The Ligase PIAS1 Restricts Natural Regulatory T Cell Differentiation by Epigenetic Repression

Abstract

CD4⁺Foxp3⁺ regulatory T (T_reg) cells are important for maintaining immune tolerance. Understanding the molecular mechanism that regulates T_reg differentiation will facilitate the development of effective therapeutic strategies against autoimmune diseases. We report here that the SUMO E3 ligase PIAS1 restricts the differentiation of natural T_reg cells by maintaining a repressive chromatin state of the Foxp3 promoter. PIAS1 acts by binding to the Foxp3 promoter to recruit DNA methyltransferases and heterochromatin protein 1 for epigenetic modifications. Pias1 deletion caused promoter demethylation, reduced histone H3 methylation at Lys⁹, and enhanced promoter accessibility. Consistently, Pias1^−/− mice displayed an increased natural T_reg cell population and were resistant to the development of experimental autoimmune encephalomyelitis. Our studies have identified an epigenetic mechanism that negatively regulates the differentiation of natural T_reg cells.

Naturally occurring thymus-derived regulatory T (nT_reg) cells play a critical role in the maintenance of self-tolerance and homeostasis within the immune system (1-4). The transcription factor Foxp3 controls the development and function of T_reg cells (5-7). Several regulatory DNA elements within the Foxp3 locus have been suggested to modulate Foxp3 expression, including the promoter region and conserved noncoding sequences (6-9). Although the transcription factors that positively regulate Foxp3 expression are well characterized (10-13), little is known about the negative regulation of Foxp3.

PIAS1 (protein inhibitor of the activated signal transducer and activator of transcription STAT1) is a SUMO E3 ligase that binds to chromatin to repress transcription (14, 15). The recruitment of PIAS1 to chromatin requires it to be phosphorylated on Ser⁹⁰. This is induced by a variety of immune regulatory stimuli, including TCR (T cell receptor) activation (16). PIAS1 was phosphorylated on Ser⁹⁰ in freshly isolated thymocytes and splenocytes (17) (Fig. 1A). There was a small increase in the percentage of thymic single-positive CD4⁺ or CD8⁺ T cells in Pias1^−/− mice, although the total T cell number was not significantly altered (fig. S1, A and B).The frequency of both thymic and splenic CD4⁺Foxp3⁺ nT_reg cells was significantly increased in Pias1^−/− mice (about 80%) (Fig. 1B). In addition, the number of thymic CD4⁺Foxp3⁺ nT_reg cells was also significantly increased in Pias1^−/− mice (about 135%) (Fig. 1B). However, the mean intensity of Foxp3 expression in Foxp3⁺ cells was not altered in Pias1^−/− cells (fig. S1C). Moreover, Pias1 disruption had no significant effect on the in vivo proliferation or survival of T_reg cells (fig. S2).

Fig. 1.

Enhanced CD4⁺ Foxp3⁺ T_reg differentiation in Pias1^−/− mice. (A) Western blot analysis of whole-cell extracts from freshly isolated thymocytes and splenocytes with an antibody specific for Ser⁹⁰-phosphorylated PIAS1, total PIAS1, or tubulin. (B) Cells from thymus or spleen of male (n = 7) wild-type and Pias1^−/− littermates were analyzed by flow cytometry to determine the percentage and the absolute cell numbers of Foxp3⁺CD4⁺ cells (gated on a CD4⁺CD8⁻ population). Similar results were obtained with female mice. (C) Bone marrow was isolated from wild-type and Pias1^−/− littermates and injected into the sublethally Rag1^−/− irradiated recipient mice (n =8). The thymic CD4⁺Foxp3⁺ population was analyzed 4 weeks after reconstitution by flow cytometry. (D) Same as in (C) except that Pias1^−/− or wild-type bone marrow (CD45.2) was mixed with wild-type C57SJL bone marrow (CD45.1) and injected into the Rag1^−/− mice (n = 3 for wild-type and n = 4 for Pias1^−/−). Experiments in (A) to (D) were performed at least three times (n = 3 to 8 for each experiment). P value was determined by unpaired t test.

