To examine axon activity further, we first looked at the mean activity across all axons as a function of normalized track position (Figure 1e). As previously reported (Krishnan et al., 2022), VTA DA axons increase activity along the track, peaking at the reward location at the end of the track. In contrast, LC input activity remains relatively constant across all positions along the track (Figure 1e,iii). To examine if this difference could be due to the lower sample size of VTA axons compared to LC axons, the LC axons were down-sampled to match the VTA sample size (n = 9 regions of interest [ROIs] in 8 mice) and the slope and intercepts of the down-sampled data were found. This was repeated 1000 times and did not generate any LC data points that overlap with VTA data demonstrating the difference in relationship between position and activity was not due to the different sample sizes (Figure 1e,iii). We also examined the position related activity of individual VTA and LC axons and observed a positive relationship between position and activity in 88.9% of VTA ROIs (8/9 in 7 mice) but only 28.7% of LC ROIs (25/87 from 12 sessions in 12 mice) while 42.5% of LC ROIs (37/87 from 16 sessions in 9 mice) had a negative relationship between position and activity (Figure 1e,iv). To account for the different track lengths between VTA and LC recordings, we also looked at the virtual distance from the rewarded end of the track as well as time from reward (Figure 1—figure supplement 1). The same trends were seen for VTA and LC axons, with 8 VTA ROIs showing positive relationships with both distance and time from reward, and 21 LC ROIs showing a positive relationship with distance from reward and 31 LC ROIs showing a positive relationship with time to reward. This suggests that track length does not influence the encoding properties of VTA and LC axons.

For a subset of VTA (n = 6) and LC ROIs (n = 26), the reward at the end of the track was removed, and the activity of these axons in the unrewarded condition was examined (Figure 2a). While the slope of the VTA population activity across positions significantly decreased (Figure 2b,i) as expected and previously reported (Krishnan et al., 2022), the LC population activity across positions did not significantly change (Figure 2c,i). This confirmed that the ramping activity in VTA axons was due to the animal’s proximity to an expected reward and this signal was not present in the average activity of LC axons .

Next, we investigated the mean activity of these axons as a function of velocity. The population mean of both VTA and LC axons increased as velocity increased (Figure 1f). This is consistent with the finding that LC inputs to dCA1 encode velocity (Kaufman et al., 2020) and the finding that some DA VTA neurons encode kinematics (Engelhard et al., 2019). Again, to account for differences in sample size, we down-sampled the LC axons 1000 times and found the slope and y-intercept of each sampling. The overlap of the VTA and LC slopes and intercepts confirms we cannot conclude any differences in velocity-related activity in the VTA and LC axon populations (Figure 1f,iii). However, analyzing individual ROIs, we observed a statistically significant positive relationship between velocity and activity in the majority, 72.4%, of LC axons (63/87 ROIs from 21 sessions in 13 mice), while only 28.6% (2/9 ROIs in 2 mice) of VTA axons showed a positive relationship with velocity (Figure 1f,iv). The strong velocity correlated activity in a small subset of VTA DA axons indicates heterogeneity in the activity of these inputs similar to what is observed in VTA DA somas (Engelhard et al., 2019). To determine if the velocity correlated activity in VTA DA axons is confounded by the ramping-to-reward activity, we examined this activity in the unrewarded condition where the ramping-to-reward activity is absent. Here, the VTA population activity across all velocities is decreased but the slope, or the relationship between velocity and activity, remains unchanged (Figure 2b,ii). This is consistent with a subset of VTA axons encoding velocity information, while the overall decrease in activity is explained by the decrease in reward related activity demonstrated in Figure 2b,i. However, the relationship between velocity and LC activity is unchanged in the unrewarded environment (Figure 2c,ii), indicating condition invariant velocity encoding in LC axons.

