Author Response
Reviewer #1 (Public Review):
Summary:
Heer and Sheffield used 2 photon imaging to dissect the functional contributions of convergent dopamine and noradrenaline inputs to the dorsal hippocampus CA1 in head-restrained mice running down a virtual linear path. Mice were trained to collect water rewards at the end of the track and on test days, calcium activity was recorded from dopamine (DA) axons originating in the ventral tegmental area (VTA, n=7) and noradrenaline axons from the locus coeruleus (LC, n=87) under several conditions. When mice ran laps in a familiar environment, VTA DA axons exhibited ramping activity along the track that correlated with distance to reward and velocity to some extent, while LC input activity remained constant across the track, but correlated invariantly with velocity and time to motion onset. A subset of recordings taken when the reward was removed showed diminished ramping activity in VTA DA axons, but no changes in the LC axons, confirming that DA axon activity is locked to reward availability. When mice were subsequently introduced to a new environment, the ramping to reward activity in the DA axons disappeared, while LC axons showed a dramatic increase in activity lasting 90 s (6 laps) following the environment switch. In the final analysis, the authors sought to disentangle LC axon activity induced by novelty vs. behavioral changes induced by novelty by removing periods in which animals were immobile and established that the activity observed in the first 2 laps reflected novelty-induced signal in LC axons.
Strengths:
The results presented in this manuscript provide insights into the specific contributions of catecholaminergic input to the dorsal hippocampus CA1 during spatial navigation in a rewarded virtual environment, offering a detailed analysis of the resolution of single axons. The data analysis is thorough and possible confounding variables and data interpretation are carefully considered.
Weaknesses:
Aspects of the methodology, data analysis, and interpretation diminish the overall significance of the findings, as detailed below.
The LC axonal recordings are well-powered, but the DA axonal recordings are severely underpowered, with recordings taken from a mere 7 axons (compared to 87 LC axons).
Additionally, 2 different calcium indicators with differential kinetics and sensitivity to calcium changes (GCaMP6S and GCaMP7b) were used (n=3, n=4 respectively) and the data pooled. This makes it very challenging to draw any valid conclusions from the data, particularly in the novelty experiment. The surprising lack of novelty-induced DA axon activity may be a false negative. Indeed, at least 1 axon (axon 2) appears to be showing a novelty-induced rise in activity in Figure 3C. Changes in activity in 4/7 axons are also referred to as a 'majority' occurrence in the manuscript, which again is not an accurate representation of the observed data.
The reviewer points out a weakness in the analysis of VTA axons in our dataset. The relatively low n (currently 7) comes from the fact that VTA axons in the CA1 region of the hippocampus are very sparse and very difficult to record from (due to their sparsity and the low level of baseline fluorescence inherent in long range axon segments). This is the reason they have not been recorded from in any other lab outside of our lab. LC axons, on the other hand, are more abundant in CA1. In the paper when comparing VTA versus LC axons we deal with the mismatch in n by downsampling the LC axons to match the VTA axons and repeated this 1000 times to create a distribution. However, because the VTA axon n is relatively low, it is possible that we have not sampled the VTA axon population sufficiently and therefore have a biased population in our dataset. The issue is that it takes months for the baseline expression of GCaMP to reach sufficient levels to be able to record from VTA axons, and it is typical to find only a single axon in a FOV per animal. There are additional reasons why mice and/or axon recordings do not reach criteria and cannot be included in the dataset (these exclusion criteria are reported in the Methods section). For instance, out of the 54 DAT-Cre mice injected, images were never conducted in 36 for lack of expression or because mice failed to reach behavioral criteria. Another 11 mice were excluded for heat bubbles that developed during imaging, z-drift of the FOV, or bleaching of the GCaMP signal.
However, we do have n=2 additional VTA axon recordings that we will add to the dataset to bring the n up from 7 to 9. We plan on re-analyzing the data with n=9 VTA axons and making comparisons to down-sampled LC axons as described above. This boost in n will increase the power of our VTA axon analysis. To more formally test whether this is sufficient for statistical tests, we plan to utilize the G*power power-analysis tool to compute statistical power for each of the different tests we use. We will report this in the next version of the paper. However, the n=2 additional axons were nor recorded in the novel environment, so the next version will remain at n=7 for the novel environment analysis. We agree with the reviewer that the lack of the novelty induced DA axon activity may be a false negative, and so we will adjust the description of our results and discussion accordingly.
