SOCS1 Is a Key Molecule That Prevents Regulatory T Cell Plasticity under Inflammatory Conditions

It has been shown that the terminally differentiated state of natural regulatory T cells (nTregs) is not defined completely by Foxp3 expression, because Foxp3⁺ Tregs are a heterogeneous population consisting of a committed Treg lineage and an uncommitted subpopulation with developmental plasticity (1). This uncommitted subset of Tregs has been shown to lose Foxp3 expression and to rapidly convert to effector/helper T cells by transfer into a lymphopenic host (1) or under inflammatory conditions (ex-Tregs) (2). Such ex-Tregs (2) have the potential to trigger autoimmunity because they show effector-memory phenotypes with pathogenic cytokine production, and their TCRs are hypothesized to be autoreactive to self-antigens (3–5). For example, Komatsu et al. (6) reported that Th17 cells that developed from Foxp3⁺ T cells play a crucial role in the pathogenesis of autoimmune arthritis. In addition, an uncommitted subset of Tregs lacks hypomethylation of the Treg-specific demethylated region (TSDR), which is responsible for stable Foxp3 expression (7–9). Even in committed Tregs, Foxp3 has been reported to be reversibly downregulated in a certain condition without losing Treg characteristics (9).

Suppressor of cytokine signaling 1 (SOCS1) inhibits the JAK–STAT pathway (10), and uncontrolled IFN-γ signaling results from a deficiency of SOCS1 (11, 12). SOCS1 is highly expressed in Tregs (13) and plays a role in maintaining suppressive functions of nTregs (13, 14). Recent research has shown that SOCS1 is a direct target of the ubiquitin-conjugating enzyme Ubc13, which is involved in the TCR-stimulated activation of the IKK–NF-κB pathway, and this targeting of SOCS1 prevents their conversion into effector-like T cells (15). We have shown that nTregs from T cell–specific Socs1-deficient mice (Socs1^fl/flLck-Cre⁺ mice) easily convert into ex-Tregs, which lose suppressive functions and Foxp3 expression in vivo with DNA hypermethylation in the conserved noncoding DNA sequence (CNS) 2 (CNS2) of the Foxp3 promoter/enhancer (16). Tregs in Socs1^fl/flLck-Cre⁺ mice are constantly exposed to a large amount of inflammatory cytokines from Socs1-deficient non-Tregs, and the Foxp3 levels in Socs1^−/− Tregs with hypomethylated TSDR are restored by deletion of the Ifng gene (16). Thus, unstable expression of Foxp3 in Socs1-deficient Tregs may be dependent on an inflammatory environment.

To evaluate the role of SOCS1 in the stability and suppressive function of Tregs, we analyzed Treg-specific Socs1-deficient mice (Socs1^fl/flFoxp3^YFP-Cre mice). In this article, we report that the conversion of Tregs to Foxp3⁻IFN-γ⁺ ex-Tregs in the tumor microenvironment was accelerated by Socs1 deficiency in Tregs, resulting in inhibition of tumor growth. Although Tregs from Socs1^fl/flFoxp3^YFP-Cre mice maintained suppressive functions and Foxp3 expression when they were transferred into Rag2^−/− mice, Tregs from Socs1^fl/flFoxp3^YFP-Cre mice produced more IFN-γ and showed stronger STAT4 phosphorylation compared with wild-type (WT) Tregs when they were cultured with IL-12–producing APCs from Socs1^fl/flLck-Cre⁺ mice. These results indicate that SOCS1 is one of the essential molecules that maintain Treg stability, particularly under the inflammatory conditions in which APCs are highly activated.

Treg-specific Socs1 conditional knockout mice (Socs1^fl/flFoxp3^YFP-Cre mice), Socs1^+/+Foxp3^YFP-Cre mice (littermate control, WT mice), and T cell–specific Socs1 conditional knockout mice (Socs1^fl/flLck-Cre⁺ mice), sex and age matched, were used. Socs1^fl/flFoxp3^YFP-Cre mice were generated by crossing Socs1^fl/fl mice with Foxp3^YFP-Cre knock-in mice, which were kindly provided by Dr. Rudensky (17). Socs1^fl/flLck-Cre⁺ mice were reported previously (13, 14) and were crossed with Foxp3-IRES-GFP knock-in mice (18) to mark Foxp3⁺ cells (16). To fate map Foxp3 in mice, Socs1^+/+Foxp3^YFP-Cre (WT) mice or Socs1^fl/flFoxp3^YFP-Cre mice were crossed with R26tdRFP mice, which are knock-in mice carrying a tandem-dimer red fluorescent protein (tdRFP) within the Rosa26 locus (Socs1^+/+Foxp3^YFP-Cre×R26tdRFP mice or Socs1^fl/flFoxp3^YFP-Cre×R26tdRFP mice) (19).

