|Home | About | Journals | Submit | Contact Us | Français|
Regulatory T (Treg) cells can express the transcription factors T-bet and GATA3 but the function of this expression and whether such cells represent stable subsets is still unknown. By using multiple reporter tools, we show that the expression of T-bet and GATA3 in Treg cells is dynamically influenced by the cytokine environment. Treg cell-specific deletion of either Tbx21 or Gata3 genes singly did not result in loss of Treg cell functions; however, mice with combined deficiency of both genes in Treg cells developed severe autoimmune-like diseases. Loss of Treg cell function was correlated with RORγt transcription factor upregulation and reduced Foxp3 expression. Thus, in the steady state, activated Treg cells transiently upregulate either T-bet or GATA3 to maintain T cell homeostasis.
Upon antigen stimulation through their T cell receptor (TCR), naive CD4+ T cells differentiate into distinct effector lineages including type 1 T helper (TH1), type 2 T helper (TH2) and interleukin-17 (IL-17)-producing T helper (TH17) cells; this process is influenced by the strength of TCR signaling as well as the cytokine environment1. The differentiation of each TH lineage is determined by the induction of specific key transcription factors: T-bet is important for the differentiation of TH1 cells2; GATA3 is indispensable for the generation of TH2 cells3; and RORγt plays a critical role in determining the fate of TH17 cells4. Not only do these transcription factors promote the differentiation toward one lineage, they also repress acquisition of other fates. For example, T-bet suppresses the expression and functions of GATA35, thus preventing the activation of an endogenous TH2 differentiation pathway during TH1 differentiation6, 7. T-bet also suppresses RORγt expression by interacting and modulating the function of Runx1, which is an important transcription factor for inducing RORγt expression during TH17 differentiation8, 9.
Regulatory T (Treg) cells, consisting of thymus-derived Treg (tTreg) cells and peripherally derived Treg (pTreg) cells, are crucial for the maintenance of immune tolerance and homeostasis10, 11, 12, 13. The transcription factor Foxp3 plays a central role in Treg generation and function. The cytokine TGF-β is required for the induction of RORγt and Foxp3 and is thus involved in the differentiation of both TH17 and Treg cells14, 15. Consequently, RORγt and Foxp3 are co-expressed at early stages of TH17 and Treg differentiation and may antagonize each other16. Indeed, in some cases, loss of Treg suppressive functions during inflammation is associated with upregulation of RORγt and IL-17 production in Treg cells17.
T-bet expression is found in a subset of Treg cells18. Although T-bet expression in these Treg cells has been shown to be important for the maintenance of Treg homeostasis during type 1 immune responses, the physiological significance of T-bet expression in Treg cells in the steady state is unknown. Furthermore, there is no report on characterizing mice with Treg cell-specific deletion of Tbx21 (encoding T-bet) even though it is known that some Treg cells express GATA3 in the steady state19, 20, 21. GATA3 can be induced when Treg cells become activated. It has been reported that Treg-specific deletion of GATA3 in mice results in spontaneous autoimmunity starting from 16 weeks of age21; however, other reports indicate that GATA3 is only critical for Treg functions during inflammation and mice with Treg-specific GATA3 deletion do not develop any disease until 6 months of age19, 20.
Although T-bet- and GATA3-expressing Treg cells have been well documented, it is not clear whether the T-bet- (TH1-) and GATA3-expressing (TH2-like) Treg cells represent stable Treg subsets. Furthermore, whether and how T-bet and GATA3 regulate the function of Treg cells, especially in the steady state, is not known. Here we report that T-bet and GATA3-expressing Treg cells could be detected in the steady state; however, their expression in Treg cells was highly dynamic. Thus, T-bet-expressing Treg cells do not represent a stable Treg subset. Single deletion of either Tbx21 or Gata3 gene specifically in Treg cells by Foxp3-Cre did not alter Treg functions. However, combined deletion of both Tbx21 and Gata3 in Treg cells allowed the development of aggressive autoimmune-like diseases in mice at very young age.
To facilitate investigation on the relationship between T-bet and GATA3-expressing Treg cells, a tri-color reporter mouse strain, in which the expression of T-bet, GATA3 and Foxp3 are depicted by different fluorescent proteins, was first constructed. Foxp3-mRFP knock-in mice22 and GATA3-GFP knock-in mice23, in which mRFP and GFP faithfully marks the expression of Foxp3 or GATA3, respectively, have been reported. A third fluorescent marker is required for reporting T-bet expression, but a previously generated T-bet-ZsGreen reporter mouse strain6 is not useful for this purpose since green fluorescence is also used to report GATA3 expression.
Utilizing a similar strategy to that previously described6, we prepared a BAC transgenic T-bet reporter mouse strain, in which AmCyan indicates T-bet expression. AmCyan-expressing cells but not AmCyan negative cells directly sorted from the spleens of the intact reporter mice stained positive for T-bet protein (Supplementary Fig. 1a). To further evaluate the faithfulness of this new T-bet-AmCyan reporter, naive CD4+ T cells (CD4+CD25−CD45RBhiAmCyan−) were isolated by cell sorting from the transgenic mice and cultured under TH1 or TH2 polarizing conditions for 4 days. Virtually all cells from the TH1 cell-polarizing culture (>90%) expressed AmCyan, whereas, under TH2 conditions, most if not all (>97%) the cells remained Amcyan negative (Supplementary Fig. 1b). These results indicate that the T-bet-AmCyan reporter faithfully reflects the expression of endogenous T-bet, just as a previous T-bet-ZsGreen reporter does. The T-bet-AmCyan reporter line was then bred with GATA3-GFP and Foxp3-RFP knock-in mice to generate a T-bet-AmCyan:GATA3-GFP:Foxp3-mRFP tri-color reporter mouse strain.
