SOCS1 is essential for regulatory T cell functions by preventing loss of Foxp3 expression as well as IFN-γ and IL-17A production

As reported previously, T cell–specific Socs1-deleted (LckCre-Socs1^f/f; cKO) mice survived >6 mo (Tanaka et al., 2008), yet all of these mice eventually developed dermatitis, splenomegaly, and lymphadenopathy with age (Fig. 1, A–C). Inflammatory cellular infiltrations in the skin, liver, and kidney were observed in these mice (Fig. 1 D). Serum Ig levels were much higher in cKO mice than in Socs1^f/f (WT) mice, and dsDNA antibodies were detected as well (Fig. 1 E). These data suggest that cKO mice developed a systemic inflammatory and autoimmune disease.

The autoimmune phenotype of T cell–specific Socs1-cKO mice has been thought to be mostly caused by hyperactivation of effector T cells and the production of a huge amount of IFN-γ (Fig. 1 F). Because Ifnγ^−/−LckCre-Socs1^f/f mice did not develop dermatitis (not depicted), these phenotypes were dependent on IFN-γ. cKO mice and Socs1^f/f littermates were crossed with knockin mice with Foxp3-IRES-GFP as reported previously (Wang et al., 2008), and the number of nT_reg cells was increased in cKO-Foxp3^GFP mice in the thymus (Fig. 1 G) and periphery (Zhan et al., 2009). Similar autoimmune phenotypes were observed in T_reg cell–specific Socs1-deleted mice, Foxp3Cre-Socs1^f/f mice (Lu et al., 2010), suggesting that SOCS1-deficient nT_reg cells have a defective suppression activity.

The defective suppression activity of SOCS1-deficient T_reg cells in vivo was confirmed by the failure to suppress colitis in Rag2^−/− mice by the cotransfer of naive T cells (Ly5.1) and T_reg cells (Ly5.2; Fig. 2 A and Fig. S1). T_reg cells were purified by FACS sorting as CD3⁺CD4⁺CD25⁺GFP⁺ (>99% Foxp3 positive). 4 wk after transfer, the fraction of Foxp3-positive cells from Ly5.2-positive cKO-Foxp3^GFP mice was significantly decreased compared with those from Ly5.2-positive WT-Foxp3^GFP mice (Fig. 2 B). IFN-γ levels in the mesenteric LN in Rag2^−/− mice cotransferred with naive T cells and Socs1^−/− T_reg cells were much higher than in Rag2^−/− mice transferred with naive T cells and WT T_reg cells (not depicted). These results indicate that the suppressive function of SOCS1-deficient T_reg cells was impaired in lymphopenic conditions, concomitant with faster loss of Foxp3 expression, compared with that of WT T_reg cells.

We then investigated the fate of Foxp3⁺ cells in lymphopenic conditions. T_reg cells from WT-Foxp3^GFP or cKO-Foxp3^GFP mice were transferred into Rag2^−/− mice, and 6 wk later, Foxp3 positivity was examined by flow cytometry. WT T_reg cells lost Foxp3 expression in the single-transfer condition more significantly than in the cotransfer with naive T cells, yet they did not develop colitis, as shown previously (Komatsu et al., 2009). Surprisingly, SOCS1-deficient T_reg cells transferred into Rag2^−/− mice developed colitis (Fig. 2 C and Fig. S2). SOCS1-deficient T_reg cells experienced a more profound loss than did WT T_reg cells (63.1% Foxp3⁺ in WT T_reg cells vs. 40.7% in SOCS1-deficient T_reg cells; Fig. 2 D).

Not only Foxp3 but also other phenotypical markers of T_reg cells, CD25 and CTLA4, which have been shown to be critical molecules in the suppression activity of T_reg cells (Wing et al., 2008), were altered in transferred SOCS1-deficient T_reg cells. Before transfer, SOCS1-deficient T_reg cells expressed CD25 and CTLA4 at significantly higher levels than did WT T_reg cells. After transfer, however, CD25 and CTLA4 were more profoundly down-regulated in Socs1^−/− T_reg cells than in WT T_reg cells, in both the Foxp3-maintaining fraction (Foxp3^+→+; Fig. 2 E) and the exFoxp3 fraction (Foxp3^+→−; Fig. 2 E, top and bottom). In addition, transferred Socs1^−/− T_reg cells produced higher amounts of IFN-γ from both the Foxp3-maintaining and the Foxp3-losing fractions (Fig. 2 F, top and bottom). The Foxp3-maintaining fraction of SOCS1-deficient Foxp3⁺ T_reg cells also produced IL-2 in addition to IFN-γ (Fig. 2 G, top and bottom). These results indicate that SOCS1-deficient T_reg cells tend to lose Foxp3 expression and produce effector cytokines, thus suggesting that SOCS1 is necessary for preventing IFN-γ–producing exFoxp3 cell development.

To exclude the possibility that the higher number of Foxp3⁻ cells in Foxp3⁺Socs1^−/− T_reg cell transfer was caused by an extraordinary expansion of contaminated Foxp3⁻Socs1^−/− cells, activated/memory Socs1^−/− or Socs1^+/+ T cells (Ly5.2⁺CD44^highCD62L⁻Foxp3⁻ cells) were mixed to 3.3% in the cotransfer of Ly5.1 WT naive T cells and T_reg cells (Fig. S3, top) or to 5% in the transfer of Ly5.1 WT T_reg cells into Rag2^−/− mice (Fig. S3, bottom). Any abnormal expansion of activated/memory Socs1^−/− T cells was not observed, suggesting that the larger number of Foxp3⁻ cells in Socs1^−/− T_reg cell transfer was not caused by rapid expansion of contaminated SOCS1-deficient Foxp3⁻ T cells.

