Single-cell RNA sequencing reveals time- and sex-specific responses of mouse spinal cord microglia to peripheral nerve injury and links ApoE to chronic pain

Activation of microglia in the spinal cord following peripheral nerve injury is critical for the development of long-lasting pain hypersensitivity. However, it remains unclear whether distinct microglia subpopulations or states contribute to different stages of pain development and maintenance. Using single-cell RNA-sequencing, we show that peripheral nerve injury induces the generation of a male-specific inflammatory microglia subtype, and demonstrate increased proliferation of microglia in male as compared to female mice. We also show time- and sex-specific transcriptional changes in different microglial subpopulations following peripheral nerve injury. Apolipoprotein E (Apoe) is the top upregulated gene in spinal cord microglia at chronic time points after peripheral nerve injury in mice. Furthermore, polymorphisms in the APOE gene in humans are associated with chronic pain. Single-cell RNA sequencing analysis of human spinal cord microglia reveals a subpopulation with a disease-related transcriptional signature. Our data provide a detailed analysis of transcriptional states of mouse and human spinal cord microglia, and identify a link between ApoE and chronic pain in humans.

but is not unique to these clusters. Microglia in clusters 7 and 8 are proliferating microglia as cell cycleassociated genes, Mki67 and Top2a are uniquely expressed in these two sub-populations and not in any other cluster (see Extended Data Fig. 6a) and show enrichment for cell division-related process in GO analysis (see Extended Data Fig. 6b). Stmn1 regulates microtubule dynamics in many cellular processes, including cell division 5,6 and cell motility. Because of its critical roles in the cell cycle, it is often referred to as the mitotic spindle regulator 7 . Stmn1 has been shown to be highly expressed in microglia (see Fig.   1 in 8 ) and its levels are particularly upregulated in proliferating cells 5,9 . As correctly indicated by the reviewer, Stmn1 is also expressed in neurons [10][11][12] , further demonstrating that this gene can be expressed in different cell types.
Similarly, PRC1 (Protein Regulator of cytokinesis) is involved in cell division (cytokinesis) 13 in a variety of cell types. PRC1 is present at high levels during the S and G2/M phases of mitosis, but its levels drop dramatically when the cell exits mitosis and enters the G1 phase. Specifically, PRC1 is a member of the MAP65/ASE1 family of nonmotor microtubule-associated proteins. PRC1 is a substrate of CDK1, which maintains PRC1 in an inactive, monomeric state. Cell-cycle dependent degradation of CDK1 leads to dephosphorylation of PRC1 and subsequent KIF4-mediated translocation to the plus ends of microtubules, where it promotes microtubule bundling by cross-linking antiparallel microtubules. The microtubule bundling functions of PRC1 play a critical role in maintaining structural integrity of the spindle midzone during cytokinesis 14,15 . Thus, the abundant expression of PRC1 in cluster 7 microglia is consistent with proliferating nature of cells in this cluster, and does not suggest that PRC1 is uniquely expressed in this subpopulation. In the revised version of the paper, we better explain the differences between uniquely expressed and abundantly expressed genes in each cluster.
We define cluster 10 as vessel-associated microglia since cells in this cluster express canonical microglia genes, along with unique expression of Cldn5. It has been previously shown that vessel-associated microglia express low levels of Cldn5 at baseline and Cldn5 expression is increased upon activation in this subpopulation 16 . As mentioned previously, removal of doublets did not affect cluster 10 markers (Rebuttal letter, Fig. 2), strengthening the notion that cells in this cluster uniquely express Cldn5 and supporting categorization of this cluster as vessel-associated microglia. In the revised version of the manuscript, we performed in situ hybridization to show that Cldn5 is present in microglia (labeled genetically in TMEM119 CreERT2 tdTomato mice) in close proximity to endothelial cells (expressing Pecam1 mRNA) (Extended data Fig. 7j), likely representing vessel-associated microglia.
Finally, we defined cluster 11 as perivascular macrophages as they express canonical microglia genes, although at reduced levels as compared to all other clusters (Fcrls, Trem2, Cx3Cr1, Tmem119, C1qa and P2ry12), along with unique expression of macrophage/monocyte markers (H2-Aa, Mrc1, Ccr2, Lyve1, Dab2, Mgl2, F13a1) (Supplementary Table 1 and Extended Data Fig. 6d). Indeed, as shown in Extended Data Fig. 6d, macrophage-specific genes, H2-Aa and Mrc1, are expressed only in cluster 11 and not in any other cluster, strongly suggesting their macrophage identity. Interestingly, this subpopulation has been recently identified to play an important role in regulation of pain after peripheral nerve injury 17 . We have expanded the discussion on these cells in the revised version of the manuscript.
