Analyzing the brainstem circuits for respiratory chemosensitivity in freely moving mice

Curation statements for this article:
  • Curated by eLife

    eLife logo

    Evaluation Summary:

    This study utilizes a miniscope approach and GCaMP6 in freely behaving conscious mice to record CO2-associated multicellular calcium responses of neurons or glia in brainstem regions implicated in CO2-dependent control of breathing. The application of this approach in this context is extremely attractive, and new to the respiratory neurobiology field. Several technical improvements could strengthen the manuscript. Foremost, the study is broad in scope and consequently not always technically rigorous in important aspects such as identification of cell types imaged. In some cases that affects interpretation of the significance of the results. Since some of the conclusions about cellular responses to CO2 are mostly at odds with a substantial literature using more established techniques, there is even greater onus on the authors to ensure reliability of the results.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

This article has been Reviewed by the following groups

Read the full article See related articles


Regulation of systemic PCO 2 is a life-preserving homeostatic mechanism. In the medulla oblongata, the retrotrapezoid nucleus (RTN) and rostral medullary Raphe are proposed as CO 2 chemosensory nuclei mediating adaptive respiratory changes. Hypercapnia also induces active expiration, an adaptive change thought to be controlled by the lateral parafacial region (pF L ). Here, we use GCaMP6 expression and head-mounted mini-microscopes to image Ca 2+ activity in these nuclei in awake adult mice during hypercapnia. Activity in the pF L supports its role as a homogenous neuronal population that drives active expiration. Our data show that chemosensory responses in the RTN and Raphe differ in their temporal characteristics and sensitivity to CO 2 , raising the possibility these nuclei act in a coordinated way to generate adaptive ventilatory responses to hypercapnia. Our analysis revises the understanding of chemosensory control in awake adult mouse and paves the way to understanding how breathing is coordinated with complex non-ventilatory behaviours.

Article activity feed

  1. Evaluation Summary:

    This study utilizes a miniscope approach and GCaMP6 in freely behaving conscious mice to record CO2-associated multicellular calcium responses of neurons or glia in brainstem regions implicated in CO2-dependent control of breathing. The application of this approach in this context is extremely attractive, and new to the respiratory neurobiology field. Several technical improvements could strengthen the manuscript. Foremost, the study is broad in scope and consequently not always technically rigorous in important aspects such as identification of cell types imaged. In some cases that affects interpretation of the significance of the results. Since some of the conclusions about cellular responses to CO2 are mostly at odds with a substantial literature using more established techniques, there is even greater onus on the authors to ensure reliability of the results.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

  2. Reviewer #1 (Public Review):

    This study by Bhandare et al. applies endoscopic fluorescence imaging with head-mounted miniscopes and genetically encoded calcium sensors in awake behaving mice to document patterns of multicellular (including neurons and glial cells) calcium activity within circumscribed regions of the medulla oblongata (retrotrapezoid nucleus-RTN, Raphe magnus and pallidus nuclei, lateral parafacial region- pFL). These regions are proposed to have important chemosensory functions for the regulation of respiratory responses to elevated systemic CO2 (hypercapnia) and are critical for homeostatic regulation of breathing in mammals. Analyzing chemosensory properties of these medullary regions has been the focus of numerous studies, but the problem of analyzing regional multi-cellular chemosensory responses in the awake freely behaving rodent has not been previously addressed, so this paper represents an advance for the field. The authors importantly demonstrate chemosensory responses of neurons and astrocytes in RTN and Raphe, and activity profiles of neurons in pFL, and they describe substantial regional heterogeneity of cellular responses to hypercapnia in the RTN and Raphe regions. From this heterogeneity and distinct regional differences in chemosensory responses, the authors propose functional roles of these regions for detecting different aspects of the hypercapnic stimulus. In general all of the results presented suggest as justified by the data presented that cellular activity profiles in these various regions are more complex and encode different features for respiratory regulation than would be predicted from neuronal activity recordings in anesthetized animals.

    Strengths: (1) The authors demonstrate the successful technical application of the deep endoscopic fluorescence imaging approach for multicellular calcium activity imaging from key medullary structures involved in respiratory control in freely behaving mice. (2) Their novel results indicate functional diversity of neuronal responses and complexity in the encoding of chemosensory signals in key regions that have not been previously described from in vivo studies in anesthetized animals.

