Control of regulatory T cell lineage commitment and maintenance

Abstract

Foxp3-expressing regulatory T (Treg) cells suppress pathology mediated by immune responses against self and foreign antigens, and commensal microorganisms. Sustained expression of the transcription factor Foxp3, a key distinguishing feature of Treg cells, is required for their differentiation and suppressor function. In addition, Foxp3 expression prevents deviation of Treg cells into effector T cell lineages and confers dependence of Treg cell survival and expansion on growth factors, foremost interleukin-2, provided by activated effector T cells. In this review we discuss Treg cell differentiation and maintenance with a particular emphasis on molecular regulation of Foxp3 expression, arguably a key to mechanistic understanding of biology of regulatory T cells.

One century ago Paul Ehrlich proposed that the immune system is programmed to avoid the generation of autoreactive immune responses and termed this aversion to autoreactivity “horror autotoxicus” (Ehrlich, 1906). Ehrlich's observations that goats could make antibodies against the blood components of other goats, but not against their own blood, represented the first evidence of immunological self-tolerance and led to his prediction that “either the disappearance of receptors or the presence of autoantitoxins is foremost among [immune] contrivances” that mediate self-tolerance. It is currently well accepted that immunological tolerance is mediated by two categories of mechanisms – recessive and dominant. Recessive tolerance refers to cell-intrinsic mechanisms that include elimination of self-reactive thymocytes or chronically stimulated peripheral T cell clones by apoptosis or their inactivation due to anergy induction. Dominant tolerance is mediated by a specialized subset of immune cells acting in trans to restrain pathogenic immune responses. Although several lymphoid cell subsets exhibit suppressive or immunomodulatory properties, Foxp3 expressing Treg cells represent the only currently known population of lymphocytes acting as dedicated mediators of dominant tolerance, whose suppressor function is vital for the maintenance of immune homeostasis. Treg cells suppress immune responses through numerous mechanisms including the production of anti-inflammatory cytokines, direct cell-cell contact, and by modulating the activation state and function of antigen presenting cells (reviewed by Shevach in this issue). It is becoming increasingly apparent that in addition to restraining autoimmunity, Treg cell suppressor function prominently features in regulation of other forms of immune-mediated, and likely, non-immune inflammation, and affects immune responses to infection and tumor growth.

Foxp3 and regulatory T cell differentiation

Initial insights that revealed the existence of a thymus-derived subset of cells capable of mediating immune tolerance through their suppression of other cells came from neonatal thymectomy (nTx) experiments performed by Nishizuka and Sakaguchi and their colleagues and in studies of tolerance in chicken-quail chimeras performed by Le Dourain and her colleagues (Ohki et al., 1987). In mouse thymectomy studies, nTx between 2 and 4 days of life resulted in T cell-mediated lesions, which could be alleviated through the transfer of thymocytes or splenocytes from adult euthymic mice (Asano et al., 1996; Bonomo et al., 1995; Nishizuka and Sakakura, 1969; Sakaguchi et al., 1982). Thus, a population of cells generated in the mouse thymus after 3 days of life mediates immune tolerance in a dominant, cell-extrinsic manner. In a landmark 1995 paper, Sakaguchi and colleagues described a population of IL-2 receptor α-chain (CD25) expressing regulatory (Treg) CD4⁺ T cells that were capable of suppressing immune responses in a variety of experimental models (Sakaguchi et al., 1995). Recent progress in understanding Treg cell biology came with the discovery of the X-chromosome encoded gene Foxp3 during efforts to identify the genetic basis for the autoimmune disorder in human patients suffering from IPEX (immune dysregulation, polyendocrinopathy, enteropathy, X-linked) syndrome and in the spontaneous mouse mutant scurfy (Bennett et al., 2001; Brunkow et al., 2001; Chatila et al., 2000; Wildin et al., 2001). Mice and humans harboring a loss-of-function mutation in the Foxp3 gene are affected by fatal early onset lymphoproliferative immune mediated disease affecting a variety of organs and tissues. Subsequent studies revealed stable expression of high amounts of Foxp3 restricted to Treg cells and its requirement for Treg cell differentiation (Fontenot et al., 2003; Fontenot et al., 2005c; Hori et al., 2003; Khattri et al., 2003; Wan and Flavell, 2007) and for their suppressor function, proliferative potential and metabolic fitness (Gavin et al., 2007; Lin et al., 2007). Furthermore, sustained Foxp3 expression in mature Treg cells is necessary for maintenance of the Treg cell phenotype and suppressor function; loss of Foxp3 or its diminished expression in Treg cells leads to acquisition of effector T cell properties including production of immune response promoting cytokines such as IL-2, IL-4, IL-17, and IFN-γ (Wan and Flavell, 2007; Williams and Rudensky, 2007) Together, these studies established a central role for Foxp3 in defining the Treg cell lineage.

