The Ets-1 transcription factor controls the development and function of natural regulatory T cells

In the C57BL/6 background, complete inactivation of the Ets-1 locus leads to >98% perinatal mortality (Eyquem et al., 2004a), whereas in mixed 129sv/C57BL/6 background, ∼12% of the mice reach adulthood. 6-wk-old viable Ets-1^−/− mice had dramatically enlarged spleens (Fig. 1 A) containing an approximately fourfold higher total cell number than WT littermate controls (Table S1). This increase was essentially accounted for by B cells (fourfold) and a high side scatter heterogeneous population (Fig. S1 A) that included eosinophils (not depicted).

In contrast, the number of T cells was reduced twofold as compared with controls (Table S1), and the reduction affected both CD4 and CD8 subsets (Fig. 1 B). Most CD4⁺ and CD8⁺ Ets-1^−/− splenocytes were large (Fig. 1 C), and an increased percentage had an activated phenotype with up-regulated levels of CD44 and down-regulation of CD62-L (Fig. 1 B). Notably, cells with a naive phenotype (CD62-L^highCD44^low) represented at best 2% and 15%, respectively, of the CD4 and CD8 T cell populations (Fig. 1 B). Furthermore, the percentages of CD69- and CD95-expressing cells were significantly increased in Ets-1^−/− T cells (Fig. S1 B). Moreover, DNA content analysis revealed a 1.5–2-fold increase of the percentage of cycling cells (Fig. 1 C).

Analysis by semiquantitative RT-PCR of the transcription status of several cytokines in freshly isolated Ets-1^−/− CD4⁺ T cells revealed high expression levels of IL-4, IL-5, IL-10, and IL–13 but not IFN-γ (Fig. 1 D). Consistently, upon in vitro stimulation, secretion of IL-4 and IL-10 was markedly increased, whereas IL-2 and IFN-γ were barely detected (Fig. S2).

These results show that inactivation of the Ets-1 transcription factor induces splenomegaly and the development of activated T cells with CD4⁺ cells expressing high levels of Th2 cytokines. These features were associated with severe naive T cell lymphopenia.

In line with previous descriptions (Eyquem et al., 2004b), histological analysis of Ets-1^−/− spleens showed a marked abnormal structure with loss of B cell follicles and marginal zones (Fig. S3). Viable mutant mice contained high numbers of activated B cells, a slight increase in the percentage of cycling cells (unpublished data), and high serum titers of IgM, as well as elevated levels of auto antibodies (Eyquem et al., 2004b).

To specifically assess a direct role of mutant T cells in the B cell immune disorder of Ets-1^−/− mice, CD3-ε–deficient mice were engrafted with Ets-1^−/− (KO/CD3) or WT Ets-1^+/+ (WT/CD3) fetal thymi under the kidney capsule. 5–6 wk after graft, both KO/CD3 and WT/CD3 mice had reconstituted CD4 and CD8 splenic T cell compartments (Fig. 2 A). Similar to Ets-1^−/− mice, the KO/CD3 chimeras had splenomegaly (see next paragraph). Although spleens from WT/CD3 bear a significant subset (∼33%) of naive CD44^lowCD62-L⁺CD4⁺ T cells, KO/CD3 chimeras, like Ets-1^−/− mice, displayed increased percentages of activated CD44^highCD4⁺ cells and very few CD44^lowCD62-L⁺CD4⁺ splenocytes (Fig. 2 B).

Whereas WT/CD3 spleens had a normal frequency of marginal zone (CD21⁺CD23⁻) B cells, this population was undetectable in KO/CD3 chimeras (Fig. 2 C). In these mice, the CD86 and MHC II on B cells was significantly higher than in control chimeras (Fig. 2 D), indicating that most B cells were in an activated state.

WT/CD3 mice had normal levels of IgM, IgG1, and IgE (Fig. 2 E). In KO/CD3 chimeras, IgM titers were similar to controls, but IgG1 and IgE titers were 10-fold increased (Fig. 2 E and not depicted). Furthermore, elevated concentrations of anti-DNA antibodies could be specifically detected in KO/CD3 chimeras (Fig. 2 E). These results demonstrate that except for IgM levels, Ets-1^−/− T cells trigger all the features of B cell dysregulation observed in Ets-1^−/− mice.

Enhanced in vivo T cell activation can originate from impaired development or loss of function of naturally occurring T reg cells. To test whether a similar process occurs in Ets-1^−/− mice, we generated Ets-1/CD3 chimeras that received (or not) WT Ly5.1 CD25⁺CD4⁺ (hereafter referred to as donor T reg cells) at the time of fetal thymus transplantation.