Mice were kept under conventional conditions at Keio University (Tokyo, Japan). All experiments involving these mice were approved by and performed according to the guidelines of the Animal Ethics Committee of Keio University.

Total RNA was prepared using the nucleospin RNA XS (MACHEREY-NAGEL). RNA was reverse transcribed to cDNA with random primers (Applied Biosystems) and a high-capacity cDNA reverse transcription kit according to the manufacturer’s protocol (Applied Biosystems). To determine the cellular expression level of each gene, quantitative real-time PCR analysis was performed using a C1000 Thermal Cycler (Bio-Rad). The PCR mixture consisted of 5 μl KAPA SYBR FAST qPCR Kits (Kapa Biosystems), 15 pmol of forward and reverse primers (Socs1, forward: 5′-CGCCAACGGAACTGCTTCTTC-3′; and reverse: 5′-TCAGGTAGTCACGGAGTACC-3′), and cDNA samples in a total volume of 10 μl. Relative RNA abundance was determined based on control–hypoxanthine phosphoribosyltransferase abundance.

Cell surface staining and flow cytometric analysis of CD3, CD4, CD25, CD62L, and CD44 (all Abs from eBioscience) expression were performed as described previously (20). For the isolation of Tregs, CD4⁺ T cells were positively selected with MACS (Miltenyi), and CD3⁺CD4⁺CD25⁺Foxp3^YFP-Cre cells or CD3⁺CD4⁺CD25⁺Foxp3^GFP cells were further purified using a FACSAria cell sorter (Becton Dickinson). For isolation of APCs, CD3⁺ T cells were depleted from splenocytes, using the FACSAria cell sorter (Becton Dickinson). The purity of the sorted populations was invariable (>99%). Intracellular staining of Foxp3, IFN-γ, and T-bet (all Abs from eBioscience) was performed following fixation and permeabilization according to the manufacturer’s instructions. To measure T cell cytokine production, cells were stimulated with PMA (50 ng/ml) and ionomycin (250 ng/ml) in the presence of GolgiPlug (BD Biosciences) for 4 h at 37°C before staining.

B16/F10 melanoma cells (2 × 10⁵ cells) were transplanted into Socs1^+/+Foxp3^YFP-Cre or Socs1^fl/flFoxp3^YFP-Cre mice. To observe the conversion of Tregs to Foxp3⁻IFN-γ⁺ ex-Tregs using a fate mapping system, 5 × 10⁵ Tregs (YFP⁺RFP⁺ cells) from Socs1^+/+Foxp3^YFP-Cre×R26tdRFP or Socs1^fl/flFoxp3^YFP-Cre×R26tdRFP mice were transferred into WT mice transplanted with 1 × 10⁶ B16/F10 melanoma cells after 2 Gy irradiation. CD3⁺CD4⁺YFP⁺RFP⁺ cells were isolated as Tregs, using a FACSAria cell sorter. Fifteen days after cell transfer, the mice were sacrificed for experiments, and cells (YFP⁺RFP⁺ or YFP⁻RFP⁺ cells) from the tumors and draining lymph nodes (LNs) were analyzed. Tregs (YFP⁺RFP⁺ cells) or ex-Tregs (YFP⁻RFP⁺ cells) were sorted using the FACSAria cell sorter, and Ifng mRNA levels were quantified using RT-PCR analysis.

Sorted CD4⁺CD25⁻CD62L⁺CD44⁻ naive T cells (4 × 10⁵ cells) from WT mice were injected i.v. into Rag2^−/− mice in combination with 2 × 10⁵ CD3⁺CD4⁺CD25⁺Foxp3^YFP-Cre cells from Socs1^+/+Foxp3^YFP-Cre or Socs1^fl/flFoxp3^YFP-Cre mice (8 wk old and sex matched). Mice were observed daily and weighed weekly. Four weeks after cell transfer, the mice were sacrificed for experiments.