Finally, we examined the activity of LC and VTA axons during rest and the transition to movement. The population of LC axons ramped up in activity during the 2 s leading up to motion onset (Figure 1g). This is consistent with reports of activity of LC axons in cortical areas (Reimer et al., 2016) showing LC activity prior to motion onset. In contrast, VTA axons show decreasing activity during the 2 s leading up to motion onset (Figure 1g). This ramping down in VTA axon activity is likely due to most periods of immobility occurring between reward delivery and the start of the next lap, during which we previously demonstrated reward related activity decays in VTA axons (Krishnan et al., 2022). Indeed, in the unrewarded condition the negative slope of the VTA population activity leading up to motion onset disappears (Figure 2b,iii), indicating this relationship is largely driven by the presence of reward rather than motion onset. However, the LC population activity remains unchanged in the unrewarded environment (Figure 2c,iii), suggesting this activity is related to the lead up to motion initiation. These differences in activity are not an artifact of lower sample size of VTA ROIs as shown by down-sampling the LC ROIs activity 1000 times and measuring the slopes and intercepts of the down-sampled data did not generate any data points that overlapped with the VTA slope and intercept (Figure 1g,iii). In further support of distinct activity profiles leading up to motion onset, we found that the majority (6/9 ROIs in 6 mice) of VTA ROIs decreased in activity leading up to motion onset but only 2/87 LC ROIs in one session decreased in activity, while the majority (56/87 ROIs in 17 sessions in 9 mice), of LC ROIs increased in activity leading up to motion onset (Figure 1g,iv). Together, this analysis demonstrates overlapping but distinct activity in VTA and LC axons during spatial navigation with VTA axons showing strong activity correlated with distance to reward and some velocity correlated activity, while LC axons demonstrate activity correlated to velocity and time to motion onset.

Because the DAT-Cre mice were injected with a virus for expression of either axon-GCaMP6s (four mice) or axon-GCaMP7b (four mice), we asked whether these two GCaMP variants led to different activity dynamics (Figure 1—figure supplement 2). The axon-GCaMP6s (five ROIs) and axon-GCaMP7b (four ROIs) both had a positive relationship with position, negative relationship with time to motion initiation, and a neutral relationship with velocity, although the GCaMP6s relationships were stronger (Figure 1—figure supplement 2a, c). Importantly, when we compared only axon-GCaMP6s expressing VTA ROIs with the axon-GCaMP6s expressing LC ROIs (Figure 1—figure supplement 3), we saw similar relationships as the comparisons using all VTA ROIs (Figure 1).

Both LC and VTA neurons have been shown to respond to novel environments (Takeuchi et al., 2016). Therefore, we aimed to examine the activity of VTA DA and LC inputs to the hippocampus during exposure to novel VR environments. Following 10 min in the familiar environment, mice were teleported to a novel VR environment of the same track length, with a reward at the same position at the end of the track. Following teleportation, we found the running speed of both DAT-Cre and NET-Cre mice transiently decreased (Figure 3a) and they spend less time immobile (Figure 3a), demonstrating mice recognize they are navigating a novel environment. While the velocity quickly recovered for both groups of mice, the freezing ratio, or amount of time spent immobile, never recovered in the first 10 laps in the novel environment for the NET-Cre mice indicating some novelty-induced changes in behavior persist.

We aligned VTA and LC axon activity to the switch to the novel environment and investigated changes in activity due to exposure to novelty. To test whether the mean axon activity is significantly elevated or lowered, we defined a baseline by generating 1000 shuffles of the axon traces across the entire recording sessions, down-sampling the shuffled data 1000 times to match the VTA (n = 7) and LC (n = 87) sample sizes, and calculating the mean and 95% CI of the shuffled data. After teleportation, the periodic activity observed in the mean of VTA axons, likely reflecting ramping-to-reward signals in each individual axon, disappeared (Figure 3b). This is evident in the traces of most of the individual VTA ROIs showing a loss of the ramping-to-reward signal (Figure 3c), and in the VTA population position binned activity showing a significant reduction in ramping activity (Supp. Fig. 1). However, one VTA ROI showed an increase in activity immediately following exposure to novelty (Figure 3c), indicating heterogeneity across VTA axons in CA1 and the lack of a novelty signal on average may be due to a small sample size.

Strikingly, LC axons show a dramatic increase in mean activity that remained elevated for >1 min following exposure to the novel environment (Figure 3b) similar with findings that LC cell body activity is elevated for minutes following exposure to environmental novelty (Takeuchi et al., 2016). A significant increase in activity above baseline activity following the switch to a novel environment can be seen in 36 LC axon ROIs (from 15/22 sessions in 10/16 mice) (Figure 3c). To further characterize this activity, we found the mean population activity for each lap and separately for 10 s time bins leading up to and following exposure to the novel environment. This analysis shows that LC activity is significantly elevated above baseline for six consecutive laps and approximately 90 s following exposure to the novel environment (Figure 3d). These findings demonstrate that LC inputs signal environmental novelty, supporting a role for these inputs in novelty encoding in the hippocampus.