We first re-examined the expression pattern of T-bet and GATA3 in Treg cells in the spleen, peripheral lymph node (PLN), and mesenteric lymph node (MLN) of the tri-color reporter mice in the steady state. About 20-30% of Treg cells expressed GATA3 in the steady state (Fig. 1a,b). In agreement with previous reports6, 18, only a small proportion of ex vivo-isolated Treg cells expressed T-bet (Fig. 1a,b). The T-bet-expressing Treg cells were enriched in the spleen; ~10% of total Treg cells in the spleen expressed T-bet but only 3-5% of Treg cells in lymph nodes did so. Thus, the existence of T-bet- and GATA3-expressing Treg cells in the steady state was confirmed in the tri-color reporter mouse model. Some Treg cells expressed both T-bet and GATA3; the percentage of GATA3+ cells among the T-bet+ population was identical to the percentage of GATA3+ cells among the T-bet− population (Fig. 1a). The vast majority of the T-bet-AmCyan positive Treg cells expressed CXCR3 whether they were GATA3-GFP positive or not, consistent with the idea that CXCR3 is a gene target of T-bet (Supplementary Fig. 2a). Treg cell subsets were then separated based on their T-bet and GATA3 expression by cell sorting (Supplementary Fig. 2b). Sorted T-bet-AmCyan+GATA3-GFP− but not the T-bet-AmCyan−GATA3-GFP− or T-bet-AmCyan−GATA3-GFP+ Treg cells expressed T-bet (Supplementary Fig. 2c). These results indicate that T-bet-AmCyan faithfully reports T-bet expression in Treg cells.
Since more than 60% of the Treg cells did not express either T-bet or GATA3, we next examined whether T-bet and GATA3 expression could be induced in such cells. Highly purified T-bet and GATA3 non-expressing Treg cells by cell sorting (Supplementary Fig. 2b) were cultured with plate-bound anti-CD3 and anti-CD28 in the presence of IL-2 with addition of either IL-4 or IFN-γ. Foxp3 expression was stably maintained in these cultures (Fig. 1c). About 16% of Treg cells upregulated GATA3-GFP+ when stimulated with IL-4, and more than 50% Treg cells became T-bet-AmCyan+ when cultured with IFN-γ (Fig. 1c).
To test whether T-bet and GATA3 expression could be induced in non-expressing Treg cells in vivo, T-bet and GATA3 non-expressing Treg cells were transferred into Rag1-deficient (Rag1 −/−) mice either alone or together with CD45.1 congenic naive (CD4+CD25−CD45RBhi) T cells. Consistent with the report that Treg cells are unstable when transferred alone into lymphopenic environments24, more than 50% of the transferred cells became Foxp3 negative T cells (data not shown). T-bet and GATA3 were induced in a portion of remaining Foxp3+ cells; some cells expressing both T-bet and GATA3 (Fig. 1d). When co-transferred with naive T cells, Foxp3+ cells retained Foxp3 expression, and T-bet and GATA3 were induced in these Treg cells whether isolated from spleen, MLN or small intestine lamina propria (siLP) (Fig. 1e). T-bet was more efficiently induced in the siLP. The proportions of T-bet- and GATA3-expressing Treg cells observed after transfer were similar to that in the steady state (Fig. 1a). These results indicate that T-bet and GATA3 expression can be induced in non-expressing Treg cells both in vitro and in vivo.
It has been previously reported that IFN-γ and STAT1 but not IL-12 and STAT4 are involved in T-bet induction in Treg cells18, 25, 26. To confirm the signaling pathways required for T-bet expression in Treg cells in our reporter system, we examined Treg cells isolated from spleen and lymph nodes of the T-bet-ZsGreen reporter mice that were wild-type, Stat1−/−, Stat4−/− or IFN-gr1−/−. In agreement with the previous studies, our results showed that deficiency in STAT1 but not STAT4, dramatically reduced T-bet expression in Treg cells (Fig. 1f and Supplementary Fig. 3). There was a partial reduction of T-bet-expressing Treg cells in IFN-gr1−/− mice. Thus, these results indicate that T-bet expression in Treg cells in the steady state depends on STAT1 activating cytokines including IFN-γ.
To assess whether the T-bet- and the GATA3-expressing Treg cells represent stable subsets, we purified T-bet−GATA3+ and T-bet+GATA3− Treg cells by cell sorting (Supplementary Fig. 2b). T-bet−GATA3+ Treg cells were cultured with plate-bound anti-CD3, anti-CD28 and IL-2, in the presence or absence of TGF-β, with addition of either IL-4 or IFN-γ (Fig. 2a). Foxp3 expression was stably maintained under all conditions. In the absence of IL-4, IFN-γ and TGF-β, some Treg cells lost GATA3 expression while others gained T-bet expression. TGF-β further reduced GATA3 expression but prevented T-bet upregulation. GATA3 expression was largely maintained when IL-4 was present, but T-bet expression was not induced. However, even in the presence of TGF-β, IFN-γ induced T-bet expression in ~40% Treg cells that had been T-bet−GATA3+, while GATA3 expression was strikingly diminished. Similarly, when T-bet+GATA3− Treg cells were cultured with plate-bound anti-CD3 and anti-CD28 with IL-2 and TGF-β, T-bet expression in the majority of the cells was lost although Foxp3 expression was retained (Fig. 2b and data not shown). IFN-γ was able to maintain T-bet expression whereas IL-4 induced GATA3 expression in T cells that originated from T-bet+GATA3− reg Treg cells.