Sorted CD3⁺CD4⁺CD25⁺Foxp3^YFP-Cre cells (2 × 10⁵ cells) from Socs1^+/+Foxp3^YFP-Cre or Socs1^fl/flFoxp3^YFP-Cre mice (8 wk old and sex matched) were injected i.v. into Rag2^−/− mice. Mice were observed daily and weighed weekly. Eight weeks after cell transfer, the mice were sacrificed for experiments.

Tregs (CD3⁺CD4⁺CD25⁺Foxp3^YFP-Cre cells or CD3⁺CD4⁺CD25⁺Foxp3^GFP cells) from WT, Socs1^fl/flLck-Cre⁺, Socs1^fl/flFoxp3^YFP-Cre (<12 wk), and Socs1^fl/flFoxp3^YFP-Cre (>12 wk) mice (age and sex matched), at a concentration of 5 × 10⁴ cells per well, were stimulated for 3 d with anti-CD3/anti-CD28 mAbs–coated beads (Dynal) at a 1:1 cell/bead ratio or with the indicated APCs together with IL-2 (20 ng/ml; PeproTech) in 96-well, flat-bottom plates. Where indicated, IL-12 (40 ng/ml; PeproTech), IFN-γ (50 ng/ml; PeproTech), phosphorothioate-modified CpG-ODN (CpG) (1 μM; GeneDesign), or anti–IL-12p40 mAb (10 μg/ml) was added.

Supernatants from cultures of APCs were harvested and pooled. IL-12p40 concentrations were measured using the commercially available Mouse ELISA Ready-SET-Go! Kit (eBioscience). All samples were run in triplicate.

Cell lysates from Tregs (2 × 10⁵ cells) were resolved by NaDodSO₄ PAGE (SDS-PAGE) and subjected to Western blot analysis. The membranes were immunoblotted with various Abs, and the bound Abs were visualized with HRP-conjugated Abs against rabbit or mouse IgG (Jackson ImmunoResearch Laboratories), using Chemi-Lumi One Super Western blotting detection reagents (Nacalai Tesque). Abs against phospho-STAT1, phospho-STAT3, and phospho-STAT4 (all from Cell Signaling Technology) and tubulin (Sigma) were used to visualize the corresponding proteins.

Paired Student t test, Mann–Whitney U tests, or Kruskal-Wallis tests with the Bonferroni test were used to determine statistical significance. Error bars show SD.

We previously reported that T cell–specific Socs1 conditional knockout mice (Socs1^fl/flLck-Cre⁺ mice) survived for >6 mo, and all eventually developed dermatitis with age (12, 16). Almost all of these Socs1^fl/flLck-Cre⁺ mice displayed splenomegaly, lymphadenopathy, and abnormal lobulation of the thymus. The autoimmune phenotype of these Socs1^fl/flLck-Cre⁺ mice has been thought to be primarily caused by hyperactivation of effector T cells, because most CD4⁺ T cells showed effector-memory phenotypes (CD44^highCD62L^low) (16) and Ifng^−/−Socs1^fl/flLck-Cre⁺ mice did not develop dermatitis (data not shown).

To examine the role of SOCS1 specifically in Tregs, we generated Treg-specific Socs1-deficient mice (Socs1^fl/flFoxp3^YFP-Cre mice) by crossing Socs1^fl/fl mice with Foxp3^YFP-Cre knock-in mice (17). We confirmed Socs1 gene deletion in Foxp3^YFP-Cre-positive cells, but not in Foxp3^YFP-Cre-negative cells. In Socs1^+/+Foxp3^YFP-Cre (WT) mice, the expression of Socs1 was higher in Tregs than in non-Tregs (Supplemental Fig. 1). Similar to the T cell–specific Socs1-deficient Socs1^fl/flLck-Cre⁺ mice, dermatitis, splenomegaly, and LN swelling were observed in the Treg-specific Socs1-deficient Socs1^fl/flFoxp3^YFP-Cre mice (Fig. 1A, 1B). However, penetration of these phenotypes was ∼40% in these mice, and dermatitis was milder and later than that observed in the Socs1^fl/flLck-Cre⁺ mice (16). Unlike in the Socs1^fl/flLck-Cre⁺ mice, abnormal lobulation of the thymus was not detected in the Socs1^fl/flFoxp3^YFP-Cre mice.