IL-10 from T_reg cells has been shown to play an important role in the suppression of naive T cell–induced colitis in Rag2^−/− mice (Maloy et al., 2003). Thus, we suspected that impaired IL-10 production from SOCS1-deficient T_reg cells could be a mechanism for SOCS1-deficient T_reg cell–mediated colitis. However, production of IL-10 was significantly higher in Socs1^−/− T_reg cells compared with WT T_reg cells both before and after transfer, suggesting that IL-10 is not involved in the SOCS1-deficient T_reg cell–mediated colitis (Fig. S4).

Like in Rag2^−/− mice, Foxp3 levels were decreased more rapidly in Socs1^−/− T_reg cells than in WT T_reg cells in vitro. CFSE dilution revealed that the loss of Foxp3 occurred mostly in divided T_reg cells, suggesting that the loss of Foxp3 is associated with cell division or that cells that lost Foxp3 expression divide faster than Foxp3⁺ cells (Fig. 3 A). It has been reported that CD4⁺CD25⁺CD62L^high T_reg cells maintained stable Foxp3 expression along with suppressive activity after in vitro expansion compared with CD4⁺CD25⁺CD62L^low T_reg cells (Scalapino and Daikh, 2009). Thus, we compared Foxp3 expression levels in Socs1^+/+ and Socs1^−/− T_reg cells between the CD62^high and CD62L^low fraction after 3-d in vitro culture. The fraction of CD62L^low was much higher than that of CD62L^high in Socs1^−/− T_reg cells (Fig. 3 B). As reported, Foxp3 expression was decreased in the CD62L^low fraction in WT T_reg cells, and Foxp3 was maintained in the CD62L^high fraction (Fig. 3 C, left). In Socs1^−/− T_reg cells, Foxp3 was also maintained in the CD62L^high fraction. However, the loss of Foxp3 expression in the CD62L^low fraction was significantly more drastic in Socs1^−/− T_reg cells than in WT T_reg cells (Fig. 3 C, right).

Recently, it has been reported that the DNA methylation status of the CNS2 region in the Foxp3 enhancer is particularly important for maintenance of Foxp3 expression (Floess et al., 2007; Kim and Leonard, 2007; Zheng et al., 2010). To examine whether SOCS1-deficient Foxp3⁺ T_reg cells contained uncommitted T_reg cells or unstable inducible T_reg cells (iT_reg cells), we determined the methylation status of the CNS2 region in SOCS1-deficient T_reg cells. As shown in Fig. 3 D, the CNS2 region of T_reg cells from the LN of SOCS1-cKO mice were fully demethylated, similar to those of WT mice (Fig. 3 D). This indicates that >95% SOCS1-deficient T_reg cells were committed and did not contain iT_reg cells. After 3-d in vitro culture, the Foxp3-maintaining fraction (Foxp3^+→+) retained full demethylation in the CNS2 region in both Socs1^+/+ and Socs1^−/− T_reg cells. However, the CNS2 region in the exFoxp3 fraction (Foxp3^+→−) was fully or partially methylated, and more methylation occurred in the SOCS1-defcient exFoxp3 cells than in the SOCS1-sufficient exFoxp3 cells (Fig. 3 D), which is consistent with a more rapid loss of Foxp3 in Socs1^−/− T_reg cells than in WT T_reg cells.

In vivo, we also found hypermethylation of the CNS2 region in Foxp3⁻ cells after transfer of Foxp3⁺ cells into Rag2^−/− mice (Fig. 3 E). More intensive hypermethylation in the CNS2 region in Foxp3⁻ cells was observed in Socs1^−/− T_reg cell transfer than in WT T_reg cell transfer (Fig. 3 E; >90% methylated regions were 90% in Socs1^−/− vs. 50% in Socs1^+/+). Thus, the CNS2 region of the Foxp3 gene in exFoxp3 cells from SOCS1-deficient T_reg cells was more easily hypermethylated not only in vitro but also in vivo than that from WT T_reg cells (Fig. 3 E).

It has been reported that T_reg cells convert into Th1- or Th17-like cells in appropriate in vitro culture conditions (Yang et al., 2008). To investigate the mechanism of the loss of Foxp3 and enhanced IFN-γ production from Socs1^−/− T_reg cells in vivo, we performed an in vitro conversion assay. As shown in Fig. 4 A, when Foxp3⁺ cells were cultured in the presence of TCR stimulation and IL-2, much higher amounts of IFN-γ as well as of IL-17 were produced from Socs1^−/− T_reg cells than from WT T_reg cells. IL-12 further up-regulated IFN-γ production, whereas IL-17 was up-regulated by IL-1 and IL-6 (Fig. 4 A). Interestingly, the production of IL-17 by IL-1 and IL-6 was dependent on TGF-β because anti–TGF-β antibody abrogated the effect of IL-1 and IL-6.

