Author response:
The following is the authors’ response to the original reviews.
Public Reviews:
Reviewer #1 (Public Review):
Summary:
The authors use an innovative behavior assay (chamber preference test) and standard calcium imaging experiments on cultured dorsal root ganglion (DRG) neurons to evaluate the consequences of global knockout of TRPV1 and TRPM2, and overexpression of TRPV1, on warmth detection. They find a profound effect of TRPM2 elimination in the behavioral assay, whereas elimination of TRPV1 has the largest effect on neuronal responses. These findings are of importance, as there is still substantial discussion in the field regarding the contribution of TRP channels to different aspects of thermosensation.
Strengths:
The chamber preference test is an important innovation compared to the standard two-plate test, as it depends on thermal information sampled from the entire skin, as opposed to only the plantar side of the paws. With this assay, and the detailed analysis, the authors provide strong supporting evidence for the role of TRPM2 in warmth avoidance. The conceptual framework using the Drift Diffusion Model provides a first glimpse of how this decision of a mouse to change between temperatures can be interpreted and may form the basis for further analysis of thermosensory behavior.
Weaknesses:
The authors juxtapose these behavioral data with calcium imaging data using isolated DRG neurons. Here, there are a few aspects that are less convincing.
(1) The authors study warmth responses using DRG neurons after three days of culturing. They propose that these "more accurately reflect the functional properties and abundance of warm-responsive sensory neurons that are found in behaving animals." However, the only argument to support this notion is that the fraction of neurons responding to warmth is lower after three days of culture. This could have many reasons, including loss of specific subpopulations of neurons, or any other (artificial?) alterations to the neurons' transcriptome due to the culturing. The isolated DRGs are not selected in any way, so also include neurons innervating viscera not involved in thermosensation. If the authors wish to address actual changes in sensory nerves involved in warmth sensing in TRPM2 or TRPV1 KO mice without disturbing the response profile as a result of the isolation procedure, other approaches would be needed (e.g. skin-nerve recordings or in vivo DRG imaging).
We agree that there could be several reasons as to why the responses of cultured DRGs are reduced compared to the acute/short-term cultures. It is possible ––and likely–– that transcriptional changes happen over the course of the culturing period. It is also possible that it is a mere coincidence that the 3-day cultures have a response profile more similar to the in vivo situation than the acute cultures. In the revised manuscript, we have therefore toned down the claim that the 3-day cultures mirror the native conditions more appropriately and included the sentence “However, whether 3-day cultures resemble native sensory neurons more closely than acute cultures in terms of their (transcriptional) identity is currently unknown.” (page 5).
We now also included a section “Limitations of the study” and bring this point up there as well and acknolwedge that longer culturing periods may cause changes in the neurons and may result in a drift away from their native state.
Nevertheless, our results clearly show that acute cultures have a response profile that is much more similar to damaged/”inflamed” neurons, irrespective of any comparison to the 3 daycultures. Therefore, we believe, it is helpful to include this data to make scientists aware that acute cultures are very different to non-inflamed native/in vivo DRG neurons that many researchers use in their experiments.
(2) The authors state that there is a reduction in warmth-sensitive DRG neurons in the TRPM2 knockout mice based on the data presented in Figure 2D. This is not convincing for the following reasons. First, the authors used t-tests (with FDR correction - yielding borderline significance) whereas three groups are compared here in three repetitive stimuli. This would require different statistics (e.g. ANOVA), and I am not convinced (based on a rapid assessment of the data) that such an analysis would yield any significant difference between WT and TRPM2 KO. Second, there seems to be a discrepancy between the plot and legend regarding the number of LOV analysed (21, 17, and 18 FOV according to the legend, compared to 18, 10, and 12 dots in the plot). Therefore, I would urge the authors to critically assess this part of the study and to reconsider whether the statement (and discussion) that "Trpm2 deletion reduces the proportion of warmth responders" should be maintained or abandoned. .
Yes, we agree that the statistical tests indicated by the referee are more appropriate/robust for the data shown in Figures 1F, 2D, and 4G.