To test whether PIAS1 directly regulates the intrinsic differentiation potential of nT_reg cells, bone marrow from wild-type and Pias1^−/− mice depleted of CD4⁺ and CD8⁺ T cells was transplanted into sublethally irradiated Rag1^−/− (recombination-activating gene 1) recipient mice, which lack both T and B cells. Pias 1 disruption caused a fourfold increase in the frequency of thymic CD4⁺Foxp3⁺ cells (Fig. 1C). Transplantation experiments were also performed by injecting Rag1^−/− mice with a mixture of bone marrow from wild-type CD45.1⁺ C57SJL mice and either wild-type or Pias1^−/− mice (CD45.2). The percentage of thymic CD4⁺Foxp3⁺ from Rag1^−/− mice reconstituted with Pias1^−/− bone marrow (CD45.2) was significantly higher than that of the wild-type controls, whereas no difference was observed in thymic CD4⁺Foxp3⁺ differentiation of the CD45.1⁺ control Tcells (Fig. 1D). These studies suggest that PIAS1 negatively regulates nT_reg differentiation.

Peripheral naïve CD4⁺ T cells can be differentiated into so-called induced T_reg cells (iT_reg) (18-20). In vitro differentiation studies indicate that transient TCR activation positively regulates iT_reg differentiation, whereas persistent TCR stimulation antagonizes Foxp3 induction (21). Pias 1 disruption enhanced iT_reg differentiation when naïve CD4⁺ Tcells were transiently stimulated by TCR alone (~70% increase), but not under the persistent TCR activation conditions (fig. S3A). In addition, Pias 1 disruption resulted in the inhibition of iT_reg differentiation when cells were exposed to a prolonged transforming growth factor-b (TGF-b) treatment under the persistent TCR-stimulating conditions (fig. S3B), possibly due to the aberrant up-regulation of negative-feedback molecules such as interferon-regulatory factor-1(22) (fig. S3C).

To test the biological significance of PIAS1-mediated regulation of nT_reg in vivo, the response of Pias1^−/− mice to myelin oligodendrocyte glyco-protein (MOG)–induced experimental autoimmune encephalomyelitis (EAE) was examined. Ten days after MOG immunization, an increased population of CD4⁺Foxp3⁺ T_reg cells in the lymph nodes of Pias1^−/− mice was detected, as compared with the wild-type controls (Fig. 2A). The antigen-specific induction of the proinflammatory cytokines inter-feron (IFN)-γ and interleukin (IL)-17 was significantly suppressed, whereas a higher amount of TGF-β, an immunosuppressive cytokine, was detected in Pias1^−/− lymphocytes (fig. S4). Consistently, intracellular staining assays showed a reduced percentage of antigen-specific IFN-γ- or IL-17–producing T cells (Fig. 2B). Furthermore, a defect in lymphocyte proliferation in response to MOG, but not polyclonal CD3-specific antibody stimulation, was observed in Pias1^−/− mice (Fig. 2C). The lack of inhibition on Pias1^−/− lymphocyte proliferation in response to polyclonal stimulation is likely due to the hyperproliferation of Pias1^−/− non-T_reg lymphocytes under such conditions (fig. S5). The differentiation of T helper T_h1orT_h17 cells was not defective in vitro in the absence of PIAS1 (fig. S6), which suggests that PIAS1 does not have an intrinsic ability to regulate the differentiation of T_h1 or T_h 17 cells. Pias1^−/− mice displayed resistance toward the development of EAE (Fig. 2D), and significantly reduced amounts of IFN-γ and IL-17, but increased levels of TGF-β, were detected in the lymph nodes of Pias1^−/− mice (Fig. 2E). These results are consistent with the notion that the increased T_reg cells in Pias1^−/− mice contribute to the observed resistance toward the development of EAE.

Fig. 2.