Thymic and peripheral differentiation of Treg cells

Given the central role of Foxp3 in establishing and maintaining the Treg transcriptional program, elucidation of the cell-extrinsic and –intrinsic cues that influence Foxp3 expression will facilitate an understanding of the differentiation, maintenance, and function of Treg cells.

Foxp3 induction leading to Treg cell differentiation occurs relatively late during thymic differentiation. In addition, peripheral naïve CD4⁺ T cells are also capable of Foxp3 up-regulation during differentiation into so-called induced Treg cells or iTreg cells (reviewed in this issue by Lafaille). Presently, it is not clear whether these two modes of Treg differentiation serve different biological needs or have partially or fully redundant functions and whether mechanistic requirements for Treg cell generation in the thymus and in the periphery are distinct. Furthermore, the overall contribution of thymically generated Treg cells (tTreg) and peripherally generated iTreg cells to the overall pool of Treg cells in secondary lymphoid organs and non-lymphoid tissues under basal conditions and in the course of immune challenge or tumor growth remain largely unknown. Below we will discuss these issues in the context of recent work related to regulation of Foxp3 gene expression in the thymus and periphery.

Role of TCR signals in Treg cell differentiation

T cell receptor (TCR) signals of distinct strength and duration have been proposed to guide CD4 vs. CD8 T cell fate decision during thymic differentiation (Germain, 2002; Singer et al., 2008). Furthermore, distinct strong TCR signaling facilitates generation of “specialized” populations of T cells such as CD1d-restricted NKT cells, MR1-restricted MAIT (mucosal associated invariant T cells) expressing semi-invariant TCRs, CD8αα T cells, and H-M3-restricted CD8⁺ T cells expressing diverse TCR. Considering that TCR-ligand interactions are central to T cell lineage decision-making, an essential role and requirement for particular TCR signaling in Foxp3 induction and Treg cell lineage commitment comes at no surprise. Early observations of high expression of molecules known to be up-regulated in T cells upon acute or chronic TCR signaling, i.e. CD25, CD5, and CTLA-4, strongly supported the idea that Treg cells are exposed to TCR signals of increased strength. CD5 expression in thymocytes and peripheral T cells is proportional to the strength of TCR signals these cells are exposed to, and thereby, CD5 acts as a rheostat attenuating TCR signaling in a tunable manner through recruitment of the tyrosine phosphatase SHP-1 to the CD5 cytoplasmic tail (Azzam et al., 2001; Wong et al., 2001). Indeed, CD5- or SHP-1-deficient mice have increased frequencies of tTreg cells (M. Gavin and A.Y.R., unpublished observations) (Carter et al., 2005).

The idea of an essential role of TCR engagement in Foxp3 induction during thymic differentiation of Treg cells gained direct experimental support from earlier observations of endogenous TCR rearrangement-dependent Treg cell generation in mice expressing a transgenic TCR (Olivares-Villagomez et al., 1998). Next, it was found that, in the absence of endogenous TCR rearrangement, Treg cells expressing a transgene-encoded TCR are generated only when a cognate ligand for the receptor, encoded by another transgene, was co-expressed (Apostolou et al., 2002; Jordan et al., 2001; Kawahata et al., 2002; Walker et al., 2003). Pronounced negative selection accompanying generation of tTreg cells in these studies (Apostolou et al., 2002; Kawahata et al., 2002; Walker et al., 2003) led to a notion of selective survival of thymic self-reactive Treg precursors expressing Foxp3 as opposed to instructive TCR signals driving Foxp3 expression (van Santen et al., 2004). Consistent with this idea is increased Foxp3-dependent expression of pro-survival molecules and known attenuators of TCR signaling such as CTLA-4. Nevertheless, as we discuss below, recent studies, although not discounting superior survival of Foxp3⁺ thymocytes, provide further support to an instructive role for TCR signaling in Treg lineage commitment in the thymus.

In addition to somewhat contrived experimental models of Treg cell differentiation in TCR transgenic mice, a direct sequence analysis of the TCR repertoires displayed by Treg vs. non-Treg cells bearing a fixed transgene-encoded TCRβ chain showed that TCRα sequences in Treg cells were diverse and only partially overlapping with non-Treg cells (Hsieh et al., 2004; Pacholczyk et al., 2006; Pacholczyk and Kern, 2008; Wong et al., 2007b). Retroviral transfer of Treg or naïve CD4⁺ T cell TCRα libraries into RAG^−/− TCR transgenic T cells followed by the analysis of reactivity of Treg- vs. non-Treg-derived TCR showed that Treg TCR exhibit increased self-reactivity. This is based on their ability to confer to T cells the capacity for robust expansion and induction of autoimmune pathology upon transfer into lymphopenic recipient mice. Nevertheless, these TCR transgenic T cells transduced with Treg cell TCR are able to mount only weak in vitro responses to syngeneic APC in comparison to the responses of these cells to the “foreign” ligand recognized by the transgenic TCR (Hsieh et al., 2004). Thus, it seems likely that although increased, the affinity of Treg cell TCR for self-antigens is well below the range of conventional T cell receptors recognizing foreign antigen in a typical immune response. These data support the idea that Treg cell selection is facilitated by TCR with affinities for self peptide-ligands falling within a range between positive selection of conventional CD4⁺ T cells and negative selection of high affinity self-reactive T cells. This notion is supported by observations that the presence or absence of the Foxp3 gene does not affect efficiency or sensitivity of negative selection of thymocytes by a high affinity TCR ligand (Chen et al., 2005; Hsieh et al., 2006) and that activated T cells in Foxp3-deficient mice displayed TCR utilized by Treg cells in Foxp3-sufficient mice (Hsieh et al., 2006). Thus, in the absence of Foxp3, T cell precursors with self-reactive TCR supporting Foxp3 induction and differentiation of Treg cells are not deleted, and instead, upon maturation, become activated and contribute to pathology in Foxp3-deficient animals.