When we compared the relative populations of Socs1-sufficient and Socs1-deficient Tregs in a single Socs1^fl/flFoxp3^YFP-Cre/+ female mouse, we observed a strong difference in their ratios and absolute cell numbers in the thymus, spleen, and LNs (Fig. 1C). Because Foxp3 is randomly expressed in each chromosome owing to lyonization of the X chromosome, the YFP⁺/YFP⁻ ratio is 1:1 in Foxp3^YFP-Cre/+ female mice. Indeed, the YFP⁺/YFP⁻ ratio in the WT Socs1^+/+Foxp3^YFP-Cre/+ females was nearly 1:1 in both the thymus and the periphery (the spleen and LNs) (Fig. 1C). However, the YFP⁺/YFP⁻ ratio in the Socs1^fl/flFoxp3^YFP-Cre/+ females was 5–7:1 in the thymus and periphery. These results indicated that Socs1 deficiency in Tregs provides a great advantage in terms of increasing their population in a cell-autonomous manner, which is consistent with our previous report that showed an expansion of Tregs in Socs1^fl/flLck-Cre⁺ mice (16).

We found that Socs1^fl/flFoxp3^YFP-Cre/+ female mice also develop splenomegaly, lymphadenopathy, and dermatitis 2.5 mo after birth (Fig. 1D), suggesting a cell-intrinsic role of SOCS1 in Tregs. However, compared with Socs1^fl/flFoxp3^YFP-Cre female mice (Fig. 1B), the onset of autoimmune symptoms of Socs1^fl/flFoxp3^YFP-Cre/+ female mice was slower and severity was lower, suggesting that normal Socs1-sufficient Tregs can partly suppress the pathogenicity of Socs1-deficient Tregs. These observations support our hypothesis that pathogenic phenotypes of Socs1-deficient Tregs are obvious only under inflammatory conditions.

Next, we examined whether Socs1 deficiency in Tregs affects naive and effector T cells. The fraction of CD62L^low cells was greatly increased in Tregs from Socs1^fl/flFoxp3^YFP-Cre mice compared with controls (Fig. 1E, upper middle panel). CD44 levels were not altered between WT and Socs1-deficient Tregs (Fig. 1E, lower middle panel). Interestingly, non-Treg (YFP−) CD4⁺ T cells from Socs1^fl/flFoxp3^YFP-Cre mice showed reduced CD62L and higher CD44 expression compared with controls (Fig. 1E, right), suggesting that activated or effector CD4⁺ T cells increased in Socs1^fl/flFoxp3^YFP-Cre mice (Fig. 1E, left). In addition, Tregs in the Socs1^fl/flLck-Cre⁺ and in the older Socs1^fl/flFoxp3^YFP-Cre mice were found to express higher levels of T-bet (Fig. 1F), suggesting that T-bet expression in Tregs correlates with severe inflammation in vivo. These data suggest that Socs1-deficient Tregs have some defects in normal Treg functions, thereby increasing the activation and/or memory status of Socs1-sufficient non-Tregs in Socs1^fl/flFoxp3^YFP-Cre mice.

To examine whether deletion of Socs1 in Tregs facilitates anti-tumor immunity, 2 × 10⁵ B16/F10 melanoma cells were transplanted into WT or Socs1^fl/flFoxp3^YFP-Cre mice. The tumor grew more slowly in the Socs1^fl/flFoxp3^YFP-Cre mice than in the WT mice (Fig. 2A). Next, to investigate the fate of Tregs in vivo, fate mapping mice were generated by crossing Socs1^fl/flFoxp3^YFP-Cre and R26tdRFP knock-in mice carrying a tdRFP within the Rosa26 locus. Tregs (YFP⁺RFP⁺ cells) from fate mapping mice were transferred into WT mice transplanted with 1 × 10⁶ B16/F10 melanoma cells after 2 Gy irradiation. Transfer of WT Tregs (Socs1^+/+ Tregs) enhanced tumor growth, whereas transfer of Tregs from Socs1^fl/flFoxp3^YFP-Cre×R26tdRFP mice (Socs1^−/− Tregs) suppressed tumor growth compared with control mice without Treg transfer (Fig. 2B). Transferred Tregs did not cause autoimmune phenotypes in these mice. In the draining LN, the percentage of the ex-Treg fraction (YFP⁻ in RFP⁺) was not different between Socs1^−/− Tregs and Socs1^+/+ Tregs (Fig. 2C, upper panels, 2D). However, within the tumor, the ex-Treg fraction was significantly increased in Socs1^−/− Tregs compared with Socs1^+/+ Tregs (Fig. 2C, lower panels, 2D). In addition, Socs1^−/− ex-Tregs (YFP⁻RFP⁺ cells) isolated from the tumors expressed higher levels of Infg than did Socs1^+/+ ex-Tregs (Fig. 2E). These data suggest that the environment within the tumor facilitated the pathogenic conversion of Socs1-deficient Tregs into Foxp3⁻IFN-γ⁺ ex-Tregs.

