MicroRNA-21 Regulates Diametrically Opposed Biological Functions of Regulatory T Cells

Regulatory T cells (Tregs) are considered important for controlling the onset and development of autoimmune disease. Although studies have shown that miR-21 is expressed at higher levels in Treg cells, it remains largely elusive whether miR-21 regulates the immune-suppressive function of Tregs. In the current study, we generated mice lacking miR-21 specifically in their Tregs and investigated the role of miR-21 in regulating Treg function both in vitro and in vivo. Our study revealed that Tregs lacking miR-21 exhibit normal phenotype and unaltered function in suppressing T cell proliferation and dendritic cell activation in vitro. However, compared with miR-21-sufficient Tregs, they produce significant more IL-17 and IL-10 when under pathogenic Th17-priming condition. Adenoviral delivery of miR-21 into Treg cells is able to reduce the expression of both IL-17 and IL-10. Mechanistic study revealed that miR-21 down-regulates IL-10 expression through direct targeting of IL-10, and suppresses reprogramming of Tregs into IL-17-secreting cells through down-regulating Stat3 activity. However, we detected no significant or marginal difference in the development of various autoimmune diseases between wild type mice and mice with Treg-specific deletion of miR-21. In conclusion, our study demonstrated that miR-21 in Tregs regulates diametrically opposed biological Treg functions and is largely dispensable for the development of autoimmune disease.


INTRODUCTION
Regulatory T (Treg) cells are a specialized subpopulation of T cells that play crucial roles in maintaining immune homeostasis and preventing autoimmunity (1)(2)(3). Dysregulation in Treg cell frequency or function may lead to the development of autoimmune disease. The Foxp3 protein has generally been described as a master regulator of Treg cell development and function. Scurfy mice and patients carrying Foxp3 gene mutation display lymphoproliferation, lymphocytic infiltration, and multiorgan autoimmune diseases (4,5). In addition, autoimmune diseases in the scurfy mice can be prevented by adoptive transfer of Treg cells (6). However, there is now substantial evidence that inflammatory conditions can cause the loss of Foxp3 expression and the generation of exTregs that adopt an effector phenotype (7)(8)(9). In addition, it has been shown that Tregs may display increased plasticity toward Th17 but retained suppressive function (10).
MicroRNAs (miRNAs) are small single-stranded non-coding RNA molecules that regulate gene expression by targeting mRNA for degradation or translational repression at the posttranscriptional level (11,12).  is the most commonly up-regulated miRNA in a variety of cancers (13)(14)(15), but emerging evidence suggests that miR-21 up-regulation is also associated with the development of a variety of autoimmune diseases, such as psoriasis (16), experimental autoimmune encephalomyelitis (EAE) (17) and experimental autoimmune uveoretinitis (EAU) (18). Moreover, miR-21 knock-out mice are resistant to EAE (17) and lupus-like autoimmunity (19) and therapeutically targeting miR-21 has been shown to successfully treat various autoimmune diseases (17,(20)(21)(22). Further study demonstrated that the change in miR-21 expression was positively correlated with the expression of IL-17 in autoimmune disease. Indeed, miR-21 promotes Th17 differentiation by targeting and depleting SMAD-7, a negative regulator of TGF-b signaling (17). Alternatively, miR-21 can promote glucose metabolism of pathogenic Th17 cells by targeting the E3 ubiquitin ligase Peli1-c-Rel pathway (23).
Although miR-21 is expressed at higher levels in Tregs compared to conventional T cells (24), it is dispensable for iTreg differentiation in vitro and the development of thymic Treg (tTreg) and peripheral Treg (pTreg) in vivo, as reflected by unaltered Foxp3 expression and no developmental defect in the generation of tTreg and pTreg in the absence of miR-21 (17). However, other studies have shown that miR-21 acts as a positive regulator of Foxp3 expression, because increasing miR-21 level in non-Treg T cells or in naïve CD4 + T cells followed by polarization under Treg-priming condition significantly increases Foxp3 level (25,26). In consistency with these results, it has been shown that miR-21 regulates the balance of Th17/Treg cells during T helper cell development in patients with rheumatoid arthritis (RA) by positively regulating Treg cell development (27). To make things even more complicated, it has been reported that treating Tregs in vitro with miR-21inhibitor significantly increases Foxp3 level and miR-21 negatively regulates the frequency of Treg cells and the expression of Foxp3 in peripheral blood mononuclear cell (28). On the other hand, whether miR-21 regulates the immune-suppressive function of Tregs remains largely elusive. The only evidence came from an in vitro study showing that Tregs treated with miR-21 inhibitor remain unchanged in their ability to suppress conventional T cell proliferation and dendritic cell activation (24).
To study the functional significance of Treg-specific expression of miR-21 under physiological conditions, we generated mice lacking miR-21 specifically in their Tregs and investigated the role of miR-21 in regulating Treg function both in vitro and in vivo. Our study revealed that, although Tregs lacking miR-21 produce significant more IL-17 and IL-10 when under pathogenic Th17-priming condition, miR-21-deficiency in Tregs is largely dispensable for the development of autoimmune disease.

