IFNγ, interferon gamma; IL, interleukin; Tfh, follicular helper T cell; Treg, regulatory T cell.
The potential for self-reactive T cells to cause autoimmune disease is held in check by Foxp3+ regulatory T cells (Tregs), essential mediators of peripheral immunological tolerance. Tregs have the capacity to suppress multiple branches of the immune system, tightly controlling the different subsets of effector T cells across multiple different tissue environments. Recent genetic experiments have found mutations that disrupt specific Treg: effector T cell relationships, leading to the possibility that subsets of Tregs are required to suppress each subset of effector T cells. Here we review the environmental factors and mechanisms that allow Tregs to suppress specific subsets of effector T cells, and find that a parsimonious explanation of the genetic data can be made without invoking Treg subsets. Instead, Tregs show a functional and chemotactic plasticity based on microenvironmental influences that allows the common pool of cells to suppress multiple distinct immune responses.
The potential for autoimmune disease is a consequence of the random process of gene rearrangement in the T cell receptor (TCR) and B cell receptor (BCR/antibody) antigen receptors. The evolution of a powerful adaptive immune system, capable of effectively clearing infections, has therefore necessitated the coevolution of immune tolerance processes to prevent the activation of those T cells and B cells which bear autoreactive antigen receptors. One of the key immune tolerance mechanisms is the Foxp3 transcriptional switch, which converts potentially autoreactive T cells into regulatory T cells (Tregs) that have the capacity to suppress remaining autoreactive T and B cells 1. In the absence of Foxp3 or the dependent Treg population, severe, spontaneous and fatal autoimmunity results due to uncontrolled activation of helper T cells, in both mice and humans.
Helper T cells are capable of differentiating into specialised subsets, each with unique effector molecules, migration profile and ability to recruit other leukocytes into an immune response. Type 1 helper T cells, ‘Th1 cells’, characteristically produce IFNγ and recruit CD8 T cells into anti-viral responses. Type 2 helper T cells, or ‘Th2 cells’, produce IL-4 and drive systemic immunity. IL-17-producing T cells, or ‘Th17’, have potent inflammatory capacity through their production of IL-17. Finally, follicular helper T cells, or ‘Tfh cells’, are able to migrate into the germinal centres, where they drive high affinity long-lived antibody responses. In addition to their functions in immunity, each lineage is associated with the potential for particular immunological disorders, with type 1 diabetes, asthma, psoriasis and lupus commonly described as Th1-, Th2-, Th17- and Tfh-associated disorders, respectively. Initial work on the mechanisms of Tregs demonstrated the broad capacity of activated Foxp3+ Tregs to suppress all the key effector T cell subsets. This suppression is evident both in vitro, where the addition of Tregs reduces effector T cell activation, and in vivo, where depletion of Tregs increases activation 2. Over the last few years, however, it has become clear that the suppressive capacity of Treg has a more complex interplay with effector T cell subsets. Genetic experiments have revealed an asymmetric regulation of tolerance when Treg are made qualitatively different through gene excision. Specifically, loss of Bcl6, T-bet, IRF4 and Stat3 results in the reduction of subset-specific regulation of Tfh, Th1, Th2 and Th17 cells, respectively 3–7. Furthermore, asymmetric regulation can also be induced through purely quantitative changes in Treg number, with differential sensitivity of subset-specific regulation to Treg quantitative change 8.
The presence of subset-specific pathways of suppression has led to the multiple subset hypothesis, where Tregs exist as differentiated functional subsets, each with a single corresponding effector T cell target (Fig. 1A). The observation that key genes required for effector T cell differentiation are the same genes required for Treg suppression of these effector T cells initially supported this hypothesis, as loss of Bcl6, T-bet, IRF4 and Stat3 in effector T cells, respectively, obliterate the Tfh, Th1, Th2 and Th17 responses 9. Thus T-bet would be the master transcription factor for both Th1 effector T cells and the corresponding subset of Tregs capable of suppressing Th1 cells, and so on. Under this hypothesis, the combination of Foxp3 and an effector T cell ‘cofactor’ would differentiate Tregs and turn on molecular mediators capable of suppressing only the effector T cell branch with which the ‘cofactor’ is associated.
