achieved through two-photon calcium imaging using the head-fixed preparation allowed us to longitudinally observe and understand the inputs that axons from the reuniens send to the hippocampus during fear memory formation and retrieval, a level of detail that is difficult to achieve with current techniques in freely moving preparations.

In another unpublished study, we used Paradigm 2 and identified co-active neurons between CA1 and CA3 subregions of the hippocampus during fear memory recall. The difference between the first and subsequent laps has helped us understand the progression of neural activity in the hippocampus as mice go from retrieving the fear memory to extinction within one session. Moreover, the variability in response across mice has allowed us to regress the neural activity to the behavior, identifying neural differences between those mice that accurately recall the fear memory and those that do not. These studies showcase the broad applicability of our different CFC paradigms.

In the laboratory, CFC provokes freezing behavior in freely moving rodents, as demonstrated by several studies that span decades (Michael S. Fanselow, 1980, 1990; Fendt & Fanselow, 1999; Maren, 2001; Maren et al., 2013; Phillips & LeDoux, 1992; Tovote et al., 2015). In our study, we found that a mild electric shock applied to the tail could also trigger freezing behavior as the conditioned response in head-fixed mice, making our paradigm more comparable to the freely moving paradigms than previously attempted methods using air-puff stimulation (Lovett-Barron et al., 2014; Rajasethupathy et al., 2015). There are numerous opportunities for future work to systematically test how different parameters—such as shock amplitude, number and spacing of shocks, duration of fear conditioning, context pre-exposure time, the configuration of Control versus CFC VR environments, and the inclusion of additional sensory cues—may influence both behavioral outcomes during recall and extinction, as well as the underlying neural representations (Michael S. Fanselow, 1990; Huckleberry et al., 2016; Ji & Maren, 2007; Maren & Holmes, 2016; Poulos et al., 2016). Our current study provides only an initial exploration of this parameter space.

In this manuscript, we focused on freezing as the primary conditioned response to align with studies of freely moving animals. However, we also see other defensive responses unique to our treadmill-based task. as well as fear-related hesitation and backward movements during memory recall—defensive behaviors that have also been observed in freely moving animals (Chu et al., 2024; Furuyama et al., 2023; Totty et al., 2021; Trott et al., 2022). Our technique enables us to conduct a more detailed analysis of the animal’s behavior. The head-fixed setup allows us to easily measure physiological responses including pupil diameter, respiration, vocalization, heart rate, and body temperature. These parameters are known to change with fear conditioning (Dupin et al., 2020; Godsil et al., 2000; Korn et al., 2017; Shionoya et al., 2013; Stiedl & Spiess, 1997). Such physiological measurements can also help differentiate between the various conditioned responses we observe. There are also opportunities to use machine-learning-empowered algorithms to parse out new behavior markers for contextual fear memory, such as gestures or facial expressions (Datta et al., 2019; Mathis & Mathis, 2020; Syeda et al., 2024; Wiltschko et al., 2020). Thus, we can understand the animal’s behavior in a well-controlled environment, which when combined with neural activity recording, can lead to new insights into the neural circuits mediating different defensive responses and emotional states.

Significant insights into memory mechanisms through the CFC task have also come from recent discoveries of memory engrams in the brain (Josselyn et al., 2015; Josselyn & Tonegawa, 2020; Liu et al., 2012; Tonegawa et al., 2018). Engram cells, identified in various brain regions with the help of genetic tagging and optogenetics, play a crucial role in encoding, consolidating, and retrieving specific memories. Typically, these cells are identified based on the expression of the immediate early gene Fos during fear conditioning. Their artificial activation with optogenetics can trigger memory recall, evidenced by increased freezing behavior. Understanding which neurons form these engram cells, how Fos expression is driven during memory formation, how memory information is stored in these cells, how they evolve with learning, and how their activation results in memory recall is an active area of research (Frankland et al., 2024; Mocle et al., 2024; Monasterio et al., 2024; Pettit et al., 2022; Suthard et al., 2024; Uytiepo et al., 2024). Incorporating our paradigm with simultaneous measurements of cellular and subcellular activity can help answer some of these questions. For instance, a recent study (Monasterio et al., 2024) tagged engram cells during fear conditioning while animals were freely moving and then imaged the spontaneous dynamics of these cells using two-photon imaging before and after fear conditioning. The study revealed that cells forming engrams were inherently more active. A limitation of this study was the separate contexts for imaging versus tagging, which can be addressed with our paradigm. Our CFC paradigm allows for direct neural recording of these cells before, during, and after tagging in the same context to examine how activity during memory formation relates to engram cell emergence, population dynamics, and memory recall. Furthermore, holographic stimulation techniques can reactivate specific cell ensembles of interest to help understand the causal relationships between engram cells and memory. Moreover, the hippocampus contains both place cells and engram cells, both of which appear to be involved in contextual memory. However, their relationship is not yet fully understood. Our CFC paradigm offers an opportunity to study both place cells, as we have shown here, and engram cells simultaneously. This can be achieved by incorporating techniques that allow for in vivo labeling of Fos (Pettit et al., 2022). Thus, our head-fixed version of the CFC paradigm opens new avenues for research into the neural underpinnings of fear learning and memory.

