Regulatory T cells promote functional recovery after spinal cord injury by alleviating microglia inflammation via STAT3 inhibition

Abstract Background Immediately after spinal trauma, immune cells, and proinflammatory cytokines infiltrate the spinal cord and disrupt the focal microenvironment, which impedes axon regeneration and functional recovery. Previous studies have reported that regulatory T cells (Tregs) enter the central nervous system and exert immunosuppressive effects on microglia during multiple sclerosis and stroke. However, whether and how Tregs interact with microglia and modulate injured microenvironments after spinal cord injury (SCI) remains unknown. Method Regulatory T cells spatiotemporal characteristics were analyzed in a mouse contusion SCI model. Microglia activation status was evaluated by immunostaining and RNA sequencing. Cytokine production in injured spinal cord was examined using Luminex. The role of STAT3 in Treg–microglia crosstalk was investigated in a transwell system with isolated Tregs and primary microglia. Results Regulatory T cells infiltration of the spinal cord peaked on day 7 after SCI. Treg depletion promoted microglia switch to a proinflammatory phenotype. Inflammation‐related genes, such as ApoD, as well as downstream cytokines IL‐6 and TNF‐α were upregulated in microglia in Treg‐depleted mice. STAT3 inhibition was involved in Treg–microglia crosstalk, and STAT3 chemical blockade improved function recovery in Treg‐depleted mice. Conclusion Our results suggest that Tregs promote functional recovery after SCI by alleviating microglia inflammatory reaction via STAT3.

the injured microenvironment and impedes neural regeneration. 4 Therefore, preserving a stable microenvironment to limit neuroinflammation and facilitate axonal outgrowth is a promising strategy for SCI treatment.
Previous studies have revealed a dramatic increase in microglial populations at the epicenter of lesions during the first 7 days following SCI. 5 SCI subacute phase is dominated by M1 microglia, which contribute to cytotoxic damage and demyelination through the generation of reactive oxygen species and proinflammatory cytokines, such as iNOS, TNFα, IL-1β, and IL-6. 6 However, during the chronic phase, M2 microglia become prominent and promote regeneration through the production of the anti-inflammatory cytokine IL- 10. 7 A recent study has shown that shifting microglia toward an antiinflammatory state improves locomotive function and electrophysiological transmission after SCI. 8 Enhancing microglial dispersion promotes wound compaction and motosensory recovery following SCI. 9 Moreover, transplantation of proteinase inhibitor-treated microglia into the spinal cord lesions significantly improves wound healing and axon regeneration. 10 These findings suggest microglia as a therapeutic target to regulate injured spinal microenvironment and promote functional recovery after SCI. 11 Tregs are a subset of CD4 + T cells expressing the transcription factor FOXP3. Upon injuries or diseases, Tregs enter the central nervous system (CNS) and preserve the immune homeostasis. 12 Emerging evidence suggests that Tregs could directly modulate microglia's activity and protect against pathological neuroinflammation. 13 The expansion of endogenous Tregs in a mouse model of amyotrophic lateral sclerosis (ALS) could reset the homeostatic functions of the microglia, thereby prolonging survival time. 14,15 Further, in mice with Tregspecific gene knockouts or Treg dysfunctions, the brain microglia appeared more activated, with higher expression of GM-CSF and TNFα during experimental autoimmune encephalomyelitis (EAE). [16][17][18] Tregs in ischemic brain polarize microglia toward reparative phenotypes that promote white matter repair during both acute and chronic stages. [19][20][21] The immunoregulatory effects of Tregs on microglia were further supported by in vitro studies where Tregs, through secretion of anti-inflammatory cytokines (IL-10, IL-35, TGFβ), were shown to directly suppress the production of proinflammatory cytokines and chemokines by cultured microglia. 22,23 Previous studies have reported that treatment with IL-10, VX-765, or CCL28 increases Treg numbers and reinforces functional recovery after SCI. [24][25][26] However, whether and how Tregs interact with microglia and modulate injured microenvironment after SCI remains unknown. Recent studies have revealed that Tregs suppress the IL-6-STAT3 signaling pathway in a mouse model of middle cerebral artery occlusion. 27 STAT3 activation promotes disease-associated microglia activation during late-onset Alzheimer's disease, and is increased after SCI. [28][29][30] Yet, only few studies tracked the association between the STAT3 pathway and Tregs' beneficial effect on microglia after SCI.
In this study, we analyzed Treg population dynamics during mouse contusion SCI. Tregs were selectively depleted to formally test whether and how Tregs regulate microglia activation and limit neuroinflammation following SCI. Our results showed that Treg depletion resulted in microglial activation and a phenotypic switch via STAT3 pathway, which eventually enhanced the proinflammatory microenvironment caused by SCI. STAT3 inhibition suppressed microglia-mediated neuroinflammation and promoted neural function recovery after SCI in Treg-depleted mice. Our findings provide insights into the pathway underlying neuronal damages after SCI and reveal STAT3 as a possible target to explain Tregs' beneficial effect on recovery.

