REVIEWS

Selection of regulatory Tcells in the thymus

Chyi-Song Hsieh1, Hyang-Mi Lee1 and Chan-Wang J.Lio1,2

Abstract | The generation of regulatoryT (TReg) cells in the thymus is crucial for immune homeostasis and self-tolerance. Recent discoveries have revealed the cellular and molecular mechanisms that govern the differentiation of a subset of developing thymocytes into natural TReg cells. Several models, centred on the self-reactivity of the Tcell receptor (TCR), have been proposed to explain the generation of a TReg cell population that is cognizant of self. Several molecular pathways link TCR and cytokine signalling with the expression of the TReg cellassociated transcription factor forkhead boxP3 (FOXP3). Moreover, interplay between thymocytes and thymic antigen-presenting cells is also involved in TReg cell generation.
During the last decade, it has become clear that regula toryT (TReg) cells, which constitute approximately 10% of peripheral CD4+ Tcells, are required for the maintenance of immune homeostasis1,2. Humans with a mutation in the forkhead boxP3 (FOXP3) gene which encodes a transcription factor that is required for TReg cell develop ment and function develop IPEX syndrome (immuno dysregulation, polyendocrinopathy, enteropathy, Xlinked syndrome). This is a severe multiorgan autoimmune disease that requires treatment with bone marrow trans plantation in early childhood3. Similarly, scurfy mice, which lack FOXP3 expression, develop a lethal auto immune syndrome4. TReg cells are needed throughout life; indeed, experimental depletion of FOXP3+ TReg cells in healthy adult mice results in overwhelming auto immunity and death in a matter of weeks5,6. Thus, the generation of TReg cells is important for immune tolerance and the prevention of spontaneous autoimmunity. The first clue that TReg cells are generated in the thy mus originated from the classic neonatal thymectomy experiment, in which the removal of the thymus on day3, but not day7, after birth results in the spontane ous development of a variety of autoimmune patholo gies7. This serendipitous observation became one of the principle foundations of the TReg cell field, as it demon strated that thymusderived TReg cells that migrate to the periphery after day3 are essential for selftolerance8,9. In addition to the identification of these thymusderived FOXP3+CD4+ TReg cells, which are also known as natural TReg cells, it has become clear that conventional naive FOXP3CD4+ Tcells can differentiate in the periphery to become FOXP3+ cells10 that are known as induced TReg cells. Induced TReg cells may have an important role in tolerance to foreign antigens, such as those derived from commensal bacteria in the gut 11. The mechanisms of development and the antigen specificities of natural TReg and induced TReg cells are likely to differ. We focus here on the development of thymic natural TRegcells. In this Review, we focus on selfreactivity as the most likely primary determinant of TReg cell differentiation in the thymus. In detail, we review current models on the role of the affinity of the Tcell receptor (TCR) for thymic self antigens in natural TReg cell selection, and we examine how the signalling events downstream of TCR stimulation lead to the subsequent activation of the Foxp3 locus in a multistep process. In addition, we dis cuss the role of thymic antigenpresenting cells (APCs), which generate and present the antigenic representation of self in TReg cell differentiation.

Department of Medicine, Division of Rheumatology, Washington University School of Medicine, St. Louis, Missouri 63110, USA. 2 Present address: Department of Signalling and Gene Expression, La Jolla Institute for Allergy and Immunology, La Jolla, California 92037, USA. Correspondence to C.-S.H. e-mail: chsieh@[Link] doi:10.1038/nri3155 Published online 10 February 2012
1

The role of TCR specificity Although the existence of cells with suppressor activity was proposed several decades ago12, the current notion of TReg cells was greatly facilitated by the identificationof CD25 as a reasonably sensitive and specific marker 13. Using this marker, it was shown that CD4+CD25+ TReg cells are present in the thymus of wildtype mice, but are absent from the thymus of transgenic mice that express only the DO11.10 transgenic TCR, which recognizes the foreign antigen chicken ovalbumin14. Thus, these data provided the first hint that TCR specificity is important for thymic TReg cell differentiation. A seminal study by Caton and colleagues suggested that the expression of a TCR specific for self antigens (selfreactivity) was the crucial requirement for thymic TReg cell differentiation, as TReg cells developed in the
VOLUME 12 | MARCH 2012 | 157

NATURE REVIEWS | IMMUNOLOGY 2012 Macmillan Publishers Limited. All rights reserved

REVIEWS
thymus of TCRtransgenic mice only when the cognate antigen was also expressed via a second transgene15. This has now been observed in several TCRcognate antigen doubletransgenic mouse systems16,17. Although in wildtype mice the TCR is first expressed by thymo cytes at the CD4+CD8+ doublepositive (DP) stage, TCR expression by most transgenic lines occurs in CD4CD8 doublenegative (DN) thymocytes, allowing for abnormally early TCRcognate antigen interactions, which may affect Tcell development 18. However, data from TCRtransgenic mouse studies formed the basis for the current notion that natural TReg cell development occurs when the TCR avidity for self antigens lies between the TCR avidities that drive positive selection and negative selection19. Further support for this hypothesis came from stud ies that used mice with a limited TCR diversity to permit analyses of TCR repertoires at the individual TCR level. In these mice, it was observed that the TCR repertoires of TReg cells and conventional CD4+ Tcells are differ ent, with only a small overlap2022. Moreover, Tcells that were engineered to express TCRs isolated from the TReg cell subset were more likely to proliferate in a lympho penic host 22, consistent with the hypothesis that there is a higher level of selfreactivity in TReg cellderived TCRs than there is in TCRs from other Tcells. Together with the TCRtransgenic mouse studies, these data suggested that TCR specificity has an important role in the thymic selection of TReg cells (FIG.1). Controversy regarding the role of TCR specificity. Although the studies mentioned above favoured an important role for TCR specificity in the thymic selec tion of TReg cells, others argued against this notion. For example, it was observed that thymocytes expressing either 3A9 or KRN transgenic TCRs underwent nega tive selection rather than differentiating into TReg cells after encounter with their cognate antigens, which were also expressed as transgenes23. Moreover, in AND TCR transgenic mice that coexpressed the cognate antigen (moth cytochrome c) in a doxycyclineinducible man ner, increasing the amount of antigen did not result in increased absolute numbers of TReg cells but in a higher proportion of TReg cells owing to the deletion of nonTReg cells24. Although the use of an endogenous TCR chain might alter TCR specificity and permit the selection of TReg cells in this model, recognition of cognate antigen by the AND TCR did not directly facilitate thymic TReg cell generation but increased the relative proportion of TReg cells, as TReg cells can resist TCRinduced clonal deletion. Another study suggested that TReg cell lineage deci sions may be affected by events at the CD4CD8 stage of thymic development 25. As this stage occurs before the genetic rearrangement of the TCR is completed, these findings implied that TReg cell selection may be at least partly independent of TCR specificity. Moreover, a TCR repertoire study of TReg cells suggested that the TCRs used by TReg cells and conventional Tcells are largely overlapping 26. Finally, hybridoma cells express ing either TReg cellderived or conventional CD4+ Tcell derived TCRs responded comparably to stimulation by
a Instructive model
Intermediate TCR selfreactivity

FOXP3+ TReg cell

Thymocyte
Low TCR self-reactivity

FOXP3 naive T cell

b Stochastic or selective model

TCR-independent signal at the DN stage

Survival

High TCR self-reactivity

FOXP3+ TReg cell

FOXP3+ thymocyte

Thymocyte

Death

High TCR self-reactivity

FOXP3 thymocyte
Negative selection

TCR avidity
The combined strength of interaction between the antigen receptors on a single Tcell and multiple peptide MHC complexes on the antigen-presenting cell. The avidity can be broadly described as a function of the TCR affinity and the number of peptideMHC complexes.