As expected, 5–6 wk after graft, KO/CD3 chimera mice had enlarged spleens compared with WT/CD3 controls (Fig. 3 A). Strikingly, injection of donor T reg cells in KO/CD3 chimeras was sufficient to block development of splenomegaly (Fig. 3 A). Interestingly, in these chimeras, donor T reg cells represented 50–60% of the CD4⁺ T cell population as compared with 13–22% in WT/CD3 controls (Fig. S4 A). In addition, Ets-1^−/− CD4⁺ T cells represented only 2% of total spleen cells in KO/CD3 chimeras that received T reg cells compared with 7% in the absence of donor T reg cells, suggesting that donor T reg cells might inhibit expansion of Ets-1^−/− T cells. Notably, the CD4 T cell compartment of KO/CD3 chimeras that received donor T reg cells was also devoid of naive (CD44^lowCD62-L⁺) cells (Fig. S4 A). To further investigate the function of WT T reg cells on Ets-1^−/− T cells, CD45.2⁺CD4⁺ was sorted from WT/CD3 or KO/CD3 chimeras that had received (or not) CD45.1 WT T reg cells. Real-time RT-PCR analysis showed a 12-fold reduction of IL-4 expression in Ets-1^−/− CD4⁺ splenocytes that were isolated from mice containing WT T reg cells compared with those that did not receive suppressive cells (Fig. S4 B).

Furthermore, whereas no CD21⁺CD23⁻ marginal zone B cells were detected in the spleen of KO/CD3 chimeras, injection of WT T reg cells clearly prevented clearance of this B cell subset (Fig. 3 B). Injection of WT T reg cells also prevented IgG1 and IgE hypergammaglobulinemia in KO/CD3 chimeras (Fig. 3 C). These results demonstrate that Ets-1^−/− T cells are responsive to T reg cell suppression and point toward a defective T reg cell compartment in Ets-1–deficient mice.

Based on the aforementioned observations, we studied the thymic and peripheral T reg cell compartments of Ets-1^−/− mice. As previously reported (Bories et al., 1995; Eyquem et al., 2004a), the thymus of these mice had 10–12 times lower cell numbers, and although CD4⁺8⁺ thymocytes represented around 50% of the cells, the percentage of CD4⁺ single-positive (SP) thymocytes was comparable to age-matched littermates (Fig. 4 A). Within this subset, cells expressing Foxp3 were detected in mutant mice at frequencies similar to those of littermates. However, the amounts of Foxp3 were clearly lower in Ets-1^−/− thymocytes (Fig. 4 A), whereas CD25 was expressed at normal levels (not depicted).

In the spleen, the percentage of Foxp3⁺ cells among Ets-1^−/− CD4⁺ splenocytes was fourfold lower than in controls (Fig. 4 A). The expression level of Foxp3 and the quantity of Foxp3 transcripts in sorted Ets-1^−/− CD25⁺CD4⁺ cells were also decreased (Fig. 4 A). In contrast, expression levels of CD25 within Foxp3⁺ were equivalent in mutants and controls (Fig. 4 A).

Because in Ets-1^−/− mice ∼98% of the peripheral CD4⁺ T cells were activated, we determined the frequency of Ets-1^−/− Foxp3⁺CD4⁺ splenocytes that expressed CD103, a marker which is normally up-regulated in activated T reg cells infiltrating nonlymphoid tissues (Sather et al., 2007). As shown in Fig. 4 B, 85% of T reg cells expressed CD103 (gate G2), compared with 28% in control and, on average, at significantly higher levels. In the thymus, 44% of Ets-1^+/+ Foxp3⁺ cells were CD103⁺ and expressed high levels of this marker (Fig. 4 B, gate G1). Strikingly, 83% of Ets-1^−/− Foxp3⁺ thymocytes expressed CD103, although at lower levels. Given the very high frequency of CD103⁺ peripheral T reg cells and the fact that this marker is up-regulated on homeostatically generated peripheral T reg cells (Feuerer et al., 2010), these results raised the possibility that the majority of thymic Ets-1^−/− Foxp3⁺CD103⁺ cells might have come from the periphery. This would imply that thymic production of T reg cells is significantly reduced in Ets-1^−/− mice.

To investigate thymic development of T reg cells in a peripheral T cell free system, we performed fetal thymus organ cultures (FTOCs) with E15 thymi of Ets-1^−/− and WT mice. At day 7 of culture, WT thymic lobes had fully developed (9 × 10⁵ ± 10⁵ cells/thymus), including 45–65% of DPs (median = 50%; n = 6) and 7–25% of CD4 SP. Foxp3⁺ cells were readily identified and represented 2.5–4.5% of CD3⁺4⁺8⁻ thymocytes (Fig. 4 C). In contrast, Ets-1^−/− lobes developed poorly (2.5 × 10⁵ ± 10⁵ cells/thymus) with DPs ranging from 10 to 50% (median = 25%; n = 6) and 3 to 12% for CD4 SPs. Within this subset, Foxp3⁺ cells represented <0.6%, the cells expressing very low amounts of the Foxp3 protein. Altogether, the results show that, despite the normal frequency of thymic Foxp3-expressing cells in adults, the development of Foxp3⁺ T cell is significantly compromised in Ets-1^−/− mice.