Induction and Clinical Evaluation of Autoimmune Disease in Mouse
For the induction of EAU in mouse, mice were anesthetized by intraperitoneal injection of pentobarbital sodium (80 mg/kg) and then immunized with 400 ug inter-photoreceptor retinoidbinding protein (IRBP) 1-20 (5 mg/ml, GPTHLFQPSLVLDM AKVLLD, purchased from China Peptides) emulsified 1:1 in complete Freund's adjuvant (Chondrex, USA) with an additional 100 µl mycobacterium tuberculosis H37R (5 mg/ml, BD biosciences, USA). An additional 200 ng bordetella pertussis toxin (Millipore, USA) was intravenously injected immediately after peptide injection. Fundoscopic exam was used to assess the severity of EAU on a scale of 0-4 using previously published criteria (29).
For the induction of EAE in mouse, 8 to 12-week old female mice were injected subcutaneously into both flanks with 100 mg myelin oligodendrocyte glycoprotein (MOG) 35 For the induction of RA in mouse, 6 to 8-week old male mice were intradermally immunized with 100 mg chicken type II collagen (4 mg/ml, purchased from China Peptides) emulsified in complete Freund's adjuvant at the base of the tail, followed by a booster immunization with 100 mg chicken type II collagen emulsified in incomplete Freund's adjuvant. The cumulative score of all four paw scores were calculated on a scale of 0-16, where each paw was scored as follows: 0, normal paw; 1, one toe inflamed and swollen; 2, more than one toe, but not entire paw inflamed and swollen, or mild swelling of entire paw; 3, entire paw inflamed and swollen; 4, very inflamed and swollen or ankylosed paw.

RNA Isolation and RT-PCR
Total RNA was isolated using TRIzol reagent following manufacturer's instructions (Life Technologies, USA). RNA samples were reversely transcribed using the primescript reverse transcription kit (Takara, JPN). The expression of mouse miR-21, IL-17A, IL-17F, IFN-g, IL-10, TGF-b, CCR-6, Ebi3, Gzmb and Prf1 was determined by quantitative RT-PCR using specific primers and the Applied Biosystems 7500 system. Relative levels of expression were determined using U6 (for miR-