When we perform 2-way repeated measures ANOVA and subsequent multiple comparison test (with Dunnets correction) against Wildtype, for data shown in Fig. 2D, both the main effect (Genotype) and the interaction term (Stimulus x Genotype) are significant. The multiple comparison yields very similar result as in the current manuscript, with the difference that the TRPM2-KO data for the second stimulus (~36°C) is borderline significant (with a p-value of p=0.050).
Due to the possible dependence of the repeated temperature stimuli and the variability of each stimulus between FOVs (Fig. 2C), it is possible that a mixed-effect model that accounts for these effects is more appropriate.
Similarly, for plots 1F and 4G, Genotype (either as main effect or as interaction with Time) is significant after a repeated measures two-way ANOVA. The multiple comparisons (with Bonferroni correction) only changed the results marginally at individual timepoints, without affecting the overall conclusions. The exception is Fig. 4G at 38°C, where the interaction of Time and Genotype is significant, but no individual timepoint-comparison is significant after Bonferroni correction.
The main difference between the results presented above and the ones presented in the manuscript is the choice of the multiple comparison correction. We originally opted for the falsediscovery rate (FDR) approach as it is less prone to Type II errors (false negatives) than other methods such as Sidaks or Bonferroni, particularly when correcting for a large number of tests.
However, we are mainly interested in whether the genotypes differ in their behavior in each temperature combination and the significant ANOVA tests for Fig. 1F and 4G support that point. The statistical test and comparison used in the original/previous version of the manuscript, comparing behavior at individual/distinct timepoints, are interesting, but less relevant (and potentially distracting), as we do not go into the details about the behavior at any given/distinct timepoint in the assay.
Therefore, and per suggestion of the reviewer, we have updated the statistics in the revised version of the manuscript. Also, we now report the correct number of FOVs in the legend. The statistical details are now found in the legends of the respective figures.
(3) It remains unclear whether the clear behavioral effect seen in the TRPM2 knockout animals is at all related to TRPM2 functioning as a warmth sensor in sensory neurons. As discussed above, the effects of the TRPM2 KO on the proportion of warmth-sensing neurons are at most very subtle, and the authors did not use any pharmacological tool (in contrast to the use of capsaicin to probe for TRPV1 in Figures S3 and S4) to support a direct involvement of TRPM2 in the neuronal warmth responses. Behavioral experiments on sensory-neuron-specific TRPM2 knockout animals will be required to clarify this important point
As mentioned above, we have toned down the correlation between the cellular and behavioral data.
In the discussion we now clearly describe three possibilities as to why the Trpm2 knockout animals only show a subtle cellular thermal phenotype but a strong behavioral thermal preference phenotype: (i) permanent deletion of Trpm2 may result in developmental defects and/or compensatory mechanisms; (ii) The DRG population expressing Trpm2 may be more relevant for autonomic thermoregulation rather than behavioral responses to temperature; (iii) Trpm2 expression outside DRGs (possibly in the hypothalamic POA) may account for the altered thermal behavior.
(4) The authors only use male mice, which is a significant limitation, especially considering known differences in warmth sensing between male and female animals and humans. The authors state "For this study, only male animals were used, as we aimed to compare our results with previous studies which exclusively used male animals (7, 8, 17, 43)." This statement is not correct: all four mentioned papers include behavioral data from both male and female mice! I recommend the authors to either include data from female mice or to clearly state that their study (in comparison with these other studies) only uses male mice.
This is a valid point -- when our study started 7-8 years ago, we only used male mice (as did many other researchers) and this we would now do differently. We have now newly included a statement concerning this limitation in the “Limitations of this study” section of the manuscript.
Nevertheless, in the studies by Tan et al. And Vandevauw et al. only male animals were used for the behavioral experiments. Yarmolinsky et al. And Paricio-Montesinons et al. used both males and females while, as far as we can tell, only Paricio-Montesions et al. Reported that no difference was observed between the sexes.
