|Home | About | Journals | Submit | Contact Us | Français|
Appropriate development of regulatory T cells (Tregs) is necessary to prevent autoimmunity. Neonatal mice, unlike adults, lack factors required for Treg development. It is unclear what these missing factors are. However, signals emanating from the TCR, CD28 and γc-dependent cytokine receptors are required for Treg development. Herein we demonstrate that expression of a constitutively-active STAT5b transgene (STAT5b-CA) allows for Treg development in neonatal mice and restores Treg numbers in CD28−/− mice. Sequence analysis of TCR genes in STAT5b-CA Tregs indicates that ectopic STAT5 activation results in a TCR repertoire that more closely resembles that of naïve T cells. Using MHCII tetramers to identify antigen-specific T cells, we demonstrate that STAT5 signals divert thymocytes normally destined to become naïve T cells into the Treg lineage. Our data support a two-step model of Treg differentiation in which TCR/CD28 signals induce cytokine responsiveness; STAT5-inducing cytokines then complete the program of Treg differentiation.
Autoimmune disease is a consequence of the generation of self-reactive T cells. While most self-reactive T cells are deleted in the thymus via negative selection, some self-reactive T cells escape this process. Therefore, a second mechanism involving the generation of CD4+ regulatory T cells (Tregs) in the thymus has evolved to prevent autoimmunity (Shevach, 2000). Although several distinct types of T cells capable of exerting suppressive activity have now been described, the most well characterized are CD4+ T cells that express the transcription factor Foxp3 (referred to hereafter as Foxp3+ Tregs). Signals required for Foxp3+ Treg differentiation have been shown to include high affinity interactions between the T cell receptor (TCR) and major histocompatability class II (MHC II):peptide complexes (Bensinger et al., 2001; Aschenbrenner et al., 2007). For example, several mouse models have been developed in which both a specific TCR and its respective antigen are co-expressed as transgenes. A large percentage of T cells that develop in those mice were found to be Foxp3+ Tregs that exhibit suppressor activity (Jordan et al., 2001; Apostolou et al., 2002; Aschenbrenner et al., 2007). Likewise, work from three different groups has demonstrated that the TCR repertoire of Foxp3+ Tregs overlaps extensively with that of self-reactive T cells; in contrast, only limited overlap was seen with naïve peripheral T cells (Hsieh et al., 2006; Pacholczyk et al., 2006; Wong et al., 2007). Other signals required for Foxp3+ Treg development include CD28:B7 costimulatory signals (Tang et al., 2003; Tai et al., 2005). This has been attributed, in part, to a role for CD28 in generating IL2 required for Foxp3+ Treg homeostasis. However, CD28 appears to play an additional role in Foxp3+ Treg development that has not yet been defined (Tai et al., 2005). Taken together, these studies have led to the conclusion that effective Foxp3+ Treg differentiation requires both relatively high affinity TCR interactions with MHCII:peptide complexes as well as CD28:B7 costimulatory signals.
In contrast to the clearly accepted importance of TCR- and CD28-dependent signals for Foxp3+ Treg differentiation, the role of IL2 in this process has been quite controversial. Early studies by Malek and colleagues using IL2Rβ transgenic mice supported a role for IL2Rβ-dependent signals in Foxp3+ Treg development in the thymus (Malek et al., 2000; Malek et al., 2002). In contrast, studies by Lafaille and colleagues demonstrated that transfer of CD4+ T cells from IL2−/− mice could prevent autoimmune disease in a mouse model that normally develops spontaneous experimental autoimmune encephalomyelitis (EAE)(Furtado et al., 2002). Transfer of IL2Rα−/− CD4+ T cells did not prevent EAE in this model, suggesting a requirement for IL2 to support Foxp3+ Treg suppressive capacity but not for Foxp3+ Treg development. Additional studies from the Rudensky and Klein labs independently demonstrated that young IL2−/− mice have relatively normal numbers of thymic and peripheral Foxp3+ T cells. These latter findings seemed to convincingly rule out a requirement for IL2 in Foxp3+ Treg differentiation (D’Cruz and Klein, 2005; Fontenot et al., 2005b). However, unlike Foxp3+ Treg development in IL2−/− or IL2Rα −/−mice, we have recently found that Foxp3+ Treg development in IL2Rβ −/− and IL2−/− x IL15−/− mice is dramatically impaired (Burchill et al., 2007). Similar findings were reported by Ziegler and colleagues (Soper et al., 2007). IL2 appears to play the primary role in this process based on competitive mixed bone marrow chimera studies using IL2Rα+/+ and IL2Rα −/− donors; in those studies 5-fold more Foxp3+ Tregs in the resulting chimeric mice were derived from IL2Rα+/+ as opposed to IL2Rα −/− donor cells (Fontenot et al., 2005b). Thus, IL2Rβ-dependent signals play a key role in effective Foxp3+ Treg development.