It is possible that the change in the amount of time the animals spend running versus immobile in the novel environment could explain the increase in LC activity in the novel environment, as LC activity is related to behavior (Figure 1). For instance, LC axons show elevated activity during motion versus rest (Figure 1g). Therefore, an increase in the time spent in motion upon exposure to the novel environment could lead to an increase in LC activity. To account for the differences in time spent running vs immobile between the two environments, we removed any periods where the mice were immobile to isolate the effects of novelty from changes in behavior (Figure 4b). When only looking at activity during running in both environments, we found that LC axon activity is elevated for two laps, or 40 s, in the novel environment (Figure 4b–e). Additionally, when we look only at activity when the mice were immobile, we see that activity is elevated for 30 s in the novel environment (Figure 3—figure supplement 2b). Together, this indicates that there are two separate components that drive LC axon activity during the initial exposure to the novel environment. One, a shorter purely novelty-induced increase in activity which occurs during the first two laps, or about 40 s, in the novel environment. Two, a behavior-induced increase in LC activity caused by an increase in the percentage of time spent running that extends beyond the increase in the novelty-induced activity for six laps or 90 s.

If the short, novelty-induced signal in LC axons is an additional signal riding on top of the behavior correlated signals – position, velocity, and motion onset – we would expect a disruption of these behavioral correlations during the initial lap in a novel environment. To test this we examined these behavioral correlations lap-by-lap following exposure to the novel environment. Indeed, the slopes of the position binned, velocity binned, and motion onset aligned data are all significantly more negative in the first lap in the novel environment than the final laps of the familiar environment (Figure 4f–h). This is consistent with a decaying novelty signal that peaks at the start of the first lap and rides on top of these behaviorally correlated signals. This produces an elevation in activity at positions near the start of the first lap that is lower at positions near the end of the first lap, causing a negative slope relationship between position and LC axon activity on the first lap (Figure 4f; light green line). Furthermore, low velocities occur at the start of each lap compared to the end of the lap. Because the novelty signal is highest when animals are running slowest, the novelty signal flattens the velocity–LC activity relationship (Figure 4g; light green line). Lastly, rest periods typically occur at the start of the track. Therefore, motion onset encoding on the first lap in the novel environment occurs when the novelty signal is highest, again, flattening the relationship (Figure 4h; light green line). By the third lap in the novel environment, where the novelty-induced signal is no longer observable, the relationships between LC activity and position is no longer different than the relationship in the familiar environment (Figure 4f, green). The relationship between LC activity and motion onset is also no longer different by the third lap in the novel environment (Figure 4h). Although the relationship between velocity and LC activity is different in the third novel lap than that of the final familiar laps, by the final lap in the novel environment it is no longer different than the familiar relationship (Figure 4g). Together this demonstrates that the relationships between LC activity and behavior in the novel environment quickly return to those in the familiar environment. Interestingly, the slope is significantly increased in the final lap of the novel environment (Figure 4f), potentially indicating the development of activity at the novel reward location as has been previously reported (Kaufman et al., 2020). Altogether, examining the lap-by-lap dynamics of the position, velocity, and motion onset activity indicates that environmental novelty induces a sharp increase in LC input activity during the first two laps in the novel environment while also inducing a change in behavior that leads to increased LC input activity in subsequent laps.

We also examined the activity of a subset of LC axons (50 ROIs from 11 sessions in 9 mice) during the transition from darkness to the familiar VR environment. These axons showed elevated activity for 20 s following initial exposure to the familiar environment (Figure 3—figure supplement 2a). While this activity did not persist as long as the activity following exposure to the novel environment, it indicates these axons may be generally responsive to abrupt and unpredicted changes to the animal’s environment. This fits with the proposed role of the LC in arousal, with novelty driving greater arousal than abrupt exposure to a familiar environment.

All experimental and surgical procedures were in accordance with the University of Chicago Animal Care and Use Committee guidelines. For this study, we used 10- to 20-week-old male Slc6a3 Cre+/− (DAT-Cre+/−) mice obtained from JAX labs and Slc6a2 Cre+/− (NET-Cre+/−) (23–33 g) mice obtained from the Tonegawa lab (Wagatsuma et al., 2018). Male mice were used over female mice due to the size and weight of the headplates (9.1 mm × 31.7 mm, 2 g) which were difficult to firmly attach on smaller female skulls. Mice were individually housed in a reverse 12 hr light/dark cycle at 72°F and 47% humidity, and behavioral experiments were conducted during the animal’s dark cycle.