Wildtypes are all C57bl/6N from the provider Janvier. Generally, all lines are backcrossed to C57bl/6 mice and additionally inbreeding was altered every 4-6 generations by crossing to C57bl/6. Exactly how many times the Trp channel KOs have been backcrossed to C57bl/6 mice we cannot exactly state.
Reviewer #2 (Public Review):
Summary:
The authors of the study use a technically well-thought-out approach to dissect the question of how far TRPV1 and TRPM2 are involved in the perception of warm temperatures in mice. They supplement the experimental data with a drift-diffusion model. They find that TRPM2 is required to trigger the preference for 31{degree sign}C over warmer temperatures while TRPV1 increases the fidelity of afferent temperature information. A lack of either channel leads to a depletion of warm-sensing neurons and in the case of TRPV1 to a deficit in rapid responses to temperature changes. The study demonstrates that mouse phenotyping can only produce trustworthy results if the tools used to test them measure what we believe they are measuring.
Strengths:
The authors tackle a central question in physiology to which we have not yet found sufficient answers. They take a pragmatic approach by putting existing experimental methods to the test and refining them significantly.
Weaknesses:
It is difficult to find weaknesses. Not only the experimental methods but also the data analysis have been refined meticulously. There is no doubt that the authors achieved their aims and that the results support their conclusions.
There will certainly be some lasting impact on the future use of DRG cultures with respect to (I) the incubation periods, (II) how these data need to be analyzed, and (III) the numbers of neurons to be looked at.
As for the CPT assay, the future will have to show if mouse phenotyping results are more accurate with this technique. I'm more fond of full thermal gradient environments. However, behavioural phenotyping is still one of the most difficult fields in somatosensory research.
We thank the referee and were happy to read that the referee finds our study valuable and insightful.
Reviewer #3 (Public Review):
Summary and strengths:
In the manuscript, Abd El Hay et al investigate the role of thermally sensitive ion channels TRPM2 and TRPV1 in warm preference and their dynamic response features to thermal stimulation. They develop a novel thermal preference task, where both the floor and air temperature are controlled, and conclude that mice likely integrate floor with air temperature to form a thermal preference. They go on to use knockout mice and show that TRPM2-/- mice play a role in the avoidance of warmer temperatures. Using a new approach for culturing DRG neurons they show the involvement of both channels in warm responsiveness and dynamics. This is an interesting study with novel methods that generate important new information on the different roles of TRPV1 and TRPM2 on thermal behavior.
Open questions and weaknesses:
(1) Differences in the response features of cells expressing TRPM2 and TRPV1 are central and interesting findings but need further validation (Figures 3 and 4). To show differences in the dynamics and the amplitude of responses across different lines and stimulus amplitudes more clearly, the authors should show the grand average population calcium response from all responsive neurons with error bars for all 3 groups for the different amplitudes of stimuli (as has been presented for the thermal stimuli traces). The authors should also provide a population analysis of the amplitude of the responses in all groups to all stimulus amplitudes. Prior work suggests that thermal detection is supported by an enhancement or suppression of the ongoing activity of sensory fibers innervating the skin. The authors should present any data on cells with ongoing activity.
We have now included grand average population analysis of the different groups in the revised version, this is found in Figure 2E and F. Based on the referee’s suggestion and the new analysis, we now can report a (subtle) cellular phenotype observed in DRG cultures of Trpm2 deficient animals: when averaging all warmth responses, the new analysis suggests that Trpm2-deficient cultures lack modulation of the response magnitude across the three increasing consecutive warmth stimuli (33°C, 36°C and 39°C).
Concerning the point about ongoing activity: We are not sure if it is possible in neuronal cultures to faithfully recapitulate ongoing activity. Ongoing activity has been mostly recorded in skinnerve preparations (or in older studies in other types of nerve recordings) and there are only very few studies that show ongoing activity in cultured neurons and in those instances the ongoing activity only starts in sensory neuron cultures when cultured for even longer time periods than 3 days (Ref.: doi: 10.1152/jn.00158.2018). We have very few cells that show some spontaneous activity, but these are too few to draw any conclusions. In any case, nerve fibers might be necessary to drive ongoing activity which are absent from our cultures.