Several pieces of evidence indicate that IL2 (and IL15) entrain Foxp3+ Treg differentiation via activation of STAT5. For example, transgenic mice expressing a constitutively active form of STAT5 (called STAT5b-CA) have a significant increase in the percentage and total number of Foxp3+ Tregs in both the thymus and spleen (Burchill et al., 2003). Furthermore, crossing the STAT5b-CA transgene onto the IL2Rβ −/− background restores Foxp3+ Treg development in the thymus. Likewise, IL2Rβ mutants that have been engineered to signal exclusively through STAT5 also restore Foxp3+ Tregs (Burchill et al., 2007). Conversely, mice lacking STAT5 in developing thymocytes fail to generate Foxp3+ Tregs (Burchill et al., 2007; Yao et al., 2007). Thus while multiple cytokines can induce Foxp3+ Treg differentiation, they all do so via STAT5 activation.
Although we now have some appreciation of the signals required to induce Foxp3+ Treg differentiation, the process this still remains somewhat enigmatic. Work from the Rudensky and Chatila labs has demonstrated that mice lacking functional Foxp3 exhibit precursors that can induce transcription of a GFP cDNA knocked into the foxp3 gene locus. These cells exhibit many other hallmarks of Foxp3+ Tregs including increased CD25 expression, but do not exhibit suppressor activity or prevent autoimmune disease in vivo (Gavin et al., 2007; Lin et al., 2007). Likewise, Fontenot et al reported that neonatal mice, which lack Foxp3+ Tregs, do express a population of CD4+CD25+Foxp3− thymocytes (Fontenot et al., 2005a). It was unclear at that time what these cells were, although it was speculated that they might be Foxp3+ Treg progenitors that failed to complete differentiation due to the absence of as yet uncharacterized factors. Thus, a key question that remains is whether high-affinity TCR/CD28 signals induce an autonomous program that leads to Foxp3+ Treg differentiation or whether this process occurs via multiple steps and requires the influence of other factors such as cytokines. Herein, we present data supporting a model of Foxp3+ Treg lineage commitment in which high affinity TCR/CD28 signaling leads to the generation of thymocytes with enhanced cytokine responsiveness, and that a second step, involving cytokine/STAT5-dependent signals completes the process of Foxp3+ Treg differentiation. Moreover, this linkage of TCR/CD28- and cytokine/STAT5-dependent signals plays an important role in shaping the TCR repertoire of Foxp3+ regulatory T cells.
To test potential cooperation between TCR and IL2R signaling pathways in promoting Foxp3+ Treg development, we utilized a transgenic mouse model in which a mutant form of the transcription factor STAT5 (referred to herein as STAT5b-CA) is expressed throughout lymphocyte development (Burchill et al., 2003). Although the STAT5b-CA construct we used is typically referred to as constitutively active, it still requires phosphorylation on Tyr-699 for activation (Onishi et al., 1998). In the absence of exogenous cytokines, the STAT5b-CA transgene results in low levels of constitutive STAT5 activity ex vivo. Phosphorylation of STAT5b-CA is further increased following exposure to STAT5 activating cytokines; thus, the transgene is best described as a weakly constitutively active, but hyperactivateable form of STAT5b ((Will et al., 2006) and Fig. 5C). Using these mice, we examined whether STAT5-dependent signals could induce Foxp3+ Treg development in one day old neonatal mice. Previous studies have demonstrated that such mice are essentially devoid of CD4+Foxp3+ Tregs (Fontenot et al., 2005a). We reasoned therefore that if STAT5 signals are sufficient to drive Foxp3+ Treg development, then we should see Foxp3+ Tregs in neonatal mice which normally lack such cells. Consistent with previous findings, we observed that newly generated CD4+HSAhiFoxp3+ Tregs are virtually absent in the thymus of LMC mice; in contrast, we observed a significant population of CD4+HSAhi Foxp3+ thymocytes in one day old STAT5b-CA mice (Fig. 1). Thus, STAT5b-CA expression promotes the development of Foxp3+ Tregs under conditions which are normally not permissive for the differentiation of such cells.
Although STAT5b-CA expression clearly induces Foxp3+ Treg differentiation in neonatal mice, it is unclear what upstream factors it is compensating for. Previous studies have demonstrated that neonatal dendritic cells express much lower levels of the costimulatory ligands B7-1 and B7-2; furthermore, these cells are quite poor at inducing immune responses (Muthukkumar et al., 2000). Thus, one possibility is that neonatal mice lack sufficient expression of B7-1/B7-2 on thymic antigen presenting cells to induce Foxp3+ Treg differentiation. An alternative possibility is that neonatal mice may lack γc-dependent cytokines (most likely IL2) that are required for Foxp3+ Treg differentiation. To test these two possibilities, we introduced the STAT5b-CA transgene onto CD28−/− and γc−/− backgrounds to examine whether constitutive STAT5 signals can drive Foxp3+ Treg development in the absence of costimulatory- or γc-cytokine-dependent signals. As illustrated in figure 2A, the STAT5b-CA transgene restored Foxp3+ Treg numbers in the absence of γc indicating that STAT5 activation can restore Foxp3+ Treg development in the absence of all γc-dependent cytokines. Importantly, the Foxp3+ Tregs in STAT5b-CA x γc−/− mice are functional suppressor cells. Specifically, in γc−/− mice, which lack Foxp3+ Tregs, the small number of T cells that remain show an activated phenotype as determined by upregulated CD69 and downregulated CD62L expression. This phenomena is reversed in STAT5b-CA x γc−/− mice (Sup. Fig. 1). Moreover, T cell numbers are restored in STAT5b-CA x γc−/− mice, and the mice remain healthy up to at least 9 months of age, suggesting that they have a functional Foxp3+ Treg population. More surprisingly, we found that thymic Foxp3+ Treg numbers were also restored in STAT5b-CA x CD28−/− mice (Fig. 2B,C). Foxp3+ thymocytes from STAT5b-CA x CD28−/− mice retained high expression of additional markers indicative of a Foxp3+ Treg phenotype such as CD25 and GITR suggesting that we had restored a normal population of Foxp3+ regulatory T cells (Fig. 2D). A potential caveat to the above rescue experiments is that it is difficult to determine whether these findings result from induced Foxp3+ Treg differentiation or just enhanced proliferation/survival of the few Tregs that emerge in CD28−/− mice. Nevertheless, these data provide evidence that CD28 signals may be upstream of STAT5 activation and suggest the possibility that CD28 signals induce cytokine responsiveness during Foxp3+ Treg differentiation.