Another advantage of our method is that it allows for studying changes in neuronal dynamics in response to an aversive stimulus, such as a shock. Our fields of view during the shock and subsequent flight-like response periods remained stable, and we could correct image movements during these periods with Suite2p (Pachitariu et al., 2017). This is a significant improvement over one-photon and electrophysiological techniques, where abrupt movements such as those induced by an electric shock can change fields of view and reduce the likelihood of recording from the same groups of cells, thereby hindering longitudinal assessments of neural dynamics. In Paradigm 1, where we imaged hippocampal cells during shock, we did not find cells responding to the shock itself, as others have reported (Barth et al., 2023; Mocle et al., 2024; Suthard et al., 2024). We suspect this could be due to the number or duration of shocks administered and our shocks not being associated with a specific cue or location. In future studies, we plan to optimize these parameters to understand better how neural dynamics change in response to an aversive cue.

Finally, we recognize that our head-fixed CFC paradigm has some limitations. Compared to the freely moving CFC, our setup is less naturalistic. We defined different contexts as two visually distinct VR environments while odor, sound, and tactile cues remained constant. This could have made it harder for mice to differentiate between the two contexts, particularly in Paradigm 2. Unlike freely moving animals in CFC studies, our mice were water-restricted throughout the experimental process. Because the animals need to get used to wearing a tailcoat and running in the VRs without a reward, we tend to ’over-train’ the mice. This criterion excludes mice that are less adept at running, limiting the number that go through the process and requiring larger cohorts of mice. This also presents scalability challenges compared to freely moving paradigms, as successful implementation of the head-fixed CFC approach requires weeks of reward-based training, larger sample sizes, and a technically demanding VR setup that integrates multiple hardware and software components to precisely monitor behavior. Additionally, while rewards were removed during experimental sessions, water rewards were still used to motivate running during training. This reward-associated training could subtly influence later behavior. For instance, the increased freezing in the Familiar VR on later recall days and remapping of place cells in that context could reflect the lingering effects of reward-associated training. While Paradigms 2 and 3 address many of these concerns, our approach still differs from freely moving paradigms due to this initial reward-based training. The impact of using rewards during training on recall behavior needs further investigation. Furthermore, despite our best efforts to include equal numbers of male and female mice, only some female mice reached the experimental stage. This could be due to the tail-coat weight and the reduced weight of females after water restriction compared to males. Female mice of the same age (10-12 weeks) weighed significantly less than males (20-23 g vs 25-30 g) and lost further weight following water restriction. Although we observed no behavioral differences between the four tested females and the male mice, we cannot guarantee similar results in other female animals without further testing. Studies indicate that different sexes may display varying conditioned responses, as seen in rats (Gruene et al., 2015; S. Maren et al., 1994). In the future, we aim to address these limitations.

In conclusion, we have developed a head-fixed version of the classic CFC paradigm. Our findings demonstrate that mice exhibit freezing behavior when fear-conditioned in a VR environment. Moreover, we observed that hippocampal place cells remap and narrow their fields in response to fear conditioning, mirroring observations in freely moving animals. This innovative paradigm bridges the behavioral task gap for techniques requiring head-fixation, paving the way for new investigations into neural circuits and deepening our understanding of the neural foundations of emotional memory.

Methods

Subjects

All experimental and surgical procedures adhered to the University of Chicago Animal Care and Use Committee guidelines. This study utilized 10-12 week-old C57BL/6J, Thy1GCaMP6s and Tg(Grik4-cre)G32-4Stl/J mice. Following the start of water restriction, the mice were individually housed in a reverse 12-hour light/dark cycle, and behavioral experiments were conducted during the animals’ dark cycle.