    Weaknesses: While the experiments are technically well executed for the most part in terms of implementing this imaging approach in freely behaving mice, important clarifications about experimental design and data analyses are required, and the authors do not fully discuss important technical limitations that may significantly affect interpretation of the results. A major concern with this imaging approach is whether the targeted regional neuronal and glial populations have been adequately sampled throughout the regions studied to reasonably understand the spatial and functional heterogeneity of the neuronal or glial hypercapnic responses. There is also the problem that cell identify has not been adequately established in some regions (e.g. RTN, pFL), which the authors need to specifically address to fully justify their conclusions about significance of their results. This is a particular problem for RTN neurons in which the chemosensory neurons that encode graded elevations of CO2, which were rarely found in this study, have an established molecular phenotype and this phenotype has not be sufficiently verified to know if the various response patterns identified in the freely behaving mouse are actually associated with this functional cell type.

    So while these new results indicate diversity of regional neuronal responses, many of these responses differ from the existing literature on how these cells are proposed to respond to CO2 from experiments with anesthetized rodents. Most notably, neurons in the RTN region were found to have an adapting excitatory response or were inhibited, rather than the expected preponderance of graded neuronal responses encoding the level of CO2; pFL neurons failed to exhibit a sustained expiratory-related oscillation predicted from studies suggesting that this area contains a conditional expiratory oscillator; and astrocytes failed to respond in terms of calcium signaling with elevated CO2. This underscores the need for additional technical clarifications to ensure reliability of the results, and for the authors to amplify discussion of the important caveats with their experimental design and functional interpretations.

  3. Reviewer #2 (Public Review):

    The approach used here is a powerful method to monitor cellular activity in vivo. As the authors point out, this method allowed them to study activity of putative central chemoreceptors in unanesthetized adult mice, providing information that has not been attainable before using the standard approach of electrophysiological recordings from neonatal or anesthetized juvenile rodents. The results they obtained were different than expected, and call into question the hypothesis of primacy of the RTN in CO2 chemoreception.

    The results provide unique information that is very valuable in understanding the respiratory response to hypercapnia, but there are three issues with interpretation that the authors should more explicitly address, and which may alter some of the conclusions.

    First, since recordings were made from neurons in an intact brain, synaptic inputs were intact and all responses that were measured were due to a combination of intrinsic responses, synaptic input and glial modulation. The authors briefly acknowledge this, but it does not always seem to be taken into account when the results are discussed. Instead, some of the text seems to imply that the responses are all intrinsic. There should be an explicit discussion of the role of synaptic connectivity, and they should discuss the relationship of their results with data from more reduced preparations (about which there is little discussion).

    Second, expression of GCaMP6 was driven by the synapsin promoter in all RTN neurons and in one-third of raphe neurons, so all nearby neurons were transfected. There was comparison to post hoc immunostaining for vGlut2, Neuromedin B+ or TPH, but it is not clear what the identity of the recorded neurons was in many cases. This is mentioned, but then the text glosses over the lack of clear identification.

    Third, GCaMP6 measures calcium levels, not membrane potential. It is a pretty safe bet that in a given neuron calcium levels are a rough surrogate for firing rate, but there may not be a linear relationship. For example, adaptation may reflect slow inactivation of calcium channels rather than a decrease in firing rate. Likewise, in some neurons calcium levels may not increase as much as in others. For example, neurons with a low dynamic range (e.g. 5-HT neurons that don't increase their firing rate very high even when maximally activated), or those without many calcium channels, may not increase their calcium levels as much as other neurons. For glia it is not clear that "activation" by hypercapnia would always cause a change in intracellular calcium or membrane potential, and yet signaling pathways could still be engaged that influence neighboring neurons.

    These issues should be discussed. However, if the results are taken at face value, there are several surprises in the results.

    A graded or sustained response was exceedingly rare (4%) among RTN neurons. Instead, 22 of 46 RTN neurons adapted to continuous exposure to hypercapnia. It is widely accepted that chemoreceptors need to maintain a sustained increase in firing that is proportional to low levels of CO2 for as long as they are exposed. An abundance of papers have shown that respiratory motor output and ventilation respond to a constant increase in CO2 without adaptation or roll-off. The finding of adaptation of the response of RTN neurons is surprising, and different than has been reported from electrophysiological recordings from neonatal or anesthetized juvenile rodents. The authors interpret these results in the following way: "Although adapting responses to hypercapnia have not been described before, the responses of the neurons matched changes in minute ventilation (VE) calculated from the WBP of the mouse (Fig 2G). Furthermore the average Ca2+ trace of all 22 EA neurons showed a consistent waveform (Fig 2H) that matched the average of the rectified and smoothed WBP trace obtained from all respective mice. This correspondence between features of the responses in these neurons and the adaptive ventilatory response, supports the hypothesis that these are a physiologically important class of chemosensitive neurons." However, on inspection of the trace of VE for a single animal shown in Fig 2G, and the trace of "WBP Av ReSm" shown in Fig 2H, there is actually not any adaptation. When both of these traces are expanded (stretched) in the vertical direction, and a horizontal line is added at the level of the baseline, it is obvious that there is a graded increase in ventilation in proportion to CO2, as expected. This observation calls into question the relevance of the adapting response of RTN neurons to chemoreception.