Here we examine recent data from us and others which suggest an alternative explanation for the genetic experiments, namely that Tregs are multipotent suppressors that have the plasticity to suppress distinct subsets (Fig. 1B). In this model each Treg has the capacity to suppress each effector T cell lineage, even though distinct molecular pathways may be involved. Thus, rather than Bcl6, T-bet, IRF4 and Stat3 expression identifying stable Treg lineages, all Tregs will express, or have the capacity to express, each of these genes, which in turn potentiate lineage-specific suppression. Two main mechanisms of plasticity explain the ability of discrete molecular pathways within all Tregs to impact only on discrete effector T cell lineages. Firstly, these molecular pathways can impart on Tregs a degree of chemotactic plasticity, whereby the potential to express Bcl6 or T-bet allows the Treg to respond to microenvironmental influences and exploit conserved molecular pathways to switch on the expression of appropriate chemokine receptors. These Tregs are then better able to colocalise with particular effector T cells, at which point the generic suppressive mechanisms function. Secondly, effector T cells vary in their capacity to be suppressed by generic Treg suppressive molecules, downstream of IRF4 and Stat3. Thus loss of a molecule such as CTLA-4, expressed by all Tregs in an IRF-4 dependent manner, would have differential effects on discrete effector T cell subsets, without invoking Treg subsets. The evidence for these two models is discussed in detail below.
Tfh cells are a specialised subset of effector T cells capable of entering the germinal centres of B cell follicles during an immune response. Differentiation of Tfh cells from naïve T cells depends on SAP-mediated interactions with activated B cells and signalling through CD28 and ICOS. The stimulated T cell then turns on the master transcription factor Bcl6 and differentiates into a Tfh cell. Loss in any of these factors results in a defect in Tfh cell generation 10. The resulting Tfh cell has a unique transcriptional profile, including a downregulation of CCR7 and the expression of CXCR5, a change in chemokine receptor expression which allows the cells to home into the germinal centre 11–13. Once in the germinal centre Tfh cells are able to stimulate B cell differentiation into effector plasma cells, promoting a high affinity long-lasting antibody response.
Tregs have previously been shown to migrate into germinal centres and suppress the germinal centre response 14, 15. Recently it was demonstrated by us and others that these properties are limited to a fraction of Foxp3+ Tregs that express Bcl6 4, 6, 16. These Tregs derive from Bcl6− Tregs (rather than Bcl6+ effector T cells), and require the same follicular differentiation pathway as Tfh cells – CD28 and ICOS costimulation, SAP-dependent interaction with B cells and expression of Bcl6 4, 6, 16. In vivo, Bcl6-deficient or SAP-deficient Tregs are able to control the non-Tfh immune responses, but are unable to regulate the scale of the germinal centre response, including the number of Tfh cells and antibody production 4, 6, 16.
Critically, Bcl6+ Tregs are not a stable subset, as they do not exist in the absence of a germinal centre reaction, only arising from Tregs with the advent of a germinal centre response and then rapidly disappearing when the immune response clears 6. Furthermore, there are several lines of evidence to suggest that the ability of Bcl6+ Tregs to suppress Tfh responses is solely due to the Bcl6-dependent expression of CXCR5. Firstly, Bcl6+ Tregs, like Bcl6+ Tfh cells, express high levels of CXCR5 4, 6, 16. Secondly, CXCR5-deficient Tregs are unable to enter germinal centres and show an inability to suppress the Tfh population in vivo 4. Thirdly, when Bcl6+ Tregs are isolated and tested in vitro they display generic T cell suppression properties, with no difference in the capacity to suppress Tfh cells or undifferentiated T cells 6. Finally, Bcl6+ Tregs have a transcriptional profile very similar to Bcl6− Tregs, with no obvious changes in expression that could account for the Tfh-specific suppression apart from that of chemotactic localisation 6. As such, rather than consider Bcl6+ Tregs a subset of Tregs, it may be more useful to consider all Tregs to have a degree of chemotactic plasticity (Fig. 2). When Tregs encounter the follicular differentiation stimuli that are the hallmark of a Tfh reaction they respond by turning on Bcl6 expression, like an effector T cell. However, unlike an effector T cell, the effect of Bcl6 expression is limited to altering the chemotactic profile of the regulatory T cell. This allows a generic Treg to transiently colocalise with Tfh cells, at which point the generic suppressive mechanisms are able to act on the Tfh response.