Head-fixation surgery

Mice were anesthetized with 1-2% isoflurane and given an intraperitoneal injection of 0.5 mL of saline and a subcutaneous injection of 0.45 mL of Meloxicam (1-2 mg/kg). A small incision was made on the skull to attach a metal head-plate (9.1 mm x 31.7 mm, Atlas Tool and Die Works) to head-restrain the mice. Some mice also had an imaging window implanted above the hippocampus, as previously described (Dong et al., 2021; Krishnan et al., 2022; Sheffield & Dombeck, 2015). Post-surgery mice were housed individually. After a recovery day, they were water-restricted to 0.8-1.0 ml/day.

Behavior training

Head-restrained mice were trained to run on a cylindrical treadmill to navigate virtual reality (VR) environments similar to the setups previously described (Dong et al., 2021; Krishnan et al., 2022; Sheffield & Dombeck, 2015). Behavioral training began seven days following the water restriction. Briefly, VR environments were created using the Virtual Reality MATLAB Engine, or VIRMEN (Aronov & Tank, 2014), and projected to five screens covering the mouse’s field of view. Mice were head restrained with their limbs resting on a freely rotating Styrofoam wheel, which acted as a treadmill. VR environments were 2 m linear tracks. Mice received a water reward (4 μl) through a waterspout upon completing each traversal of the linear track, considered a lap. This reward was followed by a 1.5-second pause in the VR to allow the mouse to consume the water. The mice were then virtually teleported to the start of the track to begin a new lap. Training sessions lasted 30 minutes per day. Mice were considered well-trained when they could run at >3 laps per minute. We found that this level of training was necessary to ensure that mice would continue to run even without rewards (Krishnan et al., 2022; Krishnan & Sheffield, 2023). Generally, about 60% of mice met our criteria, which typically took around 10-14 days to achieve. Behavior data was collected using a PicoScope Oscilloscope (PICO4824, Pico Technology, v6.13.2).

Tail-coat

A custom-designed “tail-coat” delivered mild electric shocks to a head-fixed mouse’s tail during virtual reality (VR) navigation. A pictorial representation of the tail-coat can be found in Figure 1D. The tail-coat, handmade from wearable conductive fabrics, featured a lightweight design to ensure the mouse could continue running during baseline VR exploration (Table 7). First, the mouse’s tail was measured while anesthetized for head-plate surgeries. A small denim patch cloth (Michael’s) was cut into a rectangle to fit snugly on the mouse’s tail. Marks were made where snap buttons (Michael’s) would be sewn, ensuring they clipped just above the tail. Given that the mouse’s tail narrows away from the body, the buttons closer to the body were placed slightly further apart than those closer to the tip. Typically, the buttons were set 2 cm apart on the narrow end and 2.5 cm apart on the broader end. Care was taken to adjust the placement of the buttons so they weren’t too tight or loose for the mouse. We found that these measurements were generally similar across sexes and for the age group that we tested. After measurements were taken and the cloth cut, small strips of 8mm wide conductive fabric tape (Adafruit) were attached to either side of the cloth, with some tape hanging off the edge. A metal strip (McMaster Carr) of the same width was affixed to this portion of the tape to provide a secure grip for alligator clips used in fear conditioning. Finally, snap buttons were sewn with conductive thread at the marked spots. Clear nail polish was applied to secure the threads in place. The tail-coat creation process should take less than half an hour. Dimensions of the cloth for a typical male mouse, 12 weeks of age and weighing ∼30 g, are shown in Figure 1D. The denim cloth was 4 cm in width and 3.2 cm in length; the conductive tape was 1.2 cm in width and 5.2 cm in length, with 1.5 cm hanging out at each side.

Table 7:

Contextual fear conditioning protocol in head-fixed mice

Once the mice were well-trained and met our criteria of running >3 laps per minute, they were ready for the next stage. The following day, a tail-coat was attached to the mouse’s tail. Mice that maintained consistent running in the VR, with minimal pauses, while wearing the tail-coat advanced to the next stage. For mice startled by the tail-coat, we typically attempted another day with the tail-coat, rewarding them and observing if their running behavior improved to running at least three laps per minute. If not, these mice were not advanced to the next stage. In mice that do advance, on the subsequent day, the tail-coat was kept on, and the water reward was removed. All else remained the same, including the 1.5 s pause at the end, to allow animals to distinguish the start and end of a lap. If the mice continued to run as before, they moved on to the experiment stage. From that day forward, the mice no longer received water reinforcement when running in the VR environments. Typically, about 40% of mice made it to the experiment stage. Tail-coats were only added during VR exploration and removed before returning the animals to their home cage. Once the mice advanced to the experiment stage, they were taken through the entire process and not excluded from analysis, even if they showed signs of inconsistent running behavior, which happened with some mice (for instance, see Supplementary Figure 3B-C).

