TGF-β–Mediated Foxp3 Gene Expression Is Cooperatively Regulated by Stat5, Creb, and AP-1 through CNS2

Regulatory T cells (Treg) are important in preventing autoimmune disease and other forms of immunopathology (1–3), their development and function being regulated by a transcription factor Foxp3 (1). Mutations within mouse Foxp3 (4) (scurfy mouse) or human FOXP3 (5) (Immune dysregulation polyendocrinopathy, enteropathy, X chromosome-linked syndrome) give rise to multiorgan immunopathology. To better understand the role of the Foxp3 gene in Treg development and function (6–10), the molecular mechanisms of Foxp3 gene expression have been extensively studied, yet remain incompletely understood.

Thus far, a promoter, two enhancers, and conserved noncoding sequence (CNS) 3 have been characterized as regulatory regions controlling induction and maintenance of mouse Foxp3 gene expression (11–13). The two enhancers are located in CNS1 (Enhancer 1) (12) and CNS2 (Enhancer 2) (11) in the intron of the Foxp3 gene, and CNS3 is located downstream of exon 1 (13). Many transcription factors with potential to bind the Foxp3 promoter have been identified (14–22), but promoter activity is critically dependent on enhancer function (12). No enhancer activity has been detected within the CNS3 region, although an NF-κB family c-Rel binds to this region (13). CNS3 in conjunction with c-Rel appears to have a unique role in Foxp3 gene expression, possibly opening up the Foxp3 locus in natural Treg (nTreg) precursor cells.

We previously identified Enhancer 1 in CNS1 and demonstrated that its activity is regulated by transcription factors Smad3 and NFAT (12). Subsequently, AP-1 and the retinoic acid receptor were also shown to control Enhancer 1 activity (23). CNS1-deficient mice have impaired induced Treg (iTreg) development (13), a process normally regulated by signaling through both TGF-β receptor and TCR (24, 25).

The presence of another enhancer was reported by Kim and Leonard (11) and ourselves (12) and termed Enhancer 2 (12, 22). Compared with Enhancer 1, the regulation and function of Enhancer 2 is poorly understood. This enhancer is located in a region of CNS2 that contains highly methylated CpG sequences (11) in Foxp3⁻ T cells, but is demethylated in Treg (Treg-specific demethylated region). Demethylation of CpG in CNS2 by an inhibitor (5-aza-cytidine) results in upregulation of Foxp3 gene expression (11). CNS2 seems to play two distinct roles as both a positive (enhancer) as well as negative (methylation) regulator. Transcription factors Creb (11), Foxp3-Runx1-CBFβ complex (13), NF-κB (26), and Ets-1 (26) are known to bind to CNS2, such binding being dependent on CpG demethylation. Demethylation in the CNS2 region is mediated by TGF-β (11), yet the relevant TGF-β response element has not been identified. Stat5 binding around the CNS2 region has been observed by chromatin immunoprecipitation (ChIP) assay, and three potential Stat5 binding sites have been identified (27). A role for Stat5 is implicated in Foxp3 expression, as Foxp3⁺ Treg are substantially reduced in Stat5- and IL-2Rγ–deficient mice (27, 28). This has led us to hypothesize that Stat5 activated via IL-2R signaling acts on Enhancer 2. If that were the case, it is not clear which Stat5 elements would be functional, whether they would indeed influence the enhancer activity, and whether they could function with methylated binding sites.

To determine the contribution of CNS2 in regulation of Foxp3 gene expression, we have first investigated enhancer activity using luciferase assays performed in the absence of CpG methylation to exclude any inhibitory effects of that methylation. The determined basal enhancer core region encodes a 181-bp sequence and is located in the histone H4 highly acetylated region of CNS2. The previously identified Creb binding site (11) and the potential Stat binding site (27) in CNS2 are now mapped into this 181-bp enhancer core sequence with the newly identified AP-1 binding site. We establish that these transcription factors cooperate with each other in regulating enhancer activity. We then evaluated the extent to which the transcription factors could act on methylated binding sites. Creb has previously been characterized as methylation sensitive for its action in CNS2 (11). We show that AP-1 activity is also methylation sensitive, but that Stat5 is not. Importantly, the demethylation process in CNS2 is regulated by TGF-β, and we demonstrate in this study that Stat5 activation is also regulated by TGF-β in Treg. Finally, the relative contributions of these transcription factors were defined using ChIP assays performed during TGF-β–mediated Treg induction. The results suggest that Stat5 is a key regulator for opening up the CNS2 region, whereas AP-1 and Creb maintain Enhancer 2 activity.