| Animals
Wild-type C57BL/6J mice, Foxp3-EGFP transgenic mice, and Foxp3-DTR transgenic mice were purchased from Shanghai Model Organisms Center. Foxp3-EGFP mice co-express EGFP and the regulatory T cellspecific transcription factor Foxp3, which could accurately identify the FOXP3 + T cell population. 31 Depletion of Tregs was performed in Foxp3-DTR mice by intraperitoneal injection of Diphtheria toxin (DT, 40 μg/kg body weight) 3 days prior to SCI, and was repeated every 3 days to maintain Treg cell depletion until sacrifice. 32 Female mice (12-week-old, weighing 22-25 g) were used for the experiments in vivo, given that male rodents are not often used when modeling SCI due to more-severe postoperative complications, such as urinary and bowel incontinence and urinary tract infections 33 ; these male-specific complications increase mortality and adverse health issues. 34 Analysis of NIH-funded, rodent, primary research publications demonstrated that females were the sole sex used in the majority of SCI experiments. 35

| SCI model
Prior to surgery, the mice were anesthetized with isoflurane. A T9-11 laminectomy was performed, and the spinal cord was exposed at T10. A 5-g weight was dropped from a height of 11 mm onto the exposed dorsal surface of the spinal cord using a modified NYC impactor (J$K Seiko Electronic). Sham-treated mice underwent a laminectomy at T9-11 without SCI. Basso mouse scale (BMS) was used to evaluate the recovery of motor function in mice at 1, 3, 7, 14, 21, and 28 days after injury. Detailed information about BMS testing is further described in Appendix S1.

| Flow cytometry
Briefly, the single-cell suspensions were first incubated with Fc blocker (anti-CD16/32 antibody) for 15 min. For surface antibodies, such as CD4 and CD25, isolated cells were incubated for 30 min at F I G U R E 1 Tregs infiltrated the injury core area and mainly clustered around microglia after spinal cord injury (SCI). (A) Representative images showing the injury core (a) and imaging area (b) for FOXP3 (green) imaging. The location of the injury core after SCI is indicated by white circles (a), while the imaging sites (the area surrounding the injury core) are marked by white frames (b). Scale bar: 500 μm. (B) Representative immunofluorescence images of FOXP3 (green, Tregs) staining of Foxp3-EGFP mouse spinal cord section surrounding the lesion core at different time points following SCI. Scale bars: 20 μm (upper columns); 5 μm (bottom column). (C) Representative images showing the spatial relationship between FOXP3 + (Tregs) and NeuN + (neurons) cells (top row), FOXP3 + (Tregs) and IBA1 + (microglia) cells (middle row), and FOXP3 + (Tregs) and GFAP + (astrocytes) cells (bottom row). The three-dimensional reconstruction (right column) was performed by Imaris 3D imaging. Scale bars: 20 μm (left column); 5 μm (2 right columns). (D) Time course quantification of the proportion of FOXP3 + cells surrounding the injury core at day 3, 7, 14, and 28 post-SCI. (E) Calculation from images on (B) of the colocalization coefficients between FOXP3 + staining, and, respectively, NeuN + , IBA1 + , and GFAP + staining by Imaris 3D imaging. Ten to 15 Tregs per mouse were analyzed. Data are represented as the mean ± SEM (n = 5 for each group; ****p < 0.001).
Three-dimensional reconstruction and co-localization analysis were conducted using Imaris software. Morphological changes in microglia were analyzed with the Sholl analysis plugin for ImageJ. Detailed information about three-dimensional reconstruction and measurement of the co-localization coefficient using Imaris software, and the Sholl analysis is thoroughly demonstrated in Appendix S1.