Figure 1 | Models for thymic TReg cell development. a | According to the instructive model, cell fate determination is based on theNature Reviews | Immunology strength of Tcell receptor (TCR) stimulation: intermediate levels of TCR stimulation induce forkhead box P3 (FOXP3) expression, whereas higher levels induce negative selection (not shown). Low-level TCR signalling allows the cells to mature and emigrate as conventional naive Tcells. b | According to the stochastic or selective model, the induction of FOXP3 expression is attributable to a TCR-independent signal, perhaps at the early stage of CD4CD8 double-negative (DN) thymic progenitors. Compared with FOXP3 thymocytes, FOXP3+ cells are relatively resistant to negative selection, which is induced by a high level of self-reactivity of the TCR. Thus, these self-reactive FOXP3+ thymocytes survive to generate the regulatory T (TReg) cell population.

Positive selection
The process by which immature CD4+CD8+ double-positive thymocytes expressing Tcell receptors that are able to recognize self-peptideMHC complexes can proceed during the Tcell maturation process into CD4+ or CD8+ single-positive thymocytes. This selection process is important for the generation of Tcells that are restricted to the hosts MHC molecules.

autologous APCs26. Thus, there was a period of time when there was no consensus regarding the importance of TCR specificity in thymic TReg cell differentiation, and multiple competing models were proposed (FIG.1). The niche hypothesis for the development of natural TReg cells. In the past few years, several studies have renewed support for the notion that TCR specificity is important for thymic TReg cell differentiation2730, primar ily owing to the generation of transgenic mouse lines that express TCRs derived from natural TReg cells. However, the first reported transgenic mouse line expressing a natural TReg cellderived TCR showed negative selection but no TReg cell generation in the thymus. As this particular TCR was derived from a TReg cell clone that was isolated from OTII TCRtransgenic mice31, it is likely that, when expressed as a transgene in a nonOTII trans genic mouse, its higher expression levels led to negative
[Link]/reviews/immunol

Negative selection
The process by which developing Tcells expressing Tcell receptors that are highly reactive to self antigens presented on thymic antigen-presenting cells are eliminated via apoptosis.

158 | MARCH 2012 | VOLUME 12 2012 Macmillan Publishers Limited. All rights reserved

REVIEWS
selection. Another two transgenic mouse lines were independently generated by two groups27,29 using TCRs derived from naturally arising TReg cells, as determined by TCR repertoire analyses of two different transgenic mouse strains with fixed TCR chains (described above). In contrast to what was expected, virtually no TReg cells were observed in the TCRtransgenic mice when they were bred to a recombination activating gene (RAG)deficient background 27,29. This outcome was so unexpected that experimental artefacts such as an inappropri ate transgene integration site or early expression of the transgenic TCR18 were initially suspected. After excluding these possibilities27, it became clear through the use of intrathymic injection and mixed bone marrow chimaeras that the rarity of thymic TReg cell development was due to the experimental artefact of a thymus in which all thymocytes expressed TCRs with the same antigen specificity. It was observed that the frequency of natural TReg cells was inversely related to the clonal frequency of the TCRtransgenic thymo cytes (FIG.2a), implying that intraclonal competition for a limited resource was affecting TReg cell development 29. Moreover, the absolute number of TReg cells reached a plateau that was much lower than the number of CD4+ singlepositive (SP) cells that could be generated by posi tive selection (FIG.2b), suggesting that the factors required for the development of natural TReg cells are much more limiting than the factors required for positive selec tion27,29. In addition, the number of TReg cells develop ing in the thymus varied considerably depending on the specificity of the TCR, an indication that TCRs facilitate thymic TReg cell development via a quantitative, rather than qualitative, mechanism. Although the limiting factors for TReg cell development in the thymus remain to be directly defined, the role of intraclonal competition in this process implies that the limiting factor may be related to the number of APCs with sufficient antigens to trigger TReg cell development. The plateau in the number of natural TReg cells that can be generated at high clonal frequencies suggests that thymic TReg cell differentiation depends on a single thymocyteAPC encounter, which is subject to intra clonal competition. By contrast, if thymic TReg cell differ entiation required multiple thymocyteAPC encounters that were subject to intraclonal competition, the num ber of thymic TReg cells generated would be predicted to continuously decrease with increasing clonal frequency (FIG.2b). As the niche for natural TReg cell development appears to be restricted, the ligands for natural TReg cell selection are likely to be rare and tissuespecific antigens, rather than ubiquitously expressed antigens. Consistent with this notion, a recently characterized TCR that drives thymic TReg cell development appears to be spe cific for a skin antigen32. Altogether, the results obtained using TCRs isolated from naturally occurring TReg cells support the notion that thymic TReg cell development requires distinct TCR specificities. Moreover, these data show that TReg cell development is typically governed by intraclonal competition for an antigenspecific niche, suggesting that the antigens themselves are likely to be rare and tissue specific.
NATURE REVIEWS | IMMUNOLOGY 2012 Macmillan Publishers Limited. All rights reserved

The buddy hypothesis. One consistent observation in the TCRcognate antigen doubletransgenic mouse models was that the proportion of TReg cells within the total CD4+ T cell population was commonly less than 50%, suggesting that the TCRmediated selec tion of TReg cells is relatively inefficient. This may not be a substantial issue for maintaining tolerance, as the numbers of natural TReg cells in these mice were sufficient to prevent the induction of autoimmunity by conventional CD4+ Tcells bearing the same TCR specificities15. This indicates the crucial role of natu ral TReg cells in immune tolerance, because if nega tive selection was the only tolerance mechanism then the escape of any autoreactive Tcells could result in autoimmune pathology. Thus, the data from the TCRcognate antigen doubletransgenic mouse mod els suggested that autoreactive conventional Tcells would have a TReg cell buddy with the same antigen specificity, which would prevent unwanted Tcell activation and autoimmunity. In contrast with the TCRcognate antigen double transgenic mouse studies15,16, TCR repertoire analyses suggested that most of the TCRs found on TReg cells efficiently facilitate thymic TReg cell differentiation, as many TCRs that were found in the TReg cell subset were not present (or were present at frequencies below the limit of detection) in the nonTReg cell subset 20,21,33. The niche phenomenon that was suggested by the analysis of natural TReg cell TCRs may resolve this discrepancy, as the high clonal frequencies in the early TCRcognate antigen doubletransgenic mouse studies would have lowered the frequency of cells undergoing TReg cell selec tion27,29 (FIG.2a). Although it is not possible to experimen tally achieve the low clonal frequencies found during normal Tcell development, extrapolation of the avail able data from the use of transgenic TReg cell TCRs sug gests that TReg cell differentiation can be very efficient 34. Thus, the data suggest that most TReg cells express TCRs with specificities that favour TReg cell development in the thymus and that do not overlap with the TCR specificities of conventional CD4+ Tcells. However, there are some TCRs that clearly fit the buddy hypothesis, as they are found in both TReg and nonTReg cell subsets. Although these TCR specificities that inefficiently drive natural TReg cell development in the thymus have not been studied, their characterization may provide important information regarding the crite ria for TCRdriven selection. It is likely that such TCRs recognize self antigens with low affinity, or are specific for self antigens that are poorly expressed or presented by thymic APCs. Because these TCRs are selfreactive but drive inefficient TReg cell selection, we speculate that conventional CD4+ Tcells expressing these TCRs may be the primary drivers of autoimmune disease when immune regulation is perturbed. Invivo quantification of TCR signals in thymic TReg cells. Another line of support for the notion that selfreactivity drives thymic TReg cell differentiation comes from the newly described Nur77GFP trans genic mouse line. In these mice, expression of green
VOLUME 12 | MARCH 2012 | 159

REVIEWS
fluorescent protein (GFP) is driven by the Nur77 pro moter, which is activated in response to TCR stimula tion30. Interestingly, the expression of GFP was found to be substantially higher in polyclonal thymic CD4 + Tcells that expressed FOXP3 than in those that did not, consistent with the hypothesis that increased TCR selfreactivity leads to TReg cell development. Moreover, intraclonal competition was visualized as an increase in the proportion of G113 Tcells that received strong TCR stimulation (as assessed by the levels of GFP expression) as the clonal frequency of Tcells expressing the TReg cellderived TCR G113 was decreased. This finding is consistent with the notion that competition for rare self antigens is the limiting event in the thymic selection of natural TRegcells. In summary, although once controversial, the pre ponderance of current data supports the hypothesis that selfreactivity is the primary determinant that directs developing thymocytes to undergo thymic TReg cell differentiation. The self antigens for many TReg cells

a TReg cell selecting eciency depends on TCR specicity

and clonal frequency

b Models of intraclonal competition for TReg cell selection

Number of FOXP3+ TReg cells (log scale)