During the data collection of VTA axon activity we tried two variants of GCaMP: 6s and 7b, to see if one would increase the success rate of finding and recording from VTA axons. Given the long time-course of these experiments and the low yield in success, we pooled the GCaMP variants together to increase statistical power. Because the 2 additional VTA DA axons that were recorded from expressed GCaMP6s, the next version of the paper will have n=5
GCaMP6s, and n=4 GCaMP7b VTA DA axons, which will allow us to compare the activity of the two sensors in the familiar environment. The reviewer correctly pointed out that the sensors themselves could confound our results, and so they should not be pooled unless we can show they do not produce different signals in the axons. We will make this comparison and report the findings in the next version of the paper. If we find no significant differences, we will pool the data. If differences are detected, we will keep these axons separate for subsequent analysis and comparisons to LC axons.
The authors conducted analysis on recording data exclusively from periods of running in the novelty experiment to isolate the effects of novelty from novelty-induced changes in behavior. However, if the goal is to distinguish between changes in locus coeruleus (LC) axon activity induced by novelty and those induced by motion, analyzing LC axon activity during periods of immobility would enhance the robustness of the results.
This is indeed true, and this suggested analysis could further support our conclusions regarding the LC novelty signal. For the next version of the paper, we will use the periods of immobility to analyze and isolate any novelty induced activity in LC axons. However, following exposure to the novel environment, mice spend much less time immobile, therefore there may not be sufficient periods of immobility close in time to the exposure to the novel environment (which is when the novelty signal occurs). We plan to analyze mouse behavior during the early exposure to the novel environment for immobility and check whether we have enough of this behavior to perform the suggested analysis.
The authors attribute the ramping activity of the DA axons to the encoding of the animals' position relative to reward. However, given the extensive data implicating the dorsal CA1 in timing, and the remarkable periodicity of the behavior, the fact that DA axons could be signalling temporal information should be considered.
This is a very good point. We agree that the VTA DA axons could be signaling temporal information, as we have previously shown that these axons also exhibit ramping activity when you average their activity by time to reward (Krishnan et. al., 2022). We will conduct this analysis on this dataset. We have not, however, conducted any experiments designed to separate out time from distance, such as the experiments conducted in Kim et. al., 2020. Therefore, we cannot determine whether this is due to proximity in space to reward or time to reward. We will clarify in our text that by proximity, we mean either place or time, and cannot conclude which feature of the experience drives the VTA axon signal.
Krishnan, L.S., Heer, C., Cherian, C., Sheffield, M.E. Reward expectation extinction restructures and degrades CA1 spatial maps through loss of a dopaminergic reward proximity signal. Nat Commun 13, 6662 (2022).
Kim, HyungGoo R., Athar N. Malik, John G. Mikhael, Pol Bech, Iku Tsutsui-Kimura, Fangmiao
Sun, Yajun Zhang, et al. A Unified Framework for Dopamine Signals across Timescales. Cell 183, no. 6 (2020).
The authors should explain and justify the use of a longer linear track (3m, as opposed to 2m in the DAT-cre mice) in the LC axon recording experiments.
LC axon activity was recorded on a 3m track to match the track length from an experiment we recently published (Dong et al., 2021) in which mice were exposed to a novel 3m track while populations of CA1 pyramidal cells were recorded. In that paper we described the time course of place field formation on the novel track. We wanted to test if LC axons signaled novelty (as we hypothesized) and whether the time course of LC axon activity matched the time course of place field formation. We briefly discuss this in the Discussion section of this paper and hypothesize that LC axons in CA1 could open a window of plasticity in which new place fields can form.
VTA axons were recorded on a 2m track (same VR tracks as LC axons were recorded on) to match another recent paper from our lab in which reward expectation was manipulated (Krishnan et al, 2022). In that study CA1 populations of pyramidal cells were recorded during the reward expectation experiment. To match the experience during recordings of VTA axons in CA1 to test how reward expectation may influence axon signaling along the track, we also used a 2m track. The idea was to check how VTA dopaminergic inputs to CA1 may influence CA1 population dynamics along the track.