To examine whether T-bet- and GATA3-expressing Treg cells are stable in vivo, we sorted T-bet−GATA3+ T cells were transferred into Rag1−/− reg mice. As expected, more than 50% of them became Foxp3 negative T cells (data not shown). Nevertheless, ~50% of the cells that maintained Foxp3 expression turned into T-bet+GATA3+, T-bet+GATA3− or T-bet−GATA3− Treg cells (Fig. 2c). To further test the stability of the T-bet- and GATA3-expressing Treg cells in an environment where Foxp3 expression can be maintained, T-bet−GATA3+ or T-bet+GATA3− T were transferred into Rag1−/− reg mice together with congenic CD45.1 naive CD4+ T cells (Fig. 2d). Most of the transferred Treg cells retained Foxp3 expression as expected, however, the pattern of T-bet and GATA3 expression in Treg cells became very similar in the recipient mice that received either T-bet−GATA3+ or T-bet+GATA3− Treg cells. We also transferred sorted T-bet−GATA3+ cells into wild type congenic animals. The results support the idea that the expression of T-bet and GATA3 in Treg cells are dynamic (data not shown). Overall, our results indicate that the expression of T-bet and GATA3 in Treg cells is dynamic both in vitro and in vivo, and that T-bet- and GATA3-expressing Treg cells do not represent stable Treg subsets.
To confirm the dynamic feature of T-bet expression in Treg cells in intact animals, we prepared a T-bet fate-mapping mouse strain. Offspring of T-bet-ZsGreen-T2A-CreERT2 and Rosa26-loxP-STOP-loxP-tdTomato mice (T-bet-fate-mapping mice) were treated with tamoxifen. In such mice, ZsGreen indicates T-bet expression while tdTomato reflects a cohort of cells that expressed T-bet at the time of tamoxifen treatment (Fig. 3a). Before tamoxifen injection, all the T cells including CD8 T cells were tdTomato negative as expected (Fig. 3a and data not shown). One week after a single tamoxifen injection into T-bet-fate-mapping mice, the majority of the CD8 memory-like (CD8+CD44hi) cells, which constantly express T-bet, were tdTomato positive (Fig. 3a). Only a small fraction of the ZsGreen-expressing CD8+CD44hi cells were not marked by tdTomato indicating a high efficiency of Cre-mediated reporter activation. Virtually all these tdTomato positive CD8 T cells were also expressing ZsGreen suggesting that T-bet expression in CD8 T cells is stable. However, within the CD4+CD25+ Treg compartment, the ZsGreen− tdTomato+ population, which represents T-bet− Treg cells that had previously expressed T-bet, were present at a substantial proportion. In fact, these cells outnumbered the ZsGreen+tdTomato+ population, which represents T-bet-expressing Treg cells, in the peripheral and mesenteric lymph nodes (Fig. 3b). The fact that more than 60% of lymph node T-bet-expressing Treg cells became T-bet non-expressing Treg cells within 1 week confirms that T-bet expression in Treg cells is highly dynamic. By contrast, the vast majority of the NK cells, CD8 memory-like (CD8+CD44hi) cells, or CD4+CD25− cells that were tdTomato positive continued to express ZsGreen suggesting that T-bet expression in NK, CD8 and conventional CD4 T cells is relatively stable (Fig. 3c). Although most of the ZsGreen+tdTomato+ Treg cells expressed CXCR3 as expected, a substantial proportion of ZsGreen−tdTomato+ Treg cells were also CXCR3+, suggesting that the T-bet-expressed cells had up-regulated CXCR3 expression (Fig. 3d).
To test whether ZsGreen+ Treg cells can give rise to ZsGreen−tdTomato+ Treg cells we sorted CD4+CD25+ZsGreen+ cells from the T-bet fate-mapping mice and adoptively transferred them together with effector T cells into Rag1−/− recipient mice treated with tamoxifen. Two weeks after transfer, ~ half of tdTomato+ Treg cells had lost ZsGreen expression (Fig. 3e). By using the T-bet fate-mapping reporter, we confirmed that T-bet− Treg cells were originated from T-bet+ Treg cells. Thus, our data indicate that T-bet expression in Treg cells is highly dynamic in the steady state in intact animals.
To study the functions of T-bet and GATA3 in Treg cells, we generated mice with Treg-specific deletion of either Tbx21 or Gata3 by crossing Tbx21fl/fl mice27 or Gata3fl/fl mice28 to Foxp3-Cre mice expressing yellow fluorescent protein (YFP)-Cre recombinase fusion protein under the control of Foxp3 locus29. Tbx21fl/flFoxp3-Cre and Gata3fl/flFoxp3-Cre mice had normal number of CD4+ and CD8+ T cells in spleen and lymph nodes (Fig. 4a and data not shown). The percentage of CD4+Foxp3+ Treg cells from Tbx21fl/flFoxp3-Cre and Gata3fl/flFoxp3-Cre mice were similar to Foxp3-Cre mice (Fig. 4a). Furthermore, the vast majority of the CD4+ and CD8+ T cells in all the mice displayed naive phenotype (data not shown).