Frequency of FOXP3+ TReg cells

High-eciency TReg cell TCR

No competition

Low-eciency TReg cell TCR CD4+ SP thymocyte clonal frequency (log scale)

Single competitive event (experimentally observed) Multiple competitive events CD4+ SP thymocyte clonal frequency (log scale)

c
Single competitive event Peptide APC MHC class II

Low clonal frequency

CD4+ SP thymocyte

High clonal frequency

FOXP3+ TReg cell

Multiple competitive events

FOXP3 naive T cell

FOXP3+ TReg cell

FOXP3 naive T cell

Low clonal frequency

CD4+ SP thymocyte

High clonal frequency

FOXP3+ TReg cell

FOXP3 FOXP3 FOXP3 naive T cell naive T cell naive T cell

Figure 2 | Niche hypothesis for TReg cell development. a | There is an inverse relationship between the clonal frequency of CD4+ single-positive (SP) thymocytes with the same Tcell receptor (TCR) specificity and the percentage of thymic regulatoryT (TReg) cells among this CD4+ SP thymocyte population. For example, TReg cells are rare (<0.1% of the Tcell population) in the thymus of transgenic recombination activating gene (RAG)-deficient mice in which all thymocytes express a natural TReg cell-derived TCR. Moreover, the decrease in the clonal frequency of TCR-transgenic thymocytes in mixed bone marrow chimaeras results in an increased TReg cell frequency. Note that the clonal frequencies in the normal CD4+ SP thymocyte population are extremely low and therefore TReg cell development is likely to be more efficient than what can be observed experimentally using TCR-transgenic cells. TCR specificity also influences the selection of TReg cells in the thymus. The expression of TCRs derived from non-TReg cells fails to induce forkhead boxP3 (FOXP3) expression. TReg cell-derived TCRs differ quantitatively in their ability to facilitate TReg cell development (depicted here as low- and high-efficiency TReg cell-derived TCRs), presumably owing to different sizes of the TReg cell-selecting niche. b | If TReg cell selection depended only on TCR specificity, the number of TReg cells would be directly correlated with the number of CD4+ SP thymocytes (dashed blue line), similar to what was observed for the number of positively selected thymocytes with a given TCR specificity (not shown). However, what was experimentally observed for thymocytes expressing TReg cell-derived TCRs is that the number of TReg cells reaches a plateau (green line), suggesting a role for intraclonal competition. c | One hypothesis is that CD4+ SP thymocytes compete for a single interaction with thymic antigen-presenting cells (APCs) to become TReg cells. Another hypothesis is that multiple TcellAPC events occurring in series are required for the induction of FOXP3 expression. However, this would be predicted to decrease the number of TReg cells generated with increasing clonal frequency, and this is not consistent with the experimental data.

160 | MARCH 2012 | VOLUME 12 2012 Macmillan Publishers Limited. All rights reserved

[Link]/reviews/immunol

REVIEWS
Medullary thymic epithelial cells
(mTECs). A specialized type of epithelial cell located in the thymic medulla that is capable of expressing and presenting tissue-specific antigens via an AIRE-dependent mechanism. mTECs have been implicated in the establishment of self-tolerance.

appear to be uncommon, rather than ubiquitous, based on the observation that thymocytes with the same TCR specificity undergo intraclonal competition for a small TReg cell developmental niche.

TCR affinity
The strength of interaction between the Tcell receptor and a single peptideMHC complex.

Strength of the TCR signal to self Assuming that TCR selfreactivity drives the thymic selection of TReg cells, the next question is what level of selfreactivity is required. In this scenario, the answer would have important implications regarding how TReg cellmediated recognition of self antigens in the periphery compares with the recognition of such anti gens by autoreactive naive Tcells that escape thymic negative selection or TReg cell differentiation. Although positive selection is driven by a low degree of TCR self reactivity, with evidence for peptide specificity 35,36, this level of selfrecognition does not appear to be sufficient to induce thymic TReg cell generation, based on TCR repertoire studies and analyses of several TCRtransgenic mouse lines14,15,27,29,37. By contrast, TCR interactions that lead to nega tive selection are likely to impose an upper limit on TReg cell development, as, for example, an excess of antigen appears to overcome the resistance to dele tion that is associated with the expression of FOXP3 (REFS15,23,24,38,39). Although TReg cell selection has been observed to be coincident with negative selection, it may also occur at a level of TCR selfreactivity below that driving negative selection, as determined in TCR transgenic mouse models15,16 or following intrathymic transfer of cells expressing TReg cellderived TCRs29. Moreover, experimental diminishment of MHC classII expression on medullary thymic epithelial cells (mTECs)

revealed a shift from negative selection to TReg cell dif ferentiation of DO11.10 TCRtransgenic thymocytes in mice expressing ovalbumin under the control of the autoimmune regulator (Aire) promoter 40. Similar con clusions were obtained following invivo and invitro stimulation of thymocytes expressing transgenic TCRs with titrated amounts of their cognate antigens 28,41. Interestingly, the dose response to a cognate antigen and thymic TReg cell development appear to also depend on the genetic background of the mice41,42. Nonetheless, data from these studies suggest that thymic TReg cell selection is driven by a level of TCR selfreactivity that is below the level that leads to negative selection and above the level that ensures positive selection. Affinity, per cell avidity and thymic APC numbers. Whereas the strength of the TCR signal required for negative selection has been quantitatively studied in CD8+ Tcells43,44, TCR signal requirements for thymic TReg cell development have not been examined as closely. A recent study investigated the relationship between TReg cell development and TCR affinity and avidity using two transgenic mouse lines that expressed two different haemagglutininspecific TCRs. These transgenic mouse lines also expressed two different haemagglutinin anti gens, which were recognized by the transgenic TCRs with either low or high affinity 38. Interestingly, the antigens that were recognized with low affinity by the transgenic TCRs could not induce thymic TReg cell differentiation, even when they were expressed at a level that was suf ficient to drive negative selection. Thus, similarly to in a study of TReg cell induction in the periphery 45, it was suggested that TCR avidity cannot always compensate for TCR affinity in generating appropriate TCR interactions for the induction of thymic TReg cell selection. Along with the notion of niche size discussed above, the number of thymic APCs that present self anti gens and express appropriate costimulators (such as CD80 and CD86) is another factor that is likely to deter mine TReg cell generation. TReg cell development may therefore depend on multiple levels of TCRantigen interaction (FIG.3). First, the strength of the interaction between a TCR and an antigenic peptideMHC classII complex is commonly referred to as the affinity, and this determines the signalling from one TCR complex. Second, the amount of cognate antigen per thymic APC, in conjunction with the affinity of this antigen for the TCR, would affect the overall strength of TCR signalling in one Tcell. This parameter is known as the avidity of TCR stimulation (per cell avidity). Third, the number of thymic APCs presenting the cognate antigen and expressing costimulatory molecules at sufficient levels would determine, together with the affinity and avidity, the integrated TCR signal over space and time, and this may be abstractly encompassed by the notion of a TReg cell developmental niche. In the abovementioned model, changing the amount of antigen in the thymus can affect per cell avidity as well as the frequency and quality of cognate antigen presenting thymic APCs, and all of these factors may influence TReg cell differentiation. For example, antigen
VOLUME 12 | MARCH 2012 | 161

APC MHC class II Self antigen TCR T cell Amount of Anity antigen per APC Per cell avidity Niche? Self-antigen MHC class II