Although the tracks were identical for LC and VTA recordings for both the familiar and novel tracks in terms of visual cues and design, the track lengths are different (simply modulated by gain control of the rotary encoder). To account for this we normalized the lengths for our comparison analysis. This normalization allows for a direct comparison of the patterns of activity across the two types of axons, controlling for the potential confound introduced by the different track lengths. By adjusting the data to a common scale, we could assess the relative changes in activity levels at matched spatial bins, ensuring that any observed differences or similarities are due to the intrinsic properties of the axons rather than differences in track lengths. However, the different lengths do make the animal’s experience slightly different. This is somewhat offset by the observations in our study that none of the LC or VTA axon signals would be expected to be majorly influenced by variations in track length. For instance, LC axons are associated with velocity and a pre-motion initiation signal, neither of which would be influenced by track length. VTA axons are also associated with velocity, which would not influence a direct comparison to LC axon velocity signals as mice reach maximal velocity very rapidly along the track. VTA axons do ramp up in activity as they approach the reward zone, and this signal could be modulated by track length (or maybe not if the signal is encoding time to reward rather than distance). However, LC axons show no ramping to reward signals, so a comparison across axons recorded on different track lengths for this analysis is justified.
However, to add rigor to comparisons of axon dynamics recorded along 2m and 3m tracks, we plan to plot axon activity of both sets of axons by time to reward, and actual (un-normalized) distance from reward.
Krishnan, L.S., Heer, C., Cherian, C., Sheffield, M.E. Reward expectation extinction restructures and degrades CA1 spatial maps through loss of a dopaminergic reward proximity signal. Nat Commun 13, 6662 (2022).
Dong, C., Madar, A. D. & Sheffield, M.E. Distinct place cell dynamics in CA1 and CA3 encode experience in new environments. Nat Commun 12, 2977 (2021).
Reviewer #2 (Public Review):
Summary:
The authors used 2-photon Ca2+-imaging to study the activity of ventral tegmental area (VTA) and locus coeruleus (LC) axons in the CA1 region of the dorsal hippocampus in head-fixed male mice moving on linear paths in virtual reality (VR) environments.
The main findings were as follows:
- In a familiar environment, the activity of both VTA axons and LC axons increased with the mice's running speed on the Styrofoam wheel, with which they could move along a linear track through a VR environment.
- VTA, but not LC, axons showed marked reward position-related activity, showing a ramping-up of activity when mice approached a learned reward position.
- In contrast, the activity of LC axons ramped up before the initiation of movement on the Styrofoam wheel.
- In addition, exposure to a novel VR environment increased LC axon activity, but not VTA axon activity.
Overall, the study shows that the activity of catecholaminergic axons from VTA and LC to dorsal hippocampal CA1 can partly reflect distinct environmental, behavioral, and cognitive factors. Whereas both VTA and LC activity reflected running speed, VTA, but not LC axon activity reflected the approach of a learned reward, and LC, but not VTA, axon activity reflected initiation of running and novelty of the VR environment.
I have no specific expertise with respect to 2-photon imaging, so cannot evaluate the validity of the specific methods used to collect and analyse 2-photon calcium imaging data of axonal activity.
Strengths:
(1) Using a state-of-the-art approach to record separately the activity of VTA and LC axons with high temporal resolution in awake mice moving through virtual environments, the authors provide convincing evidence that the activity of VTA and LC axons projecting to dorsal CA1 reflect partly distinct environmental, behavioral and cognitive factors.
(2) The study will help a) to interpret previous findings on how hippocampal dopamine and norepinephrine or selective manipulations of hippocampal LC or VTA inputs modulate behavior and b) to generate specific hypotheses on the impact of selective manipulations of hippocampal LC or VTA inputs on behavior.
Weaknesses:
(1)The findings are correlational and do not allow strong conclusions on how VTA or LC inputs to dorsal CA1 affect cognition and behavior. However, as indicated above under Strengths, the findings will aid the interpretation of previous findings and help to generate new hypotheses as to how VTA or LC inputs to dorsal CA1 affect distinct cognitive and behavioral functions.
(2) Some aspects of the methodology would benefit from clarification.
First, to help others to better scrutinize, evaluate, and potentially to reproduce the research, the authors may wish to check if their reporting follows the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines for the full and transparent reporting of research involving animals (https://arriveguidelines.org/). For example, I think it would be important to include a sample size justification (e.g., based on previous studies, considerations of statistical power, practical considerations, or a combination of these factors). The authors should also include the provenance of the mice. Moreover, although I am not an expert in 2-photon imaging, I think it would be useful to provide a clearer description of exclusion criteria for imaging data.
We thank the reviewer for helping us formalize the scientific rigor of our study. There are ten ARRIVE Guidelines and we have addressed most of them in our study already. However, there is an opportunity to add detail. We have listed below all ten points and how we have or will address each one.