3) Does cluster 3 has a sub-cluster? Could this be a distinct cluster?
Indeed, UMAP distribution of cells in cluster 3 suggests that this cluster might be composed of several sub-clusters. However, additional formal sub-clustering analysis showed that cluster 3 can not be further divided into additional clusters based on the difference of 30% in the percentage of cells where the gene is detected in the cluster versus the rest of the cells. 4) How are top 8 genes selected per cluster? Is it unbiased or selected list? Showing a heat map of DEG will allow better visualization.
We apologize for not showing the heat map originally. Fig. 1h in the first submission showed top (most abundant) genes in each cluster and the selection of genes is unbiased. This list was taken from the accompanied table (Supplementary Table 2 We have increased the size of the graphs in Fig. 2a and c, and made individual data points darker to improve visualization. 6) There is no data shown to claim that Lgals1 and Top2a allow discriminating cluster 9 from other clusters.
We apologize for this oversight. Microglia in cluster 9 cannot be defined by a single marker gene.
Moreover, a combination of two marker genes can only define a portion of cells in cluster 9. We have included the best 8 combinations of two cluster 9 markers that can be used to detect cells in this cluster (Extended Data Fig. 7). Moreover, we performed in situ hybridization analysis for Lgals1 and Top2a to localize cluster 9 microglia in the spinal cord (Extended Data Fig. 7h), showing that cluster 9 microglia are present in males but not in females post-SNI (Extended Data Fig. 7i). 7) Cluster 5 expresses several known DAM genes. However, this seems to be for naïve and SNI condition. Does this cluster represent microglia which are activated due to dissociation and sorting?
To address this comment and assess if DEGs in cluster 5 are enriched for extraction-related genes, we assessed whether known extraction-associated genes (Rebuttal letter, Fig. 3a) are enriched in cluster 5 or other clusters. Changes in the expression of extraction-associated genes in each cluster and in each condition (Rebuttal letter, Fig. 3b and c) showed that there is no enrichment for these genes in cluster 5, as compared to other clusters, in any condition in males and females. 8) Data for Cluster 7 and 8 are not convincing unless it is shown that these are not doublets.
As we showed in our response to point 1, removal of doublets did not affect markers for any cluster, indicating that doublets, which are formed most likely post-sorting, did not affect the definition of clusters. Moreover, microglia in clusters 7 and 8 uniquely express cell cycle-associated genes, Mki67 and Top2a (Extended Data Fig. 6a) and show enrichment for cell division-related process in GO analysis (Extended Data Fig. 6b), altogether supporting the notion that these are proliferating microglia.
Additional support for this conclusion comes from our observation that microglia in clusters 7 and 8 are present at very low level at the baseline, appear transiently at day 3 post-SNI, and are absent at later time points (day 14 and 5 months, Fig. 2a, c). This temporal profile corresponds to a) previous studies which showed that microglia proliferation peaks at day 3 post-injury and subsides thereafter 18 , and b) our own assessment of microglia proliferation dynamics at all time points (Fig. 2f-i). To support our immunohistochemistry results, we also performed in situ hybridisation for cluster 7/8 microglia using probes against Mki67 (cluster 7) and Mcm6 (cluster 8) (Extended Data Fig. 7d, e). 9) Cluster 9 is claimed to be male specific which appears at day 3. But authors do not include any data that complement RNA-seq data. Claims about the cluster 10 being vessel associated microglia are also not verified. Could authors cite studies which have shown this before and/or provide confirmation data as this cluster expresses other endothelial markers (Doublets?).
As requested by the reviewer, we performed many in situ hybridization experiments (Extended Data Fig.   7). Microglia were labelled using the expression of TdTomato under the microglia-specific promoter, TMEM119. We performed in situ hybridization on spinal cord sections from tdTomato;TMEM119 CreERT2 mice for all clusters that can be defined by one unique marker gene (cluster 4, 5, 6, 7, 8, 10, and 11) and confirmed their colocalization within microglia in the dorsal horn spinal cord.