Luciferase Assay
Mouse IL-10 3'-untranslated region (3'UTR) (+1 to +702) containing the miR-21-5p binding site was inserted into the NheI/XbaI site downstream of the luc2 gene of pmirGLO luciferase reporter plasmid (Promega, Madison, USA). Control and miR-21-5p agomir were purchased from Ruibo (Guangzhou, China). For dual-luciferase assay, CD4 + CD8 -YFP + Treg cells from the spleen of WT and Foxp3 Cre-YFP miR-21 f/f mouse were flow sorted and transfected with mouse IL-10 3'UTR reporter plasmid alone or in combination with miR-21-5p agomir using Nucleofector Kit for mouse T cells (Lonza, Switzerland) per manufacturer's instructions. Transfected cells were then cultured in the presence of 2 mg/ml anti-CD3, 2 mg/ml anti-CD28, 500 IU/ml IL-2, and 10 ng/mL TGF-b. After 48 h, the luciferase activities of the whole-cell lysate were analyzed using a dual-luciferase reporter assay system (Promega, USA). Data were normalized for transfection efficiency by dividing firefly luciferase activity by that of the Renilla luciferase.

In Vivo BrdU Assay
Mice were intraperitoneally injected with 100 µl of 10 mg/ml BrdU. Administration of BrdU was repeated every 12 h for three consecutive days. Total cells were isolated from thymus, spleen and MLN 12 h after the last treatment. Cell surface staining was first performed with PE-anti-CD4 and PerCP-Cy5.5-anti-CD8. Cells were then stained with APC-anti-BrdU according to manufacturer's instructions (BD Bioscience, USA) and analyzed by flow cytometry.

Fluorescent Dye-Based Cell Proliferation Assay
Treg cells (CD4 + CD8 -YFP + ) were flow sorted as described above and incubated with 4 mM eFluor-670 (Invitrogen, USA) at 37°C for 15 min. The reaction was stopped with the same volume of cold RPMI-1640 containing 10% heat-inactivated FBS. After washing three times with PBS, eFluor670-labeled cells were cultured at 2×10 5 cells/well (96-well plate) and treated with 100 IU/ml IL-2, 2 ug/ml anti-CD3 and 2 ug/ml anti-CD28 for 72 h. Cell proliferation was then analyzed by flow cytometry.
Apoptotic cells were defined as Annexin V and 7-AAD doublepositive cells.

Adenovirus Infection
Adenovirus over-expressing miR-21 was designed and supplied by Vigene Biosciences, China. Because of the color conflict, Treg cells (CD4 + CD25 + ) from the spleen of C57BL/6 mice were isolated using EasySep Mouse CD4 + CD25 + Treg Cell Isolation Kit (STEMCELL Technologies, Canada). Isolated Treg cells were first cultured in complete RPMI-1640 medium at a density of 2×10 5 /well in 96-well plate and stimulated with 100 IU/ml IL-2 in the presence of 2 ug/ml anti-CD3 plus 2 ug/ml anti-CD28 for 36~48 h. Then 9x10 10 pfu adenovirus was added to the plate. After 12 h, replace culture medium with fresh complete RPMI-1640 medium in the presence of 2 mg/mL anti-CD3, 2 mg/mL anti-CD28, 100 IU/mL IL-2, and continue to culture for 36 h. Transduced cells were then cultured in the presence of 50 ng/ml IL-6, 20 ng/ml IL-1b, 20 ng/ml IL-23 for 72 h. Transduction efficiency was~60-70% as determined by examination under a fluorescent microscope.

Statistical Analysis
The data were analyzed using GraphPad Prism 8.0 and expressed as mean ± SD. The student's t-test was used to compare the significance of the difference between two groups. ANOVA was utilized for multi-group comparison, with Tukey's test for subsequent post-hoc comparisons between two groups. Differences were considered statistically significant at *P<0.05, **P<0.01, and ***P<0.001.