To evaluate the cell-intrinsic role of Ets-1 in T reg cells, we cotransferred various ratios of WT (CD45.1⁺) and Ets-1^−/− (CD45.2⁺) fetal liver cells into irradiated Rag^−/−γc^−/− mice. In 1:1 WT/KO chimera, analyzed 1 mo after transfer when T cells were still poorly detected in peripheral blood, the frequency of Ets-1^−/− cells in CD4 SP was only 4.83% (Fig. 5 A). Strikingly, 2.04% of WT (CD45.2⁻) CD4⁺ SP cells expressed Foxp3 at normal levels, but only 0.47% of the Ets-1^−/− (CD45.2⁺) counterpart stained for Foxp3 and expressed very low levels of this transcription factor (Fig. 5 A). The deficit of Ets-1^−/− T reg cells appeared even more pronounced when gating on thymocytes expressing high level of Foxp3 (Fig. 5 A, middle). In contrast, in 1:1 WT/WT chimeras, normal Foxp3⁺ cells were readily detected in the CD4 SP derived from both donors. To investigate the development of T reg cells in the presence of an excess of Ets-1^−/− cells, we analyzed 1:20 WT/KO chimera 2 mo after transfer. As shown in Fig. 5 B, although >75% of the CD4⁺ SPs was derived from the Ets-1^−/− background, the development of WT Foxp3⁺ thymocytes was not impaired and represented ∼80% of the T reg cell subset. This deficit of Ets-1^−/− CD4⁺Foxp3⁺ cells, which resulted in an increased percentage of Ets-1^−/− CD4⁺Foxp3⁻ SP cells among the CD45.2 subset, appeared significant in our series of chimera mice (Fig. S5). Collectively, these results show that the thymic environment in these chimeras is capable of efficiently supporting normal but not Ets-1^−/− T reg cell development and point toward an intrinsic role for Ets-1 in this process.

As described in the previous section, 75% of CD4⁺ SP thymocytes in 1:20 WT/KO chimera were derived from the mutated background (Fig. 5 B). However, in the spleen of these animals, Ets-1^−/− cells represented <15% of CD4⁺ T cells (Fig. 6), indicating that conventional mutated CD4⁺ T cells have a significant competitive deficit compared with WT cells. Within the CD4⁺ subset, the majority of Foxp3⁺ was derived from the WT. The deficit of Ets-1^−/− CD4⁺Foxp3⁺ cells, and consequently the increased percentage of Ets-1^−/− CD4⁺Foxp3⁻ splenocytes, was significant in our series of chimera mice (Fig. S6). Furthermore, although around 60% of WT T reg cells were CD103⁺, ∼88% of Ets-1^−/− Foxp3⁺CD4⁺ expressed CD103 (Fig. 6). In contrast, ∼50% of the very few Ets-1^−/− Foxp3⁺CD4⁺ thymocytes expressed CD103 as compared with 54% in the WT counterparts (Fig. S7 A). Thus, in these lymphopenic chimeras, the severe quantitative deficit of Ets-1^−/− thymic T reg cells correlated with a significantly lower frequency of antigen-experienced T reg cells in the thymus than in the periphery. This reinforced our hypothesis that in viable Ets-1^−/− mice, recirculating Foxp3⁺ Ets-1^−/− cells were masking a quantitative deficit in the development of thymic T reg cells.

We next investigated the activation status of Ets-1^−/− CD4⁺ and CD8⁺ T cells in the spleen of fetal liver chimera mice. As described for Ets-1 viable mice, in chimeras that only received Ets-1^−/− fetal liver cells, >80% of CD4 cells were CD44^highCD62-L⁻, and <1% had a naive phenotype (CD62-L⁺CD44^low; Fig. S7). In contrast, in mixed 1:20 WT/KO chimera, the percentages of CD62-L⁺CD44^high and CD62-L⁺CD44^low within Ets-1^−/− CD4⁺ and CD8⁺ subsets were markedly increased (Fig. S7). Although the frequency of CD44^high cells did not change, the percentage of cells that expressed CD62-L was reduced by half. Interestingly, a high percentage of WT T cells in chimera mice did not express a typical naive phenotype (CD62-L⁺CD44^low), suggesting that our lymphopenic host mouse model may lead to the underestimation of the frequency of naive T cells (Fig. S7). Although spleens from 1:20 chimera contained a significant percentage of Ets-1^−/− T cells, none of them had elevated spleen cells numbers nor displayed splenomegaly (Fig. S7 B and not depicted). In contrast, mice reconstituted only with Ets-1^−/− fetal liver cells (KO chimera) were affected by both defects. Altogether, these results show that the presence of WT cells in mixed chimera is sufficient to control spleen cell numbers but to only partially rescue Ets-1^−/− naive CD4 and CD8 splenocytes.

Next, we investigated whether Ets-1^−/− CD4⁺ T cells could up-regulate Foxp3 after in vitro activation. CD4⁺CD62-L⁺ splenocytes were FACS sorted from Ets-1^−/− and WT mice and activated with anti-CD3 in IL-2– and TGF-β–containing medium. 48 h after activation, the percentage of Foxp3-expressing cells was lower in Ets-1^−/− (21%) than in WT cultures (50%; Fig. 7). We should note that, as shown in Fig. 1 B, 90% of the splenic CD4⁺CD62-L⁺ T cell population of Ets-1^−/− mice also expressed high levels of CD44, as compared with 5% in the control population. Therefore, it is not excluded that the differences observed are caused by a differential susceptibility to anti-CD3/IL-2 and TGF-β treatment among CD4 T cell subsets rather than a specific effect caused by Ets-1 inactivation. These results indicate that inactivation of the Ets-1 transcription factor reduces but does not abrogate the capacity of Ets-1^−/− CD4⁺ T cells to up-regulate Foxp3 in vitro.