Understanding of a role of TCR in Treg cell differentiation and function could have benefited from analysis of transgenic mice expressing a single monoclonal Treg cell TCR originating from Treg cells present in normal mice. An unexpected problem in these studies has been that in the absence of endogenous TCR rearrangements on a RAG-deficient background such transgenic Treg cell-derived TCR support differentiation of a very small number, if any, of Foxp3⁺ Treg cells (C.S. Hsieh, J. Marie, J.D. Fontenot, and A.Y.R., unpublished observations). However, very recent experiments by the Hsieh group demonstrated an efficient generation of Foxp3⁺ thymocytes expressing a single Treg cell-derived transgenic TCR when the number of precursor cells was dramatically reduced. Although mechanism(s) underlying this phenomenon remains uncertain, we suggest that it is likely the result of a very narrow window of affinity of TCR-ligand interactions between negative selection and positive selection of Foxp3- cells that satisfy the requirements for Foxp3 induction. Overall, these data suggest that an unprecedented intra-clonal competition limits induction of Foxp3 in thymic precursors expressing TCR of identical specificity ensuring a very broad Treg TCR specificity repertoire (Bautista et al., 2009).

In addition to TCR signals, CD28 co-stimulatory signals have an essential cell-intrinsic role in the differentiation of tTreg cells as illustrated by marked decreases in frequencies of tTreg cells in CD28-deficient and CD80-CD86-deficient mice (Salomon et al., 2000; Tai et al., 2005). Additionally, the lck-binding domain of the CD28 cytoplasmic tail is critical for the induction of Foxp3 (Tai et al., 2005), thereby suggesting a role for coordinated TCR-CD28 signaling in thymic differentiation of Treg cells.

Several transcription factors downstream of the TCR and CD28, including NFAT, NF-κB, and AP-1, have been implicated in Treg cell differentiation. In agreement with their proposed role in up-regulation of Foxp3 expression, NFAT and AP1 bind to the Foxp3 promoter (Mantel et al., 2006). In addition, CREB-ATF-1 was shown to bind an intronic regulatory region at the Foxp3 locus (Foxp3conserved non-coding sequence 2, Foxp3-CNS2) and enhance expression of a luciferase reporter driven by a Foxp3 promoter in a transient transfection assay (Kim and Leonard, 2007). Also, studies of mice with targeted ablation of genes encoding components of a TCR-dependent NF-κB signaling pathway, PKC-θ, Bcl10, CARMA1, and IκB kinase 2, showed significant impairment in tTreg differentiation (Gupta et al., 2008; Medoff et al., 2009; Schmidt-Supprian et al., 2004). Furthermore, a recent study of a CARMA1 (caspase recruitment domain 11) mutation generated in an ENU-mutagenesis screen demonstrated a complete cell-intrinsic block in Treg cell differentiation in the thymus (Barnes et al., 2009). Importantly, Allegre and colleagues found that provision of survival signals to CARMA1-deficient T cell precursors through forced expression of Bcl2 or a constitutively active form of STAT5 (STAT5-CA) fail to rescue the defect in thymic Foxp3 induction and tTreg differentiation (Molinero et al., 2009). These observations strongly support an instructive role for TCR signaling in Treg cell differentiation.

Although these studies made the importance of NF-κB signaling in thymic Treg differentiation clear, little is known about the specific NF-κB family members that are required for Treg cell differentiation. One recent study indicated that frequencies of Foxp3-positive thymocytes were reduced in mice bearing a mutation in the p105 gene that results in an inability of IKK (IκB kinase) to phosphorylate cleaved p50 (Sriskantharajah et al., 2009). In addition, it is still unclear how NF-κB family members act to mediate differentiation of Foxp3⁺ Treg cells, what NF-κB dimer(s) is critical in this process, and, if NF-κB family members directly regulate the Foxp3 gene, what sites at the Foxp3 locus they bind to and how they affect Foxp3 expression.