For cluster 10, we preformed co-labeling of spinal cord sections from tdTomato;TMEM119 CreERT2 mice with Cldn5 (cluster 10 unique marker) and Pecam1, an endothelial marker, to show close association of cluster 10 microglia with blood vessels (Extended Data Fig. 7j).
However, as correctly stated by the reviewer, our scRNA-seq analysis showed that this cluster is enriched for endothelial cell-specific markers. We therefore included a statement in the discussion section that we cannot rule out a cross-contamination of cluster 10 microglia with endothelial cells, as follows: " We propose that cluster 10 microglia (0.19% of total microglia count) represent vessel-associated microglia.
However, cells in this cluster express several endothelial gene (e.g., Pecam1). Thus, we cannot rule out the contamination of this subpopulation by endothelial cells during processing or incomplete dissociation from tightly associated endothelial cells. Alternatively, the presence of endothelial genes could be due to phagocytosis of endothelial cell fragments by microglia." For cluster 9, we used the best combination of two marker genes to detect cluster 9 microglia (Lgals1 + Top2a -) and showed that these cells are found in males but not females at day 3 post-SNI (Extended Data Fig. 7i), consistent with our scRNA-seq analyses. We hope that these comprehensive analyses and more careful wording address the concerns of the reviewer. 10) Have authors verified increased microglia numbers at later times apart from day 3? Since authors do not show increased microglia upon proliferation, either numbers are increased at later stage or there are other mechanisms involved in balancing microglia numbers. Authors have not given any explanations.
As requested by the reviewer, we have performed quantification of all microglia (using Iba1) and proliferating microglia (using colocalization of Ki67 with Iba1) for all time points (naïve, day 3, day 14 and 5 months) and both sexes (Fig. 2h, i). Consistent with the previous report 18 , microglia proliferate shortly after the injury, peaking around day 3 post-SNI, and their proliferating rate is decreased at later time points. We also now provide several potential explanations for increased proliferation in males as compared to females without changes in total cell number, including a) different temporal dynamics of proliferation in males and females, b) increased microglia apoptosis in males, and c) enhanced recruitment of microglia from other areas of the spinal cord in females. Extensive studies, which are well beyond the scope of this paper, would be required to test these options and identify the mechanism underlying this interesting phenomenon. We have initiated these studies, which might take another year or two (especially the differential migration of microglia in males and females) and hope to publish them in a separate follow-up manuscript.  Table   3). If the reviewer and editor feel these data would be better presented as several large heat maps/dot plots supplementary figures, we would be happy to do so.
As described in our paper, we found differential gene expression responses and altered cellular functions (based on GO analyses) in different clusters. Based on this information, we revealed that clusters 7 and 8 are proliferating cells, and this led us to discover that the proliferation is higher in males as compared to females. We confirmed this finding using immunohistochemistry against Ki67 in tdTomato;TMEM119 CreERT2 mice. A cluster-based analysis also allowed us to identify robust and proinflammatory responses, strongly correlated to the previously described injury responsive microglia (IRM) signature, in cluster 9 microglia in males but not in females. Finally, we also identified a small population of perivascular macrophages, and their responses to nerve injury, which have been shown to play central roles in regulating microglia responses after nerve injury 17 . In summary, we believe that our work provides the long-awaited systematic characterization of microglial transcriptional states at the single cell level in response to peripheral nerve injury in mice at acute, sub-chronic and chronic phases, identifies the association between human APOE isoforms and distinct clusters of microglia with specific pain states in humans, reveals the enhanced microglial proliferation in male mice, and identifies a malespecific subpopulation of microglia. We think that our findings are important for the field of pain and significantly advance the field toward deciphering the important and complex roles of microglia in neuropathic pain. Figure 4: A clear rational for choosing ApoE as a key gene is missing. It seems ApoE is not regulated in gender-specific manner. Also, this reviewer is confused about ApoE being the top DEG for cluster 5 ( Figure 1) as in figure 4 it is shown that ApoE is the top DEG for several clusters. Could authors clarify this?