Because SOCS1 has been shown to be an essential negative regulator for the JAK–STAT pathway, the activation status of STAT1 and STAT3 were compared using immunoblotting and intracellular staining (Fig. 4 B). STAT5 was reported to be phosphorylated in SOCS1-deficient T_reg cells at higher levels than in WT T_reg cells, which is associated with the increase in the number of T_reg cells (Lu et al., 2009). Much stronger tyrosine phosphorylation of both STAT1 and STAT3 were observed in freshly isolated SOCS1-deficient T_reg cells than in WT T_reg cells by both methods (Fig. 4 B). These data indicate that SOCS1 is necessary for the suppression of spontaneous activation of STAT1 and STAT3 in T_reg cells.

Because SOCS1 is a strong inhibitor for the IFN-γ–STAT1 pathway, we hypothesized that hyper-STAT1 activation led to abnormalities in SOCS1-deficient T_reg cells. To address this hypothesis, we used T_reg cells from Ifnγ^−/−Socs1^−/− mice.

Most of the Ifnγ^−/−Socs1^+/+ and Ifnγ^−/−Socs1^−/− T_reg cells (CD4⁺CD25^bright fraction) were CD62L^high and CD44^high, suggesting that IFN-γ deficiency diminished CD62L^low T_reg cell subsets (Fig. 5 A). Consistently, when T_reg cells from both Ifnγ^−/−Socs1^+/+ and Ifnγ^−/−Socs1^−/− mice were transferred into Rag2^−/− mice, colitis did not develop even 6 wk after transfer (Fig. 5 B). Foxp3 expression in Ifnγ^−/−Socs1^−/− T_reg cells was as stable as in Ifnγ^−/−Socs1^+/+ T_reg cells and WT T_reg cells after transfer into Rag2^−/− mice (Figs. 2 D and 5 C). After transfer, unlike in Socs1^−/− T_reg cells, CD25 and CTLA4 were maintained at significantly higher levels in Ifnγ^−/−Socs1^−/− T_reg cells (Fig. 5 D, top and bottom). We examined the methylation status of the CNS2 region in both Ifnγ^−/−Socs1^+/+ and Ifnγ^−/−Socs1^−/− T_reg cells in vitro. The CNS2 region of T_reg cells from the LN of Ifnγ^−/−Socs1^+/+ and Ifnγ^−/−Socs1^−/− mice were fully demethylated (Fig. 5 E). After 3-d culture, the Foxp3-maintaining fraction (Foxp3^+→+) retained full demethylation in the CNS2 region in both Ifnγ^−/−Socs1^+/+ and Ifnγ^−/−Socs1^−/− T_reg cells. Methylation levels of the CNS2 region in the Ifnγ^−/− exFoxp3 fraction (Foxp3^+→−) were lower than those in the Ifnγ^+/+ exFoxp3 fraction (Figs. 3 D and 5 E), suggesting that IFN-γ played a positive role in the methylation of the CNS2 region in T_reg cells. The rate of methylation in the CNS2 region in Ifnγ^−/−Socs1^−/− exFoxp3 cells was lower than that in Ifnγ^−/−Socs1^+/+ exFoxp3 cells. These results were consistent with a stable Foxp3 expression in Ifnγ^−/−Socs1^+/+ and Ifnγ^−/−Socs1^−/− T_reg cells after transfer into Rag2^−/− mice.

To test the suppression activity, T_reg cells from Ifnγ^−/−Socs1^+/+ or Ifnγ^−/−Socs1^−/− mice were then transferred into Rag2^−/− mice with naive T cells. 4 wk after transfer, T_reg cells from Ifnγ^−/−Socs1^+/+ mice suppressed colitis elicited by naive T cells (Fig. 6 A and Fig. S7). The suppression activity of Ifnγ^−/− T_reg cells was lower than that of Ifnγ^+/+ T_reg cells (Figs. 2 A and 6 A), yet recipient mice did not develop severe colitis 4 wk after T cell transfer. To our surprise, Ifnγ^−/−Socs1^−/− T_reg cells could not prevent colitis, which was confirmed by body weight loss and histological examination (Fig. 6 A and Fig. S5). Expression of Foxp3 in Ifnγ^−/−Socs1^+/+ T_reg cells as well as in Ifnγ^−/−Socs1^−/− T_reg cells was maintained in recipient Rag2^−/− mice (Fig. 6 B), just as it was in the transfer of T_reg cells alone (Fig. 2 B). Therefore, Foxp3 maintenance is not sufficient for the suppression of effector T cell functions in vivo.