Role of cytokines in thymic differentiation of Treg cells

Similar to aforementioned study of TCR utilization by peripheral Treg cells, sequence analysis of TCR repertoires displayed by thymic precursors of Treg versus non-Treg cells revealed partial overlap and bias of thymic Treg TCR towards self-recognition (Hsieh et al., 2006). Thus, the same TCR with an increased reactivity for self can be expressed by a Treg and non-Treg cell suggesting that TCR signals alone are not sufficient to drive Foxp3 up-regulation and Treg lineage commitment. Likewise, only some thymocytes in TCR transgenic RAG-deficient mice differentiate into Treg cells when confronted with the cognate ligand, whereas the rest become anergic “non-Treg” cells (Apostolou et al., 2002; Jordan et al., 2001). Additional evidence pointing to a requirement for a second signal was provided by observation of a delay in generation of Foxp3⁺ thymocytes in neonatal mice whereas their CD25⁺ Foxp3⁻ precursors (see below) are readily detectable in the thymus immediately after birth (Burchill et al., 2008; Fontenot et al., 2005a; Lio and Hsieh, 2008). Because the early wave of thymocytes is not known to be devoid of, but rather is enriched in self-reactive TCR due to lack of TdT expression (Gavin and Bevan, 1995), this delay is most likely due to the paucity of an additional factor(s) required for Treg cell differentiation.

An essential second signal for Treg cell differentiation is afforded by IL-2 and to a lesser degree two other common gamma-chain (γc) cytokines, IL-7 and IL-15. Mice lacking IL-2 or IL-2Rα chain exhibit an approximately 50% decrease in proportion and numbers of Foxp3⁺ thymocytes, whereas IL-15 or IL-7 deficiency does not affect generation of Foxp3⁺ cells. In contrast, mice lacking γc are completely devoid of Foxp3⁺ thymocytes and peripheral Foxp3⁺ T cells (Fontenot et al., 2005b) and so are mice with a combined deficiency in IL-2, IL-7, and IL-15 (Burchill et al., 2007a; Malek, 2008; Vang et al., 2008).

Building on the aforementioned observation of CD25⁺Foxp3⁻ cells preceding generation of CD25⁺Foxp3⁺CD4⁺ single positive (SP) thymocytes in the neonatal thymus and on a prominent role for IL-2 and increased strength TCR signaling in Treg cell differentiation, Lio and Hsieh proposed a two-step model of thymic Treg differentiation based on their intrathymic cell transfer studies (Burchill et al., 2008; Lio and Hsieh, 2008). According to this model, an increased TCR signal results in the up-regulation of CD25, increasing the responsiveness of tTreg precursor cells to consequent IL-2 signals that result in induction of Foxp3 (Burchill et al., 2008; Lio and Hsieh, 2008). STAT5 activated downstream IL-2 and other common γ-chain cytokine receptors represents a likely candidate transcription factor for direct regulation of Foxp3 expression (Burchill et al., 2008). Indeed, STAT5 was shown to bind the Foxp3 promoter and Foxp3-CNS2 element. Furthermore, induced ablation of a conditional Stat5 allele in DP thymocytes results in a drastic reduction in Foxp3⁺ CD4SP thymocytes with the remaining Foxp3⁺ thymocytes originating from cells that escape STAT5 deletion (Burchill et al., 2007b; Yao et al., 2007). Additionally, expression of a constitutively active STAT5 results in expansion of tTreg cells and can rescue tTreg cell numbers in the absence of IL-2 (Burchill et al., 2003; Burchill et al., 2008).

Nevertheless, the molecular mechanisms by which IL-2 facilitates Treg cell differentiation are not clear. Although there is an important function for the IL-2 receptor-STAT5 axis in the differentiation of Treg cells, it remains unclear if STAT5 directly drives Foxp3 transcription, induces changes in the chromatin characteristics at the Foxp3 locus, or promotes survival or expansion of Treg cells or their precursors. If STAT5 serves an essential non-redundant role in driving expression of Foxp3, rather than facilitating the survival of Foxp3-expressing cells, then forced expression of a pro-survival molecule like Bcl2 in STAT5-deficient cells should not rescue Treg cell deficiency. However, expression of a Bcl-2 transgene rescued the differentiation of STAT5-deficient Treg cells suggesting that Foxp3 can be induced in the absence of STAT5 in developing thymocytes (S. Malin and M. Busslinger, personal communication). Additionally, STAT5 is not required for maintenance of Foxp3 expression in Treg cells, as conditional ablation of STAT5 in Treg cells did not result in markedly lower of Foxp3 expression (Y. Zheng and A. Y. R., unpublished observations). Similarly, Foxp3 expression was still maintained in Treg cells upon conditional deletion of calcineurin B1 and impaired NFAT activation (J. Kim, G. Crabtree and A.Y.R., unpublished observations).