Extending a previous report20 showing that Gata3fl/flFoxp3-Cre mice were healthy at a young age, we found that our Gata3fl/flFoxp3-Cre mice kept for more than six months did not develop any obvious abnormality before they were euthanized (data not shown). Similarly, no apparent abnormality was noted in adult Tbx21fl/fl-Foxp3-Cre mice up to six months (data not shown). Therefore, the absence of either T-bet or GATA3 individually had no major impact on the function of Treg cells in the steady state.
To assess whether the absence of either T-bet or GATA3 could have functional consequences on the ability of Treg cells to control tissue inflammation, we utilized the well-described inflammatory bowel disease (IBD) model induced by T cell transfer into Rag1−/− recipients30. CD45.1 congenic naive CD4+ T cells were transferred into Rag1−/− mice either alone or together with Treg cells from Foxp3-Cre, Tbx21fl/flFoxp3-Cre or Gata3fl/flFoxp3-Cre mice. As expected, transferring naive CD4+ T cells alone led to severe inflammation associated with significant weight loss and colon thickening (Fig. 4b). Co-transferring Treg cells from Foxp3-Cre mice with naive CD4+ T cells prevented disease. Similarly, Treg cells from either Tbx21fl/flFoxp3-Cre or Gata3fl/flFoxp3-Cre mice were as effective as wild-type Treg cells in preventing weight loss. The total cell number of congenic CD45.1+CD4+ T cells found in the small intestine lamina propria (siLP) was strikingly reduced in all the groups in which Treg cells were co-transferred (Fig. 4c). These results indicate that Treg cells deficient in either T-bet or GATA3 have substantial suppressive functions in this IBD model.
To assess the cross-regulation between T-bet and GATA3 in Treg cells, Treg cells from Foxp3-Cre, Tbx21fl/flFoxp3-Cre and Gata3fl/flFoxp3-Cre mice were cultured with plate-bound anti-CD3 and anti-CD28 in the presence of IL-2 with addition of either IL-4 or IFN-γ. Tbx21fl/flFoxp3-Cre Treg cells expressed higher amounts of GATA3 protein in the presence of IFN-γ, whereas Gata3fl/fl-Foxp3-Cre Treg cells expressed higher levels of T-bet in the presence to IL-4 compared to WT Treg cells cultured under the same conditions (Fig. 4d). A modest staining of GATA3 was noted in the group of IL-4-stimulated Gata3-deficient cells presumably because anti-GATA3 also recognizes the truncated GATA3 protein lacking functional zinc fingers encoded by exon 4. These results suggest that cross-regulation between T-bet and GATA3 may occur in Treg cells within certain cytokine environments.
Since the expression of T-bet and GATA3 in Treg cells are both dynamic and since T-bet and GATA3 may regulate each other under certain conditions, we asked whether T-bet and GATA3 have a redundant role in modulating Treg cell functions. To test this, we generated Tbx21fl/flGata3fl/flFoxp3-Cre double knockout (DKO) mice harboring Treg-specific deletion of both T-bet and GATA3.
DKO mice were born at the expected Mendelian ratio and appeared to be as healthy as their littermates (either Tbx21fl/flGata3fl/+Foxp3-Cre or Tbx21fl/flGata3+/+Foxp3-Cre) during the first month of life. However, by 6-8 weeks of age, DKO mice displayed splenomegaly and lymphadenopathy (Fig. 5a), which was reflected by the increased total cell number (Fig. 5b). T-bet and GATA3 double ablation in Treg cells led to marked increase in activated CD4+Foxp3−CD62LloCD44hi effector T cells (Fig. 5c). Histopathological evaluation of the DKO mice showed spleen and lymph node hyperplasia and massive cell infiltration in the liver, lung, kidney, small intestine and colon (Fig. 5d and data not shown), whereas control littermates did not show any noticeable pathology.
Consistent with their activated phenotype, increased capacity to produce each of the effector cytokines, including IFN-γ, IL-4 and IL-17, was observed (Fig. 5e). To test the consequences of IL-4 up-regulation, we measured levels of serum immunoglobulin isotypes in 8-week-old control, DKO and each single knockout mice. Consistent with an uncontrolled IL-4 and IFN-γ production in these mice, IgE serum concentrations were massively elevated in the DKO mice and serum IgG1 and IgG2a level was also significantly increased (Fig. 5f). Thus, both Th1- and Th2-related antibody responses are dysregulated when both T-bet and GATA3 are deficient in Treg cells.
We next explored whether T-bet and GATA3 double deficiency in Treg cells impairs their generation, maintenance, and/or suppressive functions. DKO mice exhibited no significant alteration in the development of Foxp3+CD4+ Treg cells in the thymus compared to WT mice (Supplementary Fig. 4a and 4b). Although the frequency of splenic Foxp3+ Treg cells in the DKO was slightly reduced compared to the control mice, the frequency of Foxp3+ Treg cells was significantly increased in the lymph nodes of the DKO mice (Supplementary Fig. 4c). DKO Treg cells proliferate at a similar rate as WT Treg cells judged by Ki-67 staining and DKO Treg cells expressed normal levels of Bcl-2 (Supplementary Fig. 5). Therefore, DKO Treg cells do not have proliferative or survival defects and the dramatic phenotype observed in the DKO mice is not because of reduced Treg cells.