12)
We apologize for being unclear regarding the rationale to focus on Apoe. Apoe is the most abundant gene in cluster 5, a shown in Fig. 1h. In addition, analysis of the DEGs after peripheral nerve injury (comparison between SNI and corresponding sham control) revealed that Apoe is the most upregulated DEG after SNI at two time points and in both sexes (Fig. 4). These results, together with known polymorphisms in the APOE gene and its link to numerous diseases in humans, including Alzheimer's disease, prompted us to study the association between APOE polymorphisms and pain in humans. This led to an exciting discovery that carriers of APOE-ε4 are protected, whereas carriers of APOE-ε2 have an increased risk to develop specific pain conditions. We have now also included a separate analysis for each sex (Fig. 5) which shows that the increased risk for chronic pain in carriers of APOE-ε2 is detected only in men and not in women, indicating sex differences in humans. We are very excited about these findings and initiated a large study to investigate the role and the underlying mechanisms of each isoform (APOE-ε2 and APOE-ε4) in pain. In the revised version of the manuscript, we better explain the rationale to focus on ApoE and the differences between ApoE being the most abundant gene in cluster 5 and the most upregulated DEG in several clusters after SNI.

13) Is there a difference between males and females in expression of ApoE? and if it is not the case then
authors could simply state that ApoE is upregulated independent of gender instead of describing same thing two times.
We thank the reviewer for this suggestion. ApoE is upregulated in both male and female mice after SNI at chronic time points. We now modified the text to make it clearer.
14) ApoE immunostaining is not convincing as only one cell is shown per condition/time point. Authors should show overviews with better quality photomicrographs and provide quantifications of APoE immunoreactivity.
As requested, we have performed numerous additional experiments and quantified ApoE immunoreactivity at all time points in males and females (Fig. 4f, g). We also used AiryScan microscopy to obtain high resolution images of ApoE and demonstrate its intracellular distribution pattern in microglia in chronic phases of neuropathic pain (Fig. 4e). We also present low-and high-magnification images in which ApoE is co-stained with markers of microglia, astrocytes, and neurons to show relative expression of ApoE in different cell types (Fig. 4d).
15) The human data with regards to co-relation between apoE alleles and chronic pain as well as human spinal cord microglia single cell analysis are interesting but clear connections between mouse and human data are missing and most of all mechanisms of how ApoE could play a role chronic pain are not addressed at all. Indeed, our study identified an intriguing link between polymorphisms in the APOE gene and chronic pain in humans. Additionally, in the revised version of the manuscript, we performed a sex-based analysis and found that the association between APOE isoforms and pain in the human population is substantially stronger in men as compared to women (Fig. 5). We are very interested to understand the role of ApoE in pain and investigate the underlying mechanisms. We invested significant efforts to answer this important question within the timeframe of the review process. First, we generated AAV (AAV/TM6-EF1-mCherry-U6-LoxP-GFP-stop-LoxP-mApoe-shRNA, Vector Biolabs) with the goal to downregulate ApoE selectively in microglia (by expressing shRNA against ApoE in a Cre-dependent manner in microglia in TMEM119 CreERT2 mice). Unfortunately, we could not show an abundant expression of Apoe shRNA, as assessed by the presence GFP, in microglia. For this AAV, we used an AAV serotype (AAV/TM6) that was claimed to have stronger tropism to microglia as compared to other serotypes 19 .
Unfortunately, this approach was not efficient enough to allow strong expression of AAV in microglia and ablation of ApoE. In addition, we obtained Apoe general KO mice from The Jackson Laboratory.
Analysis of behavioural phenotypes in two mouse models of nerve injury, SNI and CCI, showed no change in mechanical sensitivity at different time points. The lack of strong phenotypes in ApoE general KO mice might be caused by several factors including developmental compensation and redundancy with other lipoproteins. Importantly, this finding is consistent with studies in the Alzheimer's disease (AD) field, where ApoE is also upregulated. These studies showed that whereas humanized APOE-ε4 mice have increased neuroanatomical (e.g., decreased hippocampus volume and dentate gyrus thickness, increased in astrocyte coverage in piriform cortex) deficits in a mouse model of AD, Apoe general KO mice do not exhibit this phenotype 20 . [REDACTED] We now include a statement that future studies on the functional role of ApoE in pain are required and c learly state that this is a limitation of the current work. We thank the reviewer for suggesting these important articles and as requested, added them to the revised version of the manuscript.