EL4 LAF (12) subclone BO2 cells were cultured in IMDM with l-glutamine, 25 mM HEPES (Cellgro), and 5% FBS. T cells were cultured in RPMI 1640 with l-glutamine (Cellgro) and 10% FBS. CD4⁺CD25⁺ and CD4⁺CD25⁻ T cells were isolated from mouse spleen by EasySep Mouse CD4⁺CD25⁺ Regulatory T Cell Isolation Kit (STEMCELL Technologies) and by a cell sorter using anti-CD4 (RM4-5; eBioscience) and anti-CD25 (PC61.5; eBioscience) from preisolated CD4⁺ T cells. Isolated CD4⁺CD25⁺ T cells were analyzed by an internal Foxp3 staining kit (eBioscience). The purity of CD4⁺CD25⁺ T cells was >85 (isolation kit) and 95% (cell sorter) and for CD4⁺CD25⁻ T cells was >90 (isolation kit) and 95% (cell sorter), respectively. To generate iTreg, CD4⁺CD25⁻ T cells were cultured with plate-coated anti-CD3 (KT3; 5 μg/ml in PBS), anti-CD28 (1 μg/ml), recombinant human TGF-β1 (5 ng/ml; PeproTech), and mouse IL-2 (20 ng/ml; PeproTech). Pharmacological inhibitors JNK Inhibitor II (0–20 μM) (SP600125; LC Laboratories), Smad3 inhibitor (10 μM; SIS3; EMD Millipore), STAT5 inhibitor (200 μM; 573180; EMD Millipore), and JAK inhibitor I (10 μM; 420099; EMD Millipore) were used where indicated.

RNA was isolated from T cells by using the RNeasy mini kit (Qiagen), and cDNA was prepared with iScript cDNA synthesis kit containing a random and oligo(dT) primer mixture (Bio-Rad). Quantitative RT-PCR was performed by using SsoAdvanced SYBR Green supermix (Bio-Rad). Socs3 expression level was normalized by 18S rRNA level. The primer sequences used were as follows: Socs3 Forward, 5′-TCAAGACCTTCAGCTCCAAAAG-3′ and Socs3 Reverse, 5′-CCCCCAGAATAGATGTAGTAAGCTC-3′; and 18S rRNA Forward, 5′-CTTAGAGGGACAAGTGGCG-3′ and 18S rRNA Reverse: 5′-ACGCTGAGCCAGTCAGTGTA-3′.

Transfectants were generated using EL4 cells. FLAG–c-Fos, Stat5a, Stat5b, and IL-2Rγ dominant-negative (DN) cDNA were cloned into the downstream of the EF-1α promoter in pMF-neo vector. IL-2Rγ DN cDNA encoded the 1–290 aa sequence based on the human IL-2Rγ DN sequence (29). To construct Luciferase reporter plasmid, pGL4 plasmid (Promega) was used. Wild-type (Wt) and mutant fragments of the CNS2 were generated by PCR and cloned into the downstream of the luciferase gene of the Foxp3 promoter (1.88 kb) reporter plasmid described previously (12). The luciferase reporter plasmids with AP-1 and Stat5 response elements and the AdML minimal promoter were generated by the procedure described previously (12).

A total of 1 × 10⁷ EL4 cells was transfected by Gene Pulser Xcell (Bio-Rad) with 8 μg luciferase reporter plasmids and 2 μg phRL-TK (Promega) as an internal control plasmid and cultured for 24 h in six-well plates. If required, cells were activated with plate-coated anti-CD3 (5 μg/ml in PBS) and anti-CD28 (1 μg/ml) with or without TGF-β (5 ng/ml). Cells were harvested, and luciferase activities were analyzed with the Dual-Luciferase Reporter Assay System (Promega). For cotransfection experiments, 5 μg Stat5a and/or Stat5b expression plasmids were transfected with 5 μg luciferase reporter plasmids. Total DNA amount was adjusted with the empty pMF-neo vector. Luciferase assays were repeated more than three times.