Under an IFNγ/STAT1-dependant signal naïve CD4+ T cells can be committed to Th1 differentiation, through the activation of the master regulator T-bet (T-box expressed in T cells)17. In turn T-bet enhances IFNγ secretion and increases expression of the chemokine receptor CXCR3, allowing cells to enter inflammatory sites and amplify the cell-mediated immune response 18. The finding that Tregs also need to express the Th1 canonical transcriptional factor T-bet in order to efficiently suppress Th1 responses was the first evidence that Tregs had differential effects over T cell subsets 5.
Only a fraction of Tregs express T-bet at any given time, although this fraction increases in size during an inflammatory reaction. T-bet+ Tregs express CXCR3 and are thus able to migrate into inflammatory sites with effector Th1 cells. Moreover T-bet+ Tregs exert suppressive activity over Th1 effector T cells at these sites. This has been well described in experiments performed in the Foxp3-deficient (Scurfy) mouse transfer model. Scurfy mice do not have any Tregs and consequently develop high numbers of auto-reactive T cells, leading to the death of the animals from massive inflammation a few weeks after birth. While wild-type Tregs are able to rescue the inflammatory phenotype of Scurfy mice by regulating Th1, Th2 and Th17 effector cells, T-bet-deficient Tregs failed to regulate Th1 responses even though they could regulate Th2 and Th17 responses 5.
Just as the ability of Bcl6+ Tregs to suppress Tfh cells is dependent on expression of CXCR5, the ability of T-bet+ Tregs to suppress Th1 cells is dependent on expression of CXCR3. A similar phenotype to T-bet-deficient Tregs is observed with CXCR3-deficient Tregs 19, 20. Thus, the inability of T-bet-deficient Tregs to control Th1 mediated inflammation is probably due to insufficient amounts of Treg accumulation at the site of Th1 responses rather than an intrinsic impairment of Treg mediated suppression. For example, in ConA-induced Th1-driven hepatitis, CXCR3-deficient Treg failed to migrate towards CXCR3 ligands CXCL9, CXCL10 and CXCL11, which are highly expressed in the inflammatory liver, resulting in massive liver damage by unregulated Th1 effectors 19. In a reciprocal study, forced expression of CXCR3 on Tregs increased their ability to decrease hepatic and gut acute graft-versus-host disease (GVHD) in a murine model of experimental GVHD 21. Importantly, CXCR3-deficient Tregs are efficient at suppressing Th1 cells in vitro 19, again supporting the contention that the primary role of T-bet expression in Tregs is to modify homing and allow them to colocalise with Th1 cells (Fig. 2).
In the case of T-bet+ Tregs, it is not known whether the expression of T-bet is stable. However the parsimonious explanation is that like Bcl6 expression in response to follicular immunity, T-bet is temporarily expressed in generic Tregs following exposure to Th1 immunity. For T-bet+ Tregs, we have some data on how T-bet expression in Tregs drives CXCR3 expression, while induction of IFNγ is kept low: unlike effector T cells where T-bet expression leads to high IFNγ production and Th1 differentiation. This resistance is mediated by expression of SOCS1 and the microRNA miR-146a, which decreases Stat1 signalling and expression, respectively 22, 23. As Stat1 signalling is responsible for turning on T-bet expression in response to IFNγ, it is likely that attenuated Stat1 allows Tregs to turn on lower levels of T-bet and gain the chemotactic capacity of Th1 cells, without actually differentiating into the Th1 lineage. Thus, loss of miR-146a or SOCS1 in Treg results in excessive T-bet expression and differentiation of Tregs into IFNγ-producing Th1 cells 22, 23. This Treg dedifferentiation has also been demonstrated during heavy Th1 responses against Toxoplasma gondii infection, where Treg produce IFN-γ and contribute to the Th1 response 24 and in human Treg co-cultured with high numbers of allogeneic B cells in vitro, where Treg promote rather than inhibit effector T cell responses 25. An additional, or related, stability mechanism appears to be the induction of GATA3 following Treg activation, which restrains Tregs in inflamed tissues from expressing effector cytokines 26. Whether this is a stability mechanism that acts to restrain Tregs from entering all potential effector T cell lineages, or just the Th1 and Th17 lineages, is yet to be determined.
