Regulatory T-cells inhibit microglia-induced pain hypersensitivity in female mice

Peripheral nerve injury-induced neuropathic pain is a chronic and debilitating condition characterized by mechanical hypersensitivity. We previously identified microglial activation via release of colony-stimulating factor 1 (CSF1) from injured sensory neurons as a mechanism contributing to nerve injury-induced pain. Here, we show that intrathecal administration of CSF1, even in the absence of injury, is sufficient to induce pain behavior, but only in male mice. Transcriptional profiling and morphologic analyses after intrathecal CSF1 showed robust immune activation in male but not female microglia. CSF1 also induced marked expansion of lymphocytes within the spinal cord meninges, with preferential expansion of regulatory T-cells (Tregs) in female mice. Consistent with the hypothesis that Tregs actively suppress microglial activation in females, Treg deficient (Foxp3DTR) female mice showed increased CSF1-induced microglial activation and pain hypersensitivity equivalent to males. We conclude that sexual dimorphism in the contribution of microglia to pain results from Treg-mediated suppression of microglial activation and pain hypersensitivity in female mice.


Introduction
Microglia are brain resident macrophages with essential roles in neural circuit function in physiology and disease (Priller and Prinz, 2019;Hammond et al., 2018;Vainchtein and Molofsky, 2020). Microglia respond in sexually dimorphic ways in a variety of contexts, including autism, stroke, neurodegenerative diseases, and interestingly in the microglial contribution to pain processing (Mogil, 2020;Villa et al., 2018;Weinhard et al., 2018;Sorge et al., 2011;Inyang et al., 2019;Rosen et al., 2019;Kodama and Gan, 2019;Guneykaya et al., 2018). For example, although male and female microglia are competent to induce pain (Yi et al., 2021), pharmacologic ablation or chemogenetic inhibition of microglia reverses peripheral nerve injury-induced mechanical hypersensitivity only in male mice (Sorge et al., 2015;Saika et al., 2020). In contrast, inhibition of microglia is sufficient to reverse injury-induced hypersensitivity in B-and T-cell deficient female mice (Sorge et al., 2015). Taken together, these data imply that there are sex-specific differences in how the innate and adaptive immune compartments interact to regulate neuropathic pain.
We previously identified microglial activation via release of the myeloid survival factor, colonystimulating factor 1 (CSF1), from injured sensory neurons as a mechanism contributing to nerve injuryinduced pain (Guan et al., 2016). Here, we show that intrathecal administration of CSF1 is sufficient to induce pain (mechanical hypersensitivity) in male, but not female mice. Transcriptomic profiling of dorsal horn microglia and morphologic analyses demonstrated that this sex-specific effect correlates with robust microglial activation in male but not female mice. Furthermore, intrathecal CSF1 markedly expanded lymphocytes and myeloid cells in the spinal cord meninges, and resulted in a preferential expansion of regulatory T-cells (Tregs), in female mice. Finally, we demonstrate that Treg depletion (FoxP3 DTR ) in female mice promotes CSF1-induced microglial activation and is sufficient to induce CSF1-induced pain hypersensitivity equivalent to males. Our findings reveal novel cross-regulatory interactions between Tregs and spinal cord microglia that modulate a sex-specific pain phenotype.