| Sequencing of microglia RNA
Microglia were separated from spinal cord as described before. 27 Briefly, the single-cell suspensions were incubated with anti-CD45-PerCP and anti-CD11b-APC antibodies for 30 min at 4°C in the dark  kits were used to quantify IL-6 and TNFα levels (Neobioscience) in cell culture supernatants.

| RNA extraction and qPCR
Total RNA from tissues or primary microglia was extracted with TRIzol (Invitrogen). cDNA was synthetized using a ReverTra Ace qPCR RT Kit (Toyobo). Quantitative real-time PCR (qRT-PCR) was performed using BioRad CFX Connect and SYBR Green PCR Master Mix (Toyobo). Expression data were normalized to the internal control GAPDH. The primer sequences are listed in Table S1.

| Western blotting
Proteins extracted from spinal cord tissues and primary microglia were separated by SDS-PAGE and transferred onto nitrocellulose filters membranes. The membranes were blocked for 1 h at room temperature using 5% nonfat milk, and incubated overnight at 4°C with primary antibodies (1:1000). The following primary antibodies were used: rabbit anti-STAT3 and anti-phospho-STAT3 (Cell Signaling), mouse anti-IL-6 and rabbit anti-TNFα (Santa Cruz), rabbit antiβ-ACTIN and rabbit anti-GAPDH (Servicebio). Next, the membranes were incubated with secondary antibodies for 1 h at room temperature (1:1000; Servicebio). Protein expression levels were normalized to β-ACTIN or GAPDH internal controls.

| Primary microglia culture
Briefly, brains from postnatal 1-to 3-day-old mice were diced after removing the meninges. The dissociated cells were plated onto PDL- streptomycin). The cells were grown to confluence for 12 days, and microglia were collected by shaking the mixed glia-containing flasks for 1 h at 180 rpm at 37°C.

| Oxygen-glucose deprivation/reoxygenation
Oxygen-glucose deprivation/reoxygenation (OGD/R) was used to mimic ischemic injury in vitro, as described previously. 6 Briefly, the culture medium of primary microglia cultures was replaced with glucose-free DMEM (Gibco). The cells were placed in a hypoxic incubator (Thermo Scientific) at 94% N 2 , 5% CO 2 , and 1% O 2 at 37°C for 3 h, and then returned to standard culture medium and incubation conditions for another 3 h (37°C, 95% air, and 5% CO 2 ).

| Luxol fast blue staining
Briefly, spinal cord tissue sections were immersed in 95% ethanol for 5 min at room temperature, followed by overnight incubation in 0.1% LFB (Servicebio) at 60°C. Observation and image acquisition were performed using a light microscope (BX51; Olympus). Kolmogorov-Smirnov test was used to assess data distribution.

| Statistical analysis
For normally distributed data, Student's t test was used to assess differences between two groups, and one-way ANOVA was used to analyze differences between three or four datasets. Data that did not exhibit a normal distribution were analyzed using Mann-Whitney U test between two groups or Kruskal-Wallis tests between three datasets. p values <0.05 were considered statistically significant.

| Tregs infiltrated the injury core area and mainly clustered around microglia after SCI
To assess the spatiotemporal patterns of Treg after SCI, we first analyzed the CD4 + CD25 + FOXP3 + Treg population in spleen and blood by flow cytometry in Foxp3-EGFP (GFP) mice. The gating strategy is shown in Figure S1A. We observed an overall increase in splenic and blood Treg number from day 3 until at least 28 days after SCI  Confocal analysis by Imaris 3D imaging was used to probe the spatial relationship between FOXP3 + (Tregs) and NeuN + (neurons), IBA1 + (microglia), and GFAP + (astrocyte) cells at 7 days after SCI ( Figure 1C). Three-dimensional reconstruction and measurement of the co-localization coefficient between NeuN + , IBA1 + , or GFAP + staining with FOXP3 + staining revealed that infiltrated Tregs mainly clustered around microglia and astrocytes (Treg/neuron coefficient: 0.0699; Treg/astrocyte coefficient: 0.2117; Treg/microglia coefficient: 0.4483; Figure 1E). These findings suggest that following SCI, Tregs infiltrate the spinal lesion and interact with CNS resident cells, including microglia and astrocytes. We then assessed potential morphological changes in IBAI + microglia caused by Treg depletion by Sholl analysis. The imaging area for Sholl analysis is shown in Figure 2A. In the presence of Tregs and after SCI, microglia adopted an amoeboid phenotype with enlarged soma and shorter processes compared with uninjured microglia (p < 0.05; Figure 2B,D-F). However, the main difference was observed upon Treg depletion, consisting of increased number of processes (p < 0.05; Figure 2D) and larger soma size (p < 0.001; Figure 2F) compared with WT mice (Figure 2B Figure 2C). In contrast, the absence of Tregs did not seem to influence process length ( Figure 2E).

