Cutting Edge: TGF-β-Induced Expression of Foxp3 in T cells Is Mediated through Inactivation of ERK

Known for their ability to suppress T cell proliferation and function, CD4⁺CD25⁺ T regulatory cells (Treg)² are critical for the maintenance of peripheral tolerance (1, 2, 3). The Forkhead transcription factor, Foxp3, has been shown to be a critical control element in the development and function of Treg cells (4). There are at least two subsets of Treg cells. One subset, known as the naturally occurring Treg cells (nTreg), is generated during the normal process of T cell maturation in the thymus. The other subset, known as the induced Treg cells (iTreg), develops as a consequence of induction of mature T cells under particular conditions in the periphery (5, 6). The peripheral induction of Treg cells can be accomplished by TGF-β (7). Under two different experimental conditions we observed that TGF-β induced expression of Foxp3 in stimulated T cells, which demonstrated the expected suppressive functions (8, 9). Such a dependence on TGF-β in the induction of Foxp3 in T cells has been confirmed by others (10). More recent findings have demonstrated that TGF-β-mediated Foxp3 expression is regulated through epigenetic events (11). However, the mechanism through which TGF-β mediates the expression of Foxp3 in induced Treg cells remains unclear. In the present study, we report that the TGF-β-induced epigenetic regulation of Foxp3 expression is mediated through inactivation of ERK and down-regulation of DNA methyltransferases (DNMTs).

BALB/c mice, NOD mice, and BDC2.5 TCR transgenic NOD mice (BDC mice) were purchased from The Jackson Laboratory. Mice were used according to guidelines approved by the Northwestern University Institutional Animal Care and Use Committee (Chicago, IL).

CD4⁺CD25⁻ T cells were isolated by negative selection using a mixture of biotinylated Abs (anti-CD8 (clone 53-6.7), CD25 (clone 7D4), Ly-76 (clone Ter-119), Gr1 (clone RB6-8C5), CD49b/Pan-NK (clone DX5), B220 (clone RA3-682), and CD11b (clone M1/70); BD Biosciences). CD4⁺CD25⁺ T cells were purified using a mouse T regulatory cell isolation kit from Miltenyi Biotec. CD4⁺CD25⁻ T cells were activated with 1 μg/ml plate-bound anti-CD3 and anit-CD28 (BD Pharmingen) for the indicated number of days. In some cases, CD4⁺CD25⁻ T cells from the BDC mice were activated with NOD splenic dendritic cells (DCs) pulsed with 100 ng/ml BDC peptide (RVRPLWVRME) at a 3:1 T cell to DC ratio. In indicated experiments, recombinant TGF-β1 (R&D Systems) at 2 ng/ml or 5-Aza-2′-deoxycytidine (5-Aza) (Sigma-Aldrich), MEK inhibitor UO126 (Promega), p38 inhibitor SB203580 (Promega), or JNK inhibitor II SP600125 (Calbiochem), each at 5 μM, was added. Anti- TGF-β1,-2,-3 mAb (clone 1D11; R&D Systems) was used at 5 μg/ml.

BDC T cells were cultured with syngeneic splenic DCs pulsed with BDC peptide in the presence of 5 μM UO126, after which DCs were removed and CD25⁺ T cells further enriched. For proliferation assays, CD4⁺CD25⁻ BDC T cells were cultured with splenocytes at a T cell to APC ratio of 1:5 and BDC peptide. [³H]Thymidine (1 μCi/well; PerkinElmer) was added for the last 18 h of a 72-h assay. For suppression assays, the UO126 induced CD25⁺ T cells were added at the indicated ratios and [³H]thymidine uptake was measured.

Cell lysates were prepared with modified radioimmune precipitation assay (RIPA) buffer. Protein extract (15 μg) was used for each sample. Membranes were incubated with phosphorylated ERK1/2 (p-ERK1/2) (1/750 dilution; Cell Signaling Technology), ERK1/2 (1/750 dilution; Cell Signaling Technology), phosphorylated p38, p38, phosphorylated JNK, JNK (all at 1/1000 dilution; Cell Signaling Technology), or GAPDH (1/300 dilution; Advanced ImmunoChemical), followed by anti-mouse-IgG-HRP (1/2000). Proteins were detected with the ECL detection kit (Amersham Biosciences).