To understand these unexpected results, we examined the production of inflammatory cytokines from T_reg cells or naive/effector cells transferred into Rag2^−/− mice. IL-10 levels were not significantly different between Ifnγ^−/−Socs1^+/+ and Ifnγ^−/−Socs1^−/− T_reg cells (Fig. S4). Because Th17 has also been implicated in colitis (Chaudhry et al., 2009; Durant et al., 2010), we then compared IFN-γ and IL-17 production from naive/effector (Ly5.1⁺) T cells (Fig. 6 C, top) and T_reg cells (Ly5.1⁻, Ly5.2⁺; Fig. 6 C, middle) in the LN of transferred recipient mice, after which percent positivity values of cytokine production were determined by flow cytometry (Fig. 6 C, bottom). As shown in Fig. 6 C, when naive T cells alone were transferred into Rag2^−/− mice, a portion of these cells converted into effector cells that produced mainly IFN-γ, but with very low levels of IL-17 (Fig. 6 C, a). When naive T cells and WT T_reg cells were cotransferred, differentiation of Th1-like cells from naive T cells was completely suppressed (Fig. 6 C, b). However, when naive T cells and Socs1^−/− T_reg cells were cotransferred, both IFN-γ– and IL-17–producing cells appeared from transferred naive T cells, which is consistent with the failure of suppression of colitis by Socs1^−/− T_reg cells (Fig. 6 C, c). Thus, SOCS1-deficient T_reg cells promoted Th1 and Th17 cell differentiation from naive T cells, whereas WT T_reg cells did not. In contrast, when naive T cells and Ifnγ^−/−Socs1^+/+ T_reg cells were cotransferred, a significant fraction of IL-17–producing cells appeared from transferred naive T cells (Fig. 6 C, d). IL-17 was also produced from transferred T_reg cells. These results could explain why Ifnγ^−/− T_reg cells did not prevent colitis very efficiently (Fig. 6 A). Much more intense IL-17 production was observed from both naive T cells and T_reg cell fractions in recipient mice cotransferred with naive T cells and Ifnγ^−/−Socs1^−/− T_reg cells (Fig. 6 C, e). These data suggest that the cytokine profile of T_reg cells instructs the direction of differentiation of Th1 or Th17 from naive T cells in lymphopenic conditions.

To confirm these instructions by T_reg cells, we first measured IL-17 from IFN-γ–deficient T_reg cells in in vitro culture experiments. As expected, Ifnγ^−/− T_reg cells produced higher amounts of IL-17 than did WT T_reg cells, and IL-17 production was further enhanced by SOCS1 deficiency (Fig. 7 A). These data suggest that IFN-γ usually suppressed IL-17 production from T_reg cells, which skewed Th1 cell differentiation from naive T cells. Coincidently, hyper-STAT1 activation was not observed in Ifnγ^−/−Socs1^−/− T_reg cells, although STAT3 activation was still high (Fig. 7 B, top and bottom). Thus, hyper STAT3 activation could account for high levels of IL-17 production from Ifnγ^−/−Socs1^−/− T_reg cells.

The effect of culture supernatants of T_reg cells on Th1/Th17 cell differentiation was then examined (Fig. 7 C). Culture supernatants of T_reg cells were added to the co-culture of naive T cells only or in the presence or absence of APCs (Fig. 7 C). In the absence of APCs, culture supernatants from Socs1^+/+ T_reg cells, Socs1^−/− T_reg cells, or Ifnγ^−/−Socs1^−/− T_reg cells did not promote Th1 or Th17 cell differentiation, suggesting that cytokines from T_reg cells modify the APCs for accelerated Th1 or Th17 cell differentiation form naive T cells (Fig. 7 C, top). Supernatants from Socs1^−/− T_reg cells instructed naive T cells to differentiate into IL-17–producing effector cells when cultured with LPS-untreated APCs (Fig. 7 C, middle; and Fig. S6, right). This finding is consistent with our observations that Socs1^−/− T_reg cells produced IL-17 (Fig. 4 A) and induced more Th17 in vivo (Fig. 6 C). This instruction of naive T cells was augmented by Ifnγ^−/−Socs1^−/− T_reg cells, which produced the largest amount of IL-17 (Fig. 7 A), consistent with the results shown in Fig. 6 C. When APCs were activated by LPS (Fig. 7 C, bottom; and Fig. S6, left), naive T cells, after the addition of supernatant of Socs1^−/− T_reg cells, were instructed to differentiate into IFN-γ–producing effector cells, which is again consistent with higher IFN-γ production from Socs1^−/− T_reg cells (Fig. 4 A), and induced more Th1 in vivo (Fig. 6 C). Ifnγ^−/−Socs1^−/− T_reg cells did not induce Th1 in LPS-treated APCs, suggesting that IFN-γ production from T_reg cells plays an important role in the Th1 instruction of naive T cells. Both IFN-γ and IL-17 antibodies abolished Th1 and Th17 instruction, respectively, confirming that IFN-γ and IL-17 from SOCS1-deficeint T_reg cells play essential roles in Th1 and Th17 induction from naive T cells (Fig. S6).

To confirm the role of APCs, we examined the effect of anticytokine antibodies on the in vitro instruction experiments (Fig. S6). Th1 instruction was eliminated by IL-12 antibodies, suggesting that IL-12 from APCs promoted Th1 cell differentiation. Consistent with this notion, Socs1^−/− T_reg cells could not suppress colitis induced by Ifnγr1^−/− naive T cells transferred into Rag2^−/− mice (Fig. S7). This experiment suggested that IFN-γ from Socs1^−/− T_reg cells still induced colitogenic T cells from Ifnγr1^−/− naive T cells. In contrast, anti–IL-6 antibody completely inhibited and anti–TGF-β antibody partially inhibited Th17 instruction. Anti–IL-23 antibody did not affect Th17 cell differentiation. These results suggest that IL-17 from cultured T_reg cells stimulated IL-6 and TGF-β production from APCs, which promoted Th17 cell differentiation from naive T cells.