To examine the in vitro suppressive function of CD4⁺CD25⁺ spleen cells that develop in the Ets-1^−/− background, naive T cells (CD4⁺CD62-L⁺CD25⁻) were sorted from WT C57BL/6 mice and were activated in vitro in the presence of various numbers of sorted 90% Foxp3-expressing T reg cells (CD4⁺CD25^high) from Rag/Ets-1^+/+ or Rag/Ets-1^−/− chimera mice. At high concentrations, both WT and mutant T reg cells inhibited anti-CD3–induced T cell proliferation (Fig. 8 A). However, the suppressive activity of mutant T reg cells was less efficient as the same level of inhibition was achieved by four- to fivefold fewer WT T reg cells (Fig. 8 A). Thus, Ets-1^−/− cells have a decreased in vitro suppressive activity.

To determine the importance of Ets-1 in T reg cell in vivo function, we used the mouse model of inflammatory bowel disease (IBD) by adoptively transferring WT naive T cells with or without cotransferring WT or Ets-1^−/− T reg cells into Rag2^−/− recipient mice. As expected, recipient mice that received WT naive CD4⁺CD25⁻CD62-L⁺ T cells (pathogenic) alone developed severe colitis 6–7 wk after cell transfer, with thickening of the colon, loss of normal, pellet-shaped stools, and marked mononuclear cellular infiltrate (Fig. 8 B). Rag2^−/− mice that were cotransferred with WT CD4⁺CD25⁺ T reg cells and WT naive T cells were protected from developing colitis with normal pellet-shaped stools and normal mucosal structure. In marked contrast, Ets-1^−/− T reg cells did not prevent the development of colitis after being cotransferred with WT naive T cells, and mice showed features of colitis, including thickening of the colon and mononuclear cellular infiltrate (Fig. 8 B).

This result demonstrates that Ets-1^−/− T reg cells do not protect from naive T cell–induced IBD. This could be because of decreased suppressive function (as shown in vitro), although decreased survival of mutant T reg cells both in vivo and in vitro cannot be excluded.

To further characterize the function of Ets-1 in T reg cells, we analyzed Foxp3 transcriptional regulatory sequences. A previous study has identified an enhancer element (CNS2) located between exon −2b and −1 of Foxp3 that contains T reg cell–specific STAT5- and CREB-binding sites (Kim and Leonard, 2007). Furthermore, this region displays a CpG island that appears to be specifically demethylated in T reg cells. Analysis of human and mouse enhancer regions with TFSEARCH software (http://www.cbrc.jp/research/db/TFSEARCH.html) confirmed the presence of two conserved ETS-binding sites located near STAT5/CREB-interacting sites that were previously suggested by Floess et al. (2007; Fig. 9 A). Ets-1 binding to the native Foxp3 gene in mouse primary cells was assessed using sorted CD4⁺CD25⁻ and CD4⁺CD25^bright splenocytes and chromatin immunoprecipitation (ChIP) with anti–Ets-1, anti-STAT5, or irrelevant (Ig) antibodies. Subsequent real-time PCR amplification of the Foxp3 gene surrounding putative binding sites in the CNS2 region showed expected STAT5 but also significant Ets-1 binding in T reg cells but not in naive cells (Fig. 9 B). Conversely, analysis of the Foxp3 promoter region did not reveal significant Ets-1 binding (Fig. 9 B). As a control, we assessed a DNA segment of the IgH constant region (Cµ) that is not expressed in T cells and found neither STAT5 nor Ets-1 binding. To investigate the functional significance of ETS-binding sites on the activity of CNS2, the minimal enhancer region (including the two ETS-binding sites) was cloned into the pGL3-promoter vector, and luciferase activity was measured into transfected Jurkat T cells. Whereas the intact enhancer exhibited an approximately sixfold increase in activity in response to PMA plus ionomycin, mutation of either ETS-binding site (or both) dramatically reduced CNS2 induced function (Fig. 9 C). Thus, ETS-binding sites are required for the enhancer function of the CNS2 region. To further investigate how Ets-1 binding impacted the Foxp3 locus, the methylation status of CNS2 was analyzed by bisulphite sequencing on DNA isolated from male T cells (see Materials and methods). Interestingly, striking differences between WT and Ets-1^−/− CD4⁺CD25^bright T cells could be observed. CpG motifs within our amplified sequence displayed a high degree of methylation (100%) within conventional CD4⁺CD25⁻ T cells from both genotypes (Fig. 9 D). However, whereas the amplified region was unmethylated within WT T reg cells, CpG dinucleotides in this region were largely methylated in Ets-1^−/− T reg cell (Fig. 9 D). This result demonstrated that Ets-1 is essential for epigenetic modulations associated with Foxp3 expression.