(2) The authors should better place their findings in context with the literature and highlight the novelty of their findings. The introduction builds a story of a 'disconnect' or 'contradictory' findings about the role of TRPV1 and TRPM2 in warm detection. While there are some disparate findings in the literature, Tan and McNaughton (2016) show a role for TRPM2 in the avoidance of warmth in a similar task, Paricio et al. (2020) show a significant reduction in warm perception in TRPM2 and TRPV1 knock out lines and Yarmolinksy et al. (2016) show a reduction in warm perception with TRPV1 inactivation. All these papers are therefore in agreement with the authors finding of a role for these channels in warm behavior. The authors should change their introduction and discussion to more correctly discuss the findings of these studies and to better pinpoint the novelty of their own work.
Paricio-Montesinos et al. argue that TRPM8 is crucial for the detection of warmth, as TRPM8KO animals are incapable of learning the operant task. TRPM2-KO animals and, to a smaller extent TRPV1-KO animals, have reduced sensitivity in the task, but are still capable of learning/performing the task. However, in our chamber preference assay this is reversed: TRPM2-KO animals lose the ability to differentiate warm temperatures while TRPM8 appears to play no major role. A commonality between the two studies is that while TRPV1 affects the detection of warm temperatures in the different assays, this ion channel appears not to be crucial.
Similarly, Yarmolinsky et al. show that Trpv1-inactivation only increases the error rate in their operant assay (from ~10% to ~30%), without testing TRPM2. And Tan et al. show the importance of TRPM2 in the preference task, without testing for TRPV1.
More generally, the choice of the assay, being either an operant task (Paricio-Montesinos et al. and Yarmolinsky et al.) or a preference assay without training of the mice (Tan et al. and our data here), might be important and different TRP receptors may be relevant for different types of temperature assays, which we have now included at the end of the discussion section in the revised manuscript. While our results generally agree with the previous studies, they add a different perspective on the analysis of the behavior (with correlation to cellular data). We now edited the manuscript to highlight the advances more clearly.
Nevertheless, we believe that a discrepancy between cellular and behavioral data in the former studies exists and we kept this in the introduction. We hope that our data and suggestions of more nuanced analysis of cellular and behavioral responses, in particular also differences in their kinetics, may be helping to guide future studies.
(3) The responses of 60 randomly selected cells are shown in Figure 2B. But, looking at the TRPM2-/- data, warm responses appear more obvious than in WTs and the weaker responders of the WT group appear weaker than the equivalent group in the TRPV1-/- and TRPM2-/- data. This does not necessarily invalidate the results, but it may suggest a problem in the data selection. Because the correct classification of warm-sensitive neurons is central to this part of the study more validation of the classifier should be presented. For example, the authors could state if they trained the classifier using equal amounts of cells, show some randomly selected cells that are warm-insensitive for all genotypes, and show the population average responses of warm-insensitive neurons.
The classifier was trained on a balanced dataset of 1000 (500 responders and 500 nonresponders), manually labelled traces across all 5 temperature stimuli. The prediction accuracy was 98%. We have now described more clearly how the classifier was trained (See Materials and Methods) and include examples of responders and non-responders, the population averages of each class as well as a confusion matrix of the classification in the revised manuscript (Suppl. Figure 4A and B).