To confirm this possibility, we used IL-17–deficient naive T cells for the transfer. As shown in Fig. 7 D, Rag2^−/− mice transferred with IL-17–deficient naive T cells developed colitis, which was less severe than in WT naive T cell transfer. However, Socs1^−/− T_reg cells caused extremely severe colitis (Fig. 7 D), which was triggered by the large levels of IFN-γ (not depicted). As we expected, Ifnγ^−/−Socs1^−/− T_reg cells as well as Ifnγ^−/−Socs1^+/+ T_reg cells suppressed colitis induced by IL-17–deficient naive T cell transfer (Fig. 7 D). These data further support our hypothesis that Ifnγ^−/− T_reg cells instruct Th17 cell differentiation of naive T cells, thereby suppressing colitis less efficiently, and these processes are accelerated when SOCS1 is deficient in Ifnγ^−/− T_reg cells.

In this study, we demonstrated that SOCS1 is necessary for T_reg stability and suppressor functions: SOCS1 protects T_reg cells from harmful effects of inflammatory cytokines, which promote the loss of Foxp3 expression and the conversion into Th1/Th17-like effector cells. Recent studies have shown that T_reg cells rapidly lose Foxp3 expression upon the transfer into a lymphopenic host (Komatsu et al., 2009) or in inflammatory conditions (Zhou et al., 2009). Such exFoxp3 cells (Zhou et al., 2009) or lapsed T_reg cells (Murai et al., 2010) develop an effector-memory phenotype, produce pathogenic cytokines, and can trigger the development of autoimmunity. Multiple studies have suggested that T_reg cells isolated from inflammatory sites express reduced amounts of Foxp3, possibly increasing susceptibility to autoimmunity (Wan and Flavell, 2007; Tang et al., 2008).

Foxp3 expression is regulated by various factors, such as Smad2/3 (Takimoto et al., 2010), Runx1 (Kitoh et al., 2009; Rudra et al., 2009), STAT5 (Yao et al., 2007), c-Rel (Hori, 2010), and DNA methylation of the CNS2 region in the Foxp3 enhancer (Zheng et al., 2010). DNA methylation of CNS2 is particularly important because CNS2 is required for Foxp3 expression in the progeny of dividing T_reg cells (Zheng et al., 2010). Surprisingly, we found that the CNS2 region of freshly isolated SOCS1-deficient T_reg cells was fully demethylated, indicating that they are committed Foxp3⁺ T_reg cells. In Rag2^−/− mice as well as in in vitro culture, part of the committed Foxp3⁺ cells expanded and converted into Foxp3⁻ cells (i.e., exFoxp3 cells). The CNS2 region in SOCS1-deficient exFoxp3 cells was more methylated than that in WT T_reg cells. Also, because the CNS2 methylation status of T_reg cells was altered after 3-d in vitro culture and there was a partially methylated CNS2 region, it is unlikely that all of the exFoxp3 cells were caused by the expansion of contaminated Foxp3⁻ effector T cells. Accordingly, we concluded that SOCS1 contributes to T_reg cell integrity by maintaining stable Foxp3 expression.

Our present study indicated that the excessive IFN-γ–STAT1 signals resulted in faster loss of Foxp3 expression in T_reg cells in lymphopenic conditions as well as in an in vitro culture system. One possibility is that the IFN-γ–STAT1 pathway increases reprogramming from SOCS1-deficient T_reg cells to effector cells. IFN-γ may promote the expansion or development of memory type CD62L^low cells because we observed that the CD62L^low fraction was at normal levels in IFN-γ/SOCS1 double-deficient T_reg cells. STAT1 may antagonize STAT5, yet this is unlikely because we could not observe the reduction of STAT5 phosphorylation in our SOCS1-deficient T_reg cells. STAT1 has also been shown to inhibit the TGF-β–Smad pathway (Tanaka et al., 2008). The phenotypes of Smad2/3-deficient T_reg cells were similar to those observed in SOCS1-deficient T_reg cells (Takimoto et al., 2010). Thus, the suppression of Smad signaling by STAT1 could be one of the mechanisms behind the reprogramming of T_reg cells. The relationship between these factors and the reprogramming of T_reg cells remain to be clarified.

Recently, Rubtsov et al. (2010) reported that highly purified T_reg cells were very stable under physiological and inflammatory conditions. Lu et al. (2010) reported that Foxp3 expression of SOCS1-deficient T_reg cells from Foxp3Cre-Socs1^f/f mice was maintained in the presence of naive and effector T cells, although the suppression activity of Socs1^−/− T_reg cells was defective and IFN-γ was produced from Socs1^−/− T_reg cells with STAT1 hyperactivation. Thus, Foxp3 expression in Socs1^−/− T_reg cells from Foxp3Cre-Socs1^f/f mice was more stable than that in Socs1^−/− T_reg cells from LckCre-Socs1^f/f mice. Because SOCS1 of LckCre-Socs1^f/f mice is deleted in the T_reg cell progenitors, whereas SOCS1 of Foxp3Cre-Socs1^f/f mice is deleted after T_reg cell maturation, this discrepancy may be explained by differences in the developmental conditions of T_reg cells. It is also possible that in LckCre-Socs1^f/f mice, cytokines from Socs1^−/− effector T cells could contribute to more drastic phenotypic changes in Socs1^−/− T_reg cells.