We next tested whether T-bet and GATA3 double deficiency in Treg cells results in changes in the expression of Treg signature genes. The expression of CD25 and CTLA-4 were normal while the expression GITR and CD103 were reduced in the DKO Treg cells compared to WT Treg cells (Supplementary Fig. 6). There was a substantial reduction of Nrp-1 expression in the DKO Treg cells. Although Nrp-1 has been reported to potentiate Treg cell functions, it is dispensable for maintaining immune homeostasis31. Thus, defective expression in Nrp-1 in DKO Treg cells cannot explain the autoimmune phenotype seen in these mice. The intensity of Foxp3 expression was substantially reduced in Treg cells from the DKO mice compared to Treg cells from the control mice (Fig. 5g). Consistent with a previous report, Foxp3 expression was only slightly reduced in Gata3fl/flFoxp3-Cre mice in comparison of Foxp3-Cre and Tbx21fl/flFoxp3-Cre mice. Since reduced Foxp3 expression often translates to Treg cell instability, we performed transfer experiments with total splenocytes from the DKO mice. Rag1−/− mice, which received splenocytes from 8-week old DKO mice, developed IBD as manifested by weight loss (Fig. 5h). As expected, the percentage of CD4+Foxp3+ Treg cells were markedly decreased upon transfer confirming the instability of the Treg cells lacking the expression of both T-bet and GATA3; higher proportion of these Treg cells expressed RORγt compared to control Treg cells (Fig. 5i). These results suggest that Treg cell-specific deficiency of both T-bet and GATA3 affects their function and stability.
To assess suppressive function of Treg cells in vivo, CD45.1 congenic naive T cells were transferred into Rag1−/− mice either alone or together with Treg cells from 8-week-old DKO or control Foxp3-Cre mice. As expected, transfer of naive T cells alone led to severe inflammation associated with significant weight loss and colon pathology including massive inflammatory cell infiltration that was prevented when control Foxp3-Cre Treg cells were co-transferred (Fig. 6a,b). However, Treg cells from the DKO mice failed to suppress the disease. Similarly, the number of congenic CD45.1+CD4+ T cells in the siLP was significantly reduced by co-transferring control Foxp3-Cre Treg cells but not DKO Treg cells (Fig. 6c). Many DKO Treg cells had lost Foxp3 protein expression whereas control Treg cells were largely stable as expected (Fig. 6d). In the DKO cells that still expressed Foxp3, a significantly elevated proportion also expressed RORγt protein (Fig. 6e). Accordingly, DKO Treg cells significantly upregulated the TH17 effector cytokine IL-17A (Fig. 6f). Thus, Treg cells deficient in both T-bet and GATA3 are defective in homeostasis and suppressive functions, associated with their abnormal TH17 phenotype.
To further assess whether the role of T-bet and GATA3 in Treg cells is cell intrinsic, we prepared mixed bone marrow chimeras using the Foxp3-Cre, DKO, Tbx21fl/flFoxp3-Cre or Gata3fl/flFoxp3-Cre donors together with the CD45.1 congenic donors. The contribution of each donor to CD4+Foxp3+ Treg cells was then examined. The frequencies of Foxp3+ Treg cells among the CD45.1+ cells were the same in all groups as expected (data not shown). The frequencies of Foxp3+ Treg cells originating from Foxp3-Cre, Tbx21fl/flFoxp3-Cre or Gata3fl/flFoxp3-Cre bone marrow were similar, but the frequency of Foxp3+ Treg cells developed from DKO bone marrow progenitors was significantly reduced (Fig. 7a). In addition, RORγt protein expression was dramatically increased in Foxp3+ Treg cells derived from the DKO donor compared to Treg cells from other donors and the MFI of Foxp3 from the DKO group was significantly lower (Fig. 7b). In fact, RORγt expression was mostly found within the Foxp3lo population in the DKO group and such cells expressed high amounts of CD44 indicating an activated Treg population (Fig. 7c). The CD44hi Treg cell population was significantly reduced in the DKO group compared to other groups indicating that without T-bet and GATA3, activated Treg cells are either inefficiently generated or poorly maintained. In addition, the vast majority of the CD44hi Treg cells in the DKO group expressed RORγt. Consequently, a substantial proportion of the Foxp3+ Treg cells from DKO donor were capable of producing IL-17A (Fig. 7d). Thus, these results indicate that the defects of DKO Treg cells are cell intrinsic.
Treg cells are critical for maintaining immune tolerance. When such cells are defective or absent, various types of autoreactive CD4+ T effectors, including TH1, TH2 and TH17 cells, become activated. The diversity of the Treg cell population has evoked general interest by researchers in the field32. Whether defined subsets of Treg cells are responsible for controlling distinct types of T cells either through a common or specific mechanism has been unclear.
It has been suggested that Treg cells might utilize some components of the transcriptional machinery operating in distinct effector cells to effectively control a particular type of immune response, for example by showing that Treg cells could use IRF4 to selectively control TH2 immune responses33. However, IRF4 is not only important for TH2 cell differentiation, but is also involved in TH17 cell differentiation. Furthermore, IRF4 is expressed in all activated Treg cells and has broader effects on Treg cell activation34. Therefore, it remains possible that IRF4 is critical for optimal Treg cell activation and that suppression appears selective because the control of the differentiation of TH2 cells requires more robust Treg cell function than does the control of TH1 or TH17 cells. The striking observation that Treg-specific deletion of Stat3 leads to uncontrolled TH17 responses at 6 weeks of age suggests that Treg cells need to receive comparable signals to those required to induce a particular type of effector cells in order to effectively suppress those effector cells35. It remains attractive to test whether STAT1 and STAT6 are indispensable for the Treg cells to suppress TH1 and TH2 responses, respectively.