Number of APCs with antigen

Figure 3 | Factors that determine TReg cell development based on self-reactivity. The Tcell receptor (TCR) interactions that determine regulatoryT (TReg) cell development Nature Reviews | Immunology may be affected by several factors. The first factor is the affinity of a single TCR molecule for a single self-peptideMHC classII complex presented by a thymic antigen-presenting cell (APC). The second is the avidity of a single TcellAPC interaction, which is determined by the number of peptideMHC ligands on the APC in conjunction with the TCR affinity. Third, the size of the antigen-specific TReg cell niche which is likely to be determined by the total number of APCs presenting a given self antigen, together with the affinity and avidity affects the probability of TcellAPC encounter and thus determines the number of TReg cells that can differentiate at a given time. Finally, it is possible that Tcell interactions with multiple APCs favour different cell fate decisions, such as negative selection rather than TReg cell differentiation.
NATURE REVIEWS | IMMUNOLOGY

2012 Macmillan Publishers Limited. All rights reserved

REVIEWS
presentation by APCs that lack costimulatory molecules may favour negative selection rather than TReg cell gen eration46. Although it remains to be tested, we specu late that, in addition to the signals derived from a single thymocyteAPC interaction, serial encounters of a single thymocyte with multiple thymic APCs may affect Tcell fate. It has been suggested using twophoton micro scopy that negative selection involves multiple Tcell APC encounters47. By contrast, our interpretation of the data supporting the niche hypothesis is that TReg cell differentiation is based on a single competitive antigen dependent event (FIG.2b). This notion is reminiscent of the observation that the induction of FOXP3 expression in peripheral Tcells invitro is facilitated by the with drawal of, rather than continuous, TCR signalling 48. Future studies will be required to quantify the affinity and the amount of antigen required for thymic TReg cell selec tion, and to determine whether the frequency of Tcell APC encounters can alter cell fate decisions between TReg cell generation and negative selection. The data from the Nur77GFP reporter mice sug gest that the signal strength required for TReg cell differ entiation is substantially higher than that experienced by most of the FOXP3 Tcell population30. Moreover, a tenfold reduction in the amount of the agonistic antigen ovalbumin presented by mTECs results in a shift from negative selection to TReg cell generation in DO11.10 thymocytes40. Thus, the range of TCR selfreactivity
a Gaussian distribution model
Positive selection Cell number

(which broadly encompasses affinity, avidity and antigenAPC availability) that is permissive for TReg cell development appears to be greater than that represented by a Gaussian distribution, which has been previously proposed as a mathematical model for Tcell develop ment 19. Moreover, as the frequency of TReg cells in the thymus is small (approximately 3% of CD4+ SP cells), the range of TCR selfreactivity that drives negative selection rather than TReg cell generation is extremely small on a Gaussian distribution when drawn to scale. Thus, we propose that a Poisson distribution of the strength of TCR selfreactivity may better approximate the TCR repertoire that is present after positive selection, as it would represent the small proportion of selfreactive thymocytes as an extended tail (FIG.4). Shaping of the TCR repertoire of natural TReg cells in the periphery. The end result of thymic TReg cell dif ferentiation based on TCR selfreactivity is the release of a population of Tcells that might be considered to have a memory of the selfantigen repertoire presented in the thymus. However, unlike conventional memory Tcells, which persist in the absence of antigens, the main tenance of FOXP3+ natural TReg cells that have left the thymus appears to depend on the presence of self anti gens in peripheral tissues4952. This notion is supported by a recent study demonstrating that antigenmediated activation of peripheral TReg cells leads to their prolifera tion and prolonged persistence in the tissue from which the antigen originated, even if antigen expression is no longer detectable53. Thus, TReg cells can develop features ascribed to conventional memory Tcells. Future studies are required to quantify the similarities and differences between conventional memory Tcells and memory TReg cells, and to determine the proportion of memory versus naive peripheral TReg cells. However, it appears that the recognition of self antigens in the periphery by natural TReg cells allows these cells to persist and respond dynami cally to self antigens that are released after injury, thereby preventing the induction of autoimmunity.

TReg cell selection

Negative selection

Relative avidity to self antigens

b Poisson distribution model

Cell number

Positive selection

TReg cell selection Relative avidity to self antigens

Negative selection

Figure 4 | Models of self-reactivity and TReg cell generation. a | The figure shows the classic Gaussian distribution model as it was first proposed by Maloy and Powrie19. The Nature Reviews | Immunology graph depicts the relationship between the relative avidity to self antigens (x axis) and the selected cell number (y axis). The distribution of positively selected CD4+ single-positive (SP) cells is shown in blue, with the alternative cell fates of regulatoryT (TReg) cell selection and negative selection shown in green and red, respectively. b | In a Poisson distribution model, most of the positively selected CD4+ SP cells have the same avidity to self antigens as in the Gaussian model, but the range of self-reactivity that induces TReg cell selection and negative selection is much greater. Thus, the difference between the levels of self-reactivity that induce positive selection and those that result in TReg cell selection or negative selection is much larger in the Poisson model than in the Gaussian model.

Molecular signals downstream of the TCR The role of the NFB pathway in natural T Reg cell development. TCR engagement stimulates a variety of downstream signalling molecules and transcription factors, including AKT, mammalian target of rapamy cin (mTOR), nuclear factor of activated Tcells (NFAT) and nuclear factorB (NFB). Although Ca2+ signalling seems to be involved in thymic TReg cell development, a role for NFAT which is activated downstream of Ca2+ signalling has not been clearly demonstrated54,55. Moreover, activation of the AKTmTOR pathway is inhibitory to TReg cell differentiation 48,56,57; this may occur via the inhibition of forkhead boxO (FOXO) transcription factors, which have recently been shown to be required for TReg cell development 5860. Because TCR activation can induce conflicting signals for FOXP3 induction, an interesting future question will be to determine whether AKTmTOR signalling is differ entially regulated so that TReg cell development occurs more efficiently in the thymus than in the periphery 61.
[Link]/reviews/immunol

162 | MARCH 2012 | VOLUME 12 2012 Macmillan Publishers Limited. All rights reserved

REVIEWS
APC Ca2+ Calcium channel Thymocyte PLC PKC DAG MHC class II Self antigen LAT TCR

CD80 or CD86 CD28 PI3K mTOR CARMA1 BCL-10 MALT1 TAK1 AKT FOXO1 or FOXO3

CD154

Ca2+ STIM1 STIM2

InsP3

InsP3R IKK

NFAT

ER

Ca2+

cREL ?

? NFAT

FOXO CNS1 CNS2 Foxp3 locus 1

cREL ? CNS3 1

Nucleus

Promoter Exons: 2a 2b

Figure 5 | Molecular mechanisms linking TCR specificity to FOXP3 expression. The recognition of a self-antigen MHC classII complex by the Tcell receptor (TCR) triggers several downstream signalling pathways. The primary Nature Reviews | Immunology mechanism for the induction of forkhead boxP3 (FOXP3) in developing thymic regulatoryT (TReg) cells appears to be the activation of the nuclear factor-B (NF-B) pathway, via the CARMA1BCL-10MALT1 complex. IB kinase- (IKK) is activated by this complex and by TGF-activated kinase1 (TAK1) and then phosphorylates NF-B inhibitor (IB), leading to the dissociation of IB from NF-B and the subsequent degradation of IB (not shown). Once released, the NF-B family transcription factor cREL translocates to the nucleus and binds to the conserved non-coding sequence3 (CNS3) region of the Foxp3 locus. Although less well characterized, the nuclear factor of activated Tcells (NFAT) pathway, which is activated downstream of Ca2+ signalling, may also have a positive role in the induction of FOXP3. Interestingly, the AKTmTOR (mammalian target of rapamycin) pathway may be a negative regulator of FOXP3 induction through the phosphorylation and inhibition of forkhead boxO (FOXO) transcription factors, which facilitate the expression of Foxp3. The role of AKTmTOR signalling in the thymus in inhibiting TReg cell differentiation remains to be clarified. APC, antigen-presenting cell; BCL-10, Bcell lymphoma10; CARMA1, CARD-containing MAGUK protein1; DAG, diacylglycerol; ER, endoplasmic reticulum; InsP3, inositol-1,4,5-trisphosphate; InsP3R, InsP3 receptor; LAT, linker for activation of Tcells; MALT1, mucosa-associated lymphoid tissue lymphoma translocation protein1; PI3K, phosphoinositide 3-kinase; PKC, protein kinase C; PLC, phospholipaseC; STIM, stromal interaction molecule.