A recent study has demonstrated that during lethal Toxoplasma gondii infection, the exposure of WT T_reg cells to high amounts of Th1 inflammatory mediators superimposes a Th1 effector program on T_reg cells of not only the Foxp3-losing portion but also the Foxp3-maintaining portion (Oldenhove et al., 2009). Thus, the conversion of T_reg cells to Th1 is one of the mechanisms for uncontrolled Th1 response in STAT1-hyperactivated T_reg cells, and this process is usually protected by SOCS1. Recently, Lu et al. (2010) reported on the defective suppression activity of SOCS1-deficient T_reg cells and IFN-γ production from these T_reg cells by STAT1 hyperactivation. In addition, our data suggest the intriguing possibility that hyperactivation of STAT3 in T_reg cells facilitates the conversion to Th17-like cells, which leads to an uncontrolled response in effector Th17 cells. STAT5, in contrast, enhances Foxp3 expression and the number of T_reg cells (Lu et al., 2009).

Like STAT1, STAT3 has been reported to inhibit Foxp3 expression during iT_reg/Th17 cell differentiation together with TGF-β (Takimoto et al., 2010). However, the role of STAT3 in Foxp3 stability in T_reg cells seems to be more complicated. STAT3-deficient T_reg cells have been shown to expand less efficiently than WT T_reg cells in lymphopenic conditions (Durant et al., 2010) and lose their suppression activity, especially for Th17 (Chaudhry et al., 2009; Durant et al., 2010). Foxp3 was stable in Ifnγ^−/−Socs1^−/− T_reg cells with hyperactivated STAT3. Thus, it is unlikely that the hyperactivation of STAT3 is linked to Foxp3 instability. Previously, we have shown that STAT3 activation is reduced in SOCS1-deficient naive T cells in response to IL-6 and TGF-β caused by SOCS3 induction (Tanaka et al., 2008). The mechanism of hyperactivation of STAT3 in SOCS1-deficient T_reg cells remains to be investigated. Although Foxp3 expression of Ifnγ^−/−Socs1^−/− T_reg cells is stabilized, these Ifnγ^−/−Socs1^−/− T_reg cells were still defective in terms of the suppression of IL-17 production from both T_reg cells and effector T cells in Rag2^−/− mice. This indicates that the stable expression of Foxp3 is not a sufficient condition for the suppression of effector Th17 cells. This may be linked to a high conversion potential of T_reg cells to Th17 cells. In a recent study using IL-17–RFP reporter mice with a Foxp3-GFP reporter, Yang et al. (2008) demonstrated the presence of a RFP⁺GFP⁺ transient phase upon T cell activation in vitro and in vivo. Thus, it is highly possible that both Th17 and T_reg cell programs could occur simultaneously in one cell, which underlies the conversion of Foxp3-positive T_reg cells to IL-17–producing cells. We observed that STAT3 was hyperactivated in Socs1^−/− T_reg cells as well as in Ifnγ^−/−Socs1^−/− T_reg cells. Although we do not have direct proof demonstrating that STAT3 hyperactivation is responsible for the conversion of T_reg cells to Th17, it is likely true because IL-17 production from nT_reg cells was enhanced by IL-6 and IL-1 and because STAT3, in addition to TGF-β signals, has been shown to be an essential factor for the generation of Th17.

SOCS1-deficient T_reg cells, which convert to Th1- and Th17-like cells, seem to instruct naive T cells to differentiate into the same cytokine-producing effector T cells. This can also explain why STAT1 or STAT3 hyperactivated T_reg cells cannot control Th1 or Th17 effectors, respectively. Intriguingly, previous studies have indicated that STAT1–T-bet (Koch et al., 2009) and STAT3 (Chaudhry et al., 2009; Durant et al., 2010) in T_reg cells are required for Th1 and Th17 suppression, respectively. Our current study, as well as a previous study (Lu et al., 2010), clearly indicates that both the lack of STAT1/3 and unrestrained STAT1/3 activation in T_reg cells leads to a severe failure of immunological tolerance.

Anti–IFN-γ, –IL-17, –IL-12, and –IL-6 antibodies nearly completely inhibited Th1 and Th17 cell differentiation (Fig. S6). Thus, we clarified that IFN-γ from Th1-converted T_reg cells and IL-17 from Th17-converted T_reg cells preferentially induced effector Th1 and Th17 cell differentiation, respectively, by modulating APCs. It has been reported that T_reg cells interact with DCs before naive T cells and down-regulate the expression of CD80/86 caused by high levels of CTLA4 on T_reg cells (Onishi et al., 2008). Because SOCS1-deficient T_reg cells lose CTLA4 expression, they may not be able to suppress the expression levels of co-stimulator ligands. It has been reported that DCs treated with IFN-γ preferentially induce Th1 (Hanada et al., 2003), whereas IL-17–treated DCs preferentially induce Th17 cell differentiation because of the higher production of IL-1β, IL-6, and IL-23 from IL-17–treated DCs (Sutton et al., 2009). IL-17 may not directly induce Th17 cell differentiation in the absence of APCs because IL-17 has been reported to actually reduce Th17 cell differentiation (Smith et al., 2008). These previous studies support our results, and IFN-γ and IL-17 from T_reg cells are thought to instruct effector T cell differentiation into Th1 or Th17 by modulating APCs.