In addition to TCR and γc cytokine receptors, a recent study by Chen and colleagues suggested that an early wave of Foxp3⁺ thymocyte generation is dependent upon TGF-β receptor (TGFβR) signaling known to drive Foxp3 expression in peripheral naïve Foxp3⁻ T cells upon their activation (see below). In support of this notion, ablation of the TGFβRI subunit in DP thymocytes resulted in a profound, but transient impairment in generation of Foxp3⁺ Treg cells during the first week of life followed by the recovery of Foxp3 thymocyte numbers to those observed in wild-type mice (Liu et al., 2008). Earlier reports also failed to find a defect in tTreg generation in week-old mice lacking of TGFβ1 or subjected to TGFβRII ablation in DP thymocytes (Li et al., 2006; Marie et al., 2005; Marie et al., 2006). It was proposed that in resemblance of peripheral generation of Foxp3⁺ iTreg cells discussed in detail below, TGF-β induced Smad-mediated activation of the Foxp3 locus through interaction with a conserved Smad-NFAT response element (Foxp3 conserved non-coding sequence 1 – Foxp3-CNS1) in thymocytes is essential for tTreg cell generation (Tone et al., 2008). The recovery of tTreg cells was explained by compensation of TGF-β signaling deficiency afforded by increasing amounts IL-2 (Liu et al., 2008). However, in addition to a proposed role for Smad-mediated induction of Foxp3 expression, there are several alternative explanations for the observed effect of TGF-β on tTreg generation including the survival of tTreg or their precursors when a relatively small cohort of thymic precursors reaches maturity in the neonatal thymus. Thus, elucidation of the transcription factors that bind to the Foxp3 gene and are required for transcriptional activation and maintenance of Foxp3 expression remains a challenge for the field.

Requirements for peripheral differentiation of Treg cells

Initial in vitro and in vivo studies showed that induction of Foxp3 expression in peripheral naïve T cells is facilitated by a sub-optimal TCR signal or by a combination of strong TCR signal with high amounts of TGFβ (Chen et al., 2003; Kretschmer et al., 2005; Selvaraj and Geiger, 2007; Zheng et al., 2004). However, these findings do not necessarily indicate that any chronic exposure of a peripheral T cell to a cognate self or foreign antigen can lead to generation of Treg cells in the periphery nor do they suggest that iTreg cells make up a large proportion of the peripheral Treg population in the secondary lymphoid organs. In opposition to the latter idea, substantial overlaps were observed between TCR repertoires displayed by thymic and peripheral Foxp3⁺ cells and by thymic and peripheral Foxp3⁻ CD4⁺ T cells, respectively. Complementing these data, a comparison of TCR utilization of thymic Foxp3⁺ versus Foxp3⁻ cell subsets and of peripheral Foxp3⁺ versus Foxp3⁻ cell subsets showed limited overlap (Hsieh et al., 2006). In agreement with these findings, in mice expressing diabetogenic transgene-encoded αβTCR BDC2.5 specific for unidentified pancreatic self antigen thymic and periperal Foxp3⁺CD4⁺ cells expressed a similar repertoire of endogenous TCRα chains, which were used in this study as unique tags of individual T cell clones. This elegant approach, made possible because generation of Foxp3⁺ cells in BDC2.5 TCR trangenic mice requires endogenous TCRα chain rearrangement, revealed that TCRα chain utilization in thymic and peripheral Foxp3⁻ non-Treg cells is distinct from Foxp3⁺ Treg cells (Wong et al., 2007a). The most remarkable observation in this study was that BDC2.5 TCR-expressing Treg and non-Treg cell clones present in the pancreatic lymph nodes and pancreas, and therefore exposed to chronic stimulation by self antigen, remained distinct with their endogenous TCRα chain usage reflecting their thymic origin (Wong et al., 2007a). Together, these results indicate that the majority of Treg cells present in the periphery are of thymic origin and that iTreg generation has specific prerequisits.

A requirement for a distinct TCR signal and specificity for iTreg generation was revealed by the analysis of the TCR repertoire of Foxp3⁺ T cells generated upon transfer of purified Foxp3⁻ CD4⁺ T cells into lymphopenic recipients in comparison to that of the progeny of divided cells that remained Foxp3 negative (Lathrop et al., 2008). The resulting TCR repertoires were distinct and only partially overlapping in resemblance of TCR utilization by Foxp3⁺Treg cells and Foxp3⁻ “non-Treg” CD4⁺ T cells present in unmanipulated mice (Hsieh et al., 2004; Lathrop et al., 2008). These results suggest that TCR of certain specificities support iTreg differentiation. A non-mutually exclusive possibility is that particular TCR specificities confer a poised state to the Foxp3 locus and thus support more efficient Foxp3 induction.