Pias1−/− mice are resistant to MOG-induced EAE. (A) The percentage of CD4⁺Foxp3⁺ cells in wild-type (LN^+/+) and Pias1^−/− (LN^−/−) lymph node cells 10 days after a single MOG_35-55 injection emulsified in Freund's complete adjuvant (n = 7). (B) Lymphocytes from Pias1−/− mice and wildtype littermates 10 days after MOG_35-55 injection as in (A) were either untreated or treated with 4 μg/ml of MOG_35-55 for 3 days in the presence of brefeldin A during the last 5 hours of culture. IFN-γ–or IL-17–producing CD4+ cells were assayed by intracellular staining followed by flow cytometry. (C) Same as in (B) except that cells were either untreated or treated with MOG_35-55 or CD3-specific antibody for 3 days. Cell proliferation was measured by onecolor cell proliferation kit. (D) Pias1−/− female mice and their wild-type littermates were immunized with MOG_35-55 and pertussis toxin to induce EAE as described in (17). The development of EAE was scored (n = 4). (E) Wildtype and Pias1−/− littermates were immunized as in (D), and lymphocytes were isolated 21 days after the first MOG_35-55 injection and cultured for 3 days in vitro (n = 4). Cytokine production in the cell supernatant was measured by enzyme-linked immunosorbent assay (ELISA). Shown in (A) to (E) is a representative of three independent experiments (n = 3 to 7 for each experiment). Error bars represent SEM. P value was determined by unpaired t test.

PIAS1 contains a chromatin-binding domain that targets PIAS1 to gene promoters to regulate transcription (14-16). We tested the hypothesis that Foxp3 may be a direct PIAS1-target gene by ChIP (chromatin immunoprecipitation) assays. PIAS1 bound to the Foxp3 promoter, but not the two conserved noncoding DNA sequences (CNS1 and CNS2), in thymic CD4⁺CD8⁺ and CD4⁺CD25⁻ T cells (Fig. 3A and fig. S7A). In contrast, the binding of PIAS1 to the Foxp3 promoter was weak in thymic and peripheral CD4⁺CD25⁺ cells, consistent with the reduced expression of PIAS1 in these cells (fig. S7B). The Foxp3 promoter is hyper-methylated in CD4⁺CD25⁻, but hypomethylated in CD4⁺CD25⁺ cells (6, 7, 23-25), which inversely correlates with the amounts of PIAS1 expression (fig. S7B). Bisulfite-sequencing analysis was performed to examine the methylation of the Foxp3 locus in various subpopulations of wild-type and Pias1^−/− cells. Several CpG sites in the Foxp3 promoter were hypermethylated in wild-type thymic CD4⁺CD8⁺,CD4⁺CD25⁻ and splenic CD4⁺CD25⁻ T cells, but were hypomethylated in thymic and splenic CD4⁺CD25⁺ T_reg cells (Fig. 3B and fig. S8A). It is noteworthy that Pias1 disruption caused a significant reduction of DNA methylation of the Foxp3 promoter (Fig. 3B and fig. S8A). In contrast, the removal of PIAS1 showed no effect on the methylation of the heavily methylated CNS2 element (fig. S8B), which is consistent with the lack of significant PIAS1 binding to the CNS2 region (Fig. 3A). Pias1 disruption also showed a modest effect on the demethylation of the Foxp3 promoter in the common lymphoid progenitors and thymic CD4⁻CD8⁻ cells (fig. S9).

Fig. 3.

PIAS1 maintains a repressive chromatin state of the Foxp3 promoter. (A) ChIP assays were performed with freshly sorted CD4⁺CD8⁺ thymocytes from male Pias1^+/+ or Pias1^−/− mice (n = 4), using PIAS1-specfic antibody or IgG. Bound DNA was quantified by quantitative real-time fluorescence polymerase chain reaction (QPCR), with specific primers against the various regions of the Foxp3 locus, and normalized with the input DNA. (B) Methylation analysis of the Foxp3 promoter was performed by bisulfite conversion of genomic DNA from various sorted T cell populations of wild-type and Pias1^−/− male mice (n = 4). The x axis represents the positions of the CpG sites relative to the transcription start site (+1) in the Foxp3 gene; the y axis represents the percentage. (C) ChIP assays were performed the same way as in (A) except that antibodies against trimethylated H3K9, histone H3 trimethylated at Lys²⁷ (H3K27), or acetylated histone H3 (AceH3) were used, and the primers against the Foxp3 promoter region were used. (D) REA assays were performed with thymic CD4⁺CD8⁺ or splenic CD4⁺CD25⁻ T cells. The data were quantified by QPCR and expressed as a ratio of digestion at the Foxp3 promoter or CNS2 region to digestion at a nondigestible CNS1 region of the Foxp3 locus. Shown in (A) to (D) is a representative of three independent experiments (n = 4 to 6 for each experiment). Error bars represent SEM. P value was determined by unpaired t test.

