scRNA-sequencing reveals subtype-specific transcriptomic perturbations in DRG neurons of PirtEGFPf mice in neuropathic pain condition
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
This is an important paper that uses state of the art technology to address the underlying neurobiology of neuropathic pain, a topic of considerable translational relevance. The study describes changes in gene expression at a single cell resolution in somatosensory neurons following peripheral nerve injury. Bioinformatics analyses were employed to segregate neurons in sub-classes and to derive predictions on potential functions of regulated genes. While the work has considerable strengths, such as the single cell approach, there are also some weaknesses, including the fact that new gene candidates did not undergo functional analysis.
(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 #2 agreed to share their name with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Neuroscience (eLife)
Abstract
Functionally distinct subtypes/clusters of dorsal root ganglion (DRG) neurons may play different roles in nerve regeneration and pain. However, details about their transcriptomic changes under neuropathic pain conditions remain unclear. Chronic constriction injury (CCI) of the sciatic nerve represents a well-established model of neuropathic pain, and we conducted single-cell RNA-sequencing (scRNA-seq) to characterize subtype-specific perturbations of transcriptomes in lumbar DRG neurons on day 7 post-CCI. By using Pirt EGFPf mice that selectively express an enhanced green fluorescent protein in DRG neurons, we established a highly efficient purification process to enrich neurons for scRNA-seq. We observed the emergence of four prominent CCI-induced clusters and a loss of marker genes in injured neurons. Importantly, a portion of injured neurons from several clusters were spared from injury-induced identity loss, suggesting subtype-specific transcriptomic changes in injured neurons. Moreover, uninjured neurons, which are necessary for mediating the evoked pain, also demonstrated cell-type-specific transcriptomic perturbations in these clusters, but not in others. Notably, male and female mice showed differential transcriptomic changes in multiple neuronal clusters after CCI, suggesting transcriptomic sexual dimorphism in DRG neurons after nerve injury. Using Fgf3 as a proof-of-principle, RNAscope study provided further evidence of increased Fgf3 in injured neurons after CCI, supporting scRNA-seq analysis, and calcium imaging study unraveled a functional role of Fgf3 in neuronal excitability. These findings may contribute to the identification of new target genes and the development of DRG neuron cell-type-specific therapies for optimizing neuropathic pain treatment and nerve regeneration.
Article activity feed
-
-
Author Response
Reviewer #2 (Public Review):
Suggestions to improve the paper:
Major Issues
- I do not think that the introduction accurately reflects the state of the field with respect to single cell omics and nerve injury. The CCI model is different than the SNI model, which has been used in most previous studies, in terms of the nature of the injury, and the resolution of pain after the injury. I do not think it is accurate to claim that the CCI model is somehow more relevant clinically, because both models are just that. It is also not really true that co-mingling, un-injured neurons have not been profiled before. The Renthal paper did this, but using a different model. There is value in what the authors have done here, but they can state it more clearly in the introduction. In particular, most published studies have only …
Author Response
Reviewer #2 (Public Review):
Suggestions to improve the paper:
Major Issues
- I do not think that the introduction accurately reflects the state of the field with respect to single cell omics and nerve injury. The CCI model is different than the SNI model, which has been used in most previous studies, in terms of the nature of the injury, and the resolution of pain after the injury. I do not think it is accurate to claim that the CCI model is somehow more relevant clinically, because both models are just that. It is also not really true that co-mingling, un-injured neurons have not been profiled before. The Renthal paper did this, but using a different model. There is value in what the authors have done here, but they can state it more clearly in the introduction. In particular, most published studies have only used male mice, so the sex differences aspect of this work is important. In that regard, the authors did not cite any of the growing literature on sex differences in neuropathic pain mechanisms.
We revised the introduction and discussion to address the comments. Specifically, we revised the related information about animal models (Page 4-5). Although Renthal et al. examined co-mingling, “un-injured” neurons using a sciatic crush injury model, they did not find cell-type specific changes in uninjured neurons. The reason for this is unclear, but we speculate that it may be partially due to differences in the techniques (e.g., tissue processing, cell sorting, sequencing depth) and animal models (CCI versus crush injury). Compared to sciatic CCI induced by loose ligation of the sciatic nerve, crush injury would injure most nerve fibers (~50% of L3-5 DRG neurons are axotomized in this model). Therefore, the remaining “uninjured’ neurons for sequencing may be much less than that in the CCI model. In addition, we used Pirt-EGFPf mice to establish a highly efficient purification approach to enrich neurons for scRNA-seq and therefore largely increased the number of genes detected in DRG neurons. Comparatively, the neuronal selectivity and number of genes detected were lower in the previous study, which may have resulted in fewer DEGs and decreased ability to detect aforementioned changes. We include a brief discussion (Page 24).