Moreover, it is still unclear whether there is dynamic var iation in the level of AKTmTOR activation depending on the strength of the TCRantigen interaction or on the activation status of theAPCs. Among the multiple pathways downstream of the TCR, the NFB pathway appears to be the main one involved in thymic TReg cell differentiation37,62,63 (FIG.5). This was first suggested by analyses of mice with muta tions in genes encoding components of this signalling cascade, such as protein kinaseC (PKC), CARD containing MAGUK protein1 (CARMA1), Bcell lym phoma10 (BCL10), TGFactivated kinase1 (TAK1) and IB kinase (IKK). These mice displayed dra matic decreases in the frequency of thymic TReg cells (reviewed in REF.64). Moreover, it was observed that
NATURE REVIEWS | IMMUNOLOGY

enforced NFB activation via overexpression of a con stitutively active form of IKK was sufficient to bypass the requirement for TCRmediated recognition of self, as it resulted in the development of FOXP3+ TReg cells in the thymus of RAGdeficient TCRtransgenic mice (expressing either the MHC classIIrestricted OTII TCR or the MHC classIrestricted P14 TCR), even though neither of these mouse lines normally generates TReg cells37. Taken together, these data suggest that NFB signalling is both necessary and sufficient for thymic TReg cell differentiation. Of the NFB family of transcription factors, it appears that cREL is the most important for the induc tion of FOXP3 expression, although there is some debate as to whether it acts as a homodimer or pairs with other
VOLUME 12 | MARCH 2012 | 163

2012 Macmillan Publishers Limited. All rights reserved

REVIEWS
NFB family members37,62,63. cREL has been reported to bind to the conserved noncoding sequence3 (CNS3) region of the Foxp3 locus, which is marked by a per missive histone modification (H3K4me1) even as early as in CD4+CD8+ DP thymocytes63. Thus, it has been proposed that cREL is a pioneer transcription factor that links TCR engagement with the opening of the Foxp3locus. The role of costimulation in natural TReg cell develop ment. CD28 has an important cellintrinsic role in the generation of thymic TReg cells, as CD28deficient mice have an approximately 80% reduction in the frequency of thymic TReg cells65,66. Although it was predicted that CD28 contributes to the overall TCR signalling strength to promote selection into the TReg cell lineage, recent studies by two independent groups have suggested that this is not the case. Instead, it was observed that the TReg cell TCR repertoires of CD28sufficient and deficient mice were not dramatically different 67. Studies using TCRcognate antigen doubletransgenic mouse mod els have supported this notion46. Thus, it was proposed that the primary role of CD28 is to enhance either the efficiency of TReg cell development and/or the survival of thymocytes undergoing TReg cell differentiation46,67. Interestingly, another costimulatory molecule, CD40, which binds to CD154 expressed on activated Tcells68, may also have a role in the expansion, rather than selection, of TReg cell populations69.
TCR dependent TCR independent IL-2 and IL-15

Immature SP CD4+ T cell

FOXP3CD25+ TReg cell precursor

FOXP3+ TReg cell

TCR Self antigen MHC class II

CD28 CD80 or CD86

APC

Figure 6 | A two-step model for TReg cell development. When thymocytes recognize self-peptideMHC classII complexes in the presence of co-stimulatory molecules Nature Reviews | Immunology (such as CD80 or CD86) and with sufficiently high per cell avidity for regulatoryT (TReg) cell selection, some of these cells are selected as FOXP3CD25+ TReg cell precursors. Presumably the Tcell receptor (TCR) signal leads to the activation of several downstream pathways, including nuclear factor-B (NF-B) activation, and results in the remodelling of the forkhead boxP3 (Foxp3) locus, rendering it permissive to the induction of FOXP3 expression by interleukin-2 (IL-2) signalling. At this point, the TReg cell precursor does not require further TCR stimulation for the expression of FOXP3. Instead, cytokine signals mediated by IL-2, or to a lesser extent IL-15, facilitate the induction of FOXP3 expression. APC, antigen-presenting cell; SP, single-positive.

Stages in natural TReg cell development and the role of cytokines. The neonatal thymectomy experiments sug gested that TReg cell differentiation is delayed compared with conventional Tcell maturation8. Interestingly, it was noted that a population of FOXP3CD25+CD4+ SP cells temporally preceded the development of FOXP3+CD25+ TReg cells9. It was subsequently found that the FOXP3 CD25+CD4+ SP cell population was enriched in cells that developed into FOXP3+CD4+ Tcells after intrathymic transfer 70. Notably, these TReg cell precursors did not require further TCR stimulation to upregulate FOXP3, suggesting that TReg cell differentiation could be divided into TCRdependent and independent steps (FIG.6). It was possible that the TCRindependent step reflected merely a lag in the time between the TCR signal and FOXP3 expression. However, shortterm cul ture of the FOXP3CD25+CD4+ SP thymocytes invitro suggested that this was not the case, and that additional signals were required for the upregulation of FOXP3. As cytokines had been shown to be important for the devel opment and maintenance of the thymic TReg cell popula tion7173, and given that the TReg cell precursors are found in a population of cells that express CD25 (the chain of the highaffinity interleukin2 (IL2) receptor), it was hypothesized that cytokines such as IL2 are involved in this second step70. Indeed, invitro culture of FOXP3 CD25+ thymocytes with IL2, and to a lesser extent with IL15, was sufficient to rapidly induce FOXP3 expres sion70, potentially through the binding of signal trans ducer and activator of transcription5 (STAT5) directly to the Foxp3 locus7375. Consistent with this invitro observation, expression of a hyperactive form of STAT5 increased the invivo fre quency of TReg cells in the thymus and skewed the TReg cell TCR repertoire to include TCRs that are normally found on conventional FOXP3 Tcells76. The level of IL2 may therefore represent an antigenindependent niche involved in TReg cell generation and survival66,72, as opposed to the antigenspecific niche discussed above. As TCR stimulation upregulates CD25 expres sion through NFB activation, the expression of TCRs with high affinity to self antigens may favour the dif ferentiation of TReg cells by facilitating the generation of CD25+ TReg cell precursors. Thus, the existence of multiple steps (TCR signalling and cytokine signalling) in TReg cell development may increase the dependence of TReg cell selection on high TCR selfreactivity (FIG.6). Whereas the current evidence supports the notion that TReg cell differentiation is a multistep process in both the thymus and the periphery 48,61,77, many ques tions remain. First, although the FOXP3CD25+ thymo cyte subset is enriched in cytokineresponsive TReg cell precursors, CD25 does not represent a specific marker for these cells. FOXP3independent TReg cellassociated genes, such as the transcription factor Helios, may be better markers for TReg cell precursors7880. Second, the relative contribution of cytokines to promoting TReg cell differentiation versus survival remains to be clarified. For example, transforming growth factor (TGF) has been suggested to be directly involved in FOXP3 induc tion during thymic TReg cell development. However, its
[Link]/reviews/immunol

164 | MARCH 2012 | VOLUME 12 2012 Macmillan Publishers Limited. All rights reserved

REVIEWS
Thymus mTEC Tissuespecic antigens

AIRE

Autophagocytosis Antigen transfer Antigen transfer Tissue antigenderived peptide

MHC class II

Migratory DC (SIRP+ DC or pDC)

Thymusresident SIRP DC

Endocytosis

Peripheral antigens

Extracellular antigens

Figure 7 | The antigenic repertoire presented by thymic medullary APCs. Multiple thymic antigen-presenting cell (APC) types are capable of facilitating thymic regulatoryT Nature Reviews | Immunology (TReg) cell differentiation in the medulla. Stromal APCs, including medullary thymic epithelial cells (mTECs), can express and present tissue-specific antigens that are induced by autoimmune regulator (AIRE). Tissue-specific antigens are processed by autophagosomes and presented on the cell surface as peptideMHC classII complexes. Haematopoietic APCs include dendritic cells (DCs), macrophages and Bcells. However, the role of macrophages and Bcells is unclear (not shown). The DC subsets include plasmacytoid DCs (pDCs) and SIRP+ conventional DCs, which both migrate from the periphery and thus could potentially present extracellular antigens captured from the peripheral microenvironment. By contrast, resident SIRP conventional DCs originate in the thymus and thus probably present antigens from the thymus. In addition to presenting extracellular antigens, DCs can present mTEC-expressed antigens following antigenic transfer.

role has been controversial, as it may relate more to sur vival than to selection81,82. Finally, formal proof is still required for the notion that cytokineresponsive FOXP3 TReg cell precursors are an essential intermediate in the process of TReg cell differentiation. Future experiments are therefore needed to define TReg cell precursors with specific markers, as well as to understand the molecular status of the TCRprimed Foxp3locus.