Reviewer #2 (Remarks to the Author):
In this paper authors undertake the first single cell RNAseq analysis of spinal microglia in mouse in the naïve state and following a nerve injury as well as comparison to human microglia. They also integrate with human genetic findings in UK Biobank. There is a focus on sex specific effects which is a really critical issue in the pain field given the sex specific contribution of microglia to neuropathic pain previously shown by these authors (and others). They show microglia can be clustered into cell types and importantly there are male specific clusters which arise after injury (a male specific cluster of microglia at day 3) and also there are sex specific changes in differentially expressed genes and for instance at day 3 very different responses in the microglial clusters. This paper demonstrates novel and important findings relevant to microglia biology and sex differences in pain, these datasets will be an extremely valuable resource for glial and pain biologists (and these are being made available in GEO). Generally experiments are performed to a high standard. I particularly commend the inclusion of a very long time point (5 months) and integration with human scRNAseq. The supplementary data includes helpful methodological details (eg. FACS gating) and the Bioinformatics pipeline for scRNAseq looked appropriate to me. There are some issues regarding integration between human genetics and microglia.
Major issues: 1) Although I like the principle of using human genetic data as presented, I think it needs more thought: a) The phenotypes used in UK-Biobank are acute and chronic pain phenotypes based on pain location with no information about aetiology of that pain. In as far as it goes I think this is reasonable, only a minority will be neuropathic. The 'meta-analysis' of conditions with a neuropathic component is misleading and should be removed. Neuropathic pain is defined by IASP as pain arising due to a lesion or disease of the somatosensory nervous system and certainly can't be inferred from location. Even if you were to accept the premise that you can infer 'neuropathic component' from pain location (which I don't) then decisions are arbitrary. All of these conditions said to NOT include a neuropathic component could have a neuropathic component: Facial pain-can be associated with trigeminal neuralgia (note these are self reported phenotypes for the most part participants won't be able to distinguish causes of facial pain), headache uncommon but can be due to C2, 3 lesion, abdominal pain due to thoracic radiculopathy or proximal diabetic neuropathy, widespread pain secondary to small fiber neuropathy. In the conditions labelled as having a neuropathic component most of the causation will be non-neuropathic. The vast majority of shoulder, back, knee and hip pain will be osteoarthritis. To summarise the results regarding chronic pain and ApoE are interesting and should be retained but the neuropathic pain meta-analysis is misleading. There are other analyses that could be more useful in connecting the human and mouse data in terms of microglia (see below).
We agree with the reviewer and in the revised version of the manuscript, we removed the neuropathic meta-analysis, leaving the individual pain sites analysis (Fig. 5). b) ApoE is not only expressed in microglia it is highly expressed by astrocytes and neurons. A more convincing argument regarding the role of microglia would be to identify microglia specific genes from the authors own (and other) scRNAseq datasets and integrate this data with GWAS signals arising from the chronic pain disorders to determine how variants in genes highly expressed in microglia are contributing to heritability. This type of workflow is now freely available in FUMA and would be a more convincing argument implicating microglia (as a cell type).
We thank the reviewer for this suggestion. Our study is focused on different microglial subtypes and their potential involvement in pain. To investigate whether genes in distinct mouse and human microglia clusters are enriched for pain-relevant genes identified in human GWAS, we substantially expanded our analyses and correlated highly expressed genes in each microglia cluster with pain GWAS UK biobank datasets. We revealed an association between transcriptional signature in several mouse and human microglia subtypes with distinct pain conditions (Extended Data Fig. 12). The strongest correlation included associations between the transcriptome of mouse microglia clusters 4 and 9 with GWAS genes associated with acute back pain (Extended Data Fig. 12b), and transcriptome of mouse cluster 2 microglia with genes linked to chronic stomach/abdominal pain. We also identified the top individual genes in several clusters contributing to the association (Extended Data Fig. 12n). c) Rightly the authors repeatedly emphasise sex specific effects. Sex is a co-variate in the genetic analysis. Were there any sex specific effects in the gene variant associations?
We thank the reviewer for this suggestion, which prompted us to perform APOE genetic analyses separately for males and females, and indeed we discovered sex differences (Fig. 5). We found that the association between APOE-ε4 and chronic pain (headache and back pain) is significant in both males and females. Surprisingly, whereas males show association between APOE-ε2 and facial, abdominal, hip, and knee pain, no such association was detected in females. Moreover, males but not females show an association between APOE-ε2 and acute headache. These data demonstrate sex-specific correlations between polymorphisms in the APOE gene and pain. We show these findings in Fig. 5 and discuss them in the revised version of the manuscript.