EMSA, competition assay, supershift EMSA, and preparation of nuclear extracts were described previously (12). Reaction buffer for Stat5 EMSA was [2 μg poly(deoxyinosinic-deoxycytidylic) acid, 12 mM HEPES (pH 7.9), 20 mM KCl, 0.5 mM EDTA, 0.5 mM DTT, and 12% glycerol]. Supershift EMSA was performed with anti-Stat5 Ab (sc-835; Santa Cruz Biotechnology). EMSA probe sequence used were as follows: Stat5 Wt, 5′-CTAGCCACTTCTCGGAACGAAACCTG-3′ and Stat5 Mutant, 5′-CTAGCCACTTGGATCCACGAAACCTG-3′.

DNA fragments with Stat5 or AP-1 binding sequence were methylated with CpG Methyltransferase (M.SssI; New England Biolabs). The AP-1 and Stat5-I fragment was treated by M.SssI with or without 160 μM substrate S-adenosylmethionine (SAM; New England Biolabs) in 50 mM NaCl, 10 mM Tris-HCl (pH 8), 10 mM EDTA, and 1 mM DTT for 2 h at 37°C. DNA methylation was monitored with pGL4 plasmid and described in Supplemental Fig. 3. After purification by NucleoSpin (Clontech), the DNA fragments were ligated into the XhoI-EcoRI–digested pGL4 carrying the AdML minimal promoter, the ligation mixture was purified using the NucleoSpin (Clontech), and 5-μg ligated samples were used for luciferase assay (Supplemental Fig. 3).

ChIP assay was performed using nonstimulated, anti-CD3 + anti-CD28–stimulated, TGF-β + anti-CD3 + anti-CD28–stimulated CD4⁺CD25⁻ T cells, and CD4⁺CD25⁺Foxp3⁺ Treg as described previously (12). Polyclonal anti–acetyl histone H4 (06-866; Millipore), anti–c-Fos (sc-52; Santa Cruz Biotechnology), anti-Stat5 (sc-835; Santa Cruz Biotechnology), anti-Creb (sc-186; Santa Cruz Biotechnology), or control rabbit IgG (sc-2027; Santa Cruz Biotechnology) were used. The ChIP primer sequences used were as follows: A Forward, 5′-AGGTATTGGTGGAAAGTGGGCTA-3′ and A Reverse, 5′-ACCCATGCACAGAGGGAATGGAAT-3′; B Forward, 5′-AAGGGCAAACTGGGCTCAGAGATGA-3′ and B Reverse, 5′-TTGCCAAAGTCAAGGTCCCACTGA-3′; C Forward, 5′-TA GCGGACCAAAGAACACTAGTAAC-3′ and C Reverse, 5′-CATCTTGCAAAGTTCGTACATACAG-3′; D Forward, 5′-ATCAATACACACAGTAAGAAGGTGGAT-3′ and D Reverse, 5′-CCT TGAGTCTTCAAATGCTGTAGTA-3′; E Forward, 5′-GGCCTGCCTAATACTCACCA-3′ and E Reverse, 5′-GCTAGTCTATCCTGTAGCCGGA-3′; F Forward, 5′-CCACTAGACAGACCATA TCCAATTC-3′ and F Reverse, 5′-GTGCTCTTTATGTTTGGTCAGAACT-3′; G Forward, 5′-CCTCTACTGCTTACCTTTGCATTTA-3′ and G Reverse, 5′-CTACAACAGGAGCATTACTGGAGAT-3′; H Forward, 5′-AAGGAGTTCCAGACTACAAGAAACC-3′ and H Reverse: 5′-TTCAGGATCCTATACTTCCAGTCAG-3′; I Forward, 5′-CAGACTC CTCTTTTCATGCTAATGT-3′ and I Reverse, 5′-CATCTTCTTTGCTATCATGTCTGG-3′; c-Fos Forward, 5′-AGAACTTGGGTTTTGCATGG-3′ and c-Fos Reverse, 5′-GGAAGAGAAGGGGGCAGATA-3′; and Creb Forward, 5′-TACAGGATAGACTAGCCACTT-3′ and Creb Reverse, 5′-AATATGTTTTCCTATCGGGGT-3′. Primers A to I were used for Fig. 1, c-Fos primers were used for Fig. 4, and c-Fos primers and Creb primers were used for Figs. 7 and 8.