(1) Experimental design - we go into great depth explaining the experimental set-up, how we used the autofluorescent blebs as imaging controls, how we controlled for different sample sizes between the two populations, and the statistical tests used for comparisons. We also carefully accounted for animal behavior when quantifying and describing axon dynamics both in the familiar and novel environments.
(2)Sample size - We state both the number of ROIs and mice for each analysis. Wherever we state how many axons had a certain kind of activity, we will also state the number of mice we saw this activity in. For the next version of the paper, we plan to conduct a
power analysis using G*power to assess the power of our sample sizes for statistical analysis.
(3) Inclusion/exclusion criteria - 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. Out of the remaining 21 NET-CRE, 5 were excluded for heat bubbles, z-drift, or bleaching, while 11 DAT-Cre were excluded for the same reasons. This was determined by visually assessing imaging sessions, followed by using the registration metrics output by suite2p. This registration metric conducted a PCA on the motion-corrected ROIs and plotted the first PC. If the PC drifted largely, to the point where no activity was apparent, the video was excluded from analysis.
(4) Randomization - Already included in the paper is a description of random down sampling of LC axons to make statistical comparisons with VTA axons. LC axons were selected pseudo-randomly (only one axon per imaging session) to match VTA sampling statistics. This randomization was repeated 1000 times and comparisons were made against this random distribution.
(5) Blinding-masking - no blinding/masking was conducted as no treatments were given that would require this. We will include this statement in the next version.
(6) Outcomes - We defined all outcomes measured, such as those related to animal behavior and related axon signaling.
(7) Statistical methods - None of the reviewers had any issues regarding our description of statistical methods, which we described in detail in this version of the paper.
(8) Experimental animals - We described that DAT- Cre mice were obtained through JAX labs, and NET-Cre mice were obtained from the Tonegawa lab (Wagatsuma et al. 2017)
(9) Experimental procedure - Already listed in detail in Methods section.
(10) Results - Rigorously described in detail for behaviors and related axon dynamics.
Wagatsuma, Akiko, Teruhiro Okuyama, Chen Sun, Lillian M. Smith, Kuniya Abe, and Susumu Tonegawa. “Locus Coeruleus Input to Hippocampal CA3 Drives Single-Trial Learning of a Novel Context.” Proceedings of the National Academy of Sciences 115, no. 2 (January 9, 2018): E310–16. https://doi.org/10.1073/pnas.1714082115.
Second, why were different linear tracks used for studies of VTA and LC axon activity (from line 362)? Could this potentially contribute to the partly distinct activity correlates that were found for VTA and LC axons?
A detailed response to this is written above for a similar comment from reviewer 1.
Third, the authors seem to have used two different criteria for defining immobility. Immobility was defined as moving at <5 cm/s for the behavioral analysis in Figure 3a, but as <0.2 cm/s for the imaging data analysis in Figure 4 (see legends to these figures and also see Methods, from line 447, line 469, line 498)? I do not understand why, and it would be good if the authors explained this.
This is an error leftover from before we converted velocity from rotational units of the treadmill to cm/s. This will be corrected in the next version of the paper.
(3) In the Results section (from line 182) the authors convincingly addressed the possibility that less time spent immobile in the novel environment may have contributed to the novelty-induced increase of LC axon activity in dorsal CA1 (Figure 4). In addition, initially (for the first 2-4 laps), the mice also ran more slowly in the novel environment (Figure 3aIII, top panel). Given that LC and VTA axon activity were both increasing with velocity (Figure 1F), reduced velocity in the novel environment may have reduced LC and VTA axon activity, but this possibility was not addressed. Reduced LC axon activity in the novel environment could have blunted the noveltyinduced increase. More importantly, any potential novelty-induced increase in VTA axon activity could have been masked by decreases in VTA axon activity due to reduced velocity. The latter may help to explain the discrepancy between the present study and previous findings that VTA neuron firing was increased by novelty (see Discussion, from line 243). It may be useful for the authors to address these possibilities based on their data in the Results section, or to consider them in their Discussion.
This is a great point. The decreased velocity in the novel environment could lead to a diminished novelty response in LC axons. We will add a discussion point on this in the next version. This could also be the case for VTA axons, so will add a discussion point that the lack of novelty signaling seen in VTA axons could be due to reduced velocity masking this signal.
(4) Sensory properties of the water reward, which the mice may be able to detect, could account for reward-related activity of VTA axons (instead of an expectation of reward). Do the authors have evidence that this is not the case? Occasional probe trials, intermixed with rewarded trials, could be used to test for this possibility.