2) Regarding APoE expression. Figure 4-includes immunostaining of ApoE. Ideally immunostaining should be performed in female as well as male mice (I appreciate at transcript level expression is increased in female but that does not mean it would be at protein level). We need lower power views and co-staining with other cell type markers-what is the pattern of ApoE expression in microglia relative to astrocytes, neurons etc. in the spinal cord. Finally what cellular compartment is it expressed in-? nuclear ?cytoplasmic (I didn't see convincing membrane/vesicular expression)?
To address this comment, we performed ApoE immunostaining in different conditions (SNI day 3, day 14 and 8 months in males and females) and quantified the expression levels of ApoE (Fig. 4f, g). The new data are consistent with our scRNA-seq results showing the upregulation of ApoE protein levels at day 14 and 8 months in both males and females.
We also performed co-immunostaining of ApoE with markers of microglia (Iba1), astrocytes (GFAP), and neurons (NeuN), and presented low magnification images to demonstrate relative expression of ApoE in different cell types (Fig. 4d).
Finally, using high-resolution AiryScan imaging, we show that ApoE is present in the cytoplasm of microglia as individual puncta (Fig. 4e), likely representing lipid particles 24 .
Minor issues 1. Would it not have been possible to compare this scRNAseq dataset on spinal microglia in the naïve state with published datasets on brain microglia at least in broad terms to see if there are any key differences in sub-types? To my eye this is in itself an important question (I appreciate that there are already some comparisons with B Stevens published work following injury and batch effects may make it difficult to compare with published data but some discussion on whether the sub-groups are broadly similar would be helpful).
We thank the reviewer for this suggestion. We indeed have compared our data with previously identified transcriptional signatures (Fig. 3j, k). We have expanded the discussion on this topic as suggested by the reviewer.
2. Figure 6-correlations in panel F, are between mouse clusters 1-6 and human clusters 1-8. Human cluster 7 and 8 don't appear to have a 'correlate' in mouse. Although Human cluster 8 appears to relate to microglial proliferation does it bear any resemblance to mouse clusters 7, 8 which albeit small numerically but also seem to relate to proliferation and are not shown in this plot?
In the revised version of the manuscript, we expanded the correlation analysis to mouse clusters 1-8 and human clusters 1-8. This analysis showed that mouse cluster 8 exhibits correlation with several human microglia clusters, including cluster 8 (Fig. 6f).
3. Can the authors confirm that in harvesting spinal cord no attempt was made to separate ipsilateral versus contralateral or dorsal form ventral? Obviously the most relevant changes to explain pain will be ipsilateral dorsal to injury but there may have been methodological reasons why the whole lumbar segment was used in which case please just make this clear in methods.
To avoid dissection of the spinal cord to ipsilateral and contralateral sides, which could increase variability and introduce artifacts, mice underwent bilateral SNI/Sham and the whole lumbar spinal cord was collected for scRNA-seq analysis. The comparison was performed between SNI to control mice that underwent bilateral sham surgery. We have clarified this important experimental detail in the revised version.

Reviewer #3 (Remarks to the Author):
The manuscript by Tansley et al uses single cell RNA-seq to study the transcriptional responses in spinal cord microglia at three different time points after spared nerve injury in both male and female mice. They find several interesting differences between male and female microglial responses, including an increased proportion of proliferative microglia in male mice and a male/injury-specific microglia subset with an interferon/cytokine-induced signature. They also find that during the late time points after injury, ApoE was one of the most highly upregulated genes in both male and female microglia across several clusters.
The authors go on to study the association of ApoE haplotype (e2, e3, e4) with chronic pain in a human cohort, finding an increased risk of long-term back and knee pain in apoe e2 carriers, and a decreased risk of back and knee pain and headaches in apoe e4 carriers. Finally the authors have performed singlecell RNA-seq of human microglia (n=3 cases) demonstrating the presence of several transcriptionally distinct clusters, including one highly expressing ApoE. The study is ambitious and well performed and there are several interesting findings that would be of interest to the field. The association between ApoE haplotype and chronic pain is particularly noteworthy. However, the study is entirely descriptive and does not address whether the sex and/or injury-specific microglia subsets contribute to different stages of pain development and maintenance (a question that is highly relevant to the field and the authors as it is asked in the second sentence of the abstract).