Treg Cells Lacking miR-21 Display Normal Phenotype
To study the functional significance of Treg-specific expression of miR-21 under physiological conditions, we generated mice with Treg specific deletion of miR-21 (Foxp3 Cre-YFP miR-21 f/f ). Deletion of miR-21 in both rested Treg (rTreg) and activated Treg (aTreg) was confirmed by RT-PCR, while miR-21 expression in freshly isolated and activated conventional T cells was not affected ( Figure 1A). Thymic and peripheral T cell development was normal in Foxp3 Cre-YFP miR-21 f/f mice ( Figure S1A). The proportions ( Figure S1A) and numbers ( Figure S1B) of CD4 + CD8 -YFP + Treg cells from thymus, spleen and mesenteric lymph node (MLN) were comparable between Foxp3 Cre-YFP miR-21 +/+ (WT) and Foxp3 Cre-YFP miR-21 f/f mice. We further confirmed normal development of Foxp3 Cre-YFP miR-21 f/f tTreg cells based on proportional expression of Nrp-1 ( Figure S1C). The phenotype of Treg cells, characterized by the expression Foxp3, CD25, cytotoxic T lymphocyte-associated protein 4 (CTLA-4) and glucocorticoidinduced TNFR-related protein (GITR), was also similar between WT and Foxp3 Cre-YFP miR-21 f/f mice ( Figure S2A). Furthermore, we observed a similar outcome for Treg cell phenotype when splenic Treg cells were activated and treated with pathogenic Th17-priming cytokines (IL-6+IL-1b+IL-23) or Th1-priming cytokine (IL-12) in vitro ( Figure S2B). CD25 expression was not examined in this experimental setting because it is also a T cell activation marker. Because it has been reported that miR-21 inhibits cell apoptosis and promotes cell proliferation, we then assessed the apoptosis and proliferation of Treg cells from WT and Foxp3 Cre-YFP miR-21 f/f mice. As shown in Figures S3A, B, the apoptosis ratio, as determined by the proportion of Annexin V + 7AAD + cells, was similar in fresh splenic Tregs and Tregs treated with inflammatory cytokine mix (TNF-a+IL-1b+IFN-g) in vitro. Next we examined the proliferation of Tregs in vitro using fluorescent dye-based cell proliferation assay. Our results showed that miR-21-deficiency in Treg cell does not affect the  proliferation of Tregs when cultured in vitro with IL-2 in the absence or presence of activation (anti-CD3 plus anti-CD28) ( Figures S4A, B). In addition, we used labeling with the thymidine analog BrdU to assess Treg cell proliferation in vivo.
We found that the proportions of BrdU-positive Treg cells in the thymus, spleen and MLN were also similar between WT and Foxp3 Cre-YFP miR-21 f/f mice (Figures S4C, D). Taking together, these results indicate that miR-21-deficiency in Treg cells doesn't affect their development, phenotype, apoptosis and proliferation.