We next wanted to determine whether STAT5 activation could restore Foxp3+ Treg development in the absence of high-affinity TCR signals. MHCII−/− mice are defective in positive selection of all CD4+ thymic T cells. Thus, crossing STAT5b-CA mice with MHCII−/− mice is likely to be uninformative. In addition, we wished to be able to unambiguously differentiate between STAT5 effects on post-selection Foxp3+ Treg expansion versus pre-selection induction of Foxp3+ Treg development. To circumvent these issues, we examined whether the repertoire of TCRs expressed in the Foxp3+ Treg pool was perturbed in STAT5b-CA mice. Specifically, recent studies have demonstrated that both thymic and peripheral Foxp3+ T cell populations have diverse TCR repertoires that exhibit very limited overlap with the TCR repertoire of CD4+Foxp3− naïve T cells (Hsieh et al., 2004; Hsieh et al., 2006; Pacholczyk et al., 2006). To test if constitutive activation of STAT5 alters the TCR repertoire of Foxp3+ Tregs we utilized a novel strategy (Moon et al., 2007) employing tetramers of the MHC class II molecule, I-Ab, covalently bound to the 2W1S variant of peptide 52–68 of the MHC class II I-Eα molecule (Crawford et al., 1998; Rees et al., 1999) (referred to hereafter as the 2W1S-tetramer). We used this 2W1S-tetramer in conjunction with magnetic bead enrichment to isolate all 2W1S-specific T cells in the spleen and lymph nodes of a single mouse (Sup. Fig. 2). Enriched cells were subsequently stained with a panel of additional markers to characterize the 2W1S-tetramer specific population (Sup. Fig. 3). Staining was specific as the 2W1S-tetramer-positive population is not observed in mice restricted to a single TCR specificity such as TEa TCR Tg x Rag−/− (Grubin et al., 1997) or SM1 TCR Tg x Rag−/− mice (McSorley et al., 2002)(Sup. Fig. 3). As expected, virtually all 2W1S-tetramer specific CD4+ T cells are naïve (CD44lo) in mice that have not been previously exposed to the 2W1S peptide (Sup. Fig. 3). Importantly, we found that CD4+2W1S+ T cells from B6 mice are almost entirely CD25− and Foxp3− (Fig 3A), demonstrating that this TCR specificity is strikingly underrepresented in the Treg pool. In contrast, approximately 46 ± 9.2 % of 2W1S-tetramer-specific T cells from STAT5b-CA mice are Foxp3+ (Fig 3A,B). Furthermore, the percentage of 2W1S-specific Foxp3+ T cells is not significantly different (p=0.24) than the average distribution of all other TCR specificities in the Foxp3+ Treg lineage (64.2% ± 5.9%) (Fig 3B). These results demonstrate that constitutive activation of STAT5 diverts 2W1S-tetramer-specific cells that would normally develop into naïve CD4+ T cells into the Foxp3+ Treg lineage. Thus, the presence of the STAT5b-CA transgene alters the TCR selection bias that typically characterizes Foxp3+ Treg differentiation.