| Treg depletion promoted microglial morphological alterations and phenotype switch after SCI
The amoeboid phenotype reflects a shift in microglial function with increased proliferation ability and neurotoxicity. 37 Therefore, we assessed microglial proliferation by co-immunostaining of IBA1 and KI67 ( Figure 2G). We observed that on day seven after SCI, microglia in Treg-depleted mice exhibited a higher proliferative ability (p < 0.001; Figure 2I). Microglial activity is typically categorized as neurotoxic (M1) or neuroprotective (M2). 38 To compare the phenotype of microglia in the presence or absence of Tregs, we performed co-immunostaining of spinal sections for IBA1 with the M1 marker CD16, or for IBA1 with the M2 marker CD206 ( Figure 2H). This analysis showed increased numbers of CD16 + IBA1 + M1 cells in Treg-depleted mice compared with WT mice (p < 0.001; Figure 2J), accompanied by a reciprocal decrease in CD206 + IBA1 + M2 cell number (p < 0.01; Figure 2K). These results showed that by day 7 post-SCI, Treg depletion had favored the acquisition of an M1 neurotoxic phenotype over a M2 protective phenotype by microglia.

| Treg depletion resulted in enhanced microglia-mediated proinflammatory microenvironment after SCI
Increasing evidence supports that a dichotomized classification of microglia may not reflect the phenotypic diversity of these cells. 38 To better understand how Tregs influence microglia's function, microglia were sorted by flow cytometry from injured spinal cords of WT and Treg-depleted DTR mice 7 days after SCI. The purity of the sorted cells was confirmed by immunofluorescence microscopy ( Figure 3A). The sorted cells+ were analyzed by RNA sequencing to To further investigate whether Treg depletion favors a neuroinflammatory microglial responses and enhances the pro-inflammatory microenvironment provoked by SCI, we measured the levels of proand anti-inflammatory cytokines linked to aforementioned differentially expressed genes using Luminex in spinal cord taken on days 7 and 28 after SCI. GM-CSF, TNFα, IL-1β, IL-6, CCL2, and MMP-9 levels were markedly increased in DTR mice at day 7 compared with WT mice. In contrast, the levels of anti-inflammatory IL-2 and IL-33 were significantly decreased (p < 0.001; Figure 3D-F). On day 28 after SCI, there was no difference in the level of these cytokines between the two groups ( Figure 3D-F). Thus, Treg depletion resulted in increased expression of inflammatory factor at day 7 post-SCI, thereby promoting a proinflammatory microenvironment.

| Treg depletion increased IL-6 and TNFα expression in microglia after SCI
ApoD upregulation in microglia after Treg depletion was of particular interest, for it has been associated with TNFα and IL-6 production and aggravated neuroinflammation. 39 Therefore, we sought to assess whether Treg depletion enhanced TNFα and IL-6 production by microglia. Quantification by western blot indicated higher levels of IL-6 (p < 0.001; Figure 4B) and TNFα (p < 0.05; Figure 4C) in DTR mice at day 7 post-SCI compared with WT mice.
However, day 14 after SCI, no significant difference was observed ( Figure 4B,C).
Next, we performed double-labeling of spinal cord sections for TNFα or IL-6 (green) and IBA1 (red) to confirm that these cytokines were derived from microglia. We found that the proportions of both IL-6 + IBA1 + (p < 0.001; Figure 4D,F) and TNFα + IBA1 + (p < 0.05; Figure 4E,G) cells were increased in Treg-depleted mice 7 days following SCI, suggesting that Tregs may directly repress IL-6 and TNFα expression in microglia after SCI.