Mice were anesthetized (1–2% isoflurane) and injected with 0.5 ml of saline (intraperitoneal injection) and 0.05 ml of Meloxicam (1–2 mg/kg, subcutaneous injection) before being weighed and mounted onto a stereotaxic surgical station (David Kopf Instruments). A small craniotomy (1–1.5 mm diameter) was made over the VTA (±0.5 mm lateral, 3.1 mm caudal of Bregma) of DAT-Cre+/− mice or over the locus coeruleus (LC) (±0.875 mm lateral, –5.45 mm caudal of Bregma). The genetically encoded calcium indicator, pAAV-hsyn-Flex-Axon-GCaMP6s (pAAV-hSynapsin1-FLEx-axon-GCaMP6s was a gift from Lin Tian (Addgene viral prep #112010-AAV5; http://n2t.net/addgene:112010; RRID:Addgene_112010)) was injected into the VTA of DAT-Cre+/− mice (200 nl at a depth of 4.4 mm below the surface of the dura) or the LC of NET-Cre+/− mice (200 nl at a depth of 3.65 mm below Bregma). For a subset (4/8) of VTA recordings, a different GCaMP variant, pAAV-Ef1A-Flex-Axon-GCaMP7b, was injected due to the difficulty finding and recording VTA axons in dCA1 (pAAV-Ef1a-Flex-Axon-GCaMP7b) (pAAV-Ef1a-Flex-Axon-GCaMP7b was a gift from Rylan Larsen – Addgene plasmid #135419; http://n2t.net/addgene:135419; RRID: Addgene 135419). Following injections, the site was covered up using dental cement (Metabond, Parkell Corporation) and a metal headplate (9.1 mm × 31.7 mm, Atlas Tool and Die Works) was also attached to the skull with the cement. To reduce bleeding and swelling during the hippocampal window implantation, mice were separated into individual cages and water restricted for 3 weeks (0.8–1.0 ml/day). Mice then underwent surgery to implant a hippocampal window as previously described (Dombeck et al., 2010). Following implantation, the headplate was attached with the addition of a head-ring cemented on top of the headplate which was used to house the microscope objective and block out ambient light. Post-surgery mice were given 2–3 ml of water/day for 3 days to enhance recovery before returning to the reduced water schedule (0.8–1.0 ml/day).

Our VR and treadmill setup were designed similar to previously described setups (Sheffield et al., 2017; Heys et al., 2014). The virtual environments that the mice navigated through were created using VIRMEn (Aronov and Tank, 2014). 2 m (DAT-Cre mice) or 3 m (NET-Cre mice) linear tracks rich in visual cues were created that evoked numerous place fields in mice as they moved along the track at all locations (Figure 1; Bourboulou et al., 2019). Mice were head restrained with their limbs comfortably resting on a freely rotating styrofoam wheel (treadmill). Movement of the wheel caused movement in VR by using a rotatory encoder to detect treadmill rotations and feed this information into our VR computer, as in Sheffield et al., 2017; Heys et al., 2014. Mice received a water reward (4 µl) through a waterspout positioned directly in front of the animal’s mouth upon completing each traversal of the track (a lap), which was associated with a clicking sound from the solenoid. Licking was monitored by a capacitive sensor attached to the waterspout. Upon receiving the water reward, a short VR pause of 1.5 s was implemented to allow for water consumption and to help distinguish laps from one another rather than them being continuous. Mice were then virtually teleported back to the beginning of the track and could begin a new traversal. Mouse behavior (running velocity, track position, reward delivery, and licking) was collected using a PicoScope Oscilloscope (PICO4824, Pico Technology, v6.13.2). Behavioral training to navigate the virtual environment began 4–7 days after window implantation (30 min/day) and continued until mice reached >2 laps/min, which took 10–14 days (although some mice never reached this threshold). Mice that reached this behavioral threshold and had adequate GCaMP expression were imaged the following day. Out of the 36 NET-Cre mice injected, 15 were never recorded for either failing to reach behavioral criteria, or a lack of visible expression in axons. Out of the 54 DAT-Cre mice injected, images were never conducted in 36 for lack of expression or failing to reach behavioral criteria.

