Results

FOXP3 Induction in T Cell Subsets
It is assumed that FOXP3 expression can be induced in nonregulatory T cells, which is an important step in iTreg cell differentiation. However, it is not known if all CD4+ T cells have the same capacity to express FOXP3. To investigate whether FOXP3 can be expressed by any T cell subset or if expression is restricted to a distinct lineage, FOXP3 mRNA expression was analyzed in freshly isolated T cells such as CD25-depleted CD4+ cells, CD45RA+ naive or CD45RO+ memory T cells (Figure 1A), as well as T cells driven in vitro toward Th1, Th2, or iTreg phenotypes (Figure 1B; phenotype on Figure S1). The CD4+CD25–, CD45RA+, CD45RO+, and CD4+CD45RO+CD25– were able to significantly induce FOXP3 mRNA up to 30-fold upon TCR activation and addition of TGF-β. Th1 cells showed only a 10-fold increase. In contrast, Th2 cells stimulated under the same conditions did not increase FOXP3 expression. The in vitro generated iTreg cells were unable to further up-regulate FOXP3, which was already at high levels under the resting conditions (Figure 1B, right panel).
Figure 1  Th2 Cells Cannot Induce FOXP3 Expression
(A) Human T cells were activated with plate-bound anti-CD3/CD28 with or without TGF-β as indicated on the x-axis of (B). Cells were harvested after 5 days and FOXP3 mRNA was quantified by real-time PCR. Bars show the mean ± SD of 4 independent experiments.
(B) In vitro differentiated Th1, Th2, or iTreg cells were activated with anti-CD3/CD28, TGF-β, or anti-IL-4 as indicated. The phenotype of these cells was confirmed by FACS and proliferation analysis (Figure S1). Bars show the mean ± SD of four independent experiments.   Th2 cells are known to produce IL-4 upon activation, which may interact with TGF-β signaling and thus prevent FOXP3 induction. However, the neutralization of IL-4 with a blocking IL-4 antibody did not rescue FOXP3 expression in the differentiated Th2 cells (unpublished data). These data demonstrated that Th2 cells have a limited capacity to express FOXP3 (Figure 1B). The inability of Th2 cells to express FOXP3 was also documented at the single-cell level, confirming that Th2 cells lack FOXP3 expression (Figure 2A). Only iTreg cells expressed FOXP3 in resting conditions. Interestingly, we observed that resting iTreg cells express FOXP3 but show low CD25 surface expression. Repeated exposure to TGF-β did not further increase the FOXP3 expression in the iTreg lineage but transiently induced FOXP3 expression in Th1 cells. Naturally occurring Th2 cells such as CRTH2+ T cells, T cells isolated according to their IL-4 secretion, or an IL-4–producing T cell clone (BR8) were also lacking FOXP3 expression (Figure 2B). Furthermore TGF-β–mediated FOXP3 induction failed in these cells in contrast to the naive T cells (Figure 2B). Because IL-4 is the key Th2 cytokine, the expression of IL-4 and FOXP3 in freshly isolated CD4+ T cells was analyzed by fluorescence activated cell sorting (FACS). IL-4–expressing cells were most abundant among CD45RO+CD25– cells, which did not co-express FOXP3 (Figure 3A, left panel). In contrast, CD45RO+CD25+ cells abundantly expressed FOXP3, while lacking IL-4 (Figure 3A, right panel). As shown in Figure 3B, the frequency of the IL-4+ cells was always below 1% in the FOXP3+ cells close to the background. The IL-4+ cells were confined to the FOXP3– cells, as shown for the CD45RO+CD25–, CD45RO+CD25+, and CD45RO+CD25+high cells (Figure 3B). In addition, neither the Th2 clone (BR8) nor CRTH2 cells significantly expressed FOXP3 (Figure 3C). Cells enriched for their IL-4 secretion using the magnetic cell isolation technology contained some FOXP3-expressing cells, but importantly, the expression did not overlap. Taken together, these data indicate that FOXP3 was not expressed by Th2 cells and was not inducible in those cells.
Figure 2  Th2 Cells Do Not Express FOXP3
(A) Intracellular FACS analysis of FOXP3 expression in Th1, Th2, or iTreg-differentiated cells (two rounds, phenotype see Figure S1), rested or activated, with or without TGF-β. FOXP3 expression was measured after 5 d in culture. The dot blots are representative of three independent experiments.