Defective suppressive activity of Socs1-deficient Tregs from Socs1^fl/flLck-Cre⁺ mice in vivo has been reported; thus, Socs1-deficient Tregs did not suppress colitis in Rag2^−/− mice transferred with naive T cells (16). In contrast, Tregs from Socs1^fl/flFoxp3^YFP-Cre mice retained suppressive activity in vivo; Rag2^−/− mice transferred with Socs1-deficient Tregs from Socs1^fl/flFoxp3^YFP-Cre mice with naive T cells did not show severe colitis, although the body weight of these mice was slightly decreased (Fig. 3A). Foxp3 expression in these Socs1^−/− Tregs was as stable as that in the Socs1^+/+ Tregs after transfer into Rag2^−/− mice (94.5% Foxp3⁺ in WT Tregs versus 88.4% in Socs1-deficient Tregs) (Fig. 3B), whereas more severe loss of Foxp3 was observed when Tregs from Socs1^fl/flLck-Cre⁺ mice were used (16). IFN-γ levels in the mesenteric LNs in Rag2^−/− mice cotransferred with naive T cells and Socs1^−/− Tregs from Socs1^fl/flFoxp3^YFP-Cre mice were nearly the same as those in Rag2^−/− mice transferred with naive T cells and WT Tregs (data not shown).

Next, we examined the fate of Foxp3⁺ Tregs under lymphopenic conditions. Tregs from the LNs of WT or Socs1^fl/flFoxp3^YFP-Cre mice were transferred into Rag2^−/− mice, and Foxp3 positivity was examined using flow cytometry. We previously reported a similar experiment using Tregs from the LNs of WT or Socs1^fl/flLck-Cre⁺ mice in which, 6 wk after transfer, Rag2^−/− mice transferred with Tregs from Socs1^fl/flLck-Cre⁺ mice developed more profound colitis than did those transferred with Tregs from the WT mice, and the Socs1^fl/flLck-Cre⁺ Tregs showed a greater loss of Foxp3 expression than did WT Tregs (16). However, Rag2^−/− mice transferred with Tregs from the Socs1-deficient Socs1^fl/flFoxp3^YFP-Cre mice showed only a slight body weight loss and did not have colitis even when observed 8 wk later (Fig. 3C). In addition, these Socs1-deficient Tregs maintained the same levels of Foxp3 expression as WT Tregs (44.3% Foxp3⁺ in WT Tregs versus 42.3% in Socs1-deficient Tregs) (Fig. 3D). Thus, in contrast to the results of Tregs from Socs1^fl/flLck-Cre⁺ mice, suppressive functions and Foxp3 expression were maintained with relative stability in SOCS1-deficient Tregs from Socs1^fl/flFoxp3^YFP-Cre mice under lymphopenic conditions.

When stimulated with anti-CD3/CD28 Abs in vitro, Tregs from Socs1^fl/flLck-Cre⁺ mice (Socs1^fl/flLck-Cre⁺ Tregs) showed a loss of Foxp3 expression and produced IFN-γ similar to that shown in the in vivo experiments in Fig. 3, whereas WT Tregs (Socs1^+/+ Tregs) or Tregs from Socs1^fl/flFoxp3^YFP-Cre mice (Socs1^fl/flFoxp3^YFP-Cre Tregs) did not (Fig. 4A, upper). The results of stimulation with APCs (CD3⁺ T cell–depleted spleen cells) from WT mice were the same as those obtained when using CD3/CD28 stimulation (Fig. 4A, middle). However, when these Tregs were cultured with APCs from Socs1^fl/flLck-Cre⁺ mice, the Socs1^fl/flFoxp3^YFP-Cre Tregs lost Foxp3 expression and produced IFN-γ to a level statistically similar to that produced by the Socs1^fl/flLck-Cre⁺ Tregs (Fig. 4A, lower, 4B). The proportions of Foxp3⁻IFN-γ⁺ and Foxp3⁺IFN-γ⁺ cells in Socs1^fl/flFoxp3^YFP-Cre Tregs were significantly increased when they were cocultured with APCs from Socs1^fl/flLck-Cre⁺ mice, compared with coculture with anti-CD3/CD28 Abs or APCs from WT mice (Fig. 4B, Supplemental Fig. 2). Thus, when Tregs were cultured with APCs from highly inflammatory Socs1^fl/flLck-Cre⁺ mice in vitro, Socs1^fl/flFoxp3^YFP-Cre Tregs also converted to ex-Tregs. The amount of IFN-γ production from Socs1-deficient Tregs was dependent on the number of APCs and was independent of the number of Tregs (data not shown).