First insights into similarities and differences between signal requirements for thymic versus peripheral Foxp3 induction came from in vitro studies. First, CTLA-4 is dispensable for tTreg cell differentiation, but is required for TGF-β mediated iTreg cell generation in vitro (Zheng et al., 2006). Second, CD28 cross-linking inhibits induction of Foxp3 in peripheral naïve CD4 T cells upon stimulation with TGF-β (Benson et al., 2007; Kim and Rudensky, 2006) consistent with a requirement for sub-optimal TCR stimulation for iTreg cell generation. As a further mechanistic insight into TCR signaling requirements for iTreg generation, Merkenschlager and colleagues have demonstrated that early withdrawal of TCR signaling through use of PI3K-mTOR signaling pathway inhibitors after 18 hours of stimulation resulted in robust induction of Foxp3 (Sauer et al., 2008). Consistent with these findings, the Mathis and Benoist group showed that sustained Akt activation inhibits stable Foxp3 induction in peripheral Foxp3⁻ CD4⁺ T cells (Haxhinasto et al., 2008). A similar trend was observed upon modulation of Akt during induction of Foxp3⁺ cells in fetal thymic organ cultures, suggesting that in this regard iTreg and tTreg induction are similar (Haxhinasto et al., 2008). Interestingly, in contrast to the lack of generation of CARMA-1 deficient Foxp3⁺ thymocytes, CARMA1-deficient peripheral CD4 T cells were able to induce Foxp3 in response to TGF-β, supporting the idea that mechanisms for differentiation of tTreg and iTreg are distinct and that TCR-CD28 associated NF-κB signals are dispensable for peripheral Foxp3 induction (Barnes et al., 2009). However, distinct requirements for CARMA-1 for generation of tTreg versus iTreg cells might be a consequence of iTreg generation in response to stronger TCR signals in the absence of CARMA-1. In contrast, such signals may lead to death of thymic precursors lacking CARMA-1.

TGF-β receptor signaling appears to be required for most, if not all, of the induction of Foxp3 among peripheral naïve CD4⁺ T cells (Chen et al., 2003; Kretschmer et al., 2005; Selvaraj and Geiger, 2007; Zheng et al., 2004). IL-2 is also required for TGF-β mediated induction of Foxp3 in peripheral T cells in vitro (Davidson et al., 2007; Horwitz et al., 2008). In addition to potential direct STAT5-dependent regulation of the Foxp3 locus and promotion of cell survival and division in the presence of high amounts of TGF-β, IL-2 opposes differentiation of activated CD4⁺ T cells into T helper 17 (Th17) cells (Laurence et al., 2007). The latter differentiation pathway is favored when TCR and TGFβR activation in naïve CD4⁺ T cells coincides with IL-6R stimulation (Bettelli et al., 2008). (A complex relationship between Th17 and iTreg cells, representing alternative, yet related CD4⁺ T cell differentiation fates, is discussed in more detail by Littman and colleagues in this issue). Another mechanism by which TGF-β may regulate Treg cell differentiation is through the repression of Gfi-1, a transcriptional repressor that inhibits the differentiation of both iTreg and Th17 cells upon activation of peripheral T cells under Th2 conditions (Zhu et al., 2009).

The induction of Foxp3 upon chronic antigen exposure in vivo also requires TGF-β receptor signaling and is inversely correlated with cellular proliferation (Kretschmer et al., 2005). One possible explanation for this phenomenon is provided by our recent findings indicating that TGF-β cooperates with TCR signals to induce Foxp3 in part by antagonizing cell cycle dependent recruitment of maintenance DNA methyltransferase I (Dnmt1) to the Foxp3 locus resulting in its inactivation (Josefowicz et al., 2009). Not mutually exclusive is the possibility that signals, capable of inducing Foxp3, initiate chromatin remodeling and establish a poised state of the Foxp3 locus, and that consequent robust proliferation accompanied by propagation of CpG methylation by Dnmt1 may erase or prevent establishment of this permissive state. Therefore, the cytostatic effects of inhibitory signals emanating from CTLA-4 and TGF-βR may be partially responsible for their effects on Foxp3 induction. Thus, TGF-β appears to mediate differentiation of Treg cells through both direct and indirect mechanisms, with Smad3 and NFAT binding to Foxp3-CNS1 and a likely role for TGF-β signaling in survival or fitness of tTregs or their precursors.

The sum of the data so far indicate that TCR-induced sustained expression of high amounts of Foxp3 in peripheral T cells is influenced by particular aspects of intracellular signaling pathways, by kinetics of cellular proliferation, and synergy with other signals, such as TGF-β and IL-2. These features imply that iTreg cell differentiation is limited to particular environments. The bulk of experimental evidence points to gut-associated lymphoid tissues (GALT) as a unique environment favoring iTreg generation. In this regard, CD103⁺ dendritic cells present in GALT or the gut-draining mesenteric lymph nodes are capable of inducing Foxp3 expression in peripheral naïve CD4⁺ T cells through production of TGF-β and retinoic acid (Annacker et al., 2005; Benson et al., 2007; Coombes et al., 2007; Sun et al., 2007). While TGF-β acts directly, retinoic acid may predominantly enhance the induction of Foxp3 through curtailing production of IL-4, IL-21, and IFN-γ by bystander CD44^hi effector memory CD4⁺ T cells (Hill et al., 2008). However, retinoic acid still exerts direct effects on iTreg cell differentiation (Elias et al., 2008; Xiao et al., 2008). Interestingly, although retinoic acid augments Foxp3 induction, it inhibits induction of IL-10, indicating that iTreg cells and IL-10 producing Tr1 cells may represent competing and alternative cell lineages (Maynard et al., 2009). The GALT and mesenteric lymph nodes represent anatomical sites amenable to the induction of Foxp3 in response to chronic antigen exposure under tolerogenic conditions (Belkaid and Oldenhove, 2008; Curotto de Lafaille et al., 2008; Hall et al., 2008; Kretschmer et al., 2005; Kretschmer et al., 2006). This notion is further supported by a distinct TCR repertoire among Treg cells present in the mesenteric lymph nodes compared to Treg cells from other lymphoid compartments (Lathrop et al., 2008), likely reflecting the efficient generation of iTreg cells induced by distinct gut associated antigens such as food and commensal microbiota derived antigens.