(B) Shows the same experimental setup, but naturally occurring Th2 cells were analyzed. Data are representative of three independent experiments.
Figure 3  Th2- or IL-4–Producing Cells Lack FOXP3
(A) FACS analysis of intracellular FOXP3 and IL-4 expression following PMA/Ionomycin stimulation. CD4+ T cells were gated on the basis of CD45RO and CD25 surface expression (upper panel), and gated cells are shown below for the CD45RO+CD25− (A, left panel), the CD45RO+CD25+ (right panel), and the CD45RO−CD25− subsets (central panel). A statistical analysis of eight independent donors after subtraction of the isotype control are shown in (B). The dotted gray line indicates the IC background level. The error bars show the error of the mean. (C) Similarly, a Th2 clone (BR8), CRTH2+ Th2 cells, IL-4-secreting cells, and memory T cells (CD45RO) were stained for FOXP3 and IL-4. Data are representative of three independent experiments.

FOXP3 and GATA3 Kinetics in Differentiating Cells
The limited capacity of differentiated effector cells to induce FOXP3 expression suggests that iTreg induction has to occur before effector T cell differentiation occurs. Therefore, we analyzed the expression of FOXP3 and GATA3 during the differentiation of naive CD4 T cells into Th0 (neutral, anti-IL-4, anti-IFN-γ, and anti IL-12), Th2, and iTreg phenotypes. After initiation of the differentiation process, FOXP3 and GATA3 showed a similar expression kinetic within the first 3 d, which are considered to be critical in T cell commitment [18]. Under Th2 differentiation conditions, FOXP3 mRNA expression increased only marginally (Figure 4A, left panel). Thus, although GATA3 and FOXP3 showed similar kinetics, their expression polarizes at the end of the differentiation process when cells were cultured towards Th2 or iTreg cells, respectively (Figure 4A and 4B). Interestingly, the Th0 cells were expressing more FOXP3 than the Th2 cells, but expressed low levels of GATA3; however, the protein expression slightly differed from mRNA expression, suggesting also posttranslational regulation and degradation as potential additional mechanisms in the differentiation process. The phenotype of iTreg cells included an anergic phenotype upon anti-CD3 re-stimulation (Figure S1A), CD103, CTLA-4, GITR, and PD-1 surface expression (Figure S1B). On the single-cell level, it can be seen that cells progress through a transition phase, where GATA3 and FOXP3 expression coexist to some degree in the same cells, which is resolved in iTreg cells after 7 d (Figure 4B). Taken together, these data demonstrated that Th2 cells have lost their capacity to express FOXP3 and showed that Th2 and iTreg cells arise from two different differentiation pathways.
Figure 4  FOXP3 Induction During the Differentiation Process
(A) Human CD4+CD45RA+ T cells were activated with plate-bound anti-CD3/CD28 in the presence of TGF-β (5 ng/ml) or IL-4 (25 ng/ml). The cells were harvested at different time points, and mRNA was quantified by real-time PCR for FOXP3 and GATA3 expression. Bars show the mean ± SD of three independent experiments. 
(B) Intracellular GATA3 and FOXP3 staining is shown after exposure of CD4+CD45RA+ T cells to differentiating conditions as in part A of the figure. Data are representative of three independent experiments.