To clarify the factors of APCs from Socs1^fl/flLck-Cre⁺ mice that were important for the conversion of Tregs into ex-Tregs, the culture supernatant from each in vitro culture of Tregs plus APCs was added to the culture of Tregs and anti-CD3/CD28 Abs or APCs (Fig. 4C). Thus, Socs1^+/+ Tregs and Socs1^fl/flFoxp3^YFP-Cre Tregs were cultured in the presence of WT culture supernatant (i.e., supernatant from a 4-d culture of Socs1^+/+ Tregs with Socs1^+/+ APCs) or Socs1^fl/flLck-Cre⁺ culture supernatant (i.e., supernatant from a 4-d culture of Socs1^fl/flLck-Cre⁺ Tregs with APCs from Socs1^fl/flLck-Cre⁺ mice). Both Socs1^+/+ Tregs and Socs1^fl/flFoxp3^YFP-Cre Tregs efficiently converted to Foxp3⁻IFN-γ⁺ cells when they were cultured in the presence of the Socs1^fl/flLck-Cre⁺ culture supernatant (Fig. 4C, lower panels). Furthermore, Socs1^fl/flFoxp3^YFP-Cre Tregs produced the highest levels of IFN-γ from the Foxp3⁻ fraction when they were cocultured with Socs1^+/+ APCs in the presence of the Socs1^fl/flLck-Cre⁺ culture supernatant (Fig. 4C, lower right). The proportions of the Foxp3⁻IFN-γ⁺ fraction and the Foxp3⁺IFN-γ⁺ fraction are summarized in Supplemental Fig. 3. There was no significant difference between Socs1^+/+ Tregs and Socs1^fl/flFoxp3^YFP-Cre Tregs in the production of IFN-γ in the Foxp3⁺ fraction. These results suggest that the Socs1^fl/flLck-Cre⁺ culture supernatant contains some soluble factors that convert Socs1^fl/flFoxp3^YFP-Cre Tregs into pathogenic ex-Tregs.

The production of IFN-γ from Socs1^+/+ or Socs1^fl/flFoxp3^YFP-Cre Tregs cultured with APCs from Socs1^fl/flLck-Cre⁺ mice was completely inhibited by the anti–IL-12p40 Ab (Fig. 5A), but not by the anti–IFN-γ Ab (data not shown). IFN-γ production from the Foxp3⁻ fraction was observed in Socs1^fl/flFoxp3^YFP-Cre Tregs, but not in Socs1^+/+ Tregs cocultured with Socs1^+/+ APCs in the presence of IL-12 (Fig. 5B, Supplemental Fig. 4). To further characterize the role of APCs in the conversion of Socs1-deficient Tregs to Foxp3⁻IFN-γ⁺ ex-Tregs, Tregs were cocultured with Socs1^+/+ APCs that were stimulated with IFN-γ or the CpG oligonucleotide. APCs activated with IFN-γ induced Foxp3⁻IFN-γ⁺ and Foxp3⁺IFN-γ⁺ cells from Socs1^fl/flFoxp3^YFP-Cre Tregs (Fig. 5C, upper); however, APCs activated with the CpG oligonucleotide induced more IFN-γ⁺ cells than did APCs activated with IFN-γ (Fig. 5C, lower). APCs from Socs1^fl/flFoxp3^YFP-Cre mice produced significantly more IL-12p40 than did APCs from Socs1^+/+ mice in response to the CpG oligonucleotide (Fig. 6A). Thus, APCs from Socs1^fl/flFoxp3^YFP-Cre mice are already somewhat activated probably because of inflammatory conditions in these mice. Regarding the culture supernatant used in Fig. 4C, IL-12p40 was also highly secreted from APCs from Socs1^fl/flLck-Cre⁺ mice when they were cocultured with Socs1^fl/flLck-Cre⁺ Tregs (Fig. 6B), suggesting that APCs were already highly activated in Socs1-deficient mice and their activation level was similar to that of APCs stimulated with the CpG oligonucleotide.