We implemented three distinct paradigms to assess contextual fear discrimination. In the first paradigm, we conducted experiments over two days (Paradigm 1). On Day 0, mice engaged with the same VR in which they were trained, now referred to as the Familiar VR. After 10 minutes, we transitioned them to a new VR for another 10 minutes; this new VR would later be associated with shocks (CFC VR). During this phase, the mice explored both VRs without any rewards while wearing the tail-coat. After the initial exploration in the CFC VR, they received six mild electric shocks, each lasting 1 second with 1-minute intervals in between. The shocks were administered using a shock generator (Colbourn Precision animal shocker, Harvard Bioscience). A TTL pulse was generated using custom-written codes in MATLAB to turn on the shock generator via an Arduino. The shock times were recorded by the PicoScope Oscilloscope (PICO4824, Pico Technology, v6.13.2). We tested different shock intensities ranging from 0.5 to 1.2 mA. After the last shock, mice were removed from the VR and returned to their home cage. The following day (Day 1), we placed the mice in the Familiar VR or CFC VR for ten minutes to test their memory recall. No rewards were delivered during this phase; however, the tail-coat was not added during memory recall. Some mice underwent additional recall tests over subsequent days to assess the time required for fear memory extinction. As a control, some mice underwent the same paradigm but without any shocks.

In the second paradigm (Paradigm 2), the experiment spanned over three days. On Day - 1, mice explored two novel VRs for ten minutes each, establishing a baseline for context exploration. One of these VRs served as a control, while the other, the CFC VR, was where the mice would later receive electric shocks. On Day 0, the mice again spent ten minutes each in the Control and CFC VRs. After this, they received six mild electric shocks, each lasting one second, with a strength of 1 mA and an inter-trial interval of one minute. The mice wore the tail-coat on both Days -1 and 0. On Day 1, the mice were placed in either the Control VR or CFC VR for ten minutes each, in a counterbalanced manner, to test for memory recall. No rewards were given on all the experimental days; however, the tail-coat was not added to the tails on the recall day (Day 1). The third paradigm (Paradigm 3) was similar to Paradigm 2 and used two novel VRs, except that a tail-coat was also added to the mice’s tails during recall tests.

Additionally, a subset of mice in Paradigms 1 and 2 and all mice in Paradigm 3 went through multiple days of recall tests. The subset of mice was randomly chosen, not based on their recall behavior from the previous day. Mice were presented with either the CFC VR or the Control VR in a counterbalanced manner.

All VRs were randomly assigned as either Control or CFC VR to prevent biasing any inherent preferences in mice. However, the Familiar VR was always kept the same. We also used both sexes in our experiments. However, we found female mice less receptive to the “tail-coat” than their male counterparts. We believe this is due to the low average weight of female mice (∼16 g) following water restriction. Female mice of the same age (10-12 weeks) weighed significantly less than males (20-23 g vs 25-30 g) and lost further weight following water restriction. As a result, only four female mice could proceed through and complete all the behavioral tests. Our ultimate goal is to compare behavior and neural activity across sexes, considering sex as a biological variable. To achieve this, we plan to adjust the age and use female mice of appropriate weight to ensure they can run on the treadmill with the tailcoat in the future.

Behavior measurement

Freezing periods were periods during which the animal’s instantaneous velocity was 0 cm/s. Freezing epochs were defined as continuous, uninterrupted freezing periods longer than 1 s. The time taken to complete a lap was calculated as the total time (in seconds) the animal took to run from 0 to 200 cm.