We appreciate the reviewer’s good suggestion, and cited sex differences studies in neuropathic pain mechanisms (Pages 5, 25). Although our findings suggest that peripheral neuronal mechanisms may also underlie sexual dimorphisms in neuropathic pain, Renthal et al. reported no differences in subtype distributions or injury-induced transcriptional changes between males and females after sciatic nerve crush injury (Renthal et al., 2020). We also discussed the differences between current findings and previous work and also emphasized the sex differences aspect of this work in the discussion (Page 25).
- I am curious about the choice to only use samples from 7 days after CCI. One of the advantages of the CCI model is that pain resolves at about 35-60 days, depending on how the ligations are done, and this allows one to look at how transcriptional programs change in DRG neurons after pain resolves. This would give some new insight, at least in comparison to the very comprehensive profiling done in the sciatic nerve crush model by Renthal and colleagues.
We thank the reviewer for this comment. We provided the rationale for day 7 post-CCI (Page 22). It is the time point when neuropathic pain-like behavior is fully developed in most animals, and the post-injury time point examined in many previous studies. The reviewer is correct, an advantage of the CCI model is that pain resolves at about 35-60 days. Although meaningful, it was not our intention to conduct a time course study to fully characterize time-dependent transcriptional changes using scRNA-seq, which is costly and requires a great effort for data analysis, etc., and is beyond the scope of the current study. We will address this in a future study, and provided a brief discussion (Page 22).
- An alternative interpretation of the ATF3 expression is that the dissociation protocol causes this upregulation. ATF3 induction may be rapid and could occur due to the technique the authors chose to use. This could be acknowledged.
We agree and acknowledged this in our original discussion (Page 22).
- I think the authors are a bit over-confident in their call of "injured" and "un-injured" neurons based on Sprr1a expression. This is really the only grounds for calling these neurons injured or uninjured. The fact is that the CCI model does not provide a clear way to determine injured and uninjured neurons contributing to neuropathic pain. This is an advantage of the SNL model, as shown in many classic papers from the Chung lab.
We included a brief discussion about Sprr1a (Page 22). Although Atf3 is a classic marker of injured neurons in some previous studies, a recent study suggested that Sprr1a may be a better standard to define “injured” neurons (Nguyen et al., 2017). Although injured and uninjured neurons can be readily separated in the SNL model, they are mostly from different DRGs, but not intermingled in the same DRG. Since glia-neuron interaction and neuron-neuron interaction may occur between cells within the same DRG after injury, these interactions may profoundly affect neuronal excitability and gene expression. Accordingly, we choose the CCI model for the current study to determine whether injured and uninjured neurons contribute to neuropathic pain. We included a brief discussion (Page 5, 23, 24).
- There are now two papers on human DRG neurons that are available. One was recently published in eLife, and the other is available on Biorxiv, and has been there since Feb 2021. I expected the authors to make some comparisons of cell types that are changing in CCI with populations that are found in humans. Would similar effects be expected? Are these cell types represented in the human DRG?
Study of human DRG is important, and recent studies elegantly characterized neurochemical and physiological properties. Previous findings have suggested some notable difference between human and rodent DRGs. Importantly, many markers and methods used for classifying subpopulations of rodent DRG neurons do not apply well to human DRG neurons. In addition, data from human DRG came from patients with different etiologies, but not due to peripheral nerve injury as in the animal study. Due to these differences, we feel that it is difficult to make direct compassion of cell types that are changing in CCI with corresponding human DRG neurons.
Minor Issues
- Does the 40 um cell strainer eliminate some larger diameter cells from the analysis?
We think this is unlikely, as large-diameter cells such as NF1 and NF2 clusters were also observed in our dataset. Importantly, we examined the cell strainer by washing it out inversely and did not find single cells. In addition, all subtypes identified in other studies were also found in our study. Nevertheless, an underrepresentation of the amount of NF neurons may be a result of the fact that not all NF neurons are GFP-positive in Pirt-EGFPf mice. In Pirt-EGFPf mice, expression of the knockin EGFPf was under the control of the endogenous Pirt promoter. Anti-GFP antibody staining revealed that GFP is widely expressed in 83.9% of all neurons. However, Pirt-negative neurons are mainly NF200+ and have large-diameter cell bodies. In addition, compared to small neurons, large neurons are also easier to lose during FACS sorting. We included a brief discussion of this potential limitation, as the NF population may be underrepresented in our sample set (Page 21).
-
Evaluation Summary:
This is an important paper that uses state of the art technology to address the underlying neurobiology of neuropathic pain, a topic of considerable translational relevance. The study describes changes in gene expression at a single cell resolution in somatosensory neurons following peripheral nerve injury. Bioinformatics analyses were employed to segregate neurons in sub-classes and to derive predictions on potential functions of regulated genes. While the work has considerable strengths, such as the single cell approach, there are also some weaknesses, including the fact that new gene candidates did not undergo functional analysis.