Results
CSF1 is de novo expressed in injured sensory neurons (Guan et al., 2016), and in the spinal cord, parenchymal microglia are the only cells expressing CSF1 receptor (CSF1R). We first analyzed injury-induced mechanical hypersensitivity in female Avil Cre :Csf1 fl/fl mice (Adv-CSF1) in which CSF1 is specifically deleted from sensory neurons. We found that female Adv-CSF1 mice developed normal mechanical hypersensitivity after peripheral nerve injury (  1A, B), in contrast to male rats and mice, in which hypersensitivity was CSF1dependent (Guan et al., 2016;Okubo et al., 2016). Thus, CSF1 is not required to induce mechanical hypersensitivity in females.
We next assessed whether selective administration of CSF1, via an intrathecal route, is sufficient to induce mechanical hypersensitivity. Three consecutive injections of CSF1 provoked profound mechanical hypersensitivity in male, but not in female mice ( Figure 1A-C), even at very high doses (30 ng; Figure 1-figure supplement 1C). Furthermore, after intrathecal CSF1, male microglia acquired a robust amoeboid morphology, characterized by loss of ramification, but in females, microglia acquired a highly ramified morphology, consistent with a persistent homeostatic phenotype ( Figure 1D-E). Fluorescence-activated cell sorting (FACS) analysis also revealed larger numbers of microglia in males and higher expression of cell surface activation markers, CD11b/CD45 ( Figure 1F-H, Figure 1-figure supplement 1D). Taken together, these data demonstrate a male-specific impact and sufficiency of CSF1 for microglia activation and pain hypersensitivity.
To determine whether there was a differential impact of CSF1 on male versus female microglia, we transcriptionally profiled flow-sorted microglia from the lumbar dorsal horn. Sex differences were modest at baseline (86 genes, p Adj <0.01), and CSF1 induced robust gene expression changes in both male and female microglia (Figure 2A, PC1, 56% of variance). However, CSF1 induced an 8.3-fold increase in differentially expressed genes (both upregulated and downregulated) in male microglia ( Figure 2B, Supplementary file 1; adjusted p-value<0.01: males 1350 genes, females 165 genes). As CSF1 is an essential survival factor for microglia and myeloid cells, these sex-specific microglia responses to CSF1 were surprising. Neither the protein nor transcriptomic CSF1R levels differed between males and females ( Figure 2-figure supplement 1A, B).
We next examined these gene expression changes in the context of published microglial transcriptomic data sets in homeostasis and disease (Friedman et al., 2018;Figure 2C). Both male and female microglia responded to CSF1 with a decrease in homeostatic gene expression and an increase in proliferative genes, which were more prominent in males than females. Most prominent in male microglia was a striking upregulation of pathology-associated microglial activation genes ( Figure 2C; Neurodegeneration module) (Friedman et al., 2018;Keren-Shaul et al., 2017). Gene ontology (GO) enrichment analysis ( Figure 2D) revealed that male microglia induced genes and GO terms that are linked to classical immune activation and recruitment pathways, including many (Itgax,Lpl,Ccl3,Cybb,Clec7a,and Ctsb) associated with the 'disease associated microglia' DAM phenotype identified in single-cell sequencing experiments (Butovsky and Weiner, 2018). Some of these genes, for example, Ctsb, have been linked to chronic pain (Sun et al., 2012). In addition, male microglia downregulated genes facilitating responsiveness to extracellular signals as well as some supportive functions, for example, extracellular matrix regulation ( Figure 2D). Taken together, intrathecal CSF1 not only triggers pain hypersensitivity in male mice, but also induces robust transcriptomic changes associated with inflammatory activation in male but not female microglia.
Our findings suggest that other immune cells contribute to amplify or suppress the microglial response to CSF1. The CNS meninges have a rich population of immune cells that mirrors the composition of tissue resident immune cells in other organs Figure 3B  Meningeal lymphocyte-derived cytokines also impact CNS function in both normal and pathologic settings Pasciuto et al., 2020;Ribeiro et al., 2019). We examined the immune cell composition of spinal cord meninges using 11-parameter flow cytometry of dissociated meninges ( Figure 3-figure supplement 1A-C, Figure 3A-C). As expected, intrathecal CSF1 expanded meningeal macrophages (Figure 3-figure supplement 1B), but we also observed a marked increase in lymphocytes, 6.5-fold in males and 9-fold in females ( Figure 3-figure supplement 1C). Further examination of lymphocyte subsets demonstrated a similar increase of CD4 + FoxP3 T cells, CD8 + T cells, B cells, and ILC2 cells in male and female meninges, but also revealed a significantly greater expansion of regulatory T cells and natural killer (NK) cells in female mice ( Figure 3B-C).
The online version of this article includes the following figure supplement(s) for figure 1:   As NK cells are traditionally considered pro-inflammatory including in the context of pain (Greisen et al., 1999;Das et al., 2018) and microglial activation (Garofalo et al., 2020), whereas Tregs are potent suppressors of inflammation, we next asked whether Tregs in females counter the CSF1induced microglial activation and pain. To acutely deplete Tregs, we administered diphtheria toxin to FoxP3 DTR mice (Sakaguchi et al., 2008;Ali et al., 2017;Da Costa et al., 2019;Kim et al., 2007;Figure 3-figure supplement 1D,E). From these mice, we transcriptionally profiled female microglia after CSF1 intrathecal injection in the control or Treg depleted setting ( Figure 3D-F/Supplementary file 3). We found that female microglia expressed many of the male-specific CSF1 induced genes, including genes involved in immune activation and recruitment (Clec7a, Il1rn, Ccl3, Ccl4, and Ctsb; Figure 3E-F). We also observed alterations of genes that are unique to the Treg-depleted context ( Figure 3-figure supplement 1F). We conclude that Treg depletion partly restores the proinflammatory microglial response to CSF1 in female mice.
Finally, we tested whether Tregs suppress CSF1-induced mechanical hypersensitivity in female mice. We depleted Tregs in FoxP3 DTR mice by administering diphtheria toxin prior to CSF1 injection ( Figure 3G). Compared to wild-type (WT) females, Treg depletion in females led to a 33% increase in mechanical hypersensitivity (Figure 3H-I; summarizes D3 timepoint from Figure 1A). This effect was phenocopied in Rag1 −/− , which lack T-and B-cells from birth but retain innate lymphocytes, such as NK cells ( Figure 3J) and the findings are reminiscent of those reported in Rag1 −/− female after peripheral nerve injury (Sorge et al., 2015). Of note, depleting Tregs in males did not alter their mechanical hypersensitivity (Figure 3-figure supplement 1G). Acute antibody blockade of CD4 + T-cells, which include both suppressive (Tregs) and inflammatory subsets (Th1/Th2), also phenocopied this increase in mechanical hypersensitivity ( Figure 3K; Figure 3-figure supplement 1H-I). Taken together, we demonstrate that this difference reflects a suppressive effect of Tregs on the CSF1-mediated immune activation in female mice, rather than a direct pain-mediating effect of T-cells on dorsal horn pain circuitry.