IL-4 Inhibits TGF-β-Mediated iTreg Commitment
IL-4 induces differentiation of naive T cells, upon antigen encounter, into the Th2 cell lineage. We therefore asked whether IL-4 is able to inhibit TGF-β induction of FOXP3 during the priming of naive T cells. Human CD4+CD45RA+ T cells were activated with plate-bound anti-CD3/CD28 in the presence of TGF-β and/or IL-4 and harvested after 5 d. IL-4 efficiently repressed the TGF-β–mediated induction of FOXP3 expression (Figure 5A) in a concentration-dependent manner (Figure 5B). Low levels of GATA3 were induced also in the absence of IL-4, as it was previously observed [3]. However, at low concentration, IL-4 was able to marginally induce FOXP3 expression. Of note, GATA3 was also induced in the presence of TGF-β at high IL-4 concentration (Figure 5A and 5B). The IL-4–mediated prevention of FOXP3 expression was not caused by interferences of the receptor signaling, because the phosphorylation of SMAD2 or STAT6 was not affected by the addition of IL-4 and/or TGF-β, which demonstrates that IL-4 as well as TGF-β signaling were functional under these conditions (Figure 5C). Increasing amounts of IL-4 increase intracellular GATA3, whereas FOXP3 decreased, which is consistent with the mRNA analysis (Figure 5D). Furthermore, injection of IL-4 into wild-type B6 mice decreased the inducible or natural Treg number in vivo. A distinction of the Treg subsets is not possible, because recently activated iTreg cells also express surface CD25. We used complexes of recombinant mouse IL-4 (rmIL-4) plus anti-IL-4 monoclonal antibodies (mAbs), which have been shown to dramatically increase the potency of the cytokine in vivo [19]. In these mice, the percentage of CD4+CD25+ and FOXP3+ T cells dramatically decreased when the antibody-cytokine immune complexes were injected (Figure S2). Upon administration of rmIL-4 plus anti-IL-4 mAb complexes, the total number of CD4+CD25+ T cell, as well as the Foxp3+ T cells diminished by half (Figure S2G and S2H), confirming that the lower percentage was not due to an increase in the CD4+CD25– cells, but a real decrease of CD4+CD25+ T cells. In conclusion, IL-4 negatively regulates the natural or inducible Treg cell turnover not only in vitro but also in vivo. To study the effects of IL-4 on already-existing human natural or inducible Treg cells, we exposed sorted CD25+ T cells (nTreg cells) to IL-4 and analyzed FOXP3 expression and suppressive capacity. In already-existing Treg cells, IL-4 failed to inhibit FOXP3 expression (Figure S3A), and the suppressive capacity was not altered (Figure S3C). Similarly pre-existing iTreg cells did not decrease FOXP3 expression upon IL-4 exposure (Figure S3B).
Figure 5  Effect of IL-4 on FOXP3 Induction
(A) A statistical analysis was performed with six donors on day 5 (TGF-β (10 ng/ml) and with or without IL-4 (100 ng/ml)); Shown is the mean, and error bars indicated the SD of six donors. Statistical analysis was performed using the Dunnett test. Statistical significance is indicated by asterisks (*p ≤ 0.05, **p ≤ 0.01, Dunnett).
(B) CD4+CD45RA+ cells were activated in the presence of a constant concentration of TGF-β (5 ng/ml) with an increasing concentration of IL-4, as indicated. Cells were harvested for mRNA quantification after 5 d.
(C) CD4+CD45RA+ cells were stimulated in vitro with plate-bound anti-CD3/CD28, TGF-β (10ng/ml), and IL-4 (100 ng/ml) as indicated. After 1 h, cell lysates were prepared and analyzed by Western blot for phosphorylated SMAD2 and STAT6. Total STAT6 and GAPDH served as internal control.
(D) Intracellular GATA3 and FOXP3 staining are shown after exposure of CD4+CD45RA+ T cells to IL-4 as described for panel B. Data are representative of three independent experiments.

The Role of TGF-β in T Cell Differentiation
Although TGF-β–reduced CD25, IL-4 expression, and CD25 expression (Figure 6B), IL-4 significantly inhibited TGF-β–mediated induction of FOXP3 in naive T cells driven toward FOXP3+ T cells, as shown by FACS analysis (Figure 6A). It is known that IL-4 is a potent growth factor and may therefore favor the proliferation of FOXP3– cells and thus decrease the relative percentage of FOXP3+ cells. However, analysis of cell division kinetics by CFSE-labeling demonstrated that IL-4 did not differentially promote cell growth of FOXP3+ over that of FOXP3–. In fact both populations showed similarly enhanced proliferation (Figure 6A). Furthermore the TGF-β–mediated induction of FOXP3 expression was not caused by overgrowth of a CD25–FOXP3+ minority, since the number of FOXP3+ cells was low/absent in the purified CD4+CD45RA+ T cells (between 0% and 1%), and the FOXP3+ cells were not confined to the highly divided cells. CD25 was down-regulated in TGF-β–treated cells compare to activated T cells, which was even more pronounced in cells cultured with TGF-β and IL-4.