A 102-bp DNA fragment containing the three copies of 32 bp with AP-1 (Wt, Fig. 4A) or mutant (double mutants, Fig. 4A) site was amplified by PCR using biotinylated primers. Purified probes (500 ng) were mixed with nuclear extract (50 μg) in 400 μl binding buffer (60 mM KCl, 12 mM HEPES [pH 7.9], 4 mM Tris-HCl [pH 8], 0.1 mM EDTA, and 5% glycerol) containing 1 mM DTT and 10 μg poly(deoxyinosinic-deoxycytidylic) acid at 4°C for 2 h. Then, DNA–protein complexes were collected with 50 μl precleared streptavidin-agarose beads (Invitrogen) and analyzed by immunoblotting.

Immunoblotting was performed with anti–c-Jun (sc-1694; Santa Cruz Biotechnology), anti-Stat5 (sc-835; Santa Cruz Biotechnology), anti–phospho-Stat5 (C11C5; Cell Signaling Technology), and anti-Flag Ab (Sigma-Aldrich). Ab binding was detected using Luminata Forte Western HRP Substrate (Millipore).

Significance was determined with Student t test.

To determine the contribution of Enhancer 2 to regulating Foxp3 gene expression, we first analyzed histone H4 acetylation around CNS2 as a marker for open chromatin using a ChIP assay. H4 molecules in the 5′ half of the CNS2 were highly acetylated in nTreg and TGF-β–mediated iTreg (for 48 h) but not CD3-activated (activated by anti-CD3 + anti-CD28) T cells (Fig. 1A, 1B), suggesting that a regulatory region is located in this region. To determine the basal enhancer core region, luciferase assays were performed in TGF-β–treated CD3-activated EL4 cells using luciferase reporter plasmids containing a DNA fragment from within CNS2 (+4082 to +5216) and a series of deletion mutants (Fig. 1B–D). Foxp3 expression is induced in the EL4 lines under these conditions just as in CD4⁺CD25⁻ T cells (12). Τhis same reporter system has been used previously to identify transcription factors binding to the Foxp3 promoter (18, 21) and the Enhancer 1 (12, 23). Because the CpG sequences in these fragments were unmethylated, we could eliminate any inhibitory effects due to methylation. Strong luciferase activity was detected with the Wt and three deletion mutants from the 3′ end (Fig. 1C), but was lost with the +4281 mutant. The same assay was performed with deletion mutants from the 5′ end (Fig. 1D), and luciferase activity was undetectable with the +4574 mutant. Taken together, these results indicate that the enhancer core is located within the 181-bp sequence between +4307 to +4488 in the histone H4 highly acetylated region (Fig. 1A, 1B).

To identify transcription factors regulating enhancer activity, a 210-bp sequence containing the 181-bp enhancer core sequence (Fig. 2) was further analyzed by luciferase assays using Wt and mutants of the fragment (Figs. 2, 3A). Strong enhancer activity was detected with the Wt fragment in CD3-activated cells both without and with TGF-β, but not in nonactivated cells (Fig. 3B, 3C), indicating that the enhancer activity is regulated by transcription factors induced in activated T cells. A 26-bp deletion from the 3′-end substantially reduced enhancer activity (Fig. 3B, +4471). This deleted sequence contains the previously reported Creb binding site (11). Creb binding appears relevant to enhancer activity, as mutation of the Creb binding sequence (Mutant 1, Figs. 2, 3A) in the Wt fragment virtually abrogated enhancer activity (Fig. 3D). The strong enhancer activity observed with the Wt fragment also disappeared by a 59-bp deletion from the 5′-end (+4347, Fig. 3A, 3C), implicating the binding of further unknown transcription factors to this 59-bp sequence. Although the 59-bp shorter fragment (+4347, Fig. 3A, 3C) lost strong enhancer activity, it still contains the Creb binding site. Consequently, it would appear that Creb and the unknown transcription factors cooperate to regulate this enhancer. A partial invert repeat sequence (GGACGT–ACATCC) was detected in the deleted sequence (Fig. 2), and enhancer activity was reduced by mutations in these repeats (Mutant 2, Figs. 2, 3A, 3E), suggesting that the additional transcription factors bind to these sequences.