Th2 cells are key coordinators of the humoral immune response. The master transcription factor in Th2 differentiation is GATA-3, equivalent to Bcl6 in Tfh differentiation and T-bet in Th1 differentiation. An additional transcription factor, IRF4, is required for late stage Th2 differentiation 27, 28, although it is not a uniquely Th2-associated transcription factor, having additional functions in Th1 and Th17 differentiation 27, 29.
Using the Cre-Lox system, it was demonstrated that mice with IRF4-deficient Tregs developed autoimmune disease due to dysregulated Th2 responses, plasma cell tissue infiltration and IL-4-dependent IgG1 and IgE production 7. IRF4-deficient Tregs fail to regulate CCR8, an important chemokine receptor involved in effector T cell and Treg migration towards the lungs and skin during a Th2 response 7, 30. It is therefore possible that IRF4 in Tregs acts in a similar manner to Bcl6 and T-bet, by transiently modifying Treg migration in response to microenvironmental signals.
Despite the potential for such pleasing symmetry, the function of IRF4 in Tregs appears to be much more complex than simply altering chemotaxis. IRF4-deficient Tregs have major defects in activation and subsequent production of a number of different regulatory molecules, including IL-10, granzyme B and CTLA-4 7, 31. As befits a general activation defect, IRF-4 deficient Tregs are also poor at controlling Th1 responses during induced inflammation 31. This result begs the question as to why deletion of a gene required for general Treg activation results in Th2-biased autoimmunity. The answer is likely to lie in the differential biology of effector T cells, rather than within the Treg.
Recently, we demonstrated that Th1 and Th2 cells differentially respond to quantitative changes in Treg number 8 and function 8. Specifically, reductions in Treg number resulted in a linear increase in Th1 number, while the Th2 response remained quenched even in the presence of very small numbers of Tregs 8 (Fig. 3A). By contrast, reductions in Treg function allowed Th2 cells to expand while Th1 cells were kept under control 8. These findings were unified by the discovery of a unique effect of CTLA-4 suppression on the Th2 population. CTLA-4 is constitutively expressed by Tregs and can compete with CD28 for CD80/CD86 signals on antigen presenting cells, impeding the development and maintenance of naïve T cells 32. While all T cells require CD28, Th2 cells have a particular dependence on CD28 signalling, due to a disproportionate reliance on IL-2 for survival 33, 34. Thus, while Tregs suppress the proliferation of different effector T cell subsets equally, they have an additional, and specific, suppressive effect on Th2 cells via CTLA-4-dependent apoptosis 8 and suppression of GATA-3 35. Interestingly, we find that CTLA-4 expression is up-regulated following a partial loss of Treg number in vivo (Fig. 3B), which explains the complex relationship between Th2 suppression and Treg number and function: while decreases in CTLA-4 expression on Tregs leads to loss of Th2 suppression, decreases in Treg number can be compensated for by increased CTLA-4 on the remaining Tregs (Fig. 3C). It is therefore likely that IRF4 functions in Tregs in a very different way to Bcl6 or T-bet. Rather than change Treg chemotaxis, IRF4 is instead critical for Treg activation and CTLA-4 expression. This defect in Treg function in turn creates a disproportionate impact on Th2 suppression, due to the unique sensitivity of this subset to CTLA-4-mediated apoptosis (Fig. 3D).
Effector Th17 cells, producers of the cytokine IL-17 and other inflammatory mediators, are a subset of T cells specialised to combat extracellular bacteria and fungi 36. The development of Th17 cells requires TGF-β and the activation of Stat3 in response to IL-6, which in turn activates the master transcription factor RORγt. Recently, activation of Stat3 in Foxp3+ Tregs was found to be required for suppression of pathogenic Th17 responses 3. Foxp3CreStat3fl/fl mice, with a Treg-specific deletion of Stat3, showed a dysregulation of Th17 responses accompanied by early colitis and massive lymphoid and neutrophilic infiltration. By contrast, Th1 and Th2 responses remained well controlled 3.