Imaging was done using a laser scanning two-photon microscope (Neurolabware). Using a 8-kHz resonant scanner, images were collected at a frame rate of 30 Hz with bidirectional scanning through a 16×/0.8 NA/3 mm WD water immersion objective (MRP07220, Nikon). GCaMP6s and GCaMP7b were excited at 920 nm with a femtosecond pulsed two photon laser (Insight DS + Dual, Spectra-Physics) and emitted fluorescence was collected using a GaAsP PMT (H11706, Hamamatsu). The average power of the laser measured at the objective ranged between 50–80 mW. A single imaging field of view between 400 and 700 µm equally in the x/y direction was positioned to collect data from as many VTA or LC axons as possible. Time-series images were collected from 3 to 5 planes spaced 2 µm apart using an electric lens to ensure axons remained in a field of view and reduce power going to an individual plane. Images were collected using Scanbox (v4.1, Neurolabware) and the PicoScope Oscilloscope (PICO4824, Pico Technology, v6.13.2) was used to synchronize frame acquisition timing with behavior.

The familiar environment was the same environment that the animals trained in. The experiment protocol for single day imaging sessions is shown in Figure 1. Each trial lasted 8–12 min and was always presented in the same order. Mice were exposed to the familiar rewarded environment for 10 min, then were immediately teleported to the start of a novel-rewarded VR environment and allowed to navigate for 10 min. Mice on average ran 19 ± 3.8 (mean ± 95% CI) laps in the familiar environment, at which point they were teleported to the novel environment and imaging continued for 30 ± 5.4 laps. The Novel-rewarded environment (N) had distinct visual cues, colors, and visual textures, but the same dimensions (2 or 3 m linear track) and reward location (end of the track) as the familiar environment. Imaging sessions with large amounts of drift or bleaching were excluded from analysis (eight sessions for NET mice, six sessions for LC mice). Sample size for NET-Cre mice was based off previous experiments examining calcium activity of individual somas and axons (Reimer et al., 2016; Engelhard et al., 2019). For DAT-Cre mice sample size was largely limited by the difficulty of recordings and fields of view only containing one axon.

We checked the expression axon-GCaMP to confirm expression was restricted to the VTA of DAT-Cre mice and the LC of NET-Cre. Mice were deeply anesthetized with isoflurane and perfused with 10 ml phosphate-buffered saline (PBS) followed by 20 ml 4%paraformaldehyde in PBS. The brains were removed and immersed in 30% sucrose solution overnight before being sectioned at 30-µm thickness on a cryostat. Brain slices were collected into well plates containing PBS. Slices were washed five times with PBS for 5 min then were blocked in 1% bovine serum albumin, 10% Normal goat serum, and 0.1% Triton X-100 for 2 hr. Brain slices were then incubated with 1:500 rabbit-α-TH (MAB318, Sigma-Aldrich) and 1:500 mouse-α-GFP (SAB2702197, Sigma-Aldrich)in blocking solution at 4°C. After 48 hr, the slices were incubated with 1:1000 goat-α-rabbit Alexa Fluor 647 nm secondary antibody (A32731, Thermo Fisher) and 1:1000 goat-α-mouse Alexa Fluor 488 nm (A32723, Thermo Fisher) for 2 hr. Brain slices were then collected on glass slides and mounted with a mounting media with 4’6-diamidino-2-phenylindole (DAPI; SouthernBiotech DAPI-Fluoromount-G Clear Mounting Media, 010020). The whole-brain slices were imaged under ×10 and ×40 with a Caliber I.D. RS-G4 Large Format Laser Scanning Confocal microscope from the Integrated Light Microscopy Core at the University of Chicago.

Time-series images were preprocessed using Suite2p (v0.10.1)79. Movement artifacts were removed using rigid and non-rigid transformations and assessed to ensure absence of drifts in the z-direction. Datasets with z-drifts were discarded, as determined by visually assessing imaging sessions, followed by using the registration metrics output by suite2p (eight sessions). For axon imaging, ROIs were first defined using Suite2p and manually inspected for accuracy. ROIs were then hand drawn over all segments of Suite2p defined active axons using ImageJ to ensure all axon segments were included for analysis. Fluorescent activity for each ROI was extracted and highly correlated ROIs (Pearson correlation coefficient ≥0.7) were combined and their fluorescent activity was extracted. To be included Baseline corrected ΔF/F traces across time were then generated for each ROI using both a small window of 300 frames for lap-by-lap analysis, and a larger sliding window of 2000 frames to avoid flattening slow signals for novelty response analysis. Additional ROIs were drawn over autofluorescent structures that were not identified by suite2p. These ‘blebs’ were processed in the same way as axon ROIs and used as controls to check for imaging and motion artifacts.