Total RNA was extracted using an RNeasy kit (Qiagen). Primers for mouse Foxp3 were 5′-CCCAGGAAAGACAGCAACCTT-3′ (sense) and 5′-CCCAGGAAAGACAGCAACCTT-3′ (antisense) (12). Specific primers for DNMT1, DNMT3a, and DNMT3b (13) were: DNMT1, 5′-GGAAGGCTACCTGGCTAAAGTCAAG-3′ (sense) and 5′-ACTGAAAGGGTGTCACTGTCCGAC-3′ (antisense); DNMT3a, 5′-TGGAGAATGGCTGCTGTGTGAC-3′ (sense) and 5′-CACTCATCCCGTTTCCGTTTG-3′ (antisense); and DNMT3b, 5′-AGTGACCAGTCCTCAGACACGAAG-3′ (sense) and 5′-ATCAGAGCCATTCCCATCATCTAC-3′ (antisense). Primers for L-19 are 5′-CCATGAGTATGCTCAGGCTTCAGA-3′ (sense) and 5′-TACAGGCTGTGATACATGTGGCGA-3′ (antisense). Real-time PCR was performed using iQ SYBR Green Supermix (Bio-Rad Laboratories) in triplicate.

TGF-β has been shown to activate p38 MAPK and ERK in some systems (14, 15, 16), whereas in others it inhibits MAP kinases (17, 18). In yet another study, an antagonistic effect was observed between TGF-β and Ras/Raf/ERK signaling (19). These seemingly conflicting observations have added complexity to the mechanism of TGF-β action. In this study, we tested whether ERK activation in TCR-activated naive CD4⁺ T cells is affected by TGF-β1. Naive CD4⁺CD25⁻ T cells from BALB/c mice were first activated with anti-CD3 and anti-CD28 overnight. TGF-β1 (2 ng/ml) was then added to the culture. Cells were harvested and lysed at 5, 15, and 30 min and analyzed by Western blot for p-ERK. UO126 (5 μM), a specific ERK inhibitor, was also added to parallel cultures and cell lysates were analyzed in parallel. As shown in Fig. 1 A and B, TGF-β1-treated cells displayed significantly diminished levels of p-ERK as early as 5 min after treatment compared with untreated cells. The trend continued to 15 min, and at 30 min there was a significant rebound of the p-ERK level in the TGF-β1-treated cells. As expected, the UO126-treated cells also displayed significantly diminished levels of p-ERK, and the effect was more prolonged compared with that of TGF-β. Total ERK levels did not show significant alterations in either the TGF-β1- or the UO126-treated cells. Therefore, TGF-β1 treatment of activated CD4⁺ T cells led to a transient down-regulation of ERK phosphorylation, mimicking that seen with the ERK inhibitor UO126.

If naive CD4⁺CD25⁻ T cells without overnight activation were treated with TGF-β1 or UO126 and analyzed at 5, 15, and 30 min posttreatment, inhibition of p-ERK was only observed by UO126 at 30 min but not by TGF-β1 (Fig. 1, C and D). Interestingly, the level of p-ERK is significantly lower in naive CD4⁺CD25⁻ T cells than in overnight activated T cells (data not shown). Therefore, while UO126 inhibits p-ERK even at low levels, TGF-β1 inhibition appears to only manifest when the p-ERK level is heightened by previous CD3/CD28 activation.

It has been demonstrated that TGF-β could induce Foxp3 expression in previously Foxp3⁻ naive CD4⁺CD25⁻ T cells (7, 8, 9, 10, 20) through promoter demethylation (11). Furthermore, it has also been shown that inhibition of the ERK/MAPK pathway decreases DNA methylation in certain cancer cells (21, 22, 23). We therefore hypothesized that the TGF-β-induced promoter demethylation and consequent Foxp3 expression were mediated through ERK inhibition during T cell activation. We first tested whether treatment with an ERK inhibitor or a DNA demethylation agent during T cell activation could directly induce Foxp3 expression in CD4⁺CD25⁻Foxp3⁻ T cells. As shown in Fig. 2 A, naive CD4⁺CD25⁻ T cells expressed minimal Foxp3 whereas purified CD4⁺CD25⁺ T cells expressed high levels of Foxp3. Activation of naive CD4⁺CD25⁻ T cells by anti-CD3 and anti-CD28 in the presence of TGF-β1 led to significant induction of Foxp3 as previously described (8, 9, 10, 11, 12). Interestingly, activation of naive CD4⁺CD25⁻ T cells in the presence of either the ERK inhibitor UO126 or the demethylation agent 5-Aza also induced Foxp3 expression to various degrees, and the expression of Foxp3 can be seen as early as day 2 of the cultures. Similar results were observed when we used BDC TCR transgenic T cells activated with DCs pulsed with the BDC peptide (data not shown).