Stat5 binding was previously detected by ChIP assay at the 5′-flanking and intron regions of the Foxp3 gene (27, 30, 31) and including CNS2 (27, 30). One of the binding sites was located within the 181-bp enhancer core. The reduction of enhancer activity with mutation of this site (Mutant 3, Figs. 2, 3A) was observed in CD3-activated cells irrespective of whether TGF-β was present (Fig. 3F). Notably, the reduction observed in the presence of TGF-β appears larger than in its absence (Fig. 3G), suggesting that TGF-β signaling regulates enhancer activity through the potential Stat5 binding site (supported by further data shown later). In summary, we have identified three transcriptional regulatory elements in the Enhancer 2 core.

As described above, unknown transcription factor(s) bind to the invert repeat sequences near the 5′ end of the enhancer core. As shown in Fig. 3B, the activity of these transcription factors was difficult to detect in the absence of the Creb binding site on Enhancer 2. We therefore generated a luciferase reporter plasmid using a minimal promoter with three copies of the 32-bp sequence containing the invert repeat sequences from Enhancer 2 (Fig. 4A). We detected strong luciferase activity with this system in CD3-activated cells in the presence and absence of TGF-β (Fig. 4B), such activity disappearing with mutations in each one of the invert repeat sequences or in both sequences (Fig. 4A, 4B). This suggests that transcription factors induced by CD3 activation bind to the invert repeat sequences and that both repeating units are required for factor binding. The 8-bp sequence containing the invert repeat sequences (GGACGTCA and CCACATCC) appears remarkably similar to the Creb binding sequence (TGACGTCA) in Enhancer 2 (Fig. 2), yet Creb binding could not be detected. We could, however, demonstrate binding of another induced transcription factor, AP-1, which is closely related to the Creb family. AP-1 is a heterodimeric transcription factor composed of c-Jun and c-Fos family members. We detected c-Jun binding to the invert repeat sequences by a DNA pulldown assay (Fig. 4C) using nuclear extracts from CD3-activated cells both in the presence and absence of TGF-β, but not in nonactivated cells, such binding being reduced by mutation of the invert repeat sequences. To assess c-Fos binding, we used a nuclear extract from a FLAG–c-Fos EL4 transfectant, as anti–c-Fos was found to have a high background of nonspecific binding (Fig. 4C). The luciferase activity generated with the Wt reporter plasmid was reduced by an inhibitor of AP-1 activation (JNK inhibitor II) in a dose-dependent manner (Fig. 4D). Inhibitory activity of the JNK inhibitor II under these conditions was also confirmed using AP-1–mediated IL-2 transcription (Supplemental Fig. 1). Importantly, reduction of iTreg induction with the same inhibitor was previously used to study the Foxp3 Enhancer 1 (23). c-Fos binding to Enhancer 2 was also detected by ChIP assay in TGF-β–induced Treg (Fig. 4E) and nTreg (Fig. 4F) but not in CD3-activated T cells (Fig. 4E). [We used anti–c-Fos to show AP-1 (c-Fos/c-Jun heterodimer) binding to this region because c-Jun homodimers and c-Jun/c-Fos heterodimers bind to the AP-1 site, whereas c-Fos alone fails to homodimerize and bind (32).] Overall, these results suggest that AP-1 binding to Enhancer 2 in Treg and regulates Foxp3 gene expression.

We highlighted three potential Stat5 binding sites within the CNS2 region (Stat5-I, Stat5-II, and Stat5-III in Fig. 5A). Although Stat 5 binding was detected around these sites by ChIP assay (27), the exact Stat binding sites could not be identified by this assay. We therefore investigated Stat5 activities with these sites. The Stat5-I site is located in the Enhancer 2 core, and this site seems to be a regulatory element for the enhancer (Fig. 3F, 3G). To analyze the activity of these Stat5 binding sites, we applied the same strategy as for the AP-1. Luciferase reporter plasmids were generated using six copies of 22-bp DNA sequences containing the potential Stat5 binding sites (Stat5-I to III) (Fig. 5A) and the minimal promoter. Strong luciferase activity was detected with the Stat5-I site in CD3-activated cells with TGF-β. We found activity using the Stat5-II and Stat5-III sites, but these were markedly weaker than the Stat5-I (Fig. 5B). Considering the strong activity and position (between AP-1 and Creb binding sites) of the Stat5-I site, this would seem to be a key regulatory element in CNS2. To examine Stat5 binding to the Stat5-I site, EMSAs were performed. A shifted band was detected with the 26-bp Stat5-I probe, and was lost with the unlabeled Wt competitor but not a mutant (Fig. 5C). Stat5 binding to this site was confirmed by presence of a supershifted band with anti-Stat5 Ab (Fig. 5D).