The data presented in this study show that Ets-1 is required for normal development and function of T reg cells and that defects in this cell subset were responsible for some of the immunological disorders in Ets-1^−/− mice. Viable young mutant animals had reduced numbers of T reg cells in the spleen, but the frequency in the thymus appeared normal. In both sites, Ets-1^−/− T reg cells had an unusual phenotype in that they expressed CD103 (Fig. 4 B), a marker typical of cells that experienced antigen under certain inflammatory conditions (Huehn et al., 2004; Suffia et al., 2005). This raised the possibility that the majority of thymic Ets-1^−/− T reg cells were antigen-experienced recirculating cells, thereby masking an important quantitative deficit in thymic development of these cells. This was indeed the case, as supported by the very low frequency of thymic T reg cells in FTOCs (Fig. 4 C), in 5-d-old newborns (not depicted), and in mixed WT/KO chimeras analyzed at relative early time points (Figs. 5 and 6).

Ets-1^−/− mice have a severe phenotype that affects other lymphocyte lineages. In such context, the T reg cell deficit could be secondary to T reg cell–independent abnormalities, such as the T cell lymphopenia, the hyperactivation status of conventional T cells, and dysfunctions resulting thereof. As demonstrated in this study using mixed WT/KO fetal liver chimeras, normal T reg cell development was achieved in a thymic environment even with an excess of Ets-1^−/− cells. In contrast, the same environment did not rescue the deficit of T reg cell development of Ets-1^−/− precursors, ruling out a potential role of extrinsic factors, e.g., IL-2 or TGF-β, in this deficit. These results strongly supported a T reg cell–intrinsic role of Ets-1.

Normal T reg cells were shown to prevent Ets-1^−/− T cell–mediated splenomegaly and B cell dysfunctions observed in viable mutants. This block correlated with decreased accumulation of Ets-1^−/− T cells, drastic reduction in the capacity of mutant CD4⁺ T cells to produce IL-4, and significant decrease of the proportion of CD44^highCD62-L⁻ cells to the benefit of activated/memory cells coexpressing high levels of these two molecules (Fig. S7). Moreover, cells with a naive phenotype (CD44^lowCD62-L^high) were now clearly detected, but their frequency was still significantly lower than in controls. Thus, Ets-1^−/− conventional T cells were highly susceptible to key T reg cell–mediated negative signals, but the failure to accumulate high numbers of naive cells could indicate that Ets-1^−/− conventional CD4 T cells did not respond to T reg cell–mediated signals inhibiting activation or that they were in an intrinsic state of activation caused by the lack of Ets-1.

Although this issue is difficult to assess given the lymphopenic context of Ets-1^−/− and mixed WT/KO chimeras, our experiments also provided evidence suggestive of cell intrinsic defects in conventional Ets-1^−/− CD4 T cells, which provide possible explanations for the distinct spectrum of immune disorders in these mice as compared with other T reg cell–defective models with partial or complete impaired Foxp3 function. First, these cells had reduced growth potential and/or survival in vitro (Bories et al., 1995; Muthusamy et al., 1995). This could explain the T cell lymphopenia, despite the severe T reg cell deficit, instead of uncontrolled T cell expansion. Second, they showed enhanced Th2 polarization but poor capacity to produce IFN-γ (Fig. 1 and Fig. S2) or to induce IBD in Rag2-deficient hosts (not depicted). Together with the absence of IBD in viable Ets-1^−/− mice, these results suggest that inactivation of Ets-1 might negatively impact Th1 polarization, which is in line with the finding that T-bet function was markedly impaired in Ets-1 mutated T cells, stimulated under Th1-skewing conditions (Grenningloh et al., 2005). Collectively, these results suggest that Ets-1 inactivation also affects conventional T cells in a cell-intrinsic manner and provide an explanation for the absence of massive T cell expansions and enhanced Th2 polarization in mutant animals.

Ets-1 may regulate Foxp3 expression trough several nonmutually exclusive ways. It could have an indirect effect by regulating the transcription of genes encoding direct regulators of Foxp3 or by controlling expression of signaling molecules acting upstream of Foxp3 (i.e., IL-2R or TCR). However, our finding that Ets-1 binds to the CNS2 region and that the enhancer activity of this element relies on ETS-binding sites rather supports a more direct effect. Histone modification and DNA demethylation have been associated with stable expression of Foxp3 in both thymic and peripherally generated T reg cells (Floess et al., 2007; Polansky et al., 2008). In this context, inactivation of nuclear factors such as Runx–CBF-β that are involved in chromatin remodeling processes markedly impaired the T reg cell compartment (Kitoh et al., 2009; Klunker et al., 2009; Rudra et al., 2009). Interestingly, Runx–CBF-β^−/− and Ets-1^−/− mice both displayed T reg cell deficit associated with hyper IgE secretion and enhanced Th2 polarization (Kitoh et al., 2009). Thus, as the Runx–CBF-β complex, Ets-1 could contribute to trigger the epigenetic changes required for stable Foxp3 expression. This hypothesis is strengthened by the highly methylated status of the CNS2 region in splenic Ets-1^−/− T reg cells that correlated with the reduced level of Foxp3 transcripts and proteins (Figs. 4 and 9 D). However, although CBF-β was required to maintain high levels of Foxp3 in the periphery, unlike Ets-1^−/− mice, CBF-β–deficient mice had normal thymic T reg cell development (Kitoh et al., 2009; Rudra et al., 2009). This suggests that Ets-1 function may not be restricted to the context of such epigenetic modifier nuclear complexes.