Treg Cells Lacking miR-21 Exhibit Normal Immune-Suppressive Function
Because Treg cells play a key role in the control of pathological immune responses, we next sought to determine whether miR-21deficiency affects the immune-suppressive function of Tregs. MiR-21-deficiency in Tregs does not cause any apparent global immune abnormalities as refl ected by no splenomegaly or lymphadenopathy was observed in any of the Foxp3 Cre-YFP miR-21 f/f mice ( Figure S5). Fractions of splenic naïve and memory T cells in naïve mice were also comparable between WT and Foxp3 Cre-YFP miR-21 f/f mice ( Figure S6). Next, we examined the in vitro suppressive activity of Treg cells. MiR-21-deficient splenic Treg cells proved as effective as WT Treg cells in inhibiting T cell proliferation ( Figures 1B, C). It has been reported that Treg cells can suppress T cell proliferation indirectly by down-regulating costimulatory molecules on antigen presenting cells (such as dendritic cells) (31). We then examined the expression of CD80 and CD86 on bone-marrow derived dendritic cells (BMDCs) cocultured with splenic Tregs from either WT or Foxp3 Cre-YFP miR-21 f/f mice. Our results showed that Tregs from WT and Foxp3 Cre-YFP miR-21 f/f mice exhibit no significant difference in reducing the expression of CD80 and CD86 on BMDCs ( Figure 1D).
Treg Cells Lacking miR-21 Produce Significant More IL-17 and IL-10 In Vitro Secreting inhibitory cytokines, such as IL-10 and TGF-b, is one of the many effector pathways utilized by Treg cells to control immune response. On the other hand, inflammatory environments can cause the loss of Foxp3 expression and the generation of exTregs that gain an effector phenotype. Therefore, we examined re-programming of miR-21-deficient splenic Treg cells treated with anti-CD3 plus anti-CD28 in the presence of Th0, Th1 or pathogenic Th17 priming cytokines. Our results showed that under pathogenic Th17 priming condition, the levels of IL-17 mRNA and protein were significantly increased, while under Th1 priming condition, the levels of IFN-g mRNA and protein were significantly increased (Figures 2A, B). Furthermore, we found that IFN-g expression was comparable between WT and miR-21-deficient Treg cells upon abovementioned destabilizing treatments. By contrast, the mRNA expression ( Figure 2A) and secretion of IL-17A and IL-17F ( Figure 2B) by miR-21-deficient Treg cells were significantly increased compared with WT Treg cells. In addition, we found that the mRNA and protein levels of IL-10, but not TGF-b, were significantly increased by miR-21-deficient Treg cells (Figures 2A, B). We further confirmed the up-regulation of IL-17 and IL-10 by miR-21-deficient Treg cells under pathogenic Th17-priming condition using flow cytometry (Figures 2C, D). However, Foxp3 was down-regulated to similar degrees in WT and miR-21-deficient Treg cells, as the proportion of exTregs (lost Foxp3 expression) and Tregs (maintained Foxp3 expression) were comparable between WT and miR-21deficient Treg cells (Figures 3A, B). We further showed that the proportion of IL-17 + and IL-10 + cells were significantly increased by both exTregs and Tregs that lacking miR-21 ( Figures 3C, D). It shall be noted that although~50% of Tregs lost Foxp3 expression when cultured in vitro under Th17inducing condition, these exTregs may not be the same as exTregs generated in vivo. For example, in vitro generated exTregs actually produce reduced IL-17 as shown in Figure 3C. We examined the expression of RORgT and found that, although in vitro generated exTregs lost Foxp3 expression, they exhibited reduced RORgT expression when compared with Tregs that retained Foxp3 expression ( Figure S7). Although this result is consistent with that of reduced IL-17 expression by exTregs generated in vitro, it remains elusive how RORgT expression was down-regulated in those Tregs.