| Tregs attenuated microglia's inflammatory response by reducing the STAT3 pathway activation in vivo and in vitro
Previous research has reported that IL-6 and TNFα activate the transcription factor STAT3 and its downstream targets. 40 Western blotting showed a significant increase in phosphorylated STAT3 (p-STAT3) level in Treg-depleted mice (p < 0.01; Figure 5A-C).
Consistently, the expression of the STAT3 target gene suppressor of cytokine signaling 3 (Socs3) was also increased in Treg-depleted mice (p < 0.001; Figure 5D). These results indicated that in the absence of Tregs, the STAT3 pathway was highly activated following SCI.
To further investigate whether Tregs regulate microglia's activity via repressing the STAT3 pathway, Tregs were cocultured with primary microglia for 24 h, and then submitted to OGD/R. To assess the effect of blocking STAT3 pathway, we made use of SH-4-54, a small-molecular STAT3 inhibitor with high blood-brain barrier permeability. 41 The microglia were treated with SH-4-54 for 3 days prior to OGD/R ( Figure 5E). The relative expression of p-STAT3 protein and Socs3 mRNA was reduced significantly in microglia cocultured with Tregs (p < 0.001; Figure 5F-I). Similarly, and as expected, SH-4-54 treatment effectively blocked OGD/R-induced STAT3 phosphorylation (p < 0.001; Figure 5F-I). Tregs or STAT3 inhibition reduced the expression of ApoD (p < 0.001; Figure 5I), TNFα (p < 0.001; Figure 5J), and IL-6 (p < 0.001; Figure 5K) upon OGD/R. Combining Treg and SH-4-54 treatment did not further downregulate the expression of these molecules in microglia, compared with Tregs or SH-4-54 used alone (p < 0.001; Figure 5F-K). Collectively, these data collected in vivo and in vitro support that the STAT3 pathway played a pivotal role in Tregs-microglia interactions.

As shown in
Following SCI, a complex cascade of oxidative stress and inflammatory responses is initiated, leading to further demyelination, which is the major pathogenesis factor after SCI. 42 To investigate whether Treg depletion promotes secondary tissue damage after SCI, we conducted LFB staining and observed that Treg depletion resulted in increased demyelination at both 7 and 28 days after SCI, while SH-4-54 treatment partly reversed this detrimental demyelination (p < 0.01; Figure 6F-H). Motor function recovery was evaluated by the BMS. 43 In our study, all of the mice lost the motility of hind limbs and were rated 0 in BMS testing immediately after SCI, while the animals subject to sham surgery were rated 9 in BMS testing. One day after contusive injury, no significant locomotor recovery was observed. We noted that 72.2% of the injured mice did not have any motor function recovery (BMS score 0), whereas 27.8% of the animals subject to SCI exhibited slight ankle movement (BMS score 1) on day one after injury. Aggravated neural function deficits were then obvious after Treg depletion from day 3 after SCI onward (**p < 0.01; Figure 6I). STAT3 inhibition significantly ameliorated locomotor recovery from day seven after SCI, as visualized by markedly elevated BMS scores in SH-4-54-treated Treg-depleted mice compared to untreated counterparts (##p < 0.01; Figure 6I). These results showed that STAT3 inhibition suppressed microglial inflammatory response and partly reversed the neural deficits caused by Treg depletion.