Figure 6  IL-4 Inhibits TGF-β–Mediated iTreg Commitment
CFSE-labeled CD4+CD45RA+ cells were activated with plate-bound anti-CD3/CD28, TGF-β, and IL-4, as indicated. After 5 d, cells were analyzed by flow cytometry (A) and results of six independent experiments are shown in the bar graph below (B). Statistical significance (one-way Anova, Newman-Keuls) is indicated by asterisks (**p ≤ 0.01, ***p ≤ 0.001). (C) Kinetic analysis of intracellular GATA3 and FOXP3 staining is shown in panel C following exposure of CD4+CD45RA+ T cells to anti-CD3/28, IL-4 and TGF-β. Data are representative of three independent experiments.   The addition of IL-4 to iTreg-driving conditions decreased the number of FOXP3+ cells (Figure 6B). In line with the previous findings, the IL-4–producing cells and the FOXP3 expressing cells are nonoverlapping populations. Since FOXP3 is known to act as a repressor of cytokine expression [20], we therefore analyzed GATA3 and FOXP3 expression. The expression kinetic of naive T cells exposed to IL-4 and TGF-β demonstrated that GATA3 and FOXP3 are initially found in separate populations (day 2), but transiently co-express both factors (days 4–8), before establishing separate populations at the end of the differentiation process (day 10; Figure 6C), suggesting that GATA3 inhibits the development of iTreg cells by repressing FOXP3.
These results showed that IL-4 acts in vitro as an inhibitor of FOXP3 expression, without interfering with TGF-β signaling, probably acting at the level of transcription factors, and possibly by a GATA3-dependent mechanism.

GATA3 Is a Negative Regulator of FOXP3 Expression
FOXP3 expression decreased once GATA3 expression is high; therefore, we hypothesized a potential role for GATA3 in repressing FOXP3. Besides GATA3′s well-known positive effect on gene regulation, GATA3′s repressive capabilities were previously shown to restrict Th1 commitment by inhibiting STAT4 expression [2,21], and therefore GATA3 prevents differentiation into Th1 cells. To investigate whether GATA3 can directly inhibit FOXP3 induction, we transduced GATA3 or a truncated GATA3 lacking the DNA-binding domain in human primary CD4+CD45RA+ T cells using a TAT-fused, recombinantly expressed GATA3. After transduction, the cells were activated with soluble anti-CD3/CD28 in the presence or absence of TGF-β. TAT-GATA3 was successfully transduced in a homogeneous and dose-dependent manner into human CD4+ T cells (Figure 7A, upper panel). TAT-GATA3 reduced FOXP3 expression in a dose-dependent manner, whereas a DNA-binding domain truncated version (TAT-ΔDBD-GATA3) did not affect FOXP3 expression as compared with expression in untransduced cells (Figure 7A). In addition, we analyzed the inhibitory effect of GATA3 on FOXP3 in transgenic DO11.10 mice, constitutively overexpressing GATA3 under the control of the CD2 locus control region (DO11.10xCD2-GATA3). The thymic selection into the CD4 lineage is largely intact in DO11.10xCD2-GATA3 (RW Hendriks, unpublished data). These mice develop lymphomas at an older age, but signs of autoimmune disease were not described [22]. To investigate the effect of GATA3 on iTreg, CD4+CD62L+CD25– cells were isolated, activated with OVA in the presence or absence of TGF-β, and Foxp3 expression was analyzed after 4 d. The naive CD4+CD25– cells were Foxp3– (unpublished data). As described for the human cells, TGF-β dramatically up-regulated Foxp3 in the DO11.10 littermate control mice. In contrast, cells from the CD2-GATA3xDO11.10 mice showed dramatically reduced Foxp3 expression when activated with TGF-β and OVA (Figure 7B). All mice produced similar amounts of TGF-β; in addition, Smad7 was equally expressed [23] in T cells of both mice strains (Figure 7C), indicating intact TGF-β signaling.
Figure 7  GATA3 Acts as a Negative Regulator of FOXP3 Expression
(A) Human naive CD4+CDRA+ T cells were transduced with 0, 20, 100, and 500 nM of TAT-GATA3 protein, and intracellular presence of GATA3 was analyzed using FACS following anti-CD3/CD28 activation of the cells. GFP-positive cells were gated and analyzed for intracellular FOXP3 expression following a 2-d incubation period (lower panel). Data are representative of four independent experiments.
(B) CD4+CD25− T cells were isolated from D011.10 and D011.10xCD2-GATA3 mice and treated with OVA and TGF-β for 96 h. Surface CD4 and intracellular FOXP3 were measured by FACS. These data are representative of three independent experiments.