Surgery and two-photon imaging

Stereotaxic surgeries and cannula window implantation above the CA1 subregion of the hippocampus were performed in the same way as previously reported (Krishnan et al., 2022). Briefly, a genetically-encoded calcium indicator, AAV1-CamKII-GCaMP6f (pENN.AAV.CamKII.GCaMP6f.WPRE.SV40 was a gift from James M. Wilson – Addgene viral prep #100834-AAV1; https://www.addgene.org/100834/; RRID: Addgene_100834) was injected (∼50 nL at a depth of 1.25 mm below the surface of the dura) using a beveled glass micropipette leading to GCaMP6f expression in a large population of CA1 pyramidal cells. Mice underwent water restriction (1.0 ml/day) following viral injection and, seven days later, were implanted with a hippocampal window and head plate (Dombeck et al., 2010). Mice were allowed to rest from the surgeries for 3-4 days, after which behavioral training began. Imaging was done using a laser scanning two-photon microscope (Neurolabware). The setup is the same as the one previously described (Krishnan et al., 2022). Time-series images were collected through Scanbox (v4.1, Neurolabware), and the PicoScope Oscilloscope (PICO4824, Pico Technology, v6.13.2) was used to synchronize frame acquisition timing with behavior. At the end of the imaging session on Day -1 or Day 0, a 1-minute time-series movie was collected at a higher magnification and then averaged to aid as a reference frame in finding the same imaging plane on subsequent days. Across-day images collected were motion-corrected using Suite2p (Pachitariu et al., 2017) and aligned using ImageJ (v1.53, NIH). Region of interests were extracted using Suite2p.

Place cell extraction and place field parameters

Place fields were identified as previously described (Dong et al., 2021; Krishnan et al., 2022). Briefly, the 2 m track was divided into 40 position bins (each 5 cm wide). The running behavior of the animal was filtered to exclude periods where the animal was immobile (speed <1 cm/s). Filtering ensured that place cells were defined only during active exploration and not during freezing bouts. We have previously found that filtering versus not filtering yields similar results (Krishnan et al., 2022). Extracted place fields satisfied the following criteria, which were used for all conditions and all mice: 1. The average ΔF/F was greater than 10% above the baseline. 2. The cell displayed significant calcium transients on >30% of laps in the field. 3. The rising phase of the mean transient was located on the track. 4. Their p-value from bootstrapping was <0.05. Multiple place fields within the same cell were treated independently. Parameters of place fields were calculated as described before (Krishnan et al., 2022); this includes the center of mass (COM), place field reliability, out/in-field firing ratio, and place field widths.

Statistics

For related samples, we performed a paired t-test. Multiple comparisons of related samples were corrected with Bonferroni post-hoc. To quantify freezing across days, we employed a linear mixed-effects model with days of recall as the fixed effect and behavior prior to fear conditioning on Day 0 as baseline. The model also included a random effect for mice to account for the nesting of observations within individual subjects. Separate models were used for each VR. Boxplots are plotted to display the entire distribution of the data. The box in the boxplot ranges from the first quartile (25th percentile) to the third quartile (75th percentile), and the box shows the interquartile range (IQR). The line across the box represents the median (50th percentile). The whiskers extend to 1.5*IQR on either side of the box, and anything above this range is defined as an outlier. A two-tailed Kolmogorov-Smirnov (KS) Test was used to compare distributions. When not using boxplots, mean and confidence intervals are displayed.

Data and code availability

Behavior data, imaging data, and scripts used for data analysis are available on GitHub at https://github.com/seethakris/CFCMethodspaper.

Acknowledgements

We thank the members of the Sheffield Lab for comments on the manuscript, helpful discussions, and feedback throughout the process.

Additional information

Funding

This work was supported by The Whitehall Foundation, The Searle Scholars Program, The Sloan Foundation, The University of Chicago Institute for Neuroscience start-up funds and the National Institute for Health (1DP2NS111657-01, 1RF1NS127123-01) awarded to M.S. and a T32 training grant (T32DA043469) and K01 (K01DA060994-01) from National Institute on Drug Abuse awarded to S.K.

Author contributions

S.K. and M.S. conceived the idea. S.K. and M.S. supervised the project. S.K. designed the experiments. S.K., C.D., H.R., D.M., and C.C. collected the data. S.K. and C.D. analyzed the data. S.K., C.D., H.R., and M.S. interpreted the data. S.K. prepared the manuscript draft. All authors reviewed the results, provided feedback, and approved the final version of the manuscript.

Additional files

Supplemental videos. Example video of a head-fixed mouse navigating on a treadmill in VR. The examples show mouse behavior on Recall Day 1 in Paradigm 1 in: 1) the Familiar VR and 2 and 3) the CFC VR. Video 2 shows freezing behavior in CFC VR and Video 3 shows another mouse displaying more avoidance-like behaviors like slowing down and backward movements. Videos were recorded for a few seconds immediately after the mouse was transitioned to each VR environment. Videos were collected at a sampling rate of 30 frames/sec.