There are two Stat5 molecules (Stat5a and Stat5b) that are 91% identical but play different roles in some gene expression systems (33). Using the Stat5-I reporter plasmids coexpressed with Stat5a and/or Stat5b expression plasmids, we observed that luciferase activity was upregulated with Stat5b, but not with the Stat5a (Fig. 5E). In contrast, enhanced luciferase activity with both the expression plasmids was observed when the Stat5 binding site was linked to the Socs3 promoter (34, 35) (Supplemental Fig. 2). These results suggest that the Enhancer 2 activity is regulated by Stat5b through the Stat5-I site in this enhancer.

As shown in Fig. 3G, the activity of the Stat5-I element in Enhancer 2 seems to be regulated by TGF-β. This finding was confirmed by a luciferase assay in which Stat5 activity was enhanced with TGF-β (Fig. 5B). To investigate this further, EMSA was performed using the Stat5-I probe with nuclear extracts from nonactivated cells and CD3-activated cells without or with TGF-β. Strong Stat5 binding was observed with only the nuclear extract from CD3-activated cells with TGF-β (Fig. 6A). We sought evidence of Stat5 phosphorylation (p-Stat5) in EL4 and CD4⁺CD25⁻ T cells. In both cells, p-Stat5 levels were elevated in the CD3-activated cells with TGF-β, but not in the absence of TGF-β (Fig. 6B). Because Stat5 activation is known to regulate JAK molecules associated with the IL-2Rγ chain (36, 37), we investigated whether Stat5 activation operates through this IL-2Rγ–JAK pathway by using an expression plasmid encoding an IL-2Rγ DN inhibitor (DN–IL-2Rγ) (lack of the cytoplasmic region containing the JAK binding site). The luciferase activity generated by the Stat5-I reporter plasmid was substantially reduced by coexpression of the DN–IL-2Rγ (Fig. 6C), suggesting that Stat5 activation in the TGF-β–conditioned cells is also regulated by the JAK/STAT pathway via IL-2Rγ. It is well established that Stat5 activation through JAK/STAT pathway is blocked by Socs3 (36, 37). We therefore analyzed Socs3 expression in CD4⁺ T cells. Socs3 mRNA expression was reduced in CD3-activated CD4⁺CD25⁻ T cells exposed to TGF-β, and this reduction was blocked by the Smad3 inhibitor SIS3 (Fig. 6D), suggesting that Socs3 expression in T cells is inhibited by the TGF-β–mediated Smad3 pathway. Put another way, Foxp3 Enhancer 2 activity is upregulated by TGF-β by increasing Stat5 activation following downmodulation of Socs3 expression.

CpG sequences in CNS2 are highly methylated in the Foxp3⁻ T cells but demethylated in Treg. Creb, AP-1, and Stat5-I binding sites in CNS2 contain CpG sequences (Fig. 2), and dysfunction of Creb with the methylated binding site was previously demonstrated (11). To investigate the contributions of the key transcription factors for Enhancer 2 activity in methylated CNS2, we analyzed the methylation sensitivity of Stat5 and AP-1 for their binding sites. Three copies of AP-1 or six copies of Stat5-I unit fragment (shown in Figs. 4A and 5A) were treated with CpG methyltransferase with or without substrate (SAM). CpG methylation was confirmed by digestion of methylation-sensitive and -insensitive restriction enzymes and shown in Supplemental Fig. 3A. The methylated and unmethylated fragments were ligated at the 5′ of the minimal promoter in the luciferase reporter plasmid (Supplemental Fig. 3A). Luciferase assays with the purified ligation samples suggested that AP-1 (Fig. 7A) was methylation sensitive but that Stat5 (Fig. 7B) was insensitive. Stat5 binding to the methylated Stat5-I probe was confirmed by EMSA (Supplemental Fig. 3B).