MiR-21 in Treg Cells Inhibits Il-17 Expression by Down-Regulating Stat3 Activity and Reduces IL-10 Expression Through Direct Targeting of IL-10
To determine how miR-21 in Treg cells inhibits IL-17 expression, we first examined miR-21 expression in both freshly isolated and activated Treg cells in the absence or presence of pathogenic Th17-priming cytokines. Our results showed that while miR-21 expression in Treg cells was increased upon activation, it is further increased when Treg cells were activated in the presence pathogenic Th17-priming cytokines ( Figure 4A). This is not surprising because pathogenic Th17-priming cytokines including IL-6, IL-1b and IL-23 can all activate Stat3, an important transcription factor that promotes miR-21 expression. Surprisingly, while phosphor-Stat3 (pStat3) level was increased in Treg cells upon treatment with pathogenic Th17-priming cytokines, miR-21-deficient Treg cells exhibit significantly elevated pStat3 level compared with WT Treg cells ( Figure 4B). These results indicate that although Stat3 in Treg cells promotes miR-21 expression, miR-21 is able to downregulate Stat3 activity in a negative feedback manner. We further confirmed this hypothesis by showing that adenoviral delivery of miR-21 into Treg cells is able to reduce pStat3 level ( Figure 4C) and the proportion of IL-17 + cells (Figures 4D, E).
To determine whether miR-21 reduces the proportion of IL-17 + cells through down-regulating Stat3 activity, we treated Treg cells with Stat3 inhibitor (Stattic) and then examined the proportion of IL-17 + cells upon treatment with pathogenic Th17-priming cytokines. Our results showed that the proportion of IL-17 + cells is no longer increased by miR-21-deficient Treg cells when they were treated with Stattic as determined by both intracellular staining ( Figure 4F) and ELSIA ( Figure 4G).
In Figure 2 we have shown that miR-21-deficient Treg cells produce significant more IL-10, this indicates that miR-21 in Treg cells may down-regulate IL-10 expression. We confirmed this hypothesis by showing that adenoviral delivery of miR-21 into Treg cells is able to reduce the proportion of IL-10 + cells ( Figures 5A, B). To determine how miR-21 in Treg cells suppresses IL-10 expression, reporter plasmid of mouse IL-10 3'UTR together with control agomir or miR-21 agomir were electroporated into WT Tregs, and then luciferase activity was measured. Our results showed that luciferase activity was significantly lower when miR-21 level was increased in Treg cells ( Figure 5C). Furthermore, reporter plasmid of mouse IL-10 3'UTR was electroporated into both WT and miR-21-deficient Treg cells, and then luciferase activity was measured. We found that luciferase activity was significantly higher in miR-21deficient Treg cells ( Figure 5D). Since it has previously been shown that the 3'UTR of IL-10 contains potential miR-21 binding site and mutagenesis analysis confirmed that IL-10 is a valid target of miR-21 (20), we believe that miR-21 in Treg cells may suppress IL-10 expression through direct targeting of IL-10. It has been reported that, although Stat3 is a critical transcription factor for Th17 differentiation, Treg cells suppresses Th17-specific immune response upon activation of Stat3 (32,33). Since we have shown that Stat3 activity is increased in miR-21-deficient Treg cells, it is possible that those Treg cells may display improved function in suppressing Th17 response. IL-10 is implicated in Stat3-dependent promotion of Treg cell function. However, we have shown in Figure 4 that miR-21-deficient Treg cells actually produce significant more IL-10. We further demonstrated that, although Stat3 inhibition leads to the down-regulation of IL-10, miR-21deficient Treg cells still produce significant more IL-10 ( Figure 5E). We next examined the expression of CCR6, Prf1, Gzmb, and Ebi3, all of which are also implicated in Stat3-dependent promotion of Treg cell function. Our results showed that none of them displayed significant difference between WT and miR-21-deficient Treg cells ( Figure S8).

MiR-21 in Tregs Is Largely Dispensable for the Development of Autoimmune Diseases
Since Treg cells that are defective in inhibiting effector T cells in vivo may display normal in vitro suppressive activity, we next examined whether miR-21 affects Treg function in vivo using three mouse models of autoimmune disease, all of which are closely related to Th17-mediated inflammatory immune pathogenesis.  Experimental Autoimmune Uveitis (EAU) is the animal model of human uveitis that most closely resembles sympathetic ophthalmia. IRBP-induced EAU model was established in both WT and Foxp3 Cre-YFP miR-21 f/f mice. Fundoscopic exam was used to assess the degree of fundus inflammation at different time-points after immunization, and we found no significant difference between WT and Foxp3 Cre-YFP miR-21 f/f mice ( Figure 6A). In consistency with the phenotypical profiles, there is no significant difference in the concentration of IL-17A, IL-17F, IFN-g and TNF-a in the eye between WT and Foxp3 Cre-YFP miR-21 f/f mice ( Figure 6B).
Experimental autoimmune encephalomyelitis (EAE) is the most commonly used rodent model of multiple sclerosis (MS). MOG peptide-induced EAE model was established in both WT and Foxp3 Cre-YFP miR-21 f/f mice. There is no significant difference in clinical scores except that Foxp3 Cre-YFP miR-21 f/f mice exhibit a slightly increased severity during the induction phase ( Figure 6C). When the concentration of inflammatory cytokines in the brain was examined, we found no significant difference in the concentration of IFN-g and TNF-a between WT and Foxp3 Cre-YFP miR-21 f/f mice. However, the concentration of IL-17 was increased in Foxp3 Cre-YFP miR-21 f/f mice ( Figure 6D).
The collagen-induced arthritis (CIA) mouse model is the most commonly studied autoimmune model of rheumatoid arthritis. Collagen II-induced CIA model was established in both WT and Foxp3 Cre-YFP miR-21 f/f mice. Mice were examined physically every other day and four-point arthritis indexes were assessed. As shown in Figure 6E, Foxp3 Cre-YFP miR-21 f/f mice exhibit a slightly increased but significant disease severity than WT mice. In consistency with the phenotypical profiles, the concentrations of inflammatory cytokines including IL-17A, IL-17F, IFN-g and TNF-a were all significantly increased in the paw of Foxp3 Cre-YFP miR-21 f/f mice ( Figure 6F).
Taking together, these data indicate that although Foxp3 Cre-YFP miR-21 f/f mice exhibit a slightly increased RA severity than WT mice, miR-21 in Treg cells is largely dispensable for the development of autoimmune disease.