(4) The interpretation of the main behavioral results and justification of the last figure is presented as the result of changes in sensing but differences in this behavior could be due to many factors and this needs clarification and discussion. (i) The authors mention that 'crucially temperature perception is not static' and suggest that there are fluctuating changes in perception over time and conclude that their modelling approach helps show changes in temperature detection. They imply that temperature perceptual threshold changes over time, but the mouse could just as easily have had exactly the same threshold throughout the task but their motivation (or some other cognitive variable) might vary causing them to change chamber. The authors should correct this. (ii) Likewise, from their fascinating and high-profile prior work the authors suggest a model of internal temperature sensing whereby TRPM2 expression in the hypothalamus acts as an internal sensory of body temperature. Given this, and the slow time course of the behavior in chambers with different ambient temperatures, couldn't the reason for the behavioral differences be due to central changes in hypothalamic processing rather than detection by skin temperature? If TRPM2-/- were selectively ablated from the skin or the hypothalamus (these experiments are not necessary for this paper) it might be possible to conclude whether sensation or body temperature is more likely the root cause of these effects but, without further experiments it is tough to conclude either way. (iii) Because the ambient temperature is controlled in this behavior, another hypothesis is that warm avoidance could be due to negative valence associated with breathing warm air, i.e. a result of sensation within the body in internal pathways, rather than sensing from the external skin. Overall, the authors should tone down conclusions about sensation and present a more detailed discussion of these points.
We are sorry that the statement including the phrase “crucially temperature perception is not static” was ambiguous; We have now deleted this statement and instead included different possibilities as to why mice may switch from one chamber to the other stochastically.
As the referee mentioned, it is possible that some other variable (motivation etc.) makes the mouse change the chamber; Nevertheless, we hypothesize that this variable (whatever it might be) is still modulated by temperature (at least this would be the likeliest explanation that we see).
As for the aspect of internal/hypothalamic temperature sensing and its dependence on Trpm2: we have included this possibility in the discussion in the manuscript.
As for the point of negative valence mediated by breathing in warm air: yes, presumably this could also be possible. The aspect of valence is in interesting aspect by itself: would the mice be rather repelled from the (uncomfortable) hot plate or more attracted to the (more comfortable) thermoneutral plate, or both? Something to elucidate in a different study.
(5) It is an excellent idea to present a more in-depth analysis of the behavioral data collected during the preference task, beyond 'the mouse is on one side or the other'. However, the drift-diffusion approach is complex to interpret from the text in the results and the figures. The results text is not completely clear on which behavioral parameters are analyzed and terms like drift, noise, estimate, and evidence are not clearly defined. Currently, this section of the paper slightly confuses and takes the paper away from the central findings about dynamics and behavioral differences. It seems like they could come to similar conclusions with simpler analysis and simpler figures.
We have now reassessed the description of the drift diffusion model and explain it more clearly, this can be found on page 5 – 8. We have considered whether it will be better to introduce the drift diffusion model at the beginning of the study, subsequent to Figure 1 but we believe this to better suited at the end, because, indeed, the cellular results (and differences in kinetic response parameters observed in DRG cultures of Trpv1 KO mice) prompted us to assess the behavior in this way. Thus, the order of experiments presented here, represents also more the natural path the study took.
(6) In Figure 2D the % of warm-sensitive neurons are shown for each genotype. Each data point is a field of view, however, reading the figure legend there appear to be more FOVs than data points (eg 10 data points for the TRPV1-/- but 17 FOVs). The authors should check this.
We have checked and corrected the number of FOVs mentioned in the legend, and the number shown in the Figure 2D and its legend are now in agreement.
(7) Can the authors comment on why animals with over-expression of TRPV1 spend more time in the warmest chamber to start with at 38C and not at 34C?
This is an interesting observation that we did not consider before. A closer look at Figure 4H reveals that the majority of the TRPV1-OX animals, have a proportionally long first visit to the 38°C room. We can only speculate why this is the case. We cannot rule out that this a technical shortcoming of the assay and how we conduced it – but we did not observe this for the wildtype mice, thus it is rather unlikely a technical problem. It is possible that this is a type of “freezing-” (or “startle-“) behavior when the animals first encounter the 38°C temperature. Freezing behaviors in mice can be observed when sudden/threatening stimuli are applied. It is possible that, in the TRPV1-overexpressing animals, the initial encounter with 38°C leads to activation of a larger proportion of cells (compared to WT controls), possibly signaling a “threatening” stimulus, and thus leading to this startle effect. However, such a claim would require additional experiments to test such a hypothesis more rigorously.