Molecular regulation of Foxp3 gene expression

One interesting characteristic of the Foxp3 gene is the weak activity of the promoter alone, as observed in luciferase reporter assays (Kim and Leonard, 2007; Tone et al., 2008). Therefore, it seems likely that expression of Foxp3 is heavily dependent on other proximal regulatory elements. As discussed above, Foxp3-CNS1, which contains binding sites for NFAT and TGF-β-activated Smad3, is likely important for the induction of Foxp3 in peripheral naïve CD4⁺ T cells. In addition to Smad3-dependent activation of Foxp3-CNS1, other Foxp3 regulatory elements are likely to impact the chromatin state thereby promoting accessibility of the Foxp3 locus and increasing probability of its induction. Predictably, permissive chromatin modifications at Foxp3-CNS1 and the Foxp3 promoter coincide with or directly precede Foxp3 expression (Kim and Leonard, 2007; Mantel et al., 2006; Sauer et al., 2008; Tone et al., 2008). Thus, it will be important to identify additional regulatory elements, transcription factors and chromatin modifying and remodeling complexes that act upon these elements to promote changes in chromatin state allowing for induction of Foxp3 expression.

Given the central role Foxp3 plays in maintaining the Treg cell transcriptional program and cellular phenotype, maintenance of Foxp3 expression is central to Treg lineage stability. Although in vitro manipulation of Treg cells such as cross-linking of the TNF receptor superfamily member OX40 results in a loss of Foxp3 expression and suppressor activity in mature Treg cells (So and Croft, 2007; Vu et al., 2007), adoptive transfers of Treg cells into Treg cell deficient mice demonstrated stability of Foxp3 expression (Komatsu et al., 2009). Additionally, cell-fate mapping studies employing inducible Cre-mediated genetic marking of Treg cells revealed heritable long-term maintenance of Foxp3 expression in the progeny of genetically tagged cells and, therefore, Treg lineage stability in unmanipulated mice (Rubtsov and A. Y. R., unpublished observations).

What are the mechanisms for stable maintenance of Foxp3 expression in mature Treg cells? Several recent studies have pointed to CpG dinucleotide methylation at the Foxp3 locus – at the promoter and at Foxp3-CNS2 – as an important determinant in regulation of Foxp3. Demethylation of CpG motifs at the Foxp3 locus has been correlated with stable Foxp3 expression in ex vivo isolated human and mouse Treg cells. In contrast, these elements remain methylated in iTreg cells generated in vitro that do not stably express Foxp3 (Baron et al., 2007; Floess et al., 2007; Polansky et al., 2008). One possible consequence of continuous methylation of these CpG containing elements in iTreg cells is that transcription factors like the methyl-sensitive CREB fail to bind to Foxp3-CNS2. It is implicit in this line of reasoning that demethylation of Foxp3-CNS2 facilitates binding of transcription factors that mediate stable heritable maintenance of Foxp3 expression.

It is noteworthy that there are differences in Foxp3 expression upon activation of conventional T cells in humans and mice (Ziegler, 2006). While stable high expression of Foxp3 is restricted to Treg cells in both species, Foxp3 is induced after stimulation of human T cells (Allan et al., 2005; Gavin et al., 2006; Morgan et al., 2005; Walker et al., 2003; Wang et al., 2007). This relatively low Foxp3 expression in activated human T cells is dependent upon TGF-β produced by activated T cells or present in the serum and it does not result in acquisition of Treg cell phenotype and suppressor function (Gavin et al., 2006; Wang et al., 2007). Furthermore, induction of Foxp3 expression in conventional human T cells in the presence of high amounts of TGF-β fails to confer suppressor function (Shevach et al., 2008; Tran et al., 2007). It is likely that transient, low expression of Foxp3 upon activation of conventional human T cells and unstable expression of Foxp3 in iTreg cells generated in vitro is the result of a lack of engagement of the highly methylated Foxp3-CNS2. Consistent with this idea, pharmacologic inhibition, knockdown, or ablation of the Dnmt1 gene and resultant CpG motif demethylation substantially increases both induction and stability of Foxp3 expression (Kim and Leonard, 2007; Polansky et al., 2008). The mechanisms responsible for establishing the appropriate chromatin characteristics, demethylation of Foxp3-CNS2, and the propagation of these states for heritable maintenance of Foxp3 expression in dividing Treg cells are largely unknown. However, Foxp3 protein itself appears to be required for the heritable maintenance of Foxp3 expression, but does not augment the expression level of Foxp3 on a per cell basis suggesting existence of a feed-forward regulatory loop (Gavin et al., 2007). Based on these observations, it is possible that Foxp3, a lineage specifying transcription factor of Treg cells, promotes its own heritable maintenance in the progeny of dividing Treg cells representing a simple mechanism of Treg lineage stability.