Finally, to examine whether Socs1-deficient Tregs are more sensitive to stimulation with IL-12 than are WT Tregs, we measured STAT phosphorylation in response to IL-12. As shown in Fig. 6C, much stronger tyrosine phosphorylation of STAT4 was observed in Socs1-deficient Tregs from Socs1^fl/flFoxp3^YFP-Cre mice than in Socs1-sufficient Tregs after 30 min of IL-12 stimulation (Fig. 6C). The activation status of STAT1 and STAT3 was not affected by IL-12. These data suggest that hyperactivation of STAT4 by Socs1 deficiency is responsible for the high levels of IFN-γ production from Socs1-deficient Tregs.

In the current study, we demonstrated that Tregs from Socs1^fl/flFoxp3^YFP-Cre mice maintained Foxp3 expression when they were transferred into Rag2^−/− mice, unlike Tregs from Socs1^fl/flLck-Cre⁺ mice. This finding is in agreement with a previous report that Foxp3 expression of Socs1-deficient Tregs from Socs1^fl/flFoxp3^YFP-Cre mice was maintained in the presence of naive and effector T cells (14). The reasons why Foxp3 expression in Socs1^−/− Tregs from Socs1^fl/flFoxp3^YFP-Cre mice is more stable than that in Socs1^−/− Tregs from Socs1^fl/flLck-Cre⁺ mice are that 1) the environment around Tregs in Socs1^fl/flLck-Cre⁺ mice include inflammatory conditions generated by Socs1-deficient non-Tregs; and 2) SOCS1 in Socs1^fl/flLck-Cre⁺ mice is deleted in the Treg progenitors, whereas SOCS1 in Socs1^fl/flFoxp3^YFP-Cre mice is removed after Treg maturation. However, when Tregs from Socs1^fl/flFoxp3^YFP-Cre mice were cultured with IL-12–producing APCs from Socs1^fl/flLck-Cre⁺ mice, STAT4 was activated in these Tregs, and, similar to Tregs from Socs1^fl/flLck-Cre⁺ mice, they lost Foxp3 expression, which was accompanied by the production of inflammatory cytokines. CpG-stimulated APCs or, in particular, IL-12–stimulated APCs, caused Socs1-deficient Tregs from Socs1^fl/flFoxp3^YFP-Cre mice to produce IFN-γ from the Foxp3⁻ fraction. IFN-γ–producing Foxp3⁺ cells were reported to be detected in the peripheral mononuclear cells of multiple sclerosis patients (21). These ex-Tregs develop an effector-memory phenotype, produce pathogenic cytokines, and cause autoimmunity. It is of note that they also inhibit tumor progression.

The Foxp3 promoter contains a newly identified Nr4a transcription factor binding site, which is critical for nTreg development in the thymus (22, 23), in addition to transcription factor NFAT and AP-1 binding sites, as well as a TATA and a CAAT box (24). Three highly conserved CNSs, CNS1, CNS2, and CNS3, were also identified. The Treg-specific CpG hypomethylation of the CNS2 TSDR has been reported as indispensable for the development of Foxp3⁺ T cells as nTregs (25), for lineage stability, and for suppressive activity (26, 27). In addition to epigenetic modification, a recent study has indicated that the coexpression of Foxp3 with at least one of the transcription factors GATA-1, IRF4, Lef1, Ikzf4, or Satb1 is responsible for some of the Treg phenotype gene expression (28). In addition, biochemical and mass-spectrometric analyses have shown that Foxp3 forms complexes with several cofactors, such as GATA3, NFAT, or Runx1/Cbfβ (29–32). These protein–protein interactions between Foxp3 and other factors influence Foxp3 stability, although it is not clear how these Foxp3 binding proteins affect Treg stability (33). To date, various factors have been reported to regulate Foxp3 expression, including Smad2/3 (34), Runx1 (30, 35), STAT5 (36), c-Rel (37), TRAF6 (38), PTEN (39), and SOCS1 (16), although they are not bound to Foxp3 directly. The Smad2/3-deficient Treg phenotypes were similar to those observed in Socs1-deficient Tregs (34).