To correlate mRNA levels with protein expression, we further ascertained the above findings with intracellular Foxp3 staining. As shown in Fig. 2,B, TGF-β1 or UO126 induced a significant population of CD25⁺Foxp3⁺ cells measured at day 6 of culture compared with the control culture. Naive T cells treated with UO126 in the absence of TCR stimulation remained unactivated with almost all of the cells being CD25⁻. The induction of Foxp3 by UO126 was dose dependent as shown in Fig. 2,C, which peaked at 5 μM. In contrast, inhibitors to the other MAP kinases, JNK (SP600125) or p38 (SB203580), did not induce significant Foxp3 expression above control (Fig. 2 B). These results suggest that both ERK inhibition (but not JNK or p38 inhibition) and DNA demethylation can directly induce Foxp3 expression in TCR-activated naive CD4⁺CD25⁻ T cells.

To ensure that the effect of ERK inhibition on the induction of Foxp3⁺ T cells is downstream of TGF-β, we added anti-TGF-β1,-2,-3 at the beginning of the cell cultures treated with UO126. As shown in Fig. 2 D, the addition of anti-TGF-β Ab did not affect Foxp3 expression induced by UO126, suggesting that UO126 induction of Foxp3 is independent of TGF-β production.

It has been shown that there is a window period for adding TGF-β to the CD3/CD28-stimulated cultures to induce Foxp3 expression in CD4⁺Foxp3⁻ T cells (10), which is the initial 0–72 h of the CD3/CD28 stimulation. We next tested whether UO126-induced Foxp3 expression has a similar window period for action. As shown in Fig. 2 E, a similar degree of Foxp3 induction was seen with addition of UO126 on either day 0 (beginning of culture) or day 1 (after overnight stimulation), as seen with TGF-β1.

To confirm that the observed accumulation of Foxp3⁺ T cells was not due to enrichment of existing Foxp3⁺ T cells resulting from apoptosis of non-Foxp3⁺ T cells, we performed apoptosis and viability staining. As shown in Fig. 2 F, the viability assessed by annexin V⁻7-aminoactinomycin D⁻ cells were 41, 61, and 48%, respectively for control, plus TGF-β, and plus UO126. Therefore, the increase of Foxp3⁺ cells by TGF-β and UO126 treatment was indeed due to induction of new Foxp3⁺ cells rather than selective preservation of existing Fopx3⁺ cells.

We have previously shown that TGF-β1-induced CD4⁺ CD25⁺Foxp3⁺ T cells are functionally suppressive as the natural Treg cells (10, 11). We next tested the suppressor function of the UO126-induced CD4⁺CD25⁺Foxp3⁺ T cells. As shown in Fig. 3, UO126-induced CD4⁺CD25⁺Foxp3⁺ T cells were anergic themselves (suppressors (S) alone) and, when added at an incremental ratio to the responders (R), were able to exert a dose-dependent suppression of proliferation. Therefore, the UO126-induced CD4⁺CD25⁺Foxp3⁺ T cells resemble natural Treg cells in their suppressive capacity.

To test whether TGF-β or ERK inhibition could have an effect on DNA demethylation, we determined the expressions of three DNMTs, DNMT1, DNMT3a, and DNMT3b, in activated T cells treated with TGF-β1 or UO126. As a control, the effect of the demethylation agent 5-Aza was also tested. As shown in Fig. 4, naive T cells activated with anti-CD3/CD28 (3/28) Abs showed significantly higher levels of all three DNMTs compared with natural CD4⁺CD25⁺ Treg cells (CD25⁺). Interestingly, treatment with TGF-β1 during TCR activation significantly inhibited all three DNMTs when measured at day 4 posttreatment. Furthermore, treatment with UO126 also led to down-regulation of all three DNMTs, similarly as with TGF-β1 treatment. As a control, 5-Aza down-regulated all three DNMTs. Therefore, both TGF-β1 and ERK inhibition led to significant down-regulation of DNMT expressions in activated CD4⁺ T cells.

The importance of inhibition of the ERK pathway in natural CD4⁺CD25⁺ Tregs on their function has been recently demonstrated (24, 25). The role of the ERK pathway in induced Tregs has not been well studied. DNA methylation is regulated in part by the ERK pathway (21, 22, 23, 26, 27). In this report, we have shown that: 1) TGF-β1 treatment of activated CD4⁺Foxp3⁻ T cells results in inhibition of ERK activation, down-regulation of DNMTs, and induction of Foxp3 expression; and 2) UO126 (ERK inhibition) treatment of activated CD4⁺Foxp3⁻ T cells also results in down-regulation of DNMTs and induction of Foxp3 expression. We therefore postulate that in the absence of TGF-β action, the promoter of the Foxp3 gene is targeted by the DNMTs, thus remaining highly methylated. Because treatment of TGF-β leads to promoter demethylation of the foxp3 locus of T cells (11), we propose that the intermediate steps of this TGF-β-dependent effect on Foxp3 expression are the TGF-β mediated inhibition of ERK activation and the consequent down-regulation of DNMT expressions (Fig. 5).

The authors have no financial conflict of interest.