Discussion
Our identification of a sex-specific interaction between spinal cord microglia and Tregs that mediates male/female differences in a model of neuropathic pain has several important implications. First, we defined the immune activation profile of CSF1 on microglia in vivo and demonstrated robust expansion of lymphocytes within the spinal cord meninges in response to CSF1. These results are consistent with a model in which one function of CSF1-stimulated myeloid cells is to recruit other immune cells that in turn release cytokines and chemokines to impact microglial function. However, the nature of this immune response is strikingly sex-specific. In males, the balance tips toward pro-inflammatory signaling. In females, Tregs suppress inflammatory activation and limit mechanical hypersensitivity development, despite expansion of the myeloid and lymphoid compartments. As intrathecal CSF1 induces mechanical hypersensitivity in Treg-depleted female mice, we concur that female microglia are indeed competent to contribute to pain hypersensitivity (Yi et al., 2021;Sorge et al., 2015). However, our results demonstrate that CSF1-mediated cross-talk between spinal cord microglia and lymphocytes can either amplify or suppress pain phenotypes.
Our findings also introduce spinal cord meninges as a potentially relevant source of immune cells that coordinate microglial responses in the setting of neuropathic pain. Importantly, in contrast with a previous report (Costigan et al., 2009), we rarely detected lymphocytes, including T-cells, in the spinal cord, even after nerve injury (Figure 3-figure supplement 2). However, we found that immune cells markedly expand within the spinal cord meninges, even when absent from the parenchyma. As lymphocytes act primarily via secreted cytokines, we suggest that release of meningeal-derived individual mice. (K) Change in mechanical hypersensitivity at day 3 after CSF1 in female mice injected with a CD4 blocking antibody 1 day prior to CSF1 injections. Dots: individual mice. In (I-K) unpaired two-tailed t-test and (C) one-way ANOVA with Tukey's multiple testing correction. *p<0.05, **p<0.01, ****p<0.0001. DEG, differentially expressed gene; WT, wild-type.
The online version of this article includes the following figure supplement(s) for figure 3:   cytokines impacts microglial function as well as directly impacts nociceptors (Liu et al., 2014). Although our report focuses on the contribution of Tregs, we also detected a female-specific increase in meningeal NK cells in response to CSF1. NK cells are classically associated with pro-inflammatory responses, however, recent studies highlight their more diverse functions. These include instruction of anti-inflammatory astrocytes from meningeal NK cells (Sanmarco et al., 2021), beneficial effects after peripheral nerve injury (Davies et al., 2019), and a negative correlation between NK cells in the cerebrospinal fluid and mechanical pain sensitivity in chronic neuropathic pain patients (Lassen et al., 2021). The function of meningeal NK cells in CSF1-induced pain in mice remains to be determined.
In the setting of injury, inflammatory signaling at multiple access points (e.g., injury site, nerve, and DRG) activates nociceptive circuits (Yu et al., 2020). However, our finding that intrathecal activation of myeloid cells is sufficient to activate meningeal immunity raises the possibility that modulating the meninges is a potential therapeutic avenue of neuropathic pain management, by suppressing meningeal Treg expansion-mediated microglial activation or by the release of intrathecal immune modulators that override peripheral inflammatory cues. Given that human genetic analyses and other studies indicate a contribution of Tregs and their dominant cytokines in neuropathic and inflammatory pain models (Davoli-Ferreira et al., 2020;Fischer et al., 2019;Milligan et al., 2006;Eijkelkamp et al., 2016;Echeverry et al., 2009;Kringel et al., 2018), further investigations of Treg localization and impact on microglia will be relevant to understanding the generation and conceivably the treatment of nerve-injury-induced chronic pain.  (Harris et al., 2012), Avil Cre (Zurborg et al., 2011), and FoxP3 DTR (B6.129(Cg)-Foxp3 tm3(DTR/GFP)Ayr/ J) (Kim et al., 2007).