Although the above findings show that expression of the STAT5b-CA transgene results in selection of a large fraction of 2W1S-tetramer specific T cells into the Foxp3+ Treg lineage, the conclusions drawn from these experiments are limited due to the fact that only one TCR specificity was assayed. Therefore, we examined the ability of the STAT5b-CA transgene to divert T cell progenitors into the Foxp3+ Treg lineage by analyzing the TCR sequence diversity in Foxp3+ and Foxp3− populations from STAT5b-CA and LMC mice. To address this issue, we made use of mice expressing both a TCRβ transgene and a Foxp3-GFP reporter (Hsieh et al., 2006); the TCRβ transgene fixes TCRβ chain diversity while the Foxp3-GFP reporter allows us to specifically isolate Foxp3+ Tregs. Previous studies have documented that Foxp3+ Tregs in the thymus and spleen show considerable overlap in TCR gene usage; in contrast, there is little overlap in TCR gene usage between naïve T cells and Foxp3+ Tregs in either the spleen or thymus (Hsieh et al., 2004). To examine the effect of ectopic STAT5 activation on the Foxp3+ Treg TCR repertoire, we sequenced over 1000 productively rearranged TCR Vα2 genes from STAT5b-CA and LMC CD4+Foxp3+ and CD4+Foxp3− thymocytes. Consistent with previous reports, we found that TCR Vα2 gene usage in CD4+Foxp3+ thymocytes from LMC mice is different from that observed in CD4+Foxp3− naïve T cells. Specifically, we found that only 5% of unique TCR Vα2 sequences in the CD4+Foxp3− thymocytes from LMC mice or STAT5b-CA mice were also present in the CD4+Foxp3+ Tregs from LMC mice (Fig. 4A, black bars). In contrast, a much greater percentage of the unique TCR Vα2 rearrangements in LMC and STAT5b-CA non-Tregs were found in CD4+Foxp3+ thymocytes from STAT5b-CA mice (15% and 20%, respectively)(Fig. 4A, gray bars). We also examined skewing of the Foxp3+ Treg TCR repertoire by comparing the relative frequency at which specific TCR Vα2 sequences could be found in a larger database of TCR sequences obtained from splenic Foxp3+ Tregs, Foxp3−CD44lo naïve T cells, and Foxp3−CD44hi memory or effector T cells (Hsieh et al., 2006). To analyze these results we plotted individual TCR sequences obtained from CD4+Foxp3+ thymocytes on the x-axis and the relative frequency at which this particular sequence could be found in wild type splenic CD4+Foxp3+ T cells (green), CD4+Foxp3−CD44lo naïve T cells (blue), and CD4+Foxp3−CD44hi memory or effector T cells (pink) on the y-axis (Fig. 4B). This analysis demonstrated an increase in the number of TCR Vα2 sequences found in STAT5b-CA versus LMC CD4+Foxp3+ thymocytes that were also observed in splenic CD4+Foxp3− naïve T cells (Fig. 4B). To quantitate the magnitude of this difference more precisely, we took the area represented by the green, blue, and pink frequency distributions in Fig. 4B and plotted them as a bar graph (Fig. 4C). From this graph it is clear that the presence of the STAT5b-CA transgene greatly expands the number of TCR sequences in the Foxp3+ Treg repertoire that are more typically found in the naïve T cell repertoire (Fig. 4C – compare blue bars for Foxp3+ Tregs in LMC versus STAT5b-CA mice). These findings demonstrate that expression of a constitutively active form of STAT5 alters the TCR selection bias that is normally associated with Foxp3+ Treg development and subsequently expands the repertoire of sequences found in Tregs to include many more non-Treg specific TCR sequences.
To further test the ability of STAT5 activation to drive Foxp3+ Treg differentiation, we made use of a recent observation by Lio and Hsieh who identified CD4+CD25+Foxp3− thymocytes as penultimate progenitors of Foxp3+ Tregs that can be rapidly differentiated into Foxp3+ Tregs by IL2 plus IL7 addition (manuscript submitted). In contrast, they found that the CD4+CD25−Foxp3− thymocyte subset could not be converted into Tregs indicating that they are not Foxp3+ Treg progenitors. We predicted that if STAT5 activation downstream of TCR-dependent selection events is a key second step in Foxp3+ Treg differentiation, then we should be able to convert CD4+CD25−Foxp3− non-Treg progenitors into Foxp3+ Tregs in STAT5b-CA but not LMC mice. To examine this issue, we used Foxp3-GFP reporter mice (Fontenot et al., 2005c) to identify Foxp3+ and Foxp3− thymocyte subsets. We then sorted CD4+Foxp3-GFP−CD25− thymocytes from STAT5b-CA and LMC mice and placed them in culture overnight in either media alone, IL7, or IL7 and IL2. Consistent with reports from Lio and Hsieh, we found that CD4+Foxp3-GFP−CD25− thymocytes neither spontaneously converted into Foxp3+ Tregs, nor could be converted into Foxp3+ Tregs by cytokine treatment. In contrast, CD4+Foxp3-GFP−CD25− thymocytes from STAT5b-CA mice could be converted into Tregs following addition of either IL7, or IL7 and IL2 (Fig. 5A,B); stimulation of these cells with IL2 alone also worked (Sup. Fig. 4) but was substantially less effective as CD4+Foxp3-GFP−CD25− thymocytes not only fail to express the IL2Rα chain but express very low levels of the IL2Rβ chain as well (KBV and MAF, unpublished results) making theses cells very poorly responsive to IL2. In contrast, the synergistic affect of IL7 and IL2 appears to be due to the ability of IL7 to modestly upregulate CD25 in CD4+Foxp3-GFP− thymocytes from STAT5b-CA mice thereby rendering them more responsive to IL2.