Cortical thymic epithelial cells

(cTECs). Epithelial cells located in the thymic cortex that are able to positively select immature double-positive thymocytes.

Thymic APC subsets in TReg cell development As both TCR specificity and costimulatory molecules have an important role in TReg cell selection, the APCs encountered by developing thymocytes may shape the resulting TReg cell population (FIG.7). The thymus con tains a complex network of APCs that supports the vari ous stages of Tcell development; these APCs include cortical thymic epithelial cells (cTECs), mTECs and dendritic cells (DCs). Macrophages and Bcells are also present, but their role in thymic TReg cell development has not been established83.

Initial studies suggested that thymic TReg cell develop ment takes place in the cortex, as restricted expression of MHC classII molecules on cTECs still resulted in the gen eration of thymic TReg cells84. This was supported by studies showing that FOXP3+ DP thymocytes could be found in wildtype mice8587 and in TCRcognate antigen double transgenic mice15,88. However, recent reports have argued that the frequency of FOXP3+ DP cells within the thymic TReg cell population is below 1%, suggesting a minor role for cTECs in TReg cell generation in a normal thymus77,89. Instead, it was proposed that most thymic TReg cells differentiate during the immature HSAhiCD4+ SPstage. Among the medullary APCs, mTECs were originally thought to be crucial for TReg cell selection, as deletion of the MHC classII locus in bone marrowderived APCs (including DCs) in bone marrow chimaeras had little effect on the number of TReg cells generated86,90. However, recent studies have suggested that bone marrowderived APCs can also generate normal numbers of thymic TReg cells. Reciprocal bone marrow chimaeras had equivalent numbers of TReg cells irrespective of whether CD80 and CD86, or CD40, were expressed only on thymic epithe lial cells or only on bone marrowderived APCs69,91. Also, the knockdown of MHC classII expression on mTECs did not affect the frequency of thymic TReg cells40. In addi tion, transgenic expression of antigens on either cell type was sufficient for selecting TCRtransgenic TReg cells15,16,90. Thus, these results indicate that both bone marrow derived APCs and mTECs can facilitate thymic TReg cell differentiation, and either subset alone may be sufficient for the generation of normal TReg cell numbers invivo. These data suggest that the various APC subsets are individually dispensable for the generation of nor mal numbers of thymic TReg cells. However, it remains unknown whether each type of thymic APC is respon sible for the generation of TReg cells with unique TCR specificities, as mTECs and the various DC subsets differ in their ability to express, capture and present antigens for thymic TReg cell differentiation92. For example, it has been shown that plasmacytoid DCs (pDCs), as well as onethird of thymic conventional DCs (SIRP+ DCs), are derived from the periphery 93, potentially allowing them to present peripheral antigens in the thymus. By contrast, twothirds of thymic conventional DCs (SIRP DCs) originate in the thymus93. In addition, these DC subsets differ in their capacity to select TReg cells invitro28,61, sup porting the notion that each DC subset may select for TReg cells with differing TCR specificities, some of which may be unique to a particular DCsubset. mTECs are also likely to present different antigens from DCs owing to the stochastic expression of tissue specific antigens regulated by the transcription factor AIRE90,94,95. However, transfer of antigens from mTECs to DCs may blur the differences in the antigen reper toire9599. Also, the importance of AIREdependent anti gens in selecting the TReg cell population has recently been questioned100. Thus, future studies are required to address the hypothesis that the various thymic APCs select for TReg cell populations with unique TCR specificities to prevent holes in the TReg cell repertoire that might lead to tissuespecific autoimmunity.
VOLUME 12 | MARCH 2012 | 165

NATURE REVIEWS | IMMUNOLOGY 2012 Macmillan Publishers Limited. All rights reserved

REVIEWS
Open questions Although the past few years have seen substantial pro gress in understanding the process of thymic TReg cell selection, there are still many questions. TCR specific ity to self antigens appears to be the primary driver for TReg cell selection, but the endogenous ligands for TReg cell TCRs are unknown. What is the threshold for self reactivity that allows escape from TReg cell selection, and does this still permit peripheral immune responses to self? Are all of the different thymic APC types required to generate an effective TReg cell population? Does genetic background affect TCR signalling, or does it affect other, as yet unknown, mechanisms involved in thymic TReg cell selection? Answering these questions would provide a deeper understanding of how thymusderived TReg cells prevent spontaneous autoimmunity in the periphery. cREL appears to be an important molecular link between TCR activation and FOXP3 expression, but NFB activation occurs in many cell types, including peripheral Tcells. One important question therefore is why TCR activation in the periphery does not typically result in the induction of FOXP3. What are the molecular differences between immature CD4+ SP thymocytes and naive peripheral Tcells? Is it the differential regulation of the AKTmTOR pathway? Why is FOXP3 predominantly
expressed by CD4+, and not CD8+, SP thymocytes? What are the molecular characteristics of TReg cell precursors that render their Foxp3 locus permissive to binding by transcription factors induced by cytokine signalling? Understanding these molecular mechanisms may allow therapeutic treatment of autoimmune or inflammatory disorders using TReg cells that are generated from periph eral Tcells but that have the stability and functional abilities of thymusderived natural TReg cells, which is currently not achievable.

Conclusion The dependence of thymic TReg cell selection on self reactivity is remarkable on many levels. First, it limits the export of selfreactive conventional CD4+ Tcells to the periphery. Second, it creates a Tcell subset that, at the population level, can distinguish self. Whereas self reactive Tcells would be biased towards the natural TReg cell subset, Tcells recognizing foreign antigens would not have that bias, resulting in a form of selfnonself discrimination. Finally, the principle of clonal expansion is likely to be as crucial for TReg cell function as it is for conventional CD4+ Tcells, allowing TReg cells to dynami cally respond to changes in selfantigen presentation owing to trauma or homeostatic cell death processes.

1. 2. 3. 4. 5. 6.

7. 8.

9.

10. 11. 12. 13.

14.

15.