Further advances in mechanistic understanding of known and elucidation of unknown signals that determine the sustained induction of Foxp3 during thymic and peripheral Treg differentiation will facilitate development of novel approaches to the therapeutic manipulation of regulatory T cells.

Figure 1. Differentiation of thymic and induced Treg cells.

A, Most Foxp3⁺ thymic Treg (tTreg) cells differentiate from Foxp3-negative CD4⁺ SP thymocytes. The process of tTreg cell differentiation as defined by induction of Foxp3 requires: 1) increased strength of T cell receptor (TCR) stimulation by self-peptide-MHC complexes presented by thymic epithelial cells (TEC) or dendritic cells (DC), 2) CD28 signaling induced by CD80 and CD86 ligand expressed on antigen presenting cells, and 3) high affinity IL-2 receptor and other γc cytokine receptor signaling. Treg cell homeostasis is dependent on exocrine IL-2 produced by effector T cells. B, Foxp3⁺ Treg cells can also be induced from peripheral naive CD4⁺ T cells (iTreg). Conditions favoring the peripheral induction of Foxp3 include chronic low dose antigen stimulation under tolerizing conditions. iTreg cells are likely prominent in the gut-associated lymphoid tissue where chronic exposure to food, commensal, or environmental antigens likely facilitates their generation. Suboptimal co-stimulation is critical for differentiation of iTreg cells with a particularly important role for the immunomodulatory cytokine TGF-β. Additionally, IL-2 and the vitamin A metabolite retinoic acid (RA) facilitate induction of Foxp3 in peripheral Foxp3⁻ CD4 T cells. CD103⁺ DC, which produce RA and TGF-β are potent inducers of Foxp3 expression in activated T cells. These dendritic cells are present in high numbers in the gut where they likely limit immune inflammation through the generation of iTreg cells.

Figure 2. Signaling pathways, transcription factors, and regulatory DNA elements that control Foxp3 expression.

A, T cell receptor (TCR) and CD28 costimulatory signals are important for induction of Foxp3. Downstream of the TCR-CD28 signal, the NFκB pathway via PDK1, PKC-θ, the BCL-10, MALT1, CARMA1 complex and IKK is critical for Foxp3 expression in the thymus, whereas CREB, NFAT, and AP1 have been demonstrated to bind to regulatory DNA elements at the Foxp3 locus. In contrast to thymic differentiation of Treg cells, weaker TCR-CD28 signals appear to favor peripheral induction of Foxp3, as evidenced by increased induction of Foxp3 among peripheral naïve CD4 T cells with early withdrawal of PI3K-Akt-mTOR signaling, and reduced CD28 signaling. TCR and CD28 signaling together with IL-2 and TGF-β are likely important for the survival and proliferation of Treg cell precursors in addition to their putative direct effects on Foxp3 expression. STAT5, activated downstream of the IL-2 receptor, binds to the promoter and an intronic regulatory DNA element within the Foxp3 locus and may have a role in transcriptional regulation of Foxp3. TGF-β dependent synergistic binding of NFAT and Smad to conserved non-coding sequence 1 (Foxp3-CNS1) is important for the peripheral induction of Foxp3. Transcription of the Foxp3 gene is likely dependent on other regulatory DNA elements in addition to the Foxp3 promoter. Foxp3-CNS2 may function in the stable maintenance of Foxp3 expression. This function of CNS2 is dependent on demethylation of the CpG island located at this region, indicating that methyl-sensitive transcription factor binding likely mediates stable expression of Foxp3. Some in vitro generated iTreg cells and activated human CD4 T cells only transiently express Foxp3 perhaps due to a heavily methylated state at CNS2. Permissive chromatin modifications such as methylation of histone 3 at lysine residue 4 (H3K4me) and acetylation of histone 3 (H3Ac) at the promoter and CNS1 associate with active transcription of Foxp3, while CNS2 chromatin is only found in a permissive state, with demethylated CpG motifs, in mature Treg cells that stably express Foxp3. The maintenance DNA methyltransferase I (DNMT1) likely represses Foxp3 expression through the propagation of methylation marks on CpG dinucleotides at CNS2.

Receptors and extracellular signals are represented in yellow boxes and signaling intermediates and transcription factors in green. Regulatory DNA elements are highlighted in yellow. Grey boxes with green dotted outlines contain factors thought to be specifically important for peripheral Foxp3 induction.