T-bet, the master regulator of TH1 cells, is expressed by a subset of Treg cells18. More importantly, it has been reported that T-bet deficient Treg cells failed to thrive under TH1 inflammatory environment and that these cells failed to suppress TH1 response in a transfer model. However, the importance of T-bet expression by Treg cells has not been directly tested in mice with Treg-specific T-bet ablation. Furthermore, whether T-bet expression by Treg cells is required for their ability to suppress TH1-mediated autoimmunity in the steady state is unknown. Here we report that mice with Treg-specific T-bet deletion did not develop any obvious autoimmunity, indicating that T-bet is dispensable for Treg cell-mediated suppression of auto-reactive TH1 cells in the steady state.
Similarly, GATA3 is expressed by a subset of Treg cells. However, GATA3 expression in Treg cells does not seem to be required for control TH2-mediated autoimmunity in the steady state. Although it was reported that Treg-specific ablation of GATA3 resulted in TH2-related diseases starting at 16 weeks of age21, two other mouse lines with Treg cell-specific Gata3 deletion failed to develop obvious abnormalities until the mice were more than half a year old19, 20.
Our experiments were performed mainly in the steady state or under mild inflammation. Since Treg cells function differently in the steady state versus under inflammatory conditions36 or during TH1 biased infections37, the phenotype and functions of antigen-specific TH1-like Treg cells and the role of T-bet in Treg cells during TH1-promoting infections require further investigation. Similarly, it will be interesting to investigate the function of GATA3 in Treg cells during robust TH2 responses in vivo.
Although ablation of T-bet or GATA3 alone in Treg cells has no major impact on Treg cell functionality in the steady state, mice in which both genes are deleted in Treg cells develop a severe autoimmune-like phenotype beginning at 6-8 weeks of age. These results suggest T-bet and GATA3 are redundant in maintaining Treg cell functions. Reduced functionality of these Treg cells is associated with upregulation of RORγt and downregulation of Foxp3 expression in a cell intrinsic manner. Foxp3 and RORγt are normally co-expressed in T cells treated with large amounts of TGF-β and in a subset of cells in the siLP16. Foxp3 antagonizes RORγt function partly by protein-protein interaction. However, it is also possible that RORγt antagonizes Foxp3 function in Treg cells. Therefore, RORγt expression in Treg cells may require careful control. One way that Treg cells could fully suppress RORγt expression is to express T-bet and/or GATA3, both of which are capable of suppressing RORγt expression in effector T cells.
Redundancy of T-bet and GATA3 in Treg cells can be explained by their dynamic expression pattern. Our current study indicates that T-bet-expressing cells can convert into GATA3-expressing cells, and vice versa, under the influence of cytokines, both in vitro and in vivo. Such plasticity or flexibility of Treg cells by changing their pattern of transcription factor expression within the Foxp3+ compartment is different from the previously reported Treg cell plasticity with loss of Foxp3 expression and gain of effector functions under inflammatory conditions37, 38, 39.
Treg cells need to be activated to acquire full suppressive activity40, 41. T-bet expression is associated with activated Treg cells since T-bet-expressing Treg cells expressed higher Treg cell markers, including CD103, GITR and CTLA4, than T-bet-non-expressing Treg cells18. We also found that T-bet is mainly expressed by CD44hi Treg cells. Moreover, T-bet and GATA3 double deficient Treg cells contain less CD44hi Treg population. Because of the dynamic expression pattern, some T-bet- and GATA3-non-expressing cells might have expressed these molecules at an earlier time and thus consist of activated Treg cells. We propose that when Treg cells are activated in a local environment, they upregulate T-bet or GATA3 to prevent RORγt expression and to maintain Foxp3 expression.
Overall, we demonstrate that unlike TH1 and TH2 effector cells, ‘TH1-like’ and ‘TH2-like’ Treg cells in the steady state are not stable subsets and that T-bet expression in Treg cells is not required for these cells to suppress TH1 effectors. Furthermore, our data demonstrate that T-bet and GATA3 have a redundant role in controlling RORγt expression in activated Treg cells indicating that Treg cells can utilize multiple cross-regulatory mechanisms found in T effector cells to overcome the close relationship between Treg and TH17 cells and thus maintain T cell homeostasis.
The C57BL/6 T-bet-AmCyan reporter mouse strain was generated in the same way as we previously prepared the C57BL/6 T-bet-ZsGreen reporter mouse strain except for using a different fluorescent marker. The coding region of the AmCyan was amplified from pAmCyan1-N1 vector (Clontech) by PCR for recombineering using the following primers: 5′- GAC CCT CGG GTC TCT TCG ACG GCT GCT GGA AGG CGC CCA GCC CGC CTC GGA TGG CCC TGT CCA ACA AGT TCA TC -3′ and 5′- CAC TGC ATT CTA GTT GTG GTT TGA TGG GCA TCG TGG AGC CGG GCT GCG GAG ACA TGC TGA CCG GCA CCG AGC C -3′. The modified BAC clone, after sequence verification of the manipulated region, was used to generate transgenic mice by pronuclear microinjection of fertilized C57BL/6 eggs. Total of 6 founders were identified to be positive for AmCyan as assessed by Southern blot and PCR. The offspring of the founder (B6-Tbet-AmCyan-B10) that carried only one copy of the transgene were selectively maintained. Foxp3-mRFP knock-in mice (C57BL/6-Foxp3tm1Flv/J)22 were purchased from the Jackson Laboratory (#008374). GATA3-GFP reporter mice (Gata3g/+) on the C57BL/6 background were provided by Dr. James Douglas Engel23. T-bet-AmCyan reporter mice were bred with Foxp3tm1Flv/J to generate T-bet-AmCyan:Foxp3-RFP mice; these mice were further bred with Gata3g/+ mice to generate T-bet-AmCyan:GATA3-GFP:Foxp3-mRFP tri-color reporter mice. The Tri-color reporter mice were genotyped using the following primers: for T-bet-AmCyan: 5′- GTA GGT GAA GGT TCT CTC GTA G -3′, 5′- GAC AAG AGA CTT ACA CTT AGG AGT G -3′; for Gata3g/+, 5′- CAG GTG ATC GGA AGA GCA AC -3′, 5′- GTT TGC AGT TAA GGG TAT AG -3′; for Foxp3tm1Flv/J, 5′- CAA AAC CAA GAA AAG GTG GGC -3′, 5′- CAG TGC TGT TGC TGT GTA AGG GTC -3′, and 5′- GGA ATG CTC GTC AAG AAG ACA GG -3′.