Because Stat5 can function with the methylated Enhancer 2, and because Stat5 activity was dependent on TGF-β, it seems likely that Stat5 regulates the opening up of the Enhancer 2 region during TGF-β–mediated Treg induction. To investigate this possibility, Stat5, AP-1, and Creb binding to Enhancer 2 were analyzed by ChIP assays. Treg were induced by TGF-β and Foxp3⁺ Treg population was detected by FACS at different time points (0 to 72 h) (Fig. 7C). Histone H4 acetylation was also analyzed by ChIP as a marker for open chromatin. Foxp3 induction, histone H4 acetylation, and AP-1 and Creb binding, but not Stat5 binding, were well correlated (Fig. 7C, 7D). Abundant Stat5 binding to this region was detected at 18 h after stimulation and then gradually diminished (Fig. 7D). Conversely, AP-1 and Creb binding were not detected at 18 h, but could be seen to increase after 48 h. Taken together, these results suggest that Stat5 opens up Enhancer 2 in conjunction with TGF-β signaling and that this enhancer activity is then predominantly maintained by AP-1 and Creb with TCR signaling.

We next investigated whether this enhancer opens without Stat5. Stat5 binding to the enhancer at 24 h after stimulation was blocked by a Stat5 inhibitor (EMD Millipore) (38) (Fig. 8A) and a Stat5 activation inhibitor (JAK inhibitor I; EMD Millipore) (Fig. 8B). Acetylation of histone H4 in this region was also reduced with these two inhibitors (Fig. 8A, 8B), suggesting inhibition of open chromatin remodeling by these inhibitors. At 48 h of stimulation, a large proportion of cells died in the presence of the Stat5 inhibitor, presumably due to its toxicity. In contrast, the JAK inhibitor–treated cells survived at similar levels to the control, and Foxp3 induction was observed to reduce (Fig. 8C). Acetylation of histone H4 in the Enhancer 2 region was inhibited, and binding of Stat5, AP-1, and Creb to this region was not observed with the JAK inhibitor (Fig. 8C). Because radical changes occur in these cells treated with these inhibitors, and given that Stat5 regulates expression of many important genes (not only Foxp3), it is difficult to conclude a clear-cut role of Stat5 in opening up the enhancer in CNS2 without a much more detailed investigation. However, our data in Figs. 7 and 8 do suggest that Stat5 plays a key role in initiating the enhancer activity in CNS2.

CNS2 in the Foxp3 gene is a unique transcriptional regulatory region that functions both as a positive (enhancer) and negative (CpG methylation) regulator. We have investigated the role of CNS2 by separating these two opposing functions. We have identified the basal Enhancer 2 core sequence without CpG methylation and demonstrated that transcription factors AP-1, Stat5, and Creb cooperate to enable enhancer activity. Previously, several transcription factors including Creb were identified as binding to CNS2, but all were methylation sensitive (11, 13, 26). We have shown in this study that AP-1 function is also methylation sensitive. Because the CNS2 region is highly methylated in Foxp3⁻ cells, it was unclear how these factors could upregulate Foxp3 gene expression through this region. In addition, although demethylation of the CpG sequences in the CNS2 region is regulated by TGF-β signaling (11), a TGF-β response element had not previously been identified in this region. We have demonstrated in this study that the TGF-β–sensitive component in CNS2 is determined through a Stat5 activation pathway via the IL-2Rγ, and importantly, Stat5 can function with the methylated binding site in CNS2. Taken together, these findings suggest that Stat5 is a key player for opening up the CNS2 region. Indeed, Stat5 bound abundantly to CNS2 at an early stage of TGF-β–mediated Treg induction, whereas AP-1 and Creb were observed to bind much later, at a point when we presume demethylation had occurred. Two transcription factors, c-Rel and Smad3, have been previously implicated as key regulators in opening up the Foxp3 locus through CNS3 (c-Rel) and CNS1 (Smad3). Given our new findings, we propose that transcription factors Smad3 (CNS1), Stat5 (CNS2), and c-Rel (CNS3) function as initiators for Foxp3 induction by opening up the three regulatory regions, each CNS being initiated by distinct signaling processes (Smad3 activated by TGF-β signaling in CNS1, Stat5 activated by IL-2Rγ, TCR, and TGF-β signaling in CNS2, and c-Rel activated by TCR signaling in CNS3). Both Smad3 and Stat5 regulate Foxp3 gene expression in a TGF-β–dependent manner. Smad3 binding to the enhancer in CNS1 was observed within 0.5 h of stimulation (12), whereas Stat5 binding to the enhancer in CNS2 was not detected even by 6 h following stimulation. The differential timing of Smad3 and Stat5 binding to these enhancers may reflect distinct signaling pathways following TGF-β signaling. Smad3 activation is directly regulated by signaling through the TGF-β receptor, whereas Stat5 activation is indirectly regulated through SOCS3 downmodulation following TGF-β signaling. Following initiation through these regulatory regions, a number of transcription factors induced by TCR signaling bind to the Foxp3 promoter and enhancers and thereafter maintain Foxp3 gene expression. Overall, Foxp3 expression is regulated by multiple induced transcription factors acting on the Foxp3 promoter [AP-1 (17), cRel (21), NFAT (17)], the Enhancer 1 [Smad3 (12), AP-1 (23) and NFAT (12)], and the Enhancer 2 [Stat5, AP-1, and Creb (11)]. Such regulation through multiregulatory regions with multiple transcription factors may offer alternative routes to achieve suppressive function in diverse circumstances.