    What the authors seem to be referring to is a brief burst of activity in the WBP trace at the onset of 3% CO2 in Figure 2G, but there are similar bursts at other times when there is no change in CO2. There is no burst when going from 3% to 6%. These very brief, high amplitude bursts look like movement artifact. Did the authors rule that out? They show that movement doesn't induce an artifact in neuronal recordings, but movement artifact is very common in WBP. Were these bursts due to a response of the animal to sound or a pressure pulse induced by the change in gas? In many of the other WBP traces shown there is rarely a burst of activity with similar timing shortly after the switch of CO2, while there are many bursts that occur at random times unrelated to gas changes.

    Page 12, Top paragraph: "plethysmographic changes in breathing are well-known to adapt to step changes in inspired CO2 (e.g. Fig 2G,H)..." Do the authors have references for this assertion?

    Many individual RTN neurons actually do have adapting responses to hypercapnia. This is not what is expected of respiratory chemoreceptors, since as above ventilation does not adapt to a continuous hypercapnic stimulus. However, the authors conclude that "detection of change in PCO2 is an important role for RTN neurons." There is no evidence for this statement. There is no supporting evidence for any functional role of such adaptation by chemoreceptors. Some pattern generating neurons and motor neurons may adapt, but if chemoreceptors have adaptation that has not been demonstrated. Rapid step changes in CO2 probably don't occur often in vivo and the system may not have evolved to respond to them.

    A substantial fraction of raphe neurons (42%) responded as expected for chemoreceptors with a graded or sustained increase in activity in response to a step increase in CO2. The existence of such a response in 5-HT neurons has previously been questioned on the basis of recordings from anesthetized juvenile rodents, so this paper adds important and novel information.

    A substantial fraction of RTN neurons (11%) and raphe neurons (38%) displayed a decrease in activity in response to hypercapnia. It is not clear if these inhibitory responses were intrinsic or synaptic. The authors provide an explanation for this related to GABA input, but another possibility is that 5-HT neurons that are stimulated by CO2 inhibit other 5-HT neurons by activation of 5-HT1a autoreceptors. Inhibited RTN and raphe neurons may also not be involved in control of breathing. Likewise, in several places the authors suggest that tonic RTN neurons "provide tonic drive to the respiratory network." However, it is possible that these neurons are not respiratory. Just because they are in/near the RTN does not mean they are chemoreceptors.

    Page 8, 2nd paragraph: "...overall glial cells did not appear to be strongly stimulated by CO2 in ways that could make a major contribution..." Is there any evidence that glial calcium levels need to rise in order for glia to influence neighboring neurons? They don't have action potentials, and calcium dependent vesicular fusion, if it occurs at all, is not the only mechanism of neurochemical release from glia. Glia may influence neurons without any change in membrane potential or a rise in calcium.

    Page 15, Concluding remarks: "Raphe neurons tended to be active during the entire CO2 stimulus and conceivably these neurons are likely to take over from RTN neurons under pathophysiological levels of CO2." It is hard to understand how the authors come to this conclusion. The logic is never explained. The data are more supportive of the opposite conclusion. 42% of raphe neurons respond with a graded or sustained increase in firing in response to a rise in CO2 of as little as 3%, whereas most RTN neurons rapidly adapt (in only 1-2 minutes) to 3% CO2 and don't increase to 6% CO2 (a level that strongly stimulates breathing). The data are more consistent with the conclusion that at low levels of CO2, RTN neurons would have little or no effect, while raphe neurons would be expected to provide continuous activation of the respiratory network across the whole CO2 range studied.

  4. Reviewer #3 (Public Review):

    The brainstem cell types that are involved in ventilatory responses to hypercapnia are a matter of great importance to our overall understanding of the control of breathing. This paper seeks to use a cutting-edge, in vivo cell imaging approach (miniscope) to detect changes in cell activity that accompany elevations in CO2 in unanesthetized mice. It samples a multitude of cell types, including some previously suggested to serve as CO2 chemosensors for respiratory control. The observations reported here run largely in opposition to the existing understanding of how those cells respond to CO2 (e.g., serotonergic neurons inhibited by CO2, "RTN" neurons inhibited by CO2 or with an adapting excitatory response, astrocytes unresponsive to CO2). Based on this, and apparently unbothered by the lack of any functional characterization of the recorded cells, the authors develop a detailed speculative model for their role in respiratory control.