We report in this study that functional Ets-binding sites within the CNS2 region are located near sequences occupied by STAT5 and CREB nuclear factors in T reg cells (Fig. 9 A; Kim and Leonard, 2007). Interestingly, Ets-1 has been shown to interact with STAT5 in T cells that were activated in vitro in the presence of IL-2 and, upon those conditions, to bind DNA target sites (Rameil et al., 2000). Thus, during thymic T reg cell development, Ets-1–STAT5 complexes could interact with Foxp3 regulatory sequences, thereby ensuring optimal expression of the gene. Furthermore, upon cellular activation, phosphorylated Ets-1 and CREB are both able to associate with CBP/p300, two proteins know to carry histone acetyl transferase activities (Chrivia et al., 1993; Yang et al., 1998; Foulds et al., 2004). Thus, for thymocytes that experience strong TCR/CD28 and IL-2 signals, Ets-1 activity could be required to efficiently target a large complex of nuclear factors to the Foxp3 intronic enhancer. However, gene-targeted deletion of the CNS2 region was shown to impair heritable Foxp3 expression in dividing cells but, unlike Ets-1 inactivation, not to affect the development of thymic T reg cells (Zheng et al., 2010). These data suggest that Ets-1 could interact with additional regulatory sequences within the Foxp3 locus.

The CNS1 sequence appears to be essential for TGF-β–mediated Foxp3 induction in conventional T cells but dispensable for T reg cell thymic development. We show in this study that TGF-β could induce Foxp3 expression in Ets-1^−/− T cells in vitro (Fig. 7), and because we failed to detect Ets-1 binding to the Foxp3 promoter (Fig. 9 B), it seems unlikely that neither the CNS1 nor the promoter would be required for Ets-1–mediated regulation of the Foxp3 gene. An additional region, CNS3, was shown to increase the frequency of T reg cells generated in the thymus and in the spleen, but its deletion left the levels of Foxp3 unaffected (Zheng et al., 2010). Thus, Ets-1 may not act through a single regulatory region but rather interact at several sites, and only from these combined activities would proper expression and epigenetic modification at the Foxp3 locus be achieved.

Altogether, our data support a model in which Ets-1 participates in a nucleoprotein complex whose function would be regulated by TCR and/or γ-chain signaling pathways to induce and lock up Foxp3 expression during the development of thymic T reg cells. These results identify Ets-1 as a key regulator of the immune system and of the biology of T reg cells.

Mice carrying the inactivated ets-1 gene have been previously described (Bories et al., 1995; Eyquem et al., 2004a). All mice used in this study, including WT C57BL/6 mice, were maintained in our specific pathogen-free breeding facility (Departement d'Experimentation Animal, Institut Universitaire d'Hématologie, Saint Louis, France) and were sacrificed for analysis between 5 and 6 wk of age. All mouse experiments were subject to approval by the Institut Universitaire d'Hémathologie Institutional Animal Care and Use Committee.

Single cell suspensions were stained with antibodies according to standard procedures and analyzed on a FACScalibur or a Canto flow cytometer (BD). FACS results were analyzed with CellQuestpro (BD) or FlowJo software (Tree Star, Inc.). The following antibodies were purchased from BD: anti–CD8-α (53-6.7), anti-CD4 (L3T4), anti-CD25 (7D4), anti–CD3-ε (clone 145-2C11), anti-CD44 (IM7), anti-CD69 (H1.2F3), anti-CD95 (Jo2), anti-CD95L (MFL3), anti–CD62-L (MEL-14), anti-CD103 (M290), anti-CD45.2 (104), anti-CD23 (B3B4), anti-CD21 (7G6), anti-CD86 (GL1), and anti–I-A/I-E (M5/114.15.2). mAbs were conjugated with FITC, phycoerythrin, biotin or allophycocyanin, Pacific blue, streptavidin PerCP-Cy5.5, and streptavidin Cy-Chrome.

Cell sorting was performed on a FACSvantage SE (BD), and in all cases, the purity of the sorted cells was estimated at >99%. FACS DNA content analysis was performed as described previously (Lagresle et al., 2002). For intracellular staining of Foxp3, single cell suspensions from spleens and thymi were stained with anti-CD4, anti-CD8, and anti-CD25 or anti-CD3. Subsequent intracellular staining for Foxp3-expressing cells was performed using the phycoerythrin anti-Foxp3 staining kit (eBioscience) according to the manufacturer's instructions.

Sorted CD4⁺CD25⁻CD62-L⁺ T cells and irradiated Rag2^−/− spleen cells (1:3 ratio) were cultured in RPMI medium containing 10% fetal calf serum, 5 µg/ml soluble anti-CD3 (2C11), 3 ng/ml TGF-β1, and 20 U/ml mouse recombinant IL-2. Foxp3 expression was measured after 72 h of activation.

Sorted naive WT CD4⁺CD25⁻ T cells (1.5 × 10⁴ cells) were set up in triplicates in 96-well round-bottom plates (Costard) with 10⁵ irradiated Rag2^−/− spleen cells and mixed or not with variable numbers of FACS sorted WT or Ets-1^−/− CD4⁺CD25^high T reg cells to obtain the indicated ratios (T reg cells/naives) in medium containing 5 µg/ml anti-CD3. Intracellular FACS analysis of sorted CD4⁺CD25^high T reg cells revealed Foxp3 expression in >90% of the cells. Proliferation was assessed after 3 d by pulsing the cells with 1 µCi/well [3H] thymidine (GE Healthcare) for the last 16 h of culture. Cells were harvested onto filter membranes, and the amount of incorporated [3H] thymidine was measured with a liquid scintillation counter Wallac1409 (PerkinElmer). Data are representative of three independent experiments.