To examine whether PIAS1 also plays a role in the methylation of other gene promoters, we performed similar studies on CD25, a key regulator of T_reg cell differentiation (4, 6). ChIP studies showed that Cd25 is a PIAS1-target gene (fig. S10A). Pias1 disruption resulted in the hypomethylation of the Cd25 gene promoter in hematopoietic stem cells and thymic CD4⁻CD8⁻ cells (fig. S10B). Consistently, an increased frequency of CD4⁺CD25⁺ T cells was observed in Pias1^−/− thymus and spleen (fig. S10C), as well as in transplantation studies (fig. S10, D and E).

We also examined the chromatin status of the Foxp3 promoter by analyzing histone modifications. ChIP analysis showed a reduction of the repressive histone H3 methylation at Lys⁹ (H3K9) methylation code in Pias1^−/− CD4⁺CD8⁺ thymocytes (Fig. 3C). In addition, restriction enzyme accessibility analysis (REA) on the Foxp3 locus (23) showed the enhanced accessibility of the Foxp3 promoter, but not the PIAS1 nonbinding CNS2 region in Pias1^−/− thymic CD4⁺CD8⁺ and splenic CD4⁺CD25⁻ T cells (Fig. 3D), which support a more open chromatin structure of the Foxp3 promoter in Pias1^−/− T cells. Furthermore, Pias1 disruption resulted in the enhanced binding of STAT5, a key transcription factor involved in Foxp3 induction (10-12), to the Foxp3 promoter (fig. S11A). The amount of STAT5 protein or STAT5 phosphorylation was not affected by Pias1 disruption (fig. S11, B and C). A significant increase of nuclear factor of activated Tcells (NFAT) binding to the Foxp3 promoter was also observed in Pias1^−/− CD4⁺CD8⁺ thymocytes (fig. S11D). Consistently, Pias1 disruption resulted in the increased frequency of Foxp3⁺ CD4⁺CD8⁺ thymocytes (fig. S12).

We examined whether PIAS1 may maintain a repressive chromatin structure through the recruitment of DNA methyltransferases (DNMTs), which promote DNA methylation of chromatin (26). PIAS1 was shown to interact with DNMT3B in a yeast two-hybrid screen and in 293T cells upon overexpression (27, 28). DNMT3A and DNMT3B, but not DNMT1, were present in PIAS1 immunoprecipitates from thymocytes or splenic CD4⁺CD25⁻ cells (Fig. 4A). Furthermore, sequential ChIP studies indicated that PIAS1 forms a complex with DNMT3A and DNMT3B on the Foxp3 gene promoter (Fig. 4B). DNMT3B and DNMT3A bound to both the Foxp3 promoter and the CNS2 region, but not the CNS1 element, in the wild-type CD4⁺CD8⁺ thymocytes (Fig. 4C). Pias1 deletion completely abolished the binding of both DNMT3A and DNMT3B to the Foxp3 promoter (Fig. 4C), but not the CNS2 region. The levels of DNMT expression were not affected by Pias1 disruption (fig. S13). In addition, a modest binding of DNMT1 to the Foxp3 promoter was observed, which was inhibited by Pias1 disruption (fig. S14). The PIAS1-dependent recruitment of DNMTs to the Foxp3 promoter was also observed in splenic CD4⁺CD25⁻ cells (fig. S15), but not in CD4⁺CD25⁺ T_reg cells (fig. S16). Collectively, these results indicate that PIAS1 is associated with DNMT3A and DNMT3B in T cells and is required for their recruitment to the Foxp3 gene promoter.

Fig. 4.