The observation that the STAT5b-CA transgene only converts CD4+Foxp3-GFP−CD25− thymocytes into Tregs in vitro following addition of cytokines suggests that a relatively high threshold of STAT5 signaling is required for this process. To test this hypothesis we measured levels of phospho-STAT5 expression in both WT and STAT5b-CA CD4+Foxp3-GFP−CD25− thymocytes in the Treg conversion assay. As shown in Fig. 5C, unstimulated LMC CD4+Foxp3− thymocytes exhibited low levels of phospho-STAT5 expression that was increased following exposure to cytokines. Unstimulated CD4+Foxp3− thymocytes from STAT5b-CA mice exhibited a higher basal level of phospho-STAT5 staining; this is consistent with the constitutively activated nature of the STAT5b-CA transgene. Following stimulation of CD4+Foxp3− T cells from STAT5b-CA mice with IL2 and IL7 we observed two distinct patterns of phospho-STAT5 expression (Fig. 5C). Cells that remained negative for Foxp3-GFP did not exhibit an increase in phospho-STAT5 relative to unstimulated cells. In contrast, cells that induced Foxp3-GFP expression demonstrated a significant increase in phospho-STAT5 staining. These findings indicate that a threshold of STAT5 activation is required for CD4+Foxp3−thymocytes to convert into CD4+Foxp3+ Tregs. Taken together, our findings strongly support a model in which high-affinity TCR/CD28 signals induce the ability to respond to γc-dependent cytokines; subsequent encounter with such cytokines results in the final conversion of Foxp3− Treg progenitors in the thymus into Foxp3+ Tregs.
The conversion of CD4+CD25+Foxp3-GFP− thymocytes into CD4+Foxp3-GFP+ Tregs resembles, at least superficially, the conversion of mature splenic T cells into Tregs following anti-CD3, TGFβ, and IL2 stimulation that has been well described by others (Chen et al., 2003; Davidson et al., 2007; Zheng et al., 2002). Therefore, we examined whether TGFβ produced by converted Tregs, or potentially present in bovine calf serum, might be required for this process. As shown in Fig. 6A, neutralizing TGFβ had only a modest effect on the conversion of CD4+CD25−Foxp3-GFP− progenitor cells derived from STAT5b-CA mice. Conversely, addition of exogenous TGF-β resulted in minimal enhancement of the conversion process (ranging from no effect up to a 10% increase; data not shown). Moreover, using CD4+CD25+Foxp3-GFP− Treg progenitor cells obtained from wild type mice Lio and Hsieh found that neither adding exogenous TGFβ or neutralizing TGFβ antibodies, affected the conversion process (manuscript submitted). Thus, the cytokine-dependent conversion of the thymic Treg progenitors that we have identified differs significantly from the conversion of splenic CD4+ T cells because it does not require TGFβ. This is consistent with other published data on the requirements for thymic development of natural Tregs which have been shown to be TGFβ independent (Marie et al., 2005).
STAT5 associates with both histone acetyltransferases (HATs) and histone deacetylases (HDACs) and these interactions have been suggested to play an important role in regulating STAT5-dependent responses (Nakajima et al., 2001; Pfitzner et al., 1998). Therefore, we examined whether blocking histone deacetylase activity would augment or inhibit the cytokine-dependent Treg conversion process. For these studies, we inhibited HDACs with either the pan-HDAC inhibitor trichostatin A (TSA), or a more selective inhibitor, apicidin, which preferentially targets HDAC2 and HDAC3 (Khan et al., 2007). TSA functioned as expected by increasing global histone H3 and H4 acetylation in CD4+ thymocytes (Sup. Fig. 5). However, when added into the conversion assay both TSA and apicidin completely blocked the ability of CD4+CD25−Foxp3-GFP− progenitors from STAT5b-CA mice to convert into CD4+CD25+Foxp3-GFP+ Tregs (Fig. 6B). These results suggest that STAT5-dependent recruitment of HDACs is required for transcription of the foxp3 gene. Moreover, given the specificity of apicidin for HDAC2 and HDAC3 it is likely that one of these two HDACs is the relevant target.
Two of the fundamental questions remaining in the field of Foxp3+ Treg development are what governs this developmental process and how is it that the Foxp3+ Treg TCR repertoire is matched to that of self-reactive T cells. Herein we provide data demonstrating that linked TCR/CD28 and STAT5-dependent signals play a key role in both of these processes. Specifically, we demonstrate that constitutive activation of STAT5b induces Foxp3+ Treg development in newborn mice; such animals typically lack Foxp3+ Tregs due to the absence of as yet unidentified factors. Both CD28 and γc-dependent cytokines are potential factors required for Foxp3+ Treg differentiation that may be absent in neonatal mice. Supporting this hypothesis we found that STAT5-dependent signals were sufficient to restore Foxp3+ Treg development in both γc−/− and CD28−/− mice. Finally, we demonstrate that STAT5 signals are sufficient to convert thymocytes that would typically be selected into the naïve T cell pool into the Foxp3+ regulatory T cell lineage. For example, we found that a subset of CD4+Foxp3−CD25−non-Treg progenitors from STAT5b-CA, but not LMC mice, can be converted into Foxp3+ Tregs in vitro. Furthermore, using a novel strategy to identify antigen-specific T cells, we demonstrate that ectopic STAT5 signals can convert cells typically destined to become naïve T cells into the Foxp3+ Treg lineage in vivo. These findings were confirmed by sequencing of the TCR repertoire in Foxp3+ Tregs from STAT5b-CA and LMC mice; the TCR repertoire of Foxp3+ Tregs from STAT5b-CA mice exhibited much greater overlap with the naïve TCR repertoire than that observed for Foxp3+ Tregs in LMC mice. These findings definitively address the ongoing controversy regarding the role of the IL2R and STAT5 signaling in Foxp3+ Treg development. Previous studies using IL2Rβ −/−, γc−/−, and STAT5a/b−/− mice have suggested a role for these molecules in Foxp3+ Treg differentiation (Burchill et al., 2007; Fontenot et al., 2005b; Malek et al., 2000; Yao et al., 2007). However, in those studies it was difficult to distinguish between a role for these factors in post-selection expansion of Foxp3+ Treg numbers versus preselection induction of Foxp3+ Treg differentiation. The findings presented herein that STAT5 signals can divert non-Treg progenitors into the Foxp3+ Treg lineage clearly demonstrate that STAT5 signals play an important role in the pre-selection differentiation Foxp3+ Tregs. Taken together our data support a two step model of Foxp3+ Treg differentiation in which TCR/CD28 signals first enhance the ability of cells undergoing negative selection to respond to IL2R/STAT5 signals; subsequent encounter with STAT5-inducing ligands (most likely IL2) completes the differentiation process that drives Foxp3+ Treg development.