In conclusion, we propose the idea that SOCS1 is an important guardian of T_reg cells. Our findings also raise the interesting possibility that up-regulation of SOCS1 in T_reg cells at appropriate levels reinforces T_reg cell functions because SOCS1 may protect T_reg cells from harmful effects of inflammatory cytokines that accelerate the conversion of T_reg cells into effectors. These findings may improve T_reg cell therapy for autoimmune diseases and organ transplantations.

LckCre-Socs1^f/f (T cell–specific Socs1-cKO; Tanaka et al., 2008) and Socs1^f/f (littermate control; WT) mice (sex and age matched) were used. cKO mice and Socs1^f/f littermates crossed with knockin mice with Foxp3-IRES-GFP (WT-Foxp3^GFP mice and cKO-Foxp3^GFP mice) were provided by B. Malissen (Université de la Méditerranée, Marseille, France; Wang et al., 2008). Ifnγ^−/−, Ifnγ^−/−Socs1^−/−, Il17^−/−, and Ifnγr1^−/− mice with the C57BL/6J background were described previously (Nakae et al., 2002; Hanada et al., 2003; Suzue et al., 2003). Mice were kept in conventional conditions in Keio University. All experiments using these mice were approved by and performed according to the guidelines of the Animal Ethics Committee of Keio University.

Tissue samples were obtained from the proximal and distal colon, skin, liver, and kidneys and then fixed in 10% neutral buffered formalin, embedded in paraffin, and stained with hematoxylin and eosin (H&E). Assessments of the severity of colitis were performed using the histopathologic score as described previously (Inagaki-Ohara et al., 2004).

Flow cytometry–purified 4 × 10⁵ CD4⁺CD25⁻CD62L⁺CD44⁻ naive T cells from WT mice were injected intravenously into Rag2^−/− mice in combination with 2 × 10⁵ CD3⁺CD4⁺CD25⁺Foxp3^GFP cells from WT-Foxp3^GFP or cKO-Foxp3^GFP mice (8 wk old and sex matched) or 2 × 10⁵ CD3⁺CD4⁺CD25^bright T_reg cells from Ifnγ^−/−Socs1^+/+ or Ifnγ^−/−Socs1^−/− mice (8 wk old and sex matched). Mice were observed daily and weighed weekly. 4 wk after cell transfer, the mice were sacrificed for experiments, and colon sections were stained with H&E. Tissues were graded semiquantitatively. Histological grades were assigned in a blinded manner.

Flow cytometry–purified 2 × 10⁵ CD3⁺CD4⁺CD25⁺Foxp3^GFP cells from WT-Foxp3^GFP or cKO-Foxp3^GFP mice (8 wk old and sex matched) or 2 × 10⁵ CD3⁺CD4⁺CD25^bright T_reg cells from Ifnγ^−/−Socs1^+/+ or Ifnγ^−/−Socs1^−/− mice (8 wk old and sex matched) were injected intravenously into Rag2^−/− mice. Mice were observed daily and weighed weekly. 6 wk after cell transfer, the mice were sacrificed for experiments.

Cell surface staining and flow cytometric analysis of CD3, CD4, CD25, CD62L, and CD44 (all from eBioscience) expression were performed as described previously (Fontenot et al., 2005). For the isolation of T_reg cells, CD4⁺ T cells were positively selected with magnetic-activated cell sorting (Miltenyi Biotec), and CD3⁺CD4⁺CD25⁺Foxp3^GFP cells or CD3⁺CD4⁺CD25^bright cells were further purified using a FACSAria cell sorter (BD). The purity of the sorted populations was invariably >99%. Intracellular staining of Foxp3, IL-2, IFN-γ, IL-17A, and CTLA4 (all from eBioscience) was performed after fixation and permeabilization according to the manufacturer’s instructions. To measure T cell cytokine production, cells were stimulated with 50 ng/ml PMA and 250 ng/ml ionomycin in the presence of Golgi Plug (BD) for 4 h at 37°C before staining. For phospho-STAT1 and phospho-STAT3 staining, Phosflow Lyse/Perm buffer and Perm Buffer III (BD) were used according to the manufacturer’s instructions.

To examine effector T cell differentiation, 10⁵ T_reg cells/well (CD3⁺CD4⁺CD25⁺ Foxp3^GFP cells or CD3⁺CD4⁺CD25^bright cells) from WT-Foxp3^GFP, cKO-Foxp3^GFP, Ifnγ^−/−Socs1^+/+, or Ifnγ^−/−Socs1^−/− mice (8 wk old and sex matched) were stimulated with anti-CD3/anti-CD28 mAb–coated beads (Dynal) at a 1:1 cell/bead ratio for 72 h, with 10 ng/ml IL-2 in 96-well, flat-bottom plates.

3 × 10⁵ WT naive T cells/well were cultured for 5 d without, with LPS-untreated, or with 1 µg/ml LPS–stimulated APCs (7 × 10⁵ T cell–depleted spleen cells/well) and supernatants from 2-d cultured 2 × 10⁵ Socs1^+/+, Socs1^−/−, and Ifnγ^−/−Socs1^−/− T_reg cells/well. Flow cytometric analysis of IFN-γ and IL-17A was performed on these cells. These were also observed with 10 µg/ml anti–IFN-γ, –IL-12, –IL-17, –IL-6, –TGF-β, or –IL-23 antibodies.