Mice receive their water reward through a waterspout that is immobile and positioned directly in front of their mouth (which is also immobile as they are head fixed) and water delivery is triggered by a solenoid when the mice reach the end of the virtual track. Therefore, because the waterspout remains in the same place relative to the mouse, and the water reward is not delivered until they reach the end of the virtual track, there is nothing for the mice to detect. We will update the paper to make this clearer.
Additionally, on the initial laps with no reward, the ramping activity is still present (Krishnan et al, 2022) indicating this activity is not directly related to the presence/absence of water but is instead caused by reward expectation.
Reviewer #3 (Public Review):
Summary:
Heer and Sheffield provide a well-written manuscript that clearly articulates the theoretical motivation to investigate specific catecholaminergic projections to dorsal CA1 of the hippocampus during a reward-based behavior. Using 2-photon calcium imaging in two groups of cre transgenic mice, the authors examine the activity of VTA-CA1 dopamine and LC-CA1 noradrenergic axons during reward seeking in a linear track virtual reality (VR) task. The authors provide a descriptive account of VTA and LC activities during walking, approach to reward, and environment change. Their results demonstrate LC-CA1 axons are activated by walking onset, modulated by walking velocity, and heighten their activity during environment change. In contrast, VTA-CA1 axons were most activated during the approach to reward locations. Together the authors provide a functional dissociation between these catecholamine projections to CA1. A major strength of their approach is the methodological rigor of 2-photon recording, data processing, and analysis approaches. These important systems neuroscience studies provide solid evidence that will contribute to the broader field of learning and memory. The conclusions of this manuscript are mostly well supported by the data, but some additional analysis and/or experiments may be required to fully support the author's conclusions.
Weaknesses:
(1) During teleportation between familiar to novel environments the authors report a decrease in the freezing ratio when combining the mice in the two experimental groups (Figure 3aiii). A major conclusion from the manuscript is the difference in VTA and LC activity following environment change, given VTA and LC activity were recorded in separate groups of mice, did the authors observe a similar significant reduction in freezing ratio when analyzing the behavior in LC and VTA groups separately?
In response to this comment, we will analyze the freezing ratios in DAT-Cre and NET-Cre mice separately. However, other members of the lab have seen the same result in other mouse strains (See Dong et al. 2021), so we do not expect to see a difference (but it is certainly worth checking).
(2) The authors satisfactorily apply control analyses to account for the unequal axon numbers recorded in the LC and VTA groups (e.g. Figure 1). However, given the heterogeneity of responses observed in Figures 3c, 4b and the relatively low number of VTA axons recorded (compared to LC), there are some possible limitations to the author's conclusions. A conclusion that LC-CA1 axons, as a general principle, heighten their activity during novel environment presentation, would require this activity profile to be observed in some of the axons recorded in most all LC-CA1 mice.
We agree with the reviewer’s point here. To help avoid this problem, when downsampling LC axons to compare to VTA axons, we matched the sampling statistics of the VTA axons/mice (i.e. only one LC axon was taken from each mouse to match the VTA dataset).
However, in the next version of the paper we will also report the number of mice that we see a significant novel response in. We will also add the number of mice with significant activity for each of the measures in the familiar environment (e.g. how many mice had axons positively correlated with velocity).
Additionally, if the general conclusion is that VTA-CA1 axons ramp activity during the approach to reward, it would be expected that this activity profile was recorded in the axons of most all VTA-CA1 mice. Can the authors include an analysis to demonstrate that each LC-CA1 mouse contained axons that were activated during novel environments and that each VTA-CA1 mouse contained axons that ramped during the approach to reward?
As stated above, we will add the number of mice that had each activity type we reported here.
(3) A primary claim is that LC axons projecting to CA1 become activated during novel VR environment presentation. However, the experimental design did not control for the presentation of a familiar environment. As I understand, the presentation order of environments was always familiar, then novel. For this reason, it is unknown whether LC axons are responding to novel environments or environmental change. Did the authors re-present the familiar environment after the novel environment while recording LC-CA1 activity?
This is an important point to address. While we never varied the presentation order of the familiar vs novel environments, we did record the activity of LC axons in some of the mice in a dark environment (no VR cues) prior to exposure to the familiar environment. We will look at these axons to address whether they respond to initial exposure to the familiar environment. This will allow us to check whether they are responding to environmental change or novelty. We will add this analysis to the next version of the paper.