All mice used for chimera were in a C57BL/6 background. CD45.2/CD45.1 Rag2^−/−γc^−/− chimera mice were generated as previously described (Eyquem et al., 2004b). For T cell reconstitution of CD3-ε–deficient mice, thymi were isolated from 18.5 pc embryos and grafted under the kidney capsule of CD3ε^−/− mice. 5–6 wk after transfer, recipient chimera mice were sacrificed for analysis. For T reg cell transfers, spleen cell suspensions were stained with anti–CD25-biotin antibody (7D4), followed by incubation with strepavidin-coupled MicroBeads (Miltenyi Biotec). The cells were passed twice through a MACS separation column (Miltenyi Biotec) to obtain a >95% pure CD25⁺CD4⁺ population. 5 × 10⁵ purified CD25⁺CD4⁺ cells were injected i.v. into Rag2^−/− recipient mice. 5 × 10⁵ of electronically sorted CD4⁺CD25⁻CD62-L⁺ naive T cells were coinjected in the indicated mice.

RT-PCR reactions were performed in a 50-µl reaction containing the indicated amounts of genomic DNA, 2 ng/µl of each primer, 0.2 µM deoxy-nucleoside triphosphate, 2 mM MgCl₂, 50 mM KCl, 10 mM Tris-HCl, pH 8.8, 0.1% Triton X-100, and 1 U/50 µl Taq polymerase. Reactions were performed for 4 min at 95°C; 35 cycles of 1 min at 95°C, 1 min at 57°C, and 1.5 min at 72°C; and 5 min at 72°C.

Real-time RT-PCR was performed using the LightCycler FastStart kit (Roche; Espeli et al., 2006) on a LightCycler system. Amounts of RNA were calculated by reference to a log-linear standard regression curve. This curve was constructed from the number of cycles necessary to detect product accumulation after amplification of 2 µl of serial dilutions of a standard cDNA solution. The calculation of the relative concentration was performed using the LightCycler software (Roche).

Primers used were hypoxanthine-guanine phosphoribosyltransferase (HPRT) forward, 5′-CACAGGACTAGAACACCTGC-3′; HPRT reverse, 5′-GCTGGTGAAAAGGACCTCT-3′; IL-2 forward, 5′-CCTGAGCAGGATGGAGAATTACA-3′; IL-2 reverse, 5′-GCCTTATGTGTTGTAAGCAGGAGG-3′; IL-4 forward, 5′-TGACGGCACAGAGCTATTGA-3′; IL-4 reverse, 5′-ATGGTGGCTCAGTACTACGA-3′; IL-5 forward, 5′-ATGAGGCTTCCTGTCCCTACTC-3′; IL-5 reverse, 5′-GTCACCATGGAGCAGCTCAGCC-3′; IL-10 forward, 5′-GGTTGCCAAGCCTTATCGGA-3′; IL-10 reverse, 5′-ACCTGCTCCACTGCCTTGCT-3′; IL-13 forward, 5′-GGTGCCAAGATCTGTGTCTCTCC-3′; IL-13 reverse, 5′-GAGATGCCCAGGGATGGTCTCC-3′; IFN-γ forward, 5′-GCTCTGAGACAATGAACGCT-3′; IFN-γ reverse, 5′-AAAGAGATAATCTGGCTCTGC-3′; Foxp3 forward, 5′-CCAAGGATCCTACCCACTGCTGG-3′; and Foxp3 reverse, 5′-CCCAGAGGTGCCTCCGCACTGC-3′.

The minimal CNS2 region of the Foxp3 gene was PCR amplified and cloned into pGL3-Promoter (Promega) to yield pFoxp3CNS2. PCR primers used were 5-EFoxp, 5′-CACACAGTAAGAAGGTGGATCCATGC-3′; and 3-EFoxp 5′-CTGGGCTGGC­CAGCCAGCTTC­CTGCACTGTC-3′.

The site-directed mutagenesis kit (Agilent Technologies) was used to mutate the GGA motif into CCA in the targeted ETS sites. Primers for mutagenesis were as follows: FwMt E1, 5′-ACGTCACCTA­CCTCTAGGGCT­AGCACCCAC-3′; RvMt E1, 5′-GTGGGTGCTA­GCCCTAGAGGT­AGGTGACGT-3′; FwMt E2, 5′-AGCCCAATCCATGGGGCCGCCATGACG-3′; and RvMt E2, 5′-CGTCATGGCGGCCCCATGGATTGGGCT-3′. The integrity of all the CNS2 regions inserted in the pGL3-Promoter vector was established by sequencing.

Jurkat cells were transiently transfected with 1 µg of the indicated pGL3-luciferase plasmids using electroporation with the Cell Line Nucleofector kit V (Lonza). For each transfection, 0.5 µg of β-galactosidase–expressing plasmid (Invitrogen) was added for normalization. Activation and luciferase assays were performed as previously described (Kim and Leonard, 2007).