PIAS1 is required for the promoter recruitment of DNMT3A and DNMT3B. (A) Coimmunoprecipitation assays. Protein extracts from wild-type thymocytes or splenic CD4⁺CD25⁻ T cells were subjected to immunoprecipitation with PIAS1-specific antibody or IgG, followed by immunoblotting with the indicated antibodies. (B) ChIP assays were performed with wild-type thymocytes by using PIAS1-specific antibody or IgG. The presence of the Foxp3 promoter region in the precipitates was quantified by QPCR. In the re-ChIP experiments, PIAS1-specific antibody precipitates were released, reimmunoprecipitated with an antibody against DNMT3A or DNMT3B, and analyzed for the presence of the Foxp3 promoter sequence. (C) ChIP assays were performed with freshly sorted CD4⁺CD8⁺ thymocytes from male Pias1^+/+ or Pias1^−/− mice (n = 4), using an antibody against DNMT3A or DNMT3B. Bound DNA was quantified by QPCR with the specific primers against Foxp3 promoter, CNS1 or CNS2 region. (D) Methylation analysis of the Foxp3 promoter was performed by bisulfite conversion of genomic DNA from thymocytes of Rag1^−/− mice reconstituted with either control GFP (Con) or Cre-GFP (Cre) retrovirus-infected Dnmt3a^2lox/2lox/Dnmt3b^2lox/2lox bone marrow. The x axis represents the positions of the CpG sites relative to the transcription start site (+1) in the Foxp3 gene; the y axis represents the percentage. Shown in (A) to (D) is a representative of three independent experiments (n = 4 to 6 for each experiment). Error bars represent SEM. P value was determined by unpaired t test.

To test if DNMT3A and DNMT3B play a role in the Foxp3 promoter methylation, we performed transplantation experiments using mice in which the functional domains of both Dnmt3a and Dnmt3b genes are flanked by two loxP sites (Dnmt3a^2lox/2lox/Dnmt3b^2lox/2lox) (29, 30). Bone marrow cells from Dnmt3a^2lox/2lox/Dnmt3b^2lox/2lox mice were infected with green fluorescent protein (GFP) or Cre-GFP retrovirus, and the deletion of Dnmt3a and Dnmt3b was confirmed (fig. S17A). Sorted GFP⁺ bone marrow was transplanted into sublethally irradiated Rag1^−/− mice. Decreased Foxp3 promoter methylation was observed in thymocytes from the Rag1^−/− mice reconstituted with the Cre retrovirus–transduced bone marrow (Fig. 4D). ChIP assays indicated that the decreased DNMT3A and DNMT3B expression by the Cre transduction had no substantial effect on the binding of PIAS1 to the Foxp3 promoter (fig. S17B).

Heterochromatin protein 1 (HP1) plays an important role in promoting H3K9 methylation and is known to interact with DNMTs (31). HP1γ was strongly associated with the Foxp3 promoter in wild-type, but not Pias1^−/−, thymocytes (fig. S18). The PIAS1-mediated epigenetic gene regulation is selective, because the chromatin status of genes such as Ctla4 was not affected by Pias1 disruption (fig. S19).

Our studies have identified an epigenetic control mechanism in the negative regulation of Foxp3⁺ nT_reg differentiation. PIAS1 acts by maintaining a repressive chromatin state of the Foxp3 promoter, at least partly through the recruitment of DNMTs and HP1 to promote epigenetic modifications (fig. S20). Pias1 disruption results in the formation of a permissive chromatin structure of the Foxp3 promoter and enhanced promoter accessibility to transcription factors such as STAT5 and NFAT, which lead to the increased probability that precursor cells will differentiate into Foxp3⁺ T_reg cells (fig. S20). The physiological role of PIAS1 in the regulation of Foxp3 gene is supported by the observed increase of the Foxp3⁺ nT_reg population in Pias1^−/− mice and the resistance of PIAS1^−/− mice toward the development of EAE. Thus, the PIAS1 pathway may represent a therapeutic target for the treatment of auto-immune diseases. DNMTs have no intrinsic sequence specificity (26). Our finding that PIAS1 regulates the binding of DNMTs to the Foxp3 promoter, but not the CNS2 element, suggests that PIAS1 may be an important cofactor that confers specificity in the DNMT-mediated chromatin methylation. The PIAS1-mediated DNA methylation and histone modifications may serve as a fine-tuning mechanism in the control of epigenetic modifications during Tcell differentiation.