| DISCUSS ION
Following SCI, the inflammatory microenvironment becomes dominant and damages neural regeneration and functional recovery. 4,44 New repair strategies for SCI have highlighted the importance of creating a microenvironment that prevents neuroinflammation and facilitates axonal outgrowth. Tregs are involved in maintaining immune homeostasis and suppressing neuroinflammation in the pathophysiological conditions affecting the CNS. 45 Previous studies have reported that increasing Treg population promoted locomotor recovery after SCI. [24][25][26] However, whether and how Tregs regulate the injured microenvironment after SCI has remained largely unknown.
Our data showed that the number of Tregs increased at least until day 28 after SCI, with a peak of infiltrate in peri-injury areas on day 7. Previous studies have reported that Treg accumulation in ischemic brain is essential for white matter integrity after stroke. 21 Our study showed that infiltrated Treg clustered around microglia and regulated microglia activation after SCI. In the context of SCI where Tregs were depleted, microglia acquired an amoeboid phenotype characterized by enlarged and densely stained soma with few short processes.
Furthermore, Treg-depletion resulted in increased proliferation capability by microglia and a switch toward a more proinflammatory phenotype on day 7 after SCI. These findings are consistent with previous reports in ALS and stroke, which indicated that Tregs regulate microglia activation and skews their differentiation toward an anti-inflammatory and reparative phenotype. 21,[46][47][48] A study in vitro demonstrated that Tregs suppress microglial synthesis and release of reactive oxygen species induced by misfolded α-Synuclein, a protein associated with Parkinson's disease. 49 Altogether, these results support that Tregs inhibit microglia proinflammatory activation after SCI.
Microglia play a central role in neuroinflammation through the release of multiple cytokines and chemokines. 50 Importantly, our findings indicated that Tregs may improve the injured spinal microenvironment through regulating microglia activation. Several inflammation-related genes, including ApoD, Elane, Ctsg, and Cd200r3, were upregulated in microglia in Treg-depleted mice.
These differentially expressed genes have been reported to promote the release of multiple cytokines, including CCL2, GM-CSF, IL-1β, IL-2, IL-6, IL-33, MMP-9, and TNFα, and therefore, to perpetrate an inflammatory microenvironment. 39,51-56 Luminex analysis showed a robust increase in some of these proinflammatory cytokines (GM-CSF, TNFα, IL-1β, IL-6, CCL2, and MMP-9) and a reciprocal decrease in anti-inflammatory cytokines (IL-2 and IL- 33) in the injured environment after Treg depletion. These findings implied that Treg depletion promoted the development of a microgliamediated inflammatory environment after SCI. Similarly, recent studies have reported that Treg depletion almost completely abrogated the protective mechanisms against injury, through enabling TNFα, IL-1β, IL-6, and IL-17 levels to increase rapidly. 57 Conversely, the expansion of brain-resident Tregs protects against pathological neuroinflammation. 58 Collectively, these results suggest that Tregs inhibit microglia inflammatory response after SCI.
We next investigated the potential mechanism of microglia immunoregulation by Tregs. Transcriptome analysis showed that ApoD was a major gene significantly upregulated in microglia after Treg depletion. ApoD is mainly expressed in the central nervous system, and transcriptionally modifies the responses to oxidative stress and demyelination. 59 Recent studies have shown that ApoD enhances inflammation by promoting the production of TNFα and IL-6. 39,60 Interestingly, Tregs depletion led to TNFα and IL-6 upregulation in injured spinal cord. Moreover, immunofluorescence histology showed that TNFα and IL-6 were mostly colocalized with microglia, indicating that Tregs may mainly regulated the expression of TNFα and IL-6 by microglia.
TNFα and IL-6 are involved in canonical STAT3 activation signaling, which eventually impairs function recovery in numerous neuroinflammatory diseases. 29,40 In our study, we first noted a significant upregulation of phosphorylated STAT3 and Socs3 mRNA in Tregdepleted mice, suggesting an alteration of STAT3 activation status after Treg depletion. In vitro, OGD/R-induced STAT3 activation was abolished in microglia cocultured with activated Tregs and by treatment with the STAT3 inhibitor SH-4-54. Our results strongly support that Tregs regulated microglial inflammatory response by suppressing the STAT3 pathway. This hypothesis was further supported in vivo, as STAT3 inhibition decreased microglia immunoreactivity and demyelination, and improved motor function in Treg-depleted mice.
Indeed, astrocytes also play diverse roles in neuroinflammation after SCI. 66 In our study, there was no significant change in astrocytes activation in Treg-depleted mice, while other studies have elucidated that Tregs are involved in inhibiting astrocyte activation during sepsis-associated encephalopathy and stroke. 27,67,68 This discrepancy might be ascribed to different disease model and the time point we chose to observe, since astroglia rapidly increased and peaked around day one post-SCI, followed by a decrease and steady level maintenance until day 42. 5 Treg-astrocytes interaction after SCI in acute phase needs to be further investigated.
In summary, we showed that Treg-microglia crosstalk, likely in-

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare no competing interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.