The above model is supported by significant amounts of other data. First, we and others have found that STAT5 binds to the foxp3 promoter in Foxp3+ Tregs, suggesting that STAT5 may induce foxp3 expression (Burchill et al., 2007; Yao et al., 2007). Second, previous reports have documented that thymocytes undergoing negative selection upregulate expression of the IL2Rα and IL2Rβ chains (Bassiri and Carding, 2001). Importantly, IL2Rα upregulation occurs only in CD4 but not CD8 destined thymocytes as the IL2Rα chain is upregulated during negative selection in transgenic mice expressing a MHC class II (HEL) but not a MHC class I (OT-I) restricted TCR (Baldwin and Hogquist, 2007). This may explain why CD4+ thymocytes can differentiate into Tregs but CD8+ thymocytes typically do not. Third, IL2 has also been found selectively in the medulla of the thymus (Yang-Snyder and Rothenberg, 1998), a region which correlates with Treg development ( Fontenot et al., 2005a ; Aschenbrenner et al., 2007). Finally, in work by Lio and Hsieh (manuscript submitted), a CD4+CD25+Foxp3− Treg precursor has been identified that can be converted to CD4+CD25+Foxp3+ Tregs by administration of IL2 or IL15 in vitro. Taken together with our in vivo data, these studies support a role for IL2/STAT5-dependent signals in driving Treg development.
A remaining question is why all CD4+ thymocytes in STAT5b-CA mice are not converted into Foxp3+ Tregs. One potential explanation for this observation is provided by recent studies examining the methylation status of the foxp3 gene. For example, changes in DNA CpG methylation and histone H4K3-trimethylation have been found in developing CD4+Foxp3+ thymocytes (Floess et al., 2007). Likewise, TCR/CREB-dependent signals induce changes in the methylation status of histones associated with the foxp3 gene locus (Kim and Leonard, 2007). A second potential explanation is that relatively strong TCR/CD28 signals may be required to induce a state of heightened cytokine responsiveness. Only cells in which both the IL2Rα and IL2Rβ chain have been upregulated, as well as other positive regulators of IL2R signaling such as Socs2 (Tannahill et al., 2005), will likely be capable of augmenting STAT5 signals to the threshold level required to complete Treg differentiation in WT mice. We suggest that in STAT5b-CA mice, the other factors contributing to enhanced cytokine signaling play a lesser but still important role. Based on these findings we propose a model in which high-affinity TCR signals both open the foxp3 gene locus and induce the potential for cytokine responsiveness; only cells that subsequently encounter a STAT5 inducing cytokine will induce foxp3 expression thereby completing the program of Foxp3+ Treg differentiation.
An important question is how STAT5 signals result in Treg differentiation. STAT5 binds to several regions within the foxp3 gene (Burchill et al., 2007; Yao et al., 2007) and is known to interact with both histone acetyltransferases and histone deacetylases. A common perception is that histone deacetylases (HDACs) inhibit gene transcription by deacetylating histones and thereby reducing gene accessibility (Kuo and Allis, 1998). However, mounting evidence suggests that HDACs can also deacetylate other factors (Nusinzon and Horvath, 2005). Along these lines, recent work from at least two groups has demonstrated that inhibition of HDACs prevents the induction of STAT5-dependent gene transcription (Rascle et al., 2003; Xu et al., 2003). Perhaps the best example of this is work by Sun and colleagues in which they characterized transcription of the id1 gene in pro-B cells. Id1 expression is regulated by a pro-B cell specific enhancer that contains critical binding sites for C/EBPβ and STAT5. In this system STAT5 was shown to recruit a histone deacetylase that deacteylated C/EBPβ. Deacetylation of C/EBPβ resulted in enhanced DNA binding and subsequent induction of id1 gene transcription. Importantly, inhibiting HDAC activity blocked the ability of STAT5 to cooperate with C/EBPβ to induce id1 transcription; conversely, introduction of a mutant form of C/EBPβ that could no longer be acetylated removed the requirement for STAT5 in the induction of id1 transcription (Xu et al., 2003). Although we cannot rule out all other interpretations, at this point the simplest explanation for our findings is that STAT5 recruits a HDAC that acts on a second factor to induce foxp3 transcription.