Although Stat5 seems to play an important role in initiating Enhancer 2 activity with TGF-β signaling, strong enhancer activity is mainly enforced by AP-1 and Creb activated by TCR signaling. AP-1 binding to the Foxp3 promoter (17) and Enhancer 1 (23) has also been observed previously. AP-1 and Creb are closely related transcription factors, and members of these two families can form selective cross-family heterodimers (39, 40). It is likely that these transcription factors interact each other on the Enhancer 2 and perhaps in conjunction with AP-1 on the promoter and Enhancer 1.

The activation of Stat5 (p-Stat5) is regulated by JAK molecules associated with IL-2Rγ, and such activation in cells destined to be Treg depends on TGF-β signaling. Inhibition of the JAK/Stat pathway by Socs3 is well established (36, 37), and in this study, we have discovered that Socs3 mRNA levels can be dampened through a Smad3-mediated pathway that is currently uncharacterized. Socs3 is, as a result, kept at low levels in Treg by at least two mechanisms, a transcriptional and/or a posttranscriptional mechanism as shown in this study, and perhaps a translational mechanism as previously described (41). These combined observations suggest that Foxp3 gene expression can be upregulated by inhibition of Socs3 expression. This is consistent with the observation that Socs3 induction in Treg by a retroviral expression system reduced Foxp3 expression (41).

Yao et al. (27) have suggested the existence of five potential Stat5 binding sites in the Foxp3 gene and its surrounding regions, three of the sites being within CNS2, whereas the strongest binding is shown in this study as the Stat5-I site in the Enhancer 2 core. Two more Stat5 binding sites are located in the 5′-flanking regions (−0.7 and −2.56 kb) of the Foxp3 gene, and Stat5 binding around these sites was also shown by ChIP assay. The −0.7-kb Stat5 binding site is actually present in the Foxp3 promoter fragment of our luciferase reporter plasmid, but we could not detect promoter activity through this Stat5 binding site. It may be that this Stat5 element has an alternative role to upregulate Foxp3 gene expression via chromatin remodeling and/or demethylation of CpG in the Foxp3 promoter region. By analogy with CNS2, a CpG sequence in the Foxp3 promoter region is also highly methylated in Foxp3⁻ cells, such demethylation also being regulated by TGF-β signaling (11). The demethylation process in the promoter may be regulated by TGF-β–mediated Stat5 activity through the −0.7-kb Stat5 site. Although Stat5 seems to be involved in the demethylation process in the Foxp3 gene, it is still not known how Stat5 regulates the opening up and demethylation of CpG in CNS2. Perhaps this transcription factor forms a complex with other factors such as histone acetyltransferases and DNA demethylases on CNS2 to induce Enhancer 2 activity. Indeed, Stat5 interacts with p300/CBP proteins (42) that are endowed with histone acetyltransferase activity (43). Interestingly, CBP was originally identified as a Creb binding protein (43). AP-1 (c-Jun and c-Fos) belongs to a Creb-related family, and p300/CBP also interacts with c-Jun and c-Fos (43). It is conceivable that Stat5, AP-1, Creb, and p300/CBP form a large complex able to regulate Foxp3 gene expression through CNS2. The inclusion of demethylases within this complex might provide a possible explanation of how demethylation of CNS2 is achieved.

We thank Drs. J. Kaye, L. Kedes, and W. Sabbagh for review of the manuscript.

The authors have no financial conflicts of interest.