Tbx21fl/fl mice27 (kindly provided by Dr. Steve Reiner, University of Columbia) and Gata3fl/fl mice28 were on the C57BL/6 background. Tbx21fl/fl mice or Gata3fl/fl mice were crossed with Foxp3-IRES-YFP-Cre mice29 (kindly provided by Dr. Alexander Rudensky, Memorial Sloan Kettering Cancer Center, New York) to generate Tbx21fl/fl- Foxp3-Cre or Gata3fl/fl-Foxp3-Cre mice. Tbx21fl/fl-Gata3fl/fl-Foxp3-Cre (DKO) mice were generated by crossing Tbx21fl/fl-Foxp3-Cre with Gata3fl/fl-Foxp3-Cre mice. All animals were genotyped using the following primers: for Foxp3-Cre, primer pair-1: 5′- CCA GAT GTT GTG GGT GAG TG -3′, 5′- TGG ACC GTA GAT GAA TTT GAG TT -3′ and primer pair-2: 5′- AGG ATG TGA GGG ACT ACC TCC TGT A -3′, 5′- TCC TTC ACT CTG ATT CTG GCA ATT T -3′; for Tbx21fl/fl, 5′- TAT GAT TAC ACT GCA GCT GTC TTC AG -3′, 5′- CAG GAA TGG GAA CAT TCG CCT GTG -3′, and 5′- CTC TGC CTC CCA TCT CTT AGG AGC -3′; for Gata3fl/fl, 5′-TCA GGG CAC TAA GGG TTG TTA ACT T-3′, 5′-GAA TTC CAT CCA TGA GAC ACA CAA-3′.
C57BL/6 T-bet-ZsGreen reporter (TBGR, Taconic Line 8419), IFN-gr1−/−-T-bet-ZsGreen (Taconic Line 8457) and Stat4−/−-T-bet-ZsGreen mice (Taconic Line 8452) were previously described 6 and deposited into the NIAID-Taconic repository. CD45.1 congenic mice (Line 7) and Rag1−/− mice (Line 146) were also from the NIAID-Taconic repository. Stat1−/−-T-bet-ZsGreen mice were generated by crossing Stat1−/− mice42 (kindly provided by Dr. Dragana Jankovic of the NIAID, originally from Dr. Joan Durbin) with TBGR mice.
The generation of T-bet-ZsGreen-T2A-CreERT2 mouse line will be described in detail in another study. Briefly, ZsGreen-T2A-CreERT2, a DNA cassette containing sequences encoding ZsGreen, a “self-cleaving” T2A peptide43, and a fusion Cre recombinase with mutated human estrogen receptor ligand-binding domain (CreERT2)44 obtained from Addgene, was inserted into the T-bet translational start site in the BAC clone RP23-237M14. In the resulting BAC transgenic mice, T-bet-expressing cells will express both ZsGreen and CreERT2 as separated proteins. The T-bet-ZsGreen-T2ACreERT2 mice were then bred with the reporter mice carrying a loxP site–flanked STOP cassette and a DNA sequence encoding red fluorescent protein variant, tdTomato, under the control of ubiquitously expressed ROSA26 locus (JAX mice line #007914)45. Offspring of the T-bet-ZsGreen-T2A-CreERT2 and Rosa26-loxP-STOP-loxP-tdTomato mice were designated as T-bet-fate-mapping mice. Cre-mediated recombination was induced by tamoxifen treatment. Tamoxifen (Sigma) dissolved in corn oil (Sigma) was injected intraperitoneally (i.p.) into T-bet-fate-mapping mice at day 0 (3mg tamoxifen in 150ul corn oil per mouse per injection). Mice were analyzed at day 7 after the initial injection.
All the mice were bred and/or maintained in the NIAID specific pathogen free animal facility and the experiments were done when mice were 6 to 16 weeks of age under protocols approved by the NIAID Animal Care and Use Committee.
Single cell suspension was prepared directly from different lymphoid organs of mice including lymph nodes, spleen, thymus and bone marrow. Cells from small intestinal lamina propria were prepared as previously described46. Lymph node cells and splenocytes from T-bet-AmCyan:GATA3-GFP:Foxp3-RFP tri-color reporter mice were stained with APC-anti-CD4, and then sorted for CD4+RFP+AmCyan−GFP−, CD4+RFP+AmCyan+GFP− and/or CD4+RFP+AmCyan−GFP+ populations using FACSAria (BD Biosciences). Lymph node cells from Foxp3-Cre, Tbx21fl/fl-Foxp3-Cre, Gata3fl/fl-Foxp3-Cre and DKO mice were stained with APC-anti-CD4 and PE-anti-CD25, and then sorted for CD4+CD25+ YFP+ population. Naive CD4+ T cells from CD45.1 congenic mice (Taconic Line 7) were purified by cell sorting for the CD4+CD25−CD45Rbhi population after staining lymph node cells with FITC-anti-CD4, PE-anti-CD25 and APC-anti-CD45Rb.