We believe our data also provide an explanation for several other outstanding questions in the field. For example, work from several labs has demonstrated that transgenic mice co-expressing both a single TCR and its respective ligand generate a significant percentage of Foxp3+ Tregs (Apostolou et al., 2002; Jordan et al., 2001). These results suggest that agonist ligand is required to induce Treg differentiation. In contrast, work by van Santen et al using mice in which the amount of agonist ligand could be carefully titrated found that increasing amounts of agonist ligand did not result in an increase in Foxp3+ Treg numbers. Rather, they found that Foxp3+ Tregs were more resistant than non-Tregs to negative selection and that this resulted in an increase in the percentage of Foxp3+ Tregs, but not an actual increase in their numbers (van Santen et al., 2004). Their conclusion was that enhanced survival rather than induced differentiation accounted for the results in the previous studies. We suggest that our model may account for these latter findings. Notably, in the studies by van Santen et al, the agonist ligand was expressed in both the thymic cortex and medulla. Based on our findings that IL2Rβ-induced STAT5 activation is required to complete Foxp3+ Treg differentiation, we would predict that increased agonist ligand in the cortex would result primarily in negative selection due to the limited amounts of IL2 available there. In contrast, we predict that forced expression of IL2 in the cortex would rescue these negatively selected cells by converting them into Foxp3+ Tregs.
A related question concerns the dual role that TCR/CD28 signals play in both negative selection and Foxp3+ Treg development. Work from many labs have established that TCR/CD28 signals are required for both of these processes (Aschenbrenner et al., 2007; Bensinger et al., 2001; Tai et al., 2005; Tang et al., 2003). However, it is unclear how this common signaling mechanism can lead to such dramatically different cell fates. One possibility is that this simply reflects the overall strength of signal with the strongest TCR/CD28-dependent signals leading to cell death while weaker TCR/CD28-dependent signals promote Foxp3+ Treg differentiation. Although this is an attractive explanation, formal proof for such a model is lacking. We propose an alternative model in which high-affinity TCR stimulation induces a program of negative selection that includes enhanced ability to respond to IL2Rβ-dependent ligands. Whether these thymocytes die, or emerge as Foxp3+ Tregs, depends on whether they subsequently encounter STAT5-inducing ligands in the thymus (Sup Fig. 6). Intriguingly, this model would potentially allow for enrichment of Tregs specific for Aire-induced self-antigens which are localized in the medulla (Anderson et al., 2002; Aschenbrenner et al., 2007), and thus have an increased probability of encountering IL2. In contrast, abundant self-antigens, which would be predicted to induce negative selection at a high frequency in the cortex, would preferentially promote negative selection-induced cell death due to the absence of IL2. Such a mechanism would result in greater numbers of Foxp3+ Tregs directed to low abundance self-antigens presented in the thymic medulla, and thus the very cells that might be expected to escape negative selection at higher frequency.
STAT5b-CA transgenic mice were previously described (Burchill et al., 2003). C57BL/6 mice were purchased from the Jackson Laboratories (Bar Harbor, ME). Foxp3-GFP x TCR TCli Tg x TCRα+/− mice were provided by Dr. Alexander Rudensky (University of Washington) (Hsieh et al., 2006). Mice used for analysis were 4–6 weeks old, unless otherwise noted. The University of Minnesota Institutional Animal Care and Use Committee approved all animal experiments.
Thymii and peripheral lymphoid organs were isolated from STAT5b-CA and C57BL/6 mice between the ages of 4–6 wks of age and analyzed for 2W1S-specific T cell populations (Moon et al., 2007). Pooled spleen and lymph nodes were incubated with PE or APC labeled 2W1S:I-Ab tetramers followed by incubation with anti-PE or anti-APC microbeads (Millteni Biotech, Auburn, CA). Tetramer specific T cells were then enriched using magnetic LS-columns (Millteni Biotech) and enumerated using flow cytometric counting beads (Caltag, Carlsbad, CA). Enriched T cell populations were stained with CD3, CD4, CD8, CD11b, CD11c, B220, F4/80, GR1 and NK1.1 and intracellular Foxp3 as previously described (Burchill et al., 2007).
Foxp3+ and Foxp3− populations were isolated from Foxp3-GFP x TCRβ6 Tg x TCRα+/− STAT5b-CA and C57BL/6 mice using a Facs Aria cell sorter. TCR Vα2 sequences were cloned and analyzed as previously described (Hsieh et al., 2006).