To remove low signal-to-noise axons, we defined the SNR of each ROI using the power spectrum of their fluorescent activity similar to Reimer et al., 2016. For frequencies above 1 Hz, the power was defined as noise because this sits outside of the range of frequencies possible for GCaMP6s fluorescence. The SNR ratio was then defined as the ratio of the peak power between 0.5 and 1 Hz over the average power between 1 and 3 Hz. The SNR of ‘blebs‘ was also determined and any axon with an SNR greater than 1.5 std from the mean of the ‘blebs’ SNR was used for analysis (110/231 LC ROIs, 9/9 VTA ROIs).

Additionally, it was observed that a subset of axon ROIs would greatly increase fluorescence at seemingly random time points and remain elevated for the rest of the trial. This activity could be due to the axons being unhealthy and filling with calcium. Therefore, we identified these axons using the cusum function in matlab to detect changes in mean activity that remained elevated for at least 2000 frames or at least 500 frames if they were still elevated at the end of the recording session and removed them from analysis (23/110 LC ROIs, 0/7 VTA ROIs).

Mouse velocity was calculated as the change in VR position divided by the sampling rate and smoothed using a Savitzky-Golay filter with a 7 frame window and 5 degree polynomial. To find the lap mean velocity, periods where the mice were immobile (velocity <5 cm/s) were removed and the average velocity during the remaining frames was calculated. The lap mean freezing ratio was calculated as the number of frames spent immobile (velocity <5 cm/s) divided by the total number of frames for each lap.

For the three measures below, to avoid weighting axons with a high SNR more than others each ROI was normalized by (ΔF/F−ΔF/Fmin)/(ΔF/Fmax−ΔF/Fmin) where ΔF/Fmin is the 1st quantile and ΔF/Fmax was the 99 quantile for each ROI. The 1st and 99th quantiles were used in order to avoid normalize to noisy outlier data points.

To find the position binned fluorescent activity of each ROI, the track was divided into 5 cm bins. For each lap, the average fluorescence in each bin was calculated for each ROI. The position binned fluorescence was then averaged across all laps in each environment to find the mean position binned activity in the familiar and novel environments.

To find the velocity binned fluorescent activity for each ROI, the velocity was divided into 1 cm/s bins from 1 to 30 cm/s. Velocities above this 14 cm/s were excluded from figures because not all mice ran faster than 14 cm/s. For each lap, the ROIs average fluorescence in each velocity bin was calculated and then averaged across all laps in each environment to find the velocity binned activity in the familiar and novel environments.

Periods where mice were immobile (velocity <5 cm/s) for at least 1.5 s then proceed to run (velocity ≥5 cm/s) for at least 3 s were identified. The fluorescent activity for ROIs for these periods was aligned to the frame mice began running (velocity crossed above 5 cm/s). The average aligned fluorescent activity of each ROI was then determined for each environment.

To assess dynamics between each of the above measures and calcium activity of LC and VTA axons, we performed linear regression on the population’s familiar environment data and significance was assessed with an F test. To compare the dynamics between LC and VTA axons, we performed exact testing based on Monte-Carlo resampling (1000 resamples with sample size matching the lower sample size condition) as detailed in legends (Figure 2e).

To assess the changing position and velocity encoding of LC axons following exposure to a novel environment, we performed linear regression on the population fluorescence data of the average of the last four laps in the familiar environment, and each of the first three laps in the novel environment for each measure. The significance for the fit of each line was assessed with an F test, and an ANCOVA was conducted to test for differences in slope between the four laps. The same process was conducted for the motion initiation dynamics, but only using ROIs in mice who paused within the first two laps and 30 s following exposure to the novel environment.

To examine the response of LC and VTA axons to the novel VR environment, the fluorescence data were normalized by the mean for each ROI and aligned to the frame where the mice were switched to the novel environment and the mean normalized F for LC and VTA ROIs at each time point was calculated. Baseline fluorescent activity was then calculated for LC and VTA ROIs separately by generating 1000 shuffled traces of the ROIs calcium activity and subsampling down to the sample size (90 for LC; 7 for VTA) 1000 times and finding the mean of the subsampled shuffles. The mean and 95% CI of all 1000 subsamples were found and the mean activity of LC and VTA ROIs was considered significantly elevated when it passed above the 95% CI of the shuffled data. The same process was repeated to define a baseline for the time binned data (fluorescent activity divided into 50 frame bins) and the lap binned data (mean activity for each lap).

Additionally, to account for changes in behavior between the familiar and novel environments, periods where the animals were immobile (velocity ≤0.2 cm/s) were removed and running periods were concatenated together and aligned to the switch to the novel environment. Here, we again defined a baseline for the time mean traces, time binned activity, and the lap binned activity using the above bootstrapping approach.