(C) The cells treated as in (B) were harvested and mRNA was quantified by real-time PCR for SMAD7 and TGF-β expression. Bars show the mean ± SD of three independent experiments.   Taken together, these results demonstrated a repressive role of IL-4-induced GATA3 transcription factor in the generation of iTreg cells.

GATA3 Represses the FOXP3 Promoter
To investigate the molecular mechanism of GATA3-mediated repression of human FOXP3, the human FOXP3 promoter was studied and a palindromic binding site for GATA3 was discovered. The GATA-binding site is located 303 bp upstream from the transcription start site (TSS) [24]. This site is highly conserved between humans, mice, and rats (Figure S4) and may therefore play an important role in FOXP3 regulation. The functional relevance of this site was studied using a FOXP3-promoter construct [24]. We transfected human primary CD4+ T cells, in vitro differentiated Th2 cells, and Jurkat cells (Jurkat cells are known to constitutively express GATA3 [25,26]), and we measured FOXP3 promoter activity. The promoter was not active in the GATA3-expressing cell line Jurkat or in the in vitro-differentiated Th2 cells, whereas the construct was active in the CD4 cells, which express a lower amount of GATA3 (Figure 8A). Overexpression of GATA3 in naive T cells diminished luciferase activity of the FOXP3 promoter compared with the control vector (Figure 8B). To further address the function of the GATA3 site, we inserted a site-specific mutation deleting the GATA3-binding site. This mutation increased luciferase activity by 3-fold in memory CD4+CD45RO+ T cells, whereas no difference was observed in naive (GATA3–) CD4+CD45RA+ T cells, revealing a repressor activity of GATA3 on the FOXP3 promoter (Figure 8C). Furthermore GATA3 binds directly to the FOXP3 promoter as investigated by pull-down assay. HEK cells were transiently transfected with GATA3 or a control vector, and increasing amounts of lysates were incubated with oligonucleotides containing the GATA3 site of the FOXP3 promoter or a control oligonucleotide with a mutated GATA3-binding site. After the pull-down, GATA3 binding was detected by Western blot. Similarly, GATA3-expressing Th2 cells and iTreg cells were subjected to this approach. Only HEK cells overexpressing GATA3 and Th2 cells showed GATA3-binding activity (Figure 8D and 8E). These experiments demonstrated that GATA3 binds the palindromic FOXP3 promoter. To gain insights into the in vivo situation, we performed a chromatin immunoprecipitation (ChIP) using an anti-GATA3 antibody and showed that GATA3 binds to the FOXP3 promoter region in Th2 cells, but not in iTreg cells (Figure 7F). Taken together these data demonstrate that the GATA3-binding to the FOXP3 promoter is repressing FOXP3 expression.
Figure 8  GATA3 Represses the Human FOXP3 Promoter
(A) Jurkat, Th2 cells, and human primary CD4 cells were transfected with an empty vector (pGL3 basic) or a vector containing the putative FOXP3 promoter region fused to the luciferase reporter gene. Bars show the mean ± SD of arbitrary light units normalized for renilla luciferase of four independent experiments; samples were measured in triplicates.
(B) Naive CD4 T cells were transfected with the FOXP3 promoter reporter construct together with a GATA3 expression vector or an empty vector. Bars show the mean ± SD of three independent experiments.
(C) Naive (left panel) or memory (right panel) CD4 T cells were transfected with wild-type or a GATA3 mutated 511-FOXP3 promoter reporter construct and activated with PMA and ionomycin. Bars show the mean ± SD of arbitrary light units normalized for renilla luciferase of eight independent experiments; samples were measured in triplicates.
(D) Nuclear extracts were prepared from HEK cells transfected with GATA3 or an empty vector, (E) Th1, Th2, or iTreg cells and binding factors precipitated using biotinylated oligonucleotides. The oligonucleotides–transcription factor complexes were separated on a SDS-PAGE gel. The amounts of GATA3 protein in the precipitates were assessed by immunoblotting with anti-GATA3 mAb. Total nuclear extracts were also run as controls. Data are representative of three different experiments.
(F) iTreg or Th2 cells were analyzed by ChIP for GATA3 binding to the FOXP3 promoter. The “input” represents PCR amplification of the total sample, which was not subjected to any precipitation. Results are representative of three independent experiments.

Di