There are plausible pathways through which Stat3 could function in both the chemotactic model of T-bet and Bcl6 and the disproportionate effector response model of IRF4. The chemotactic model is the simplest, as Stat3 activation is known to cause upregulation of the chemokine receptor CCR6. In Th17 cells, CCR6 allows preferential migration to microenvironments such as intestine, lung and liver 37. Moreover, IL-17 produced by Th17 cells promotes epithelial cell expression of the CCR6 ligand CCL20 38, which can facilitate the migration of both Th17 cells and CCR6+ Tregs to sites such as the inflamed spinal cord or kidney 38, 39. However, contrary to the large proportion of CCR6+ Tregs found in colon and intestine 38, 40, Tregs have the same distribution in the intestine of CCR6−/− and wild-type mice 37, which suggests that CCR6 expression is not required for Treg trafficking to the gut and demonstrates the need for an alternative function of Stat3 in Tregs.
Recent data suggests that Stat3 in Tregs may actually be functioning in a similar manner as IRF4, i.e. Stat3 is required for the expression of a common suppressive mediator, IL-10, which Th17 cells are disproportionally sensitive to. While Stat3 mediates signalling in response to IL-6, it is also activated in response to the immunoregulatory cytokine IL-10. In contrast to initial expectations, activation of Stat3 in Tregs requires the IL-10 receptor but not the IL-6 receptor 41. As expression of the IL-10 receptor is required for efficient expression of IL-10 by Tregs, Stat3 drives an amplification loop for IL-10 expression 41. In this regard, it is notable that Treg expression of IL-10 is highest in the intestines 42 and that IL-10-deficiency in Tregs results in Th17-associated inflammation of the gut 43. Just as Th2 cells are disproportionately sensitive to CTLA-4, Th17 cells are disproportionately sensitive to IL-10, as IL-10 receptor blockade in T cells results in specific expansion of Th17 cells in vivo 44. Thus Stat3 and IL-10 receptor are required by Tregs to amplify IL-10 expression in the gut and the expansion of Th17 cells in mice with a Treg-specific deletion of Stat3 or IL-10 receptor is due to the hyper-sensitivity of Th17 cells to IL-10-mediated suppression (Fig. 2).
Foxp3+ Tregs have a well-appreciated capacity to inhibit the activation and proliferation of effector T cells. The concept of Tregs as a general brake on the immune system has shifted in recent years, with the observations that loss of Bcl6, T-bet, IRF4 and Stat3 in Tregs results in the reduction of subset-specific regulation of Tfh, Th1, Th2 and Th17 cells, respectively 3–7. These results have been interpreted as evidence for subsets of Tregs, each specialised for the suppression of a single different effector T cell subset. However we would contend that these results are not sufficient to meet the standard criteria of a true cellular subset, which should show stable existence and major changes in transcriptional profile, including the expression of a unique functional mediator. Rather, these results reflect plasticity wired into Tregs through Foxp3 expression and implemented upon appropriate stimuli. Thus a single population of Tregs can express, or have the capacity to express, Bcl6, T-bet, IRF4 and Stat3. Each of these transcription factors could enable the suppression of a specific effector T cell type without invoking the need for each to be restricted to a stable subset of Tregs.
Here we have described plausible mechanisms by which loss of these four transcription factors results in loss of immune regulation over a specific effector cell type, simply by causing relatively small changes in Treg plasticity. In the case of Bcl6 and T-bet, expression of these factors is stimulated in Tregs following exposure to follicular and cell-mediated immune signals, respectively. Stability mechanisms allow Tregs to turn on these transcription factors without entering the Tfh and Th1 lineages, instead largely limiting transcriptional change to the chemokine receptors CXCR5 and CXCR3. This expression allows a generic Treg to transiently colocalise with Tfh and Th1 cells, respectively, at which point generic regulatory mechanisms suppress the response. By contrast, IRF4 and Stat3 are expressed by all Tregs, and serve to amplify the expression of CTLA-4 upon activation and IL-10 upon gut localisation. The activation of Th2 and Th17 cells in response to loss of these genes in Tregs is therefore a reflection of enhanced sensitivity of Th2 cells to CTLA-4 8 and Th17 cells to IL-10 44. Together these results present a previously unappreciated complexity in the relationship between Tregs and effector T cell subsets.