In a previous study, we showed that the CNS2 TSDR of freshly isolated Socs1-deficient Tregs was fully demethylated, even in Socs1^fl/flLck-Cre⁺ mice under inflammatory conditions, indicating that they are committed Foxp3⁺ Tregs (16). It has been shown that purified Tregs are very stable, even under inflammatory conditions (40), and that uncommitted, not fully demethylated Foxp3⁺ Tregs become ex-Tregs (9). In contrast, consistent with our results, Tregs with a fully TSDR were found to lose Foxp3 expression in an experimental autoimmune encephalomyelitis model (41). These ex-Tregs with an Ag-specific TCR appeared only when they were adoptively transferred into mice with ongoing experimental autoimmune encephalomyelitis. We also found that Tregs from Socs1^fl/flFoxp3^YFP-Cre mice lost Foxp3 expression that was accompanied by activated STAT4 when they were cultured with APCs from Socs1^fl/flLck-Cre⁺ mice. In support of these results, almost all previous studies related to Treg plasticity have been studied under lymphopenic or inflammatory conditions (1, 2, 9, 16, 42–48). Although we must exclude the experimental technological limitation of methylation sequencing read depth by the Sanger sequencing method from our results, committed Socs1-deficient or STAT4-activated Tregs were determined to convert to ex-Tregs under inflammatory conditions. Additional experiments are needed to examine how SOCS1 or STAT4 is associated with the Treg fate of lineage commitment status. It has recently been reported that a TCR-mediated IKK–NF-κB signaling pathway including Ubc13 maintains the suppressive activity of Tregs by controlling the expression of SOCS1 (15), and the necessity of TCR signaling for Treg-suppressive function was shown to be independent of Foxp3 expression and IL-2 receptor signaling (49). STAT4 has been reported to be a molecule that is crucial for the onset or pathogenesis of lupus (50, 51), and it should be clarified whether STAT4 expression is involved in Treg development or Treg-suppressive functions in lupus.

It has also been shown that committed Tregs are memory Tregs that can re-express Foxp3 and maintain their suppressive function, even if they transiently downregulate Foxp3 expression (9). CNS2 contains binding sites for the transcription factors CREB (8) and STAT5 (52), whose engagement with the TSDR is considered necessary for Foxp3 reinduction in addition to TCR and/or IL-2 signaling (9). However, in our B16/F10 melanoma model, Socs1-deficient ex-Tregs effectively inhibited tumor growth. Using a fate mapping system, we determined that Socs1-deficient Tregs in the tumor lost Foxp3 expression compared with Socs1-sufficient (WT) Tregs. Both Socs1-deficient and Socs1-sufficient Tregs maintained Foxp3 expression in the draining LN. Moreover, in the tumor, it was found that IFN-γ was produced from the Foxp3⁻ population in the Socs1-deficient Tregs. It is assumed that these ex-Tregs were retained to produce tumor-suppressive cytokines. For plasticity of Socs1-deficient Tregs, IL-12 produced by APCs was needed to downregulate Foxp3 expression in conjunction with activation of STAT4. IL-6, which is an inflammatory cytokine produced by APCs, has also been reported to downregulate Foxp3 expression (46, 47, 53). Understanding how SOCS1 is associated with the induction and/or maintenance of the conversion of Tregs to Foxp3⁻IFN-γ⁺ ex-Tregs is fundamental to the management of immunological settings for autoimmunity or cancer.

Our studies demonstrated that SOCS1 is indispensable for the suppression of Tregs in vivo, which is consistent with previously reported results from T cell–specific SOCS1-deficient mice (16). Excessive activation of STAT4 in Tregs under the inflammatory conditions in which APCs are highly activated is kept in check by SOCS1 to prevent the deviation of activated Tregs into Th1-like cells. Our findings also raise the intriguing possibility that the upregulation of SOCS1 in Tregs at appropriate levels reinforces Treg functions because SOCS1 may protect Tregs from the harmful effects of inflammatory cytokines that accelerate Treg plasticity. These findings may improve Treg therapy for autoimmune diseases and organ transplantation. Furthermore, the knockdown of SOCS1 in Tregs may be helpful in providing stronger anti-tumor immunity by reprogramming Tregs into effector cells.

We thank T. Takimoto, T. Kobayashi, S. Hidano, and N. Shiino for providing technical assistance and Y. Ushijima for manuscript preparation.

The authors have no financial conflicts of interest.