    1. A major technical issue with this work is the inability to adequately rule out movement artifacts. The authors provide some indirect evidence that the measured Ca transients are independent of movement, but these are not convincing. For example, the GFP signal measured in different mice cannot be an adequate control for movement artifacts in other GCaMP-expressing mice. My understanding is that Inscopix miniscopes are capable of dual color imaging to detect a co-expressed fluorescent marker in the same cell, a better control. It is also worth noting that even this control does not rule out a movement-evoked change in GCaMP fluorescence for reasons other than the cell moving in and out of the plane of focus, e.g., from effects of movement elsewhere that are transmitted to the recorded cell. This may be especially an issue for astrocytes that are notoriously prone to producing calcium transients when mechanically disturbed (Marina et al., Nature Comm, 2020). It is also surprising that the CO2 challenges, interpreted to reflect a wide variety of cellular response patterns, were not applied in repeated or alternately ordered fashion (0-3-6, 0-6-3). This experimental protocol could have more convincingly verified a consistent response in each individual cell, supporting the designation of these as true activity patterns and, at the same time, dissociated common CO2-associated response patterns from any more randomly associated movement artifacts. A time-honored approach in electrophysiology, especially in vivo, is the use of peristimulus averaging to tease out signal from noise.

    2. Another major technical issue deals with cell identification. For astrocytes and serotonin neurons (at least in some cases), an attempt was made to restrict GCaMP expression to the targeted cell groups by using viruses including GFAP (gfaABC1D) or SERT promoters. Not so for the RTN or the pFL. In these cases, a non-specific neuronal promoter (synapsin) was used to drive GCaMP expression, with the obvious problem that the recordings were made from unidentified cells. Although this might be overlooked for the pFL, a theoretical cell group for which no cell-specific marker has been identified, it is impossible to justify for the RTN. The molecular phenotype of CO2-sensitive RTN neurons has been well-established, and viral vectors using a Phox2b promoter have been employed for many years to preferentially target those RTN neurons. There are not a lot of RTN neurons (maybe 700-800 in mice) and there are many other neurons in the same brainstem region. Thus, the likelihood that some of these other neurons were transduced with GCaMP and accounted for the measured responses to CO2 seems high. Notably, a recent paper from the Mulkey group (eLife, 2021) demonstrated that Sst-expressing parafacial interneurons in the same region are activated, inhibited and unaffected by CO2.

    It is important to acknowledge that the authors tried, post hoc, to identify some of their GCaMP-expressing parafacial cells as bona fide RTN neurons, based on immunostaining for NMB and VGLUT2. Unfortunately, the immunohistochemistry presented is not believable. First, there are no controls provided for the specificity of the antibody staining, and no references attesting to the quality of the antibody - this is unacceptable (see Rhodes and Trimmer, J Neuroscience 2006; Saper, JCN, 2005; Saper & Sawchenko, ibid, 2003). This is especially problematic for this antibody-based detection of somatic staining for a vesicular transporter and a neuropeptide, both of which are typically detected in nerve terminals and undetectable in cell bodies. A further concern is the floccular appearance of the NMB staining. Finally, even if the staining were validated as specific for the relevant antigens, the issue with cell identification would persist. That is, demonstrating that some of the GCaMP-expressing neurons were VGLUT2- and NMB-positive RTN neurons provides no guarantee that the recorded cells were also RTN neurons.

    (A further note on the antibody staining issue. There are also no quality control details provided for any of the other antibodies that were employed in this work.)

    3. For the raphe, it is not clear what information is added from the experiments from unidentified cells obtained with the synapsin-promoter driven GCaMP. For the SERT-driven GCaMP-expressing cells, the preponderance of CO2-inhibited cells is surprising given the extensive work from Richerson and colleagues.

    4. There was very little effect of CO2 on calcium in RTN astrocytes, a finding again at odds with much compelling work, e.g., from Gourine and colleagues, among others, who have found that CO2 activates those glial cells. Was the camera able to capture the astocytes near the ventral medullary surface that have been most associated with CO2 activation and control of breathing? Do the authors have independent histochemical evidence that the gfaABC1D promoter indeed restricted GCaMP expression to glial cells? Also, how does one conclude based on the strength of the calcium signal whether or not the astrocytes could play a functional role - at what threshold would they decide the signal is big enough? and on what would they base that decision.

    5. The pFL section is also problematic since again we have no idea about the identification of the neurons expressing GCaMP using the non-specific synapsin promoter. This is really an exploration of the unknown with the unknown, and the suppositions advanced regarding causality for control of active expiration based only on time-associated Ca signals in unidentified cells.