(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. …
Evaluation Summary:
This is an important paper that uses state of the art technology to address the underlying neurobiology of neuropathic pain, a topic of considerable translational relevance. The study describes changes in gene expression at a single cell resolution in somatosensory neurons following peripheral nerve injury. Bioinformatics analyses were employed to segregate neurons in sub-classes and to derive predictions on potential functions of regulated genes. While the work has considerable strengths, such as the single cell approach, there are also some weaknesses, including the fact that new gene candidates did not undergo functional analysis.
(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 #2 agreed to share their name with the authors.)
-
Reviewer #1 (Public Review):
The manuscript describes changes in single cell RNA sequencing signatures of dorsal root ganglion neurons over the development of neuropathic pain in the murine chronic constriction injury (CCI) model. Bioinformatic algorithms were employed to cluster neurons into sub-classes described previously, based on transcriptomic signatures. The authors report emergence of 4 new clusters, resulting from loss of cellular identity of neurons in known clusters and induction of inflammatory and hyperexcitability-associated genes. Moreover, they segregated neurons between injured and uninjured subclasses and observed induction of genes in both categories as well as differences between these categories. Some injured neurons maintained cellular identity. conversely, non-injured neurons were also found to show significant …
Reviewer #1 (Public Review):
The manuscript describes changes in single cell RNA sequencing signatures of dorsal root ganglion neurons over the development of neuropathic pain in the murine chronic constriction injury (CCI) model. Bioinformatic algorithms were employed to cluster neurons into sub-classes described previously, based on transcriptomic signatures. The authors report emergence of 4 new clusters, resulting from loss of cellular identity of neurons in known clusters and induction of inflammatory and hyperexcitability-associated genes. Moreover, they segregated neurons between injured and uninjured subclasses and observed induction of genes in both categories as well as differences between these categories. Some injured neurons maintained cellular identity. conversely, non-injured neurons were also found to show significant transcriptional plasticity in clusters with a prominent role in pain sensitivity. Sexual dimorphism was noted, particularly with respect to the c-LTMR class of sensory neurons.
The results of the current study are interesting, and the study is very well-performed. The fact that fluorescently labelled DRG neurons were employed here is an advantage since it led to lower representation of non-neuronal genes and better representation of neuronal genes expressed at low levels. However, it is largely descriptive and the level of advance beyond recent single cell transcriptomics studies on DRG neurons as well as older studies on bulk sequencing in models of neuropathic pain is debatable.
-
Reviewer #2 (Public Review):
Suggestions to improve the paper:
Major Issues
- I do not think that the introduction accurately reflects the state of the field with respect to single cell omics and nerve injury. The CCI model is different than the SNI model, which has been used in most previous studies, in terms of the nature of the injury, and the resolution of pain after the injury. I do not think it is accurate to claim that the CCI model is somehow more relevant clinically, because both models are just that. It is also not really true that co-mingling, un-injured neurons have not been profiled before. The Renthal paper did this, but using a different model. There is value in what the authors have done here, but they can state it more clearly in the introduction. In particular, most published studies have only used male mice, so the sex …
Reviewer #2 (Public Review):
Suggestions to improve the paper:
Major Issues
- I do not think that the introduction accurately reflects the state of the field with respect to single cell omics and nerve injury. The CCI model is different than the SNI model, which has been used in most previous studies, in terms of the nature of the injury, and the resolution of pain after the injury. I do not think it is accurate to claim that the CCI model is somehow more relevant clinically, because both models are just that. It is also not really true that co-mingling, un-injured neurons have not been profiled before. The Renthal paper did this, but using a different model. There is value in what the authors have done here, but they can state it more clearly in the introduction. In particular, most published studies have only used male mice, so the sex differences aspect of this work is important. In that regard, the authors did not cite any of the growing literature on sex differences in neuropathic pain mechanisms.
- I am curious about the choice to only use samples from 7 days after CCI. One of the advantages of the CCI model is that pain resolves at about 35-60 days, depending on how the ligations are done, and this allows one to look at how transcriptional programs change in DRG neurons after pain resolves. This would be give some new insight, at least in comparison to the very comprehensive profiling done in the sciatic nerve crush model by Renthal and colleagues.
- An alternative interpretation of the ATF3 expression is that the dissociation protocol causes this upregulation. ATF3 induction may be rapid and could occur due to the technique the authors chose to use. This could be acknowledged.
- I think the authors are a bit over-confident in their call of "injured" and "un-injured" neurons based on Sprr1a expression. This is really the only grounds for calling these neurons injured or uninjured. The fact is that the CCI model does not provide a clear way to determine injured and uninjured neurons contributing to neuropathic pain. This is an advantage of the SNL model, as shown in many classic papers from the Chung lab.
- There are now two papers on human DRG neurons that are available. One was recently published in eLife, and the other is available on Biorxiv, and has been there since Feb 2021. I expected the authors to make some comparisons of cell types that are changing in CCI with populations that are found in humans. Would similar effects be expected? Are these cell types represented in the human DRG?
Minor Issues
- Does the 40 um cell strainer eliminate some larger diameter cells from the analysis?
-