Injury, injections, and behavioral analysis
Spared Nerve Injury (SNI) was performed by ligation and transection of the sural and superficial peroneal branches of the sciatic nerve, leaving the tibial nerve intact (Shields and Eckert, 2003). CSF1 (Life Technologies; PMC2044) was injected intrathecally at low dose (15 ng) or high dose (30 ng) in a total volume of 5 µl for three times over 3 days (24 hr between injections). Behavioral analysis was done 2 hr after injections; mice were euthanized for analysis about 4 hr after the last injection. All Von Frey behavioral experiments were performed during the light cycle as previously reported (Guan et al., 2016) in a blinded manner. Intraperitoneal injection of anti-CD4 (250 µg) (InVivoPlus; Bio X Cell) and Diphtheria toxin (30 ng/g) (Sigma-Aldrich) were all in a volume of 200 µl per injection. Anti-CD4 was given 1 day prior to the start of CSF1 injections, and on day 2 of the CSF1 injections. Diphtheria toxin was given 2 days (two subsequent injections) before the start of the CSF1 injections, and on day 2 of the CSF1 injections.

Immunohistochemistry and analysis
Avertin-anesthetized mice were transcardially perfused with 1× phosphate-buffered saline (PBS) (~10 ml) followed by 4% (weight/volume) paraformaldehyde (PFA) diluted in PBS (~10 ml). Spinal cord tissue was dissected out and post-fixed in 4 % PFA for 4 hr and then transferred to a 30% sucrose solution overnight. Subsequently, spinal cords were sectioned coronally at 25 µm using a cryostat (Thermo Fisher Scientific). Spinal cord sections were incubated in a blocking solution consisting of 10% normal goat (Thermo Fisher Scientific) and 0.4% Triton (Sigma-Aldrich) diluted in 1× PBS. Primary antibodies included: rabbit anti-mouse Iba1 (WAKO, 1:2000); Alexa 647-coupled mouse anti-CD45 (BioLegend, 1:200); and hamster anti-CD3 (BD BioScience, 1:200). Antibodies were diluted in 10% normal goat with 0.4% Triton in PBS and incubated on a shaker overnight at 4 o C. Secondary antibodies (Thermo Fisher Scientific, 1:1000) were diluted in 0.4% Triton in PBS and spinal cord sections were incubated on a shaker for 2 hr at room temperature. Spinal cord sections were mounted on coverslips with DAPI containing Fluoromount-G (Thermo Fisher Scientific). Slides were imaged on an LSM700 (Zeiss) confocal microscope using 63× objectives and z-stacks with a step size of 1 µm were collected. In Fiji (Schindelin et al., 2012) (ImageJ), maximum intensity images were generated and binary, thresholded images for morphology analysis were created. Subsequently, Scholl analysis (Ferreira et al., 2014) was done in Fiji (ImageJ) on microglia from the binary images with a step size of 2.5 µm.