RPMI 1640 media (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone), 50μM β-mercaptoethanol (Sigma), 1% sodium pyruvate, 1% nonessential amino acids, 1% HEPES, 100U/ml penicillin and 100μg/ml streptomycin, and 2mM L-glutamine (all from Invitrogen) was used as culture medium. The cells were incubated at 37°C and under 5% CO2 for the indicated time. Sorted Treg cells were stimulated with plate-bound anti-CD3 (2C11) and anti-CD28 (37.51) in IL-2 (100U/ml) containing RPMI 1640 media alone or with different combinations of cytokines: TGF-β1 (5ng/ml), IL-4 (1ng/ml) or IFN-γ (1ng/ml) were added as indicated. All cytokines used in cell culture were purchased from PeproTech.
Staining of cell surface molecules was carried out using PBS with 2% FBS. For intracellular staining of cytokines, cells were first stimulated with 10 ng/ml phorbol 12-myristate 13-acetate (PMA) and 500 nM ionomycin for 4 hours, or with plate-bound anti-CD3/anti-CD28 for 5 hours, in the presence of 2 mM monensin, and then stained with a cocktail of fixable viability dye (eBioscience) and antibodies to various cell surface markers. They were then fixed with 4% paraformaldehyde for 10 min at room temperature and permeabilized in PBS containing 0.5% Triton X-100 and 0.1% BSA before staining for cytokines. Staining for transcription factors was performed with Foxp3 Staining Buffer Set (eBioscience) according to the manufacturer’s instructions. Flow cytometry data were collected with LSR II (BD Biosciences) and results were analyzed by using FlowJo software (Tree Star). Antibodies specific for mouse CD4 (RM4-5), CD8 (53-6.7), CD25 (PC61.5), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD45Rb (C363.16A), CD357 (GITR, DTA-1), CD152 (CTLA-4, UC10-4B9), Ki-67 (SolA15), IL-4 (11B11), IFN-γ (XMG1.2), IL-17A (TC11-18H10), Foxp3 (FJK-16s), RORγt (AFKJS-9), T-bet (eBio4B10), and Fixable Viability Dye eFluor® 506 were purchased from eBioscience; antibodies specific for CD103 (M290), Bcl-2 (3F11), IFN-γ (XMG1.2), and GATA3 (L50-823) were purchased from BD Biosciences; antibody specific for mouse neuropilin-1 (761705) was purchased from R&D systems; antibody specific for mouse FCγII/III (2.4G2) was prepared by Harlan.
Serum IgE concentrations were measured using Mouse IgE ELISA Ready-SET-Go kit (eBioscience). Serum IgG1 and IgG2b were measured using Mouse Ig Isotyping ELISA Ready-Set-Go kit (eBioscience). Experiments were performed following the manufacturer’s instructions.
CD4+CD25+ YFP+ T cells from Foxp3-Cre, Tbx21fl/fl reg -Foxp3-Cre, Gata3fl/fl-Foxp3-Cre and DKO mice (CD45.2) and naive CD4+CD25−CD45RBhi T cells from CD45.1 congenic mice (Line 7) were prepared by cell sorting and then washed in sterile PBS. Each Rag1−/− recipient mouse (Line 146) was injected intravenously with either 2 × 105 naive CD4+CD25−CD45RBhi T cells alone, or together with 5 × 104 CD4+CD25+YFP+ Treg cells. Treg cells purified from tri-color mice were transferred alone into Rag1−/− mice for 2 weeks, or co-transferred with CD45.1 congenic naive CD4+CD25−CD45RBhi T cells into Rag1−/− mice for 8 weeks before analysis. For splenocytes transfer experiments, 10 × 106 splenocytes from either Foxp3-Cre or DKO mice were injected into each Rag1−/− recipient (Line 146). Body weight of mice was monitored weekly in all experiments.
Bone marrow cells from femurs of Foxp3-Cre, Tbx21fl/fl-Foxp3-Cre, Gata3fl/fl-Foxp3-Cre and DKO mice (CD45.2) were mixed with those from CD45.1 congenic mice (Line 7) at 1:1 ratio and then injected retro-orbitally (10 million cells per mouse) into Rag1−/− mice (Line 146) after lethal irradiation (450 Rads twice with 2-3 hours apart).
Groups were compared with Prism 6 software (GraphPad) using a 2-tailed unpaired Student’s t test or an ordinary one-way ANOVA. Data are presented as mean ± SEM. P < 0.05 was considered significant.
We thank Drs. Ethan Shevach, Yasmine Belkaid and William Paul for their critical reading of our manuscript and helpful discussions. We also thank Dr. Liying Guo for the help on ELISA experiments, Drs. Nezih Cereb and Soo Young Yang at Histogenetics (Ossining, NY) and Dr. Steven Reiner of Columbia University (NY, NY) for constructing and providing the floxed Tbx21 mice, and the NIAID flow cytometry core facility for cell sorting. The work is supported by the Division of Intramural Research, NIAID, National Institutes of Health, USA.
COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests.
Author contributions: F. Y. performed all the experiments; S. S. and J. E. helped in some experiments and made suggestions to the manuscript; L. F. made the transgenic mouse strains; F. Y. and J. Z. designed the experiments, analyzed the data and wrote the paper.