Major points: 1) The presence of a male and injury specific microglia subset (cluster 9) is probably the most interesting finding in the mouse data. Have the authors attempted to stain this subset in tissue using one or a combination of protein or mRNA-markers to address where it is distributed anatomically? Can this subset be found in other studies using single cell seq to study microglial responses to injury/inflammation (for We thank the reviewer for their thoughtful suggestions. We performed in situ hybridization (RNAscope) experiments to localize clusters that can be defined by uniquely expressing markers in the spinal cord (Extended Data Fig. 7). Cluster 9 microglia can not be identified by one unique gene and a combination of inclusion and exclusion of gene markers (Lgals1 + Top2a -) is required, making the analysis challenging.
Nevertheless, the RNAscope experiments confirmed our scRNA-seq results by showing that cluster 9 microglia are present in the dorsal horn at day 3 post-SNI in males but not in females. Importantly, we also show that gene expression pattern in cluster 9 microglia exhibits a very strong similarity to the previously defined transcriptional signature of injury-responsive microglia (IRM) (Fig. 3k). IRM were identified in the subcortical white matter in response to lysolecithin (LPC)-induced demyelination, a common model of multiple sclerosis (MS) 25 . This finding suggests that cluster 9 microglia are not specific to peripheral nerve injury and can be also induced via direct damage to the central nervous system. As requested by the reviewer, we performed trajectory analysis for clusters 5 and 9 (Rebuttal letter, Fig. 4a-c) as well as for all clusters together (Rebuttal letter, Fig. 4d-f) for three post-injury time points. Despite significant efforts, unfortunately we were unable to identify clear patterns and come up with a coherent and clear picture to show transitions between subpopulations in different temporal phases. Since cluster 9 microglia are induced at day 3 post-SNI only in males, but ApoE is upregulated in numerous clusters at day 14 and 5 months in both sexes, we think it is unlikely that cluster 9 microglia transition to ApoE-positive microglia. We can not, however, rule out the possibility that cluster 9 microglia transition to other less reactive state or alternatively, undergo apoptosis.
2) While the authors have certainly recorded important upregulations of ApoE across several microglia clusters at day 14 and 5 months post injury, they have not addressed whether this is a protective or detrimental response as it relates to chronic pain. Have the authors attempted to address the role of ApoE itself in the maintenance of chronic pain? Conditional deletion of ApoE in microglia would be the most relevant and interesting experiment, but this would be asking too much. Have the authors tested other approaches?
We agree with the reviewer that this is a very interesting and important scientific question that might ha ve broad scientific and even clinical implications. As detailed in our response to the reviewer #1 (p oint# 15), we invested significant efforts and resources to study the functional role of ApoE in pain. Unf ortunately, our studies using a viral approach and general Apoe KO mice were inconclusive. [REDACTED] 3) It has been proposed that Trem2 and ApoE collaborate to drive the differentiation from a homeostatic We thank the reviewer for this suggestion. To address this comment using our datasets, we compared changes in the expression of Apoe and Trem2 at different time points post-SNI (day 3, day 14 and 5 months) in each cluster in males and females. We found that changes in Apoe and Trem2 are positively correlated in some clusters and conditions but negatively in others (Rebuttal letter, Fig. 5a for males and 6a for females). We also analysed Trem2 and Apoe expression in different clusters (Rebuttal letter, Fig.   5b for males and 6b for females). This analysis showed that whereas Trem2 is found in all microglia clusters, Apoe has a very distinct expression pattern. Thus, based on these analyses, we were unable to unequivocally determine if the interaction between Trem2 and Apoe drives microglia transition to disease states after nerve injury. 4. Cluster 10 microglia appear to be contaminated by endothelial cells or cell fragments. All top markers for this cluster found in Fig 1h (and   To address the concern of the reviewer experimentally, we performed in situ hybridization for Cldn5 in mice in which microglia are genetically labeled (tDTomato;TMEM CreERT2 ) and demonstrated the presence of Cldn5 + , TMEM119 + microglia, consistent with our scRNA-seq data and the work of Haruwaka et al. 16 . However, we do agree with the reviewer that considering the presence of some other endothelial markers in cluster 10 microglia (that have not been described in vessel-associated microglia), we can not rule out cross-contamination by closely associated endothelial cells. We therefore indicate this limitation in the revised manuscript and mention that a potential contamination of cluster 10 microglia by endothelial cell or endothelial cell fragments (because of