Thymi from either Foxp3-GFP or STAT5b-CA x Foxp3-GFP male mice were disrupted with ground glass slides. Cells were subsequently stained with the cell surface antibodies CD8 for enrichment of CD4 T cells using magnetic bead columns (Miltenyi Biotech). Purified CD4+CD8-Foxp3−CD25+ and CD4+CD8-Foxp3−CD25+ thymocytes were purified using a Facs Aria cell sorter (Becton Dickinson). Isolated thymocyte subsets were resuspended in DMEM containing 10% FCS (Atlas Biologicals), 50μM β-mercaptoethanol (Sigma), 1% L-glutamine (Mediatech Cellgrow), 1mM sodium pyruvate (Mediatech Cellgrow), 10 mM Hepes (Mediatech Cellgrow), and 1% streptomycin/penicillin (Mediatech Cellgrow). Cells were either seeded into 96-well round bottom plates in the presence of media alone, IL7 (5nM), or IL2 (10u/ml) plus IL7 (5nM). Some cultures were treated with either 10μg/mL anti-TGFβ (R&D systems, St. Paul, MN), 10 μg/mL isotype control IgG1 antibody (GIR-208; kindly provided by Dr. Robert Schreiber, Washington University), 100 nM Trichostatin A (Sigma, St. Louis, MO), 800 nM Apicidin (Sigma, St. Louis, MO) or DMSO (0.2%). IL2 and IL7 were obtained from the Biologicals Resource Branch, National Cancer Institute Preclinical Repository. Cells were incubated for 24 hours at 37°C in a 5% CO2 incubator. Foxp3 expression was analyzed using an LSR II (Becton Dickinson) and Flow Jo Software (Treestar).
Sorted CD4+ CD25− Foxp3-GFP− thymocytes were stimulated for 24 hours with 10 U/mL IL2 and 5 ng/mL IL7 (PeproTech). Cells were then harvested and stained for phospho-STAT5 as previously described (Will et al, 2006). For acetyl-H3 and H4 western blots CD4+ T cells were stimulated overnight with IL2 plus IL7 and either TSA or DMSO. Nuclear lysates were run on a 12% SDS-polyacrylamide, transferred to a nitrocellulose membrane and blocked for one hour with 5% non fat dry milk in 1 X PBS and 0.1% Tween. Blots were incubated with primary anti-acetyl H3 and H4 antibodies (Upstate Biology) and secondary goat anti-Rabbit IgG followed by visualization with an Odyssey Infrared Imaging System.
Supplemental Fig. 1: Foxp3+ Tregs in STAT5b-CA x γc−/− mice are functional suppressor cells. Shown are gated CD8 T cells from littermate control, γc−/− mice and STAT5b-CA x γc−/− mice stained for CD69 (left panel) and CD62L (right panel) expression. CD8+ T cells from γc−/− mice all exhibit an activated phenotype as illustrated by high CD69 and low CD62L expression. CD69 expression is reduced to background levels while CD62L remains high on CD8+ T cells from STAT5b-CAx γc−/− mice indicating that the Foxp3+ Tregs generated in those mice are capable of blocking T cell activation.
Supplemental Fig. 2: Illustration of the 2W1S Tetramer purification procedure.
Supplemental Fig. 3: Gating Strategy used to analyze 2W1S-specific T cells. Following magnetic bead enrichment samples were stained with antibodies to CD11b, CD11c, F4/80, NK1.1, B220, CD4, CD8, CD25, and intracellular Foxp3. Cell populations were fractionated by flow cytometry first using cell size and granularity followed by elimination of doublet events using FSC-A and FSC-W. T cells were then identified by high levels of CD3 expression and low levels of Dump (CD11b, CD11c, F4/80, NK1.1, B220) staining. The T cell population was then fractionated into CD4 and CD8 populations and the 2W1S-positive gate was set at a level such that gating on CD8 T cells resulted in <0.1% 2W1S-positive events. Magnetic bead enrichment results in the specific pull down of 2W1S positive cells. Neither TEa x Rag−/− nor SM1 x Rag−/− TCR transgenic T cells which recognize Ea:I-Ab and FliC:I-Ab respectively bind to the 2W1S:I-Ab tetramer complex.
Supplemental Fig. 4: Conversion of thymic Foxp3+ Tregs in the presence of IL-2 alone. Purified CD4+Foxp3-GFP−CD25− thymocytes from STAT5b-CA mice were stimulated in vitro with media alone, or increasing doses of IL-2 (10U/ml, 100U/ml and 1000U/ml). After twenty-four of stimulation, cells were stained for CD4 expression and the percentage of Foxp3-GFP+ Tregs determined by flow cytometry. Shown is a representative example of two experiments.
Supplemental Fig. 5: Trichostatin A induces a general increase in histone H3 and H4 acetylation. CD4+CD8− thymocytes from STAT5b-CA mice were stimulated overnight with IL2 plus IL7 and either DMSO or TSA. Nuclear lysates were blotted with antibodies to acetylated histone H3 and H4. Blots were reprobed with an antibody to β-actin; twice as much lysate was loaded from DMSO treated cells.
Supplemental Fig. 6: Model of IL2R/STAT5-dependent Foxp3+ Treg development.
We thank Rachel Agneberg, Jessica Oehrlein and Jared Liebelt for assistance with animal husbandry, Paul Champoux for assistance with flow cytometry, and Dan Mueller and Laura Ramsey for review of the manuscript. This work was supported by a Pew Scholar Award, a Cancer Investigator Award, a Leukemia and Lymphoma Society Scholar award, and an NIH grant (AI061165) to M.A.F. The authors declare that they have no financial conflicts of interest. Correspondence and requests for materials should be addressed to MAF at farra005/at/umn.edu.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.