Wing, K. & Sakaguchi, S. Regulatory Tcells exert checks and balances on self tolerance and autoimmunity. Nature Immunol. 11, 713 (2010). Rudensky, A.Y. Regulatory Tcells and Foxp3. Immunol. Rev. 241, 260268 (2011). dHennezel, E. etal. FOXP3 forkhead domain mutation and regulatory Tcells in the IPEX syndrome. [Link]. [Link]. 361, 17101713 (2009). Ramsdell, F. Foxp3 and natural regulatory Tcells: key to a cell lineage? Immunity 19, 165168 (2003). Lahl, K. etal. Selective depletion of Foxp3+ regulatory Tcells induces a scurfy-like disease. [Link]. Med. 204, 5763 (2007). Kim, J.M., Rasmussen, J.P. & Rudensky, A.Y. Regulatory Tcells prevent catastrophic autoimmunity throughout the lifespan of mice. Nature Immunol. 8, 191197 (2007). Nishizuka, Y. & Sakakura, T. Thymus and reproduction: sex-linked dysgenesia of the gonad after neonatal thymectomy in mice. Science 166, 753755 (1969). Asano, M., Toda, M., Sakaguchi, N. & Sakaguchi, S. Autoimmune disease as a consequence of developmental abnormality of a Tcell subpopulation. [Link]. Med. 184, 387396 (1996). Fontenot, J.D., Dooley, J.L., Farr, A.G. & Rudensky, A.Y. Developmental regulation of Foxp3 expression during ontogeny. [Link]. Med. 202, 901906 (2005). Apostolou, I. & von Boehmer, H. Invivo instruction of suppressor commitment in naive Tcells. [Link]. Med. 199, 14011408 (2004). Haribhai, D. etal. A requisite role for induced regulatory Tcells in tolerance based on expanding antigen receptor diversity. Immunity 35, 109122 (2011). Gershon, R.K. & Kondo, K. Cell interactions in the induction of tolerance: the role of thymic lymphocytes. Immunology 18, 723737 (1970). Sakaguchi, S., Sakaguchi, N., Asano, M., Itoh, M. & Toda, M. Immunologic self-tolerance maintained by activated Tcells expressing IL-2 receptor -chains (CD25). [Link]. 155, 11511164 (1995). Itoh, M. etal. Thymus and autoimmunity: production of CD25+CD4+ naturally anergic and suppressive Tcells as a key function of the thymus in maintaining immunologic self-tolerance. [Link]. 162, 53175326 (1999). Jordan, M.S. etal. Thymic selection of CD4+CD25+ regulatory Tcells induced by an agonist self-peptide. Nature Immunol. 2, 301306 (2001).

16. Apostolou, I., Sarukhan, A., Klein, L. & von Boehmer, H. Origin of regulatory Tcells with known specificity for antigen. Nature Immunol. 3, 756763 (2002). 17. Knoechel, B., Lohr, J., Kahn, E., Bluestone, J.A. & Abbas, A.K. Sequential development of interleukin 2dependent effector and regulatory Tcells in response to endogenous systemic antigen. [Link]. Med. 202, 13751386 (2005). 18. Baldwin, T.A., Sandau, M.M., Jameson, S.C. & Hogquist, K.A. The timing of TCR expression critically influences Tcell development and selection. [Link]. Med. 202, 111121 (2005). 19. Maloy, K.J. & Powrie, F. Regulatory Tcells in the control of immune pathology. Nature Immunol. 2, 816822 (2001). 20. Pacholczyk, R., Ignatowicz, H., Kraj, P. & Ignatowicz, L. Origin and Tcell receptor diversity of Foxp3+CD4+ CD25+ Tcells. Immunity 25, 249259 (2006). 21. Wong, J. etal. Adaptation of TCR repertoires to self-peptides in regulatory and nonregulatory CD4+ Tcells. [Link]. 178, 70327041 (2007). 22. Hsieh, C.-S. etal. Recognition of the peripheral self by naturally arising CD25+ CD4+ Tcell receptors. Immunity 21, 267277 (2004). 23. Shih, F.F., Mandik-Nayak, L., Wipke, B.T. & Allen, P.M. Massive thymic deletion results in systemic autoimmunity through elimination of CD4+ CD25+ T regulatory cells. [Link]. Med. 199, 323335 (2004). 24. Van Santen, H.-M., Benoist, C. & Mathis, D. Number of T reg cells that differentiate does not increase upon encounter of agonist ligand on thymic epithelial cells. [Link]. Med. 200, 12211230 (2004). 25. Pennington, D.J. etal. Early events in the thymus affect the balance of effector and regulatory Tcells. Nature 444, 10731077 (2006). 26. Pacholczyk, R. etal. Nonself-antigens are the cognate specificities of Foxp3+ regulatory Tcells. Immunity 27, 493504 (2007). 27. Leung, M.W., Shen, S. & Lafaille, J.J. TCR-dependent differentiation of thymic Foxp3+ cells is limited to small clonal sizes. [Link]. Med. 206, 21212130 (2009). 28. Atibalentja, D.F., Byersdorfer, C.A. & Unanue, E.R. Thymusblood protein interactions are highly effective in negative selection and regulatory Tcell induction. [Link]. 183, 79097918 (2009). 29. Bautista, J.L. etal. Intraclonal competition limits the fate determination of regulatory Tcells in the thymus. Nature Immunol. 10, 610617 (2009).

30. Moran, A.E. etal. Tcell receptor signal strength in Treg and iNKTcell development demonstrated by a novel fluorescent reporter mouse. [Link]. Med. 208, 12791289 (2011). By using Nur77GFP as a reporter for TCR stimulation, this study correlated the level of TCR signal received with the potential for TReg cell differentiation. 31. Dipaolo, R.J. & Shevach, E.M. CD4+ T-cell development in a mouse expressing a transgenic TCR derived from a Treg. Eur. [Link]. 39, 234240 (2008). 32. Killebrew, J.R. etal. A self-reactive TCR drives the development of Foxp3+ regulatory Tcells that prevent autoimmune disease. [Link]. 187, 861869 (2011). 33. Hsieh, C.S., Zheng, Y., Liang, Y., Fontenot, J.D. & Rudensky, A.Y. An intersection between the selfreactive regulatory and nonregulatory Tcell receptor repertoires. Nature Immunol. 7, 401410 (2006). 34. Lio, C.W. & Hsieh, C.S. Becoming self-aware: the thymic education of regulatory Tcells. Curr. Opin. Immunol. 23, 213219 (2011). 35. Lo, W.L. etal. An endogenous peptide positively selects and augments the activation and survival of peripheral CD4+ Tcells. Nature Immunol. 10, 11551161 (2009). 36. Ebert, P.J., Jiang, S., Xie, J., Li, Q.J. & Davis, M.M. An endogenous positively selecting peptide enhances mature Tcell responses and becomes an autoantigen in the absence of microRNA miR-181a. Nature Immunol. 10, 11621169 (2009). 37. Long, M., Park, S.G., Strickland, I., Hayden, M.S. & Ghosh, S. Nuclear factor-B modulates regulatory Tcell development by directly regulating expression of Foxp3 transcription factor. Immunity 31, 921931 (2009). 38. Cozzo Picca, C. etal. CD4+CD25+Foxp3+ regulatory Tcell formation requires more specific recognition of a self-peptide than thymocyte deletion. Proc. Natl Acad. Sci. USA 108, 1489014895 (2011). This study demonstrated that, in addition to avidity, the quality or affinity of the interaction between a TCR and a peptideMHC complex is important for thymic TReg cell selection. 39. Riley, M.P. etal. Graded deletion and virus-induced activation of autoreactive CD4+ Tcells. [Link]. 165, 48704876 (2000).

166 | MARCH 2012 | VOLUME 12 2012 Macmillan Publishers Limited. All rights reserved