phagocytosis or experimental artifacts) can not be ruled out (lines 415-420). We followed the reviewer's suggestion and analysed the number of CD45 high cells in our FACS data. We did not find significant differences in the number of CD45 high cells between males and females in any condition (males versus females: Day 3, p = 0.6429, Day 14, p = 0.9275, 5 months, p = 0.962, unpaired t-test). Regarding the infiltration of peripheral macrophages, we performed our analysis using two different microglia labeling approaches: 1) Iba1, and 2) tdTomato;TMEM119 CreERT2 mice. While Iba1 would also mark infiltrating microphages, tdTomato;TMEM119 CreERT2 is expressed selectively in spinal cord microglia and not in macrophages. The increased proliferation in males as compared to females was detected with both approaches, suggesting that the increased proliferation of microglia occurs in males, but the total number of microglia does not change. There are several potential explanations for this phenomenon. First, different dynamics of microglia proliferation, if faster in females, could lead to this outcome. An additional explanation is enhanced apoptosis of male microglia. Indeed, male microglia react to nerve injury more robustly, potentially leading to their cell death. A third potential explanation is increased recruitment of microglia in females as compared to males. Female microglia show enhanced expression of motility genes. We now include all these potential scenarios in the discussion section of the revised manuscript.
6. It would be important that the authors more carefully compare their identified human microglial clusters to those already published, to better understand whether there may be spinal cord specific We first generated a heatmap (Extended Data Fig. 11c) considering the top 1000 highly expressed genes of our human spinal cord microglia and comparing the expression of these genes across all four datasets.
In this preliminary analysis, we revealed that our spinal cord dataset is most similar to that of the We believe these comparisons provide insights into the similarities and differences between human spinal cord and brain microglia, however, there are several important caveats to consider. First, microglia analyzed across the studies have come from a variety of sources including autopsy and resected surgical tissue. Some of these microglia were sorted by flow cytometry prior to sequencing while others were sequenced as part of a mixed cellular suspension and the microglia were 'isolated' in silico. In addition to the varying numbers of cells analyzed and the depth of sequencing applied to each project, these cells were sequenced using different sequencing platforms (Extended Data Fig. 11e). The most appropriate experimental setup to answer this important question of a unique spinal cord microglia signature would involve the analysis of spinal cord and brain (non-spinal) tissue by the same group on the same sequencing platform. This is something we plan to do in the future.
Minor points: 1. It is difficult to assess how the cell proportions change across sex and injury from Fig 2a, b and c.
Putting time on the x-axis and making combined or individual plots for each cluster is one suggestion.
This could be combined with representative cluster color-coded UMAPs for each group, to see how cell densities change in clusters across the different conditions. If sham-groups are similar, it would be preferable to combine them.
We followed the reviewer's suggestion and made changes to improve readability of the graphs in Fig.   2a-c. First, we made the graphs bigger and individual data points darker. Since our sham conditions are unique for each post-SNI time point and sex, we need to present them all. Additionally, we included in Extended Data Fig. 3 and 4, UMAP plots for each cluster and each condition. Since we have numerous conditions (7 for males and 7 for females for 11 clusters), including them in the main figures would be challenging.
2. The quantification of Ki67+ proliferative microglia in Figure 2f-g lacks a naive control. Ideally, day 14 and 5 month timse points should also be included.
As requested, in the revised version of the manuscript, we quantify Ki67 in microglia in naïve control mice and at all three time points (day 3, day 14 and 5 months) (Fig. 2h).
3. It is unclear why Figure 4a, d and e are not presented separately for males and females.
We have now quantified ApoE levels separately in males and females and significantly expanded the IHC for ApoE, showing its expression at different time points (Fig. 4f, g). 4. Figure 4d. Please provide lower-power images so that apoe expression across several microglia and non-microglial cells can be visualized.

Karolinska Institutet
We have significantly expanded the analysis of ApoE protein distribution. First, we show low magnification images of ApoE co-stained with a marker of microglia (Iba1), astrocytes (GFAP) and neurons (NeuN) (Fig. 4d). We also show ApoE immunostaining and quantification at all time points and both sexes ( Fig. 4f. g). Finally, we present ApoE distribution in microglia using high resolution AiryScan imaging (Fig. 4e).