MiR-21-Deficient Treg Cells That
Have Experienced Inflammatory Environment In Vivo Produce Significant More IL-17 and IL-10 Since we have shown that IL-17 and IL-10 production was increased by miR-21-deficient Treg cells upon treatment in vitro with pathogenic Th17-priming cytokines, we then examined whether they also produce more IL-17 and IL-10 after experiencing inflammatory environment in vivo. In all three autoimmune disease models studied, the proportion of CD4 + CD8 -YFP + cells in the draining lymph node was comparable between WT and Foxp3 Cre-YFP miR-21 f/f mice ( Figures 7A, B). In consistency with the in vitro results from Figure 4, we found that the mRNA ( Figure 7C) and protein levels ( Figure 7D) of IL-17A, IL-17F, and IL-10 were significant increased by Treg cells from Foxp3 Cre-YFP miR-21 f/f mice in all three autoimmune disease models studied.

DISCUSSION
The expression of miR-21 is enhanced in multiple solid tumors and lymphomas where it regulates transformation by limiting the expression of various tumor suppressor genes (34). Outside of tumors, miR-21 is also highly expressed in Th17 and Treg cells (17). While studies have shown that miR-21 promotes Th17 differentiation and glucose metabolism of pathogenic Th17 cells, our current study suggests that miR-21 in Treg cells suppresses the expression of IL-17 and IL-10 under inflammatory condition but is largely dispensable for the development of autoimmune disease. Since Stat3 is a direct target of miR-21 (35,36) and plays a pivotal role in determining Th17 differentiation (37,38), it is not surprising that we found miR-21 suppresses IL-17 production by Treg cells through down-regulating Stat3 activity. However, studies have also found that miR-21 in naïve CD4 + T cells promotes Th17 differentiation through the down-regulation of Smad7 and Peli1, as miR-21-deficient CD4 + T cells display impaired Th17 differentiation (17,23). We hypothesize that whether miR-21 suppresses or promotes IL-17 production depends on the level of miR-21, and excessive miR-21 in conventional CD4 + T cells may also suppress Th17 differentiation. Our un-published data partially confirmed this hypothesis by showing that TNF-a in combination with pathogenic Th17-priming condition will significantly increase miR-21 expression and down-regulate Th17 differentiation. Although further study is needed, this preliminary data indicate that miR-21 may also suppress Th17 differentiation when under strong miR-21-inducing condition.
The plasticity and functional adaptability of Tregs under an inflammatory microenvironment has been demonstrated in autoimmunity (9). Inflammatory condition can cause the loss of Foxp3 expression and the generation of exTregs that adopt a phenotype that is more characteristic of effector CD4 + T cells (7). Foxp3 + cells are comprised of Foxp3-stable CD25 hi and Foxp3unstable CD25 lo populations, the former of which is composed of bona fide Treg cells with sustained Foxp3 expression. Th17 cells originating from CD25 lo Foxp3 + T cells have a key role in the pathogenesis of autoimmune diseases (8). Alternatively, a faction of Tregs may display increased plasticity toward effector cells but retained Foxp3 expression and suppressive function (10). In the current study, we showed that miR-21-deficient and sufficient Tregs exhibit similar degree of conversion to exTregs when cultured in vitro under pathogenic Th17-priming condition. Furthermore, we demonstrated that miR-21-deficient Tregs, whether lost or retained Foxp3 expression, produced