ChIP assays were performed using the Imprint ChIP kit (Sigma-Aldrich) as previously described (Eyquem et al., 2002). CD4⁺CD25^bright and CD4⁺CD25⁻ cells were sorted from spleens and stimulated with 100 U/ml IL-2 for 1 h. Formaldehyde (final concentration 1%) was then added to cross-link proteins and DNA. The cell lysates were sonicated and immunoprecipitated with normal rabbit serum (Millipore), α-STAT5 (Santa Cruz Biotechnology, Inc.), and α–Ets-1 (Santa Cruz Biotechnology, Inc.). The immunoprecipitated DNA was eluted and amplified by real-time SYBR green PCR using an ABI 7300 (Applied Biosystems). Values were normalized to serial dilution of corresponding input controls and are expressed as fold enrichment relative to normal rabbit serum for each experiment. The sequence-specific primers used for amplification of the Foxp3 gene are CNS2, 5′-ATCTGGCCAAGTTCAGGTTG-3′ and 5′-GGAACTGGTGTGACTGTGTGA-3′; and Foxp3 promoter, 5′-CACTCAGAGACTCGCAGCAG-3′ and 5′-GGGGTAGTGCTCTGTCTCCA-3′. Primers for IgH Cµ irrelevant region are 5′-GGCCTCGCAGATGAGTTTAG-3′ and 5′-GTGCCCATTCCAGGTAAGAA-3′.

Concentrations of serum polyclonal IgM, IgG1, IgE, and IgG2b were determined using isotype-specific ELISA (SouthernBiotech). The concentrations were determined by comparing test sample dilution series with isotype control standards (SouthernBiotech). For IgE detection, the capturing antibody was anti-IgE (R35-72; BD). For detection of autoantibodies, the capturing agent used was double-stranded sheared salmon sperm DNA (Eppendorf) at 10 µg/ml.

Genomic DNA was isolated from electronically sorted male T cells as previously described (Eyquem et al., 2002). Sodium bisulphite treatment of genomic DNA was performed using the QIAGEN kit according the manufacturer's protocol. In a subsequent PCR amplification, uracils were replicated as thymidines. PCRs were performed on Eppendorf thermocyclers in a final volume of 25 µl containing 1× PCR buffer, 1 U Taq DNA polymerase (Promega), 200 µM deoxy-nucleoside triphosphate, 12.5 pmol each of forward and reverse primers, and 7 ng of bisulphate-treated genomic DNA. The amplification conditions were 95°C for 15 min and 40 cycles of 95°C for 1 min, 55°C for 45 s, and 72°C for 1 min, and a final extension step of 10 min at 72°C. PCR products were purified and cloned into Topo vectors. Plasmid from individual clones was purified and sequenced in both directions applying the PCR primers and the ABI Big Dye Terminator v1.1 cycle sequencing chemistry (Applied Biosystems), followed by capillary electrophoresis on an ABI 3100 genetic analyzer (Applied Biosystems). Trace files were interpreted using BiQ Analyzer (Max-Planck-Institut, Saarbrucken, Germany), which allows for quantification of methylation signals. For each sample, both PCR amplification and sequencing were repeated once. The following primers (5′ to 3′ direction) were used for both PCR amplification of bisulphite-converted genomic DNA and sequence reactions: Amp 2 forward, 5′-ATTTGAATTGGATATGGTTTGT-3′; and reverse, 5′-AACCTTAAACCCCTCTAACATC-3′.

Fig. S1 shows size and phenotypic analysis of splenic T cells from Ets-1^−/− mice. Fig. S2 shows in vitro cytokine secretion by WT and Ets-1^−/− T cells. Fig. S3 depicts the organization of spleens from Ets-1^+/+ and Ets-1^−/− mice. Fig. S4 shows that Ets-1^−/− T cells are responsive to suppression from WT T reg cells. Figs. S5 and S6 show an intrinsic defect of Ets-1^−/− thymic T reg cell development and splenic T reg cells, respectively. Fig. S7 shows that WT cells partially rescue Ets-1^−/− naive CD4 and CD8 splenocytes. Table S1 shows the numbers of B and T cells (CD4⁺ and CD8⁺) in spleens from Ets-1^+/+ and Ets-1^−/− mice.

We thank M. Goodhardt and R.S. Allan for criticism of the manuscript.

This work was supported by a grant from the Ligue Nationale Contre le Cancer (RS 06/75-16). K. Chemin and E. Mouly were supported by fellowships from the Fondation pour la Recherche Médicale, and O. Burlen-defranoux and A. Bandeira were supported by a grant from the Association pour la Recherche sur le Cancer (ARC #1007).

The authors have no conflicting financial interests.

Author contributions: E. Mouly, A. Bandeira, and J.-C. Bories designed the research and analyzed the data. E. Mouly, K. Chemin, O. Burlen-defranoux, M. Chopin, L. Mesnard, M. Leite-de-Moraes, and H.V. Nguyen performed experiments. A. Bandeira had major contributions to the work. A. Bandeira and J.-C. Bories wrote the manuscript.