Fluorescence-activated cell sorting of microglia
To isolate microglia, we followed a previously described method (Galatro et al., 2017). Briefly, lumbar dorsal horn spinal cords were mechanically dissociated in isolation medium (HBSS, 15 mM HEPES, 0.6% glucose, 1 mM EDTA pH 8.0) using a glass tissue homogenizer (VWR). Next, the suspension was filtered through a 70 µm filter and then pelleted at 300 ×g for 10 min at 4 o C. The pellet was resuspended in 22% Percoll (GE Healthcare) and centrifuged at 900×g for 20 min (acceleration set to 4 and deceleration set to 1). The myelin free pelleted cells were then incubated in blocking solution consisting of anti-mouse CD16/32 antibody (eBioscience) for 5 min on ice, followed for 30 min in a mix of PE or PE/Cy7-conjugated anti-mouse CD11b (eBioscience), FITC or PE/Cy7-conjugated anti-mouse CD45 (eBioscience), and APC or APC/Cy7-conjugated anti-mouse Ly-6C (BioLegend) in isolation medium that did not contain phenol red. For flowcytometric analysis of CSF1R expressed by microglia, PE-conjugated anti-mouse CSF1R (BioLegend) was added. The cell suspension was centrifuged at 300 ×g for 10 min at 4 o C and the pellet was incubated with DAPI (Sigma-Aldrich) before sorting. Microglia were sorted on a BD FACS Aria III and gated on forward/side scatter, live cells by DAPI, and CD11b high , CD45 low , and Ly-6C neg . After sorting, cells were spun down at 500 ×g, 4 o C for 10 min and the pellet was lysed with RLT+ (Qiagen).

Flow cytometry of spinal cord meningeal cells
To exclude dead cells from the analysis, single-cell suspensions were stained with a fixable viability dye (Zombie NIR, BioLegend), followed by staining for surface antigens with a combination of the following fluorescence-conjugated mAbs:

RNA sequencing Analysis
Quality of reads was assessed using FastQC (http://www. bioinformatics. babraham. ac. uk/ projects/ fastqc) and all samples passed quality control. Subsequently, reads were aligned to mm10 (GRCm38; retrieved from Ensembl) using STAR (version 2.5.4b) (Dobin et al., 2013) without FilterMultimapNmax one so as to only keep reads that map one time to the reference genome. Uniquely mapped reads were counted using HTSeq (version 0.9.0) (Anders et al., 2015) and the DESeq2 package (version 1.24.0) (Love et al., 2014) in R was used to normalize the raw counts and perform differential gene expression analysis (using the apeglm method [Zhu et al., 2019] for effect size shrinkage). One CSF1treated WT female sample was subsequently removed from the analysis as its counts significantly deviated from the rest. Specifically, its gene expression pattern resembled severe injury, potentially due to damage to the spinal cord during the mouse experimental procedures. Batch correction was done using the Limma package (Ritchie et al., 2015) in R. Volcano plot was generated using the EnhancedVolcano package (version 1.2.0), and the heatmap using ComplexHeatmap (Gu et al., 2016) in R. Metascape was used for GO analysis (Zhou et al., 2019). FPKM values were generated using Cufflinks (version 2.2.1) (Trapnell et al., 2010).

Statistical analysis
For most statistical analyses, we used Graphpad Prism 8. Figure legends identify the specific statistical test used and additional details are provided in Table 1. RNA-sequencing data were analyzed in R as described in Materials and methods section.  Unpaired t-test between males and females for each cell type

Additional files
Supplementary files • Supplementary file 1. Transcriptomic profiling of CSF1-induced genes in microglia in males and females (Excel file).
• Supplementary file 2. FPKM values for each sample from transcriptomic profiling of males and females after saline or CSF1 (Excel file).
• Supplementary file 3. Transcriptomic profiling of CSF1-induced genes in microglia from females with and without Treg depletion (Excel file).

• Transparent reporting form
Data availability RNA sequencing data are available through GEO accession #GSE 184801 All data generated or analysed during this study and required for conclusions to be drawn are included in the manuscript and supporting files. The upload can be identified at the following link: https://www. ncbi. nlm. nih. gov/ geo/ query/ acc. cgi? acc= GSE184801.
The following dataset was generated: Author (