significant more IL-17. A fraction of Treg cells may lose YFP . Cells shown were gated on CD4 + CD8cells. The quantification of the percentage of CD4 + CD8 -YFP + Treg cells was shown in (B). (C, D) CD4 + CD8 -YFP + cells were flow sorted from the CLN (for EAU and EAE) and ILN (for RA). Total RNA was extracted and the mRNA expression of IL-17A, IL-17F, and IL-10 was determined by quantitative RT-PCR (C). In addition, sorted CD4 + CD8 -YFP + cells were cultured with a-CD3 plus a-CD28 and IL-2 for 48 h. Culture supernatants were then collected and the protein level of IL-17A, IL-17F, and IL-10 was determined by ELISA (D). Data are representative of two independent experiments. CLN, cervical lymph node; ILN, inguinal lymph node; ns, no significance. *p < 0.05, **p < 0.01.
expression and convert to exTregs during the development of autoimmune disease. In our in vivo studies we only examined the percentage of YFP + cells in the draining lymph nodes and the production of IL-17 and IL-10 by them, whether miR-21 regulates the stability of Tregs in vivo warrants further investigation. We are currently working on this using double transgenic mouse model that permits Foxp3 lineage tracing. The pathological importance of Foxp3 instability in the generation of pathogenic Th17 cells in autoimmunity has been well established. However, although our study showed that miR-21-deficient Tregs produce significant more IL-17 compared with miR-21-sufficient Tregs, miR-21 in Tregs is dispensable for the development of EAU and EAE, and miR-21-deficiency in Tregs only slightly increased the disease severity in RA mouse model. The development of inflammatory autoimmune disease often depends on the balance between Th1/Th17 cells and Tregs. It is likely that increased IL-10 expression by miR-21-deficient Tregs may offset the pathogenic effects mediated by increased Th17 response ( Figure S9). However, when under certain inflammatory condition, such as that in the RA model, miR-21-deficiency in Tregs may cause the imbalance between Th17 cells and Tregs. There is strong evidence showing that plastic Foxp3 + T cells contribute to the pathogenesis of RA. When under arthritic conditions, Treg cells lose Foxp3 expression and undergo transdifferentiation into Th17 cells. These IL-17-expressing exTregs accumulated in inflamed joints and were even more potent osteoclastogenic T cells than were naive CD4 + T cell-derived Th17 cells (8). It is worth noting that increased Stat3 in Tregs may promote the expression of other genes implicated in Treg suppressive function. However, we examined the expression of several key genes (CCR6, Ebi3, Gzmb and Prf1) that have been implicated in Stat3-dependent promotion of Treg function and found none of them displayed significant difference between WT and miR-21-deficient Treg cells.
In conclusion, our study demonstrated that miR-21 in Tregs regulates diametrically opposed biological Treg functions and whether it regulates the development of autoimmune disease depends on the specific inflammatory environment of different types of disease.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

ETHICS STATEMENT
The animal study was reviewed and approved by Shandong Institute of Ophthalmology, Shandong First Medical University.

AUTHOR CONTRIBUTIONS
QR designed research. JS, RL, XH, JB, and WZ performed research. RL contributed reagents/analytic tools. JS, WS, and QR analyzed data. JS and QR wrote the paper. All authors contributed to the article and approved the submitted version.