Supernatants from cultures of T_reg cells with or without the indicated cytokines were harvested and pooled. Where indicated, 20 ng/ml IL-1α, 20 ng/ml IL-1β, 40 ng/ml IL-6, 50 ng/ml IL-12, 50 ng/ml IL-23, and 10 µg/ml anti–TGF-β mAb (1D11) were added. IFN-γ and IL-17A concentrations were measured using commercially available mouse ELISA Ready-S (eBioscience). All samples were run in triplicate.

Cell lysates from T_reg cells (2 × 10⁵ cells) from WT-Foxp3^GFP, cKO-Foxp3^GFP, Ifnγ^−/−Socs1^+/+, and Ifnγ^−/−Socs1^−/− mice were resolved by SDS-PAGE subjected to Western blot analysis. The membranes were immunoblotted with various antibodies, and the bound antibodies were visualized with horseradish peroxidase–conjugated antibodies against rabbit or mouse IgG (Jackson ImmunoResearch Laboratories, Inc.), using Chemi-Lumi One Super Western blotting detection reagents (Nacalai Tesque). Antibodies against STAT1, phospho-STAT1, STAT3, phospho-STAT3 (all from Cell Signaling Technology), and tubulin (Sigma-Aldrich) were used to visualize the corresponding proteins.

Total RNA was prepared using a nucleospin RNA XS (MACHEREY-NAGEL). RNA was reverse transcribed to cDNA with random primers (Applied Biosystems) and a high capacity cDNA reverse transcription kit (Applied Biosystems) in accordance with the manufacturer’s protocol. To determine the cellular expression level of each gene, quantitative real-time PCR analysis was performed using a C1000 Thermal Cycler (Bio-Rad Laboratories). The PCR mixture consisted of 5 µl of KAPA SYBR FAST qPCR kits (Kapa Biosystems), 15 pmol of forward and reverse primers, and the cDNA samples in a total volume of 10 µl. Relative RNA abundance was determined based on control-HPRT abundance.

Genomic DNA extracted from freshly isolated or 3-d cultured 2 × 10⁵ T_reg cells from male WT-Foxp3^GFP, cKO-Foxp3^GFP, Ifnγ^−/−Socs1^+/+, and Ifnγ^−/−Socs1^−/− mice (8 wk old and sex matched) was digested with BamHI. 2 × 10⁵ T_reg cells from male WT-Foxp3^GFP and cKO-Foxp3^GFP mice before transfer or 4 wk after transfer into Rag2^−/− mice were also examined. H₂O was added to 500 ng of the digested DNA to a volume of 19 µl, and 1.2 µl of 6N NaOH was added; the mixture was then incubated at 37°C for 15 min. Next, 120 µl of 3.6N Na bisulfite, 0.57 mM hydroquinone, and 0.3N NaOH were added. Samples were then treated with 15 cycles of 95°C for 30 s to 50°C for 15 min. The reaction was then desalted using the Wizard DNA cleanup system (Promega) and eluted with 50 µl TE (Tris-EDTA). 3 µl of 5N NaOH was added, and the samples were incubated for 5 min at room temperature. The products were then ethanol precipitated and dissolved with 50 µl of TE buffer. The CNS2 region was PCR amplified with the primer set (forward, 5′-TTTTGGGTTTTTTTGGTATTTAAGA-3′; reverse, 5′-TTAACCAAATTTTTCTACCATTAAC-3′) and T/A cloned into a pGEM-T-Easy vector (Promega). 10 inserted plasmids from each condition were purified and sequenced. 2 × 10⁵ T_reg cells before transfer and 4 wk after transfer into Rag2^−/− mice were also analyzed.

For statistical analysis, we used the Student’s t test.

Fig. S1 shows that the suppressive activity of SOCS1-deficient T_reg cells is attenuated in Rag2^−/− mice. Fig. S2 shows colitis caused by SOCS1-deficient T_reg cells in Rag2^−/− mice. Fig. S3 shows that the loss of Foxp3 expression in T_reg cells is not caused by outgrowth of activated memory T cell contamination of injected donor cells in Rag2^−/− mice. Fig. S4 shows expression levels of IL-10 of T_reg cells before and after transfer in Rag2^−/− mice. Fig. S5 shows that the suppressive activity of Ifnγ^−/−Socs1^−/− T_reg cells is attenuated in Rag2^−/− mice. Fig. S6 shows that anti–IFN-γ, –IL-17, –IL-12, and –IL-6 antibodies nearly completely inhibit Th1 and Th17 cell differentiation through APCs. Fig. S7 shows that Socs1^−/− T_reg cells cannot suppress colitis induced by the transfer of Ifnγr1^−/− naive T cells into Rag2^−/− mice.

We thank N. Shiino for providing technical assistance and N. Aizawa for manuscript preparation.

This work was supported by special grants-in-aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan, the Japan Society for the Promotion of Science, the Takeda Science Foundation, the Uehara Memorial Foundation, the Mitsubishi Pharma Research Foundation, the Kanae Foundation for the Promotion of Medical Science, the Mochida Memorial Foundation, and the Program for the Promotion of Fundamental Studies in Health Science of the National Institute of Biomedical Innovation (to A. Yoshimura).

The authors have no conflicting financial interests.