[Link]/reviews/immunol

REVIEWS
40. Hinterberger, M. etal. Autonomous role of medullary thymic epithelial cells in central CD4+ Tcell tolerance. Nature Immunol. 11, 512519 (2010). By diminishing antigen presentation by mTECs using shRNA-mediated knockdown of CIITA expression, this group showed that decreasing the avidity of the mTECthymocyte interaction can shift the cell fate from deletion to TReg cell selection. 41. Feuerer, M. etal. Enhanced thymic selection of FoxP3+ regulatory Tcells in the NOD mouse model of autoimmune diabetes. Proc. Natl Acad. Sci. USA 104, 1818118186 (2007). 42. Romagnoli, P., Tellier, J. & van Meerwijk, J.P. Genetic control of thymic development of CD4+CD25+FoxP3+ regulatory T lymphocytes. Eur. [Link]. 35, 35253532 (2005). 43. Daniels, M.A. etal. Thymic selection threshold defined by compartmentalization of Ras/MAPK signalling. Nature 444, 724729 (2006). 44. Palmer, E. & Naeher, D. Affinity threshold for thymic selection through a T-cell receptorco-receptor zipper. Nature Rev. Immunol. 9, 207213 (2009). 45. Gottschalk, R.A., Corse, E. & Allison, J.P. TCR ligand density and affinity determine peripheral induction of Foxp3 invivo. [Link]. Med. 207, 17011711 (2010). 46. Hinterberger, M., Wirnsberger, G. & Klein, L. B7/CD28 in central tolerance: costimulation promotes maturation of regulatory Tcell precursors and prevents their clonal deletion. Front. Immunol. 2, 112 (2011). 47. Le Borgne, M. etal. The impact of negative selection on thymocyte migration in the medulla. Nature Immunol. 10, 823830 (2009). 48. Sauer, S. etal. Tcell receptor signaling controls Foxp3 expression via PI3K, Akt, and mTOR. Proc. Natl Acad. Sci. USA 105, 77977802 (2008). 49. Lathrop, S.K., Santacruz, N.A., Pham, D., Luo, J. & Hsieh, C.S. Antigen-specific peripheral shaping of the natural regulatory Tcell population. [Link]. Med. 205, 31053117 (2008). 50. Taguchi, O. etal. Tissue-specific suppressor Tcells involved in self-tolerance are activated extrathymically by self-antigens. Immunology 82, 365369 (1994). 51. Seddon, B. & Mason, D. Peripheral autoantigen induces regulatory Tcells that prevent autoimmunity. [Link]. Med. 189, 877882 (1999). 52. Wheeler, K.M., Samy, E.T. & Tung, K.S. Cutting edge: normal regional lymph node enrichment of antigenspecific regulatory Tcells with autoimmune diseasesuppressive capacity. [Link]. 183, 76357638 (2009). 53. Rosenblum, M.D. etal. Response to self antigen imprints regulatory memory in tissues. Nature 480, 538542 (2011). This study showed that TReg cells can differentiate into cells with memory characteristics that reside in the tissue and diminish recurrent autoimmune responses. 54. Bopp, T. etal. NFATc2 and NFATc3 transcription factors play a crucial role in suppression of CD4+ Tlymphocytes by CD4+ CD25+ regulatory Tcells. [Link]. Med. 201, 181187 (2005). 55. Oh-Hora, M. etal. Dual functions for the endoplasmic reticulum calcium sensors STIM1 and STIM2 in Tcell activation and tolerance. Nature Immunol. 9, 432443 (2008). 56. Haxhinasto, S., Mathis, D. & Benoist, C. The AKTmTOR axis regulates denovo differentiation of CD4+Foxp3+ cells. [Link]. Med. 205, 565574 (2008). 57. Delgoffe, G.M. etal. The kinase mTOR regulates the differentiation of helper Tcells through the selective activation of signaling by mTORC1 and mTORC2. Nature Immunol. 12, 295303 (2011). 58. Ouyang, W. etal. Foxo proteins cooperatively control the differentiation of Foxp3+ regulatory Tcells. Nature Immunol. 11, 618627 (2010). 59. Kerdiles, Y.M. etal. Foxo transcription factors control regulatory Tcell development and function. Immunity 33, 890904 (2010). 60. Harada, Y. etal. Transcription factors Foxo3a and Foxo1 couple the E3 ligase Cbl-b to the induction of Foxp3 expression in induced regulatory Tcells. [Link]. Med. 207, 13811391 (2010). References 5860 were three independent studies that demonstrated the importance of FOXO in the thymic development of TReg cells. 61. Wirnsberger, G., Mair, F. & Klein, L. Regulatory Tcell differentiation of thymocytes does not require a dedicated antigen-presenting cell but is under Tcellintrinsic developmental control. Proc. Natl Acad. Sci. USA 106, 1027810283 (2009). 62. Ruan, Q. etal. Development of Foxp3+ regulatory Tcells is driven by the c-Rel enhanceosome. Immunity 31, 932940 (2009). 63. Zheng, Y. etal. Role of conserved non-coding DNA elements in the Foxp3 gene in regulatory T-cell fate. Nature 463, 808812 (2010). References 37, 62 and 63 demonstrated that the NF-B factor cREL is necessary and sufficient for inducing the differentiation of thymic TReg cells. In addition, the analysis of the Foxp3 locus in this study revealed a conserved non-coding sequence (CNS3) as an important cis-element for the induction of Foxp3 expression during thymic TRegcell development. 64. Feuerer, M., Hill, J.A., Mathis, D. & Benoist, C. Foxp3+ regulatory Tcells: differentiation, specification, subphenotypes. Nature Immunol. 10, 689695 (2009). 65. Lohr, J., Knoechel, B., Kahn, E.C. & Abbas, A.K. Role of B7 in Tcell tolerance. [Link]. 173, 50285035 (2004). 66. Tai, X., Cowan, M., Feigenbaum, L. & Singer, A. CD28 costimulation of developing thymocytes induces Foxp3 expression and regulatory Tcell differentiation independently of interleukin 2. Nature Immunol. 6, 152162 (2005). 67. Lio, C.W., Dodson, L.F., Deppong, C.M., Hsieh, C.S. & Green, J.M. CD28 facilitates the generation of Foxp3 cytokine responsive regulatory Tcell precursors. [Link]. 184, 60076013 (2010). 68. Grewal, I.S. & Flavell, R.A. CD40 and CD154 in cell-mediated immunity. Annu. Rev. Immunol. 16, 111135 (1998). 69. Spence, P.J. & Green, E.A. Foxp3+ regulatory Tcells promiscuously accept thymic signals critical for their development. Proc. Natl Acad. Sci. USA 105, 973978 (2008). 70. Lio, C.W. & Hsieh, C.S. A two-step process for thymic regulatory Tcell development. Immunity 28, 100111 (2008). 71. Fontenot, J.D., Rasmussen, J.P., Gavin, M.A. & Rudensky, A.Y. A function for interleukin 2 in Foxp3-expressing regulatory Tcells. Nature Immunol. 6, 11421151 (2005). 72. Malek, T.R., Yu, A., Vincek, V., Scibelli, P. & Kong, L. CD4 regulatory Tcells prevent lethal autoimmunity in IL-2R-deficient mice. Implications for the nonredundant function of IL-2. Immunity 17, 167178 (2002). 73. Burchill, M.A., Yang, J., Vogtenhuber, C., Blazar, B.R. & Farrar, M.A. IL-2 receptor -dependent STAT5 activation is required for the development of Foxp3+ regulatory Tcells. [Link]. 178, 280290 (2007). Together with reference 70, this study demonstrates that the development of thymic TReg cells can be divided into at least two steps based on the dependence on TCR stimulation: a proper TCR signal probably instructs thymocytes to develop into FOXP3 TReg cell precursors, which then express FOXP3 after acquiring an IL-2 signal without a continuous TCR signal. 74. Zorn, E. etal. IL-2 regulates FOXP3 expression in human CD4+CD25+ regulatory Tcells through a STATdependent mechanism and induces the expansion of these cells invivo. Blood 108, 15711579 (2006). 75. Yao, Z. etal. Nonredundant roles for Stat5a/b in directly regulating Foxp3. Blood 109, 43684375 (2007). 76. Burchill, M.A. etal. Linked Tcell receptor and cytokine signaling govern the development of the regulatory Tcell repertoire. Immunity 28, 112121 (2008). 77. Schallenberg, S., Tsai, P.Y., Riewaldt, J. & Kretschmer, K. Identification of an immediate Foxp3 precursor to Foxp3+ regulatory Tcells in peripheral lymphoid organs of nonmanipulated mice. [Link]. Med. 207, 13931407 (2010). 78. Gavin, M.A. etal. Foxp3-dependent programme of regulatory T-cell differentiation. Nature 445, 771775 (2007). 79. Hill, J.A. etal. Foxp3 transcription-factor-dependent and -independent regulation of the regulatory Tcell transcriptional signature. Immunity 27, 786800 (2007). 80. Thornton, A.M. etal. Expression of Helios, an Ikaros transcription factor family member, differentiates thymic-derived from peripherally induced Foxp3+ Tregulatory cells. [Link]. 184, 34333441 (2010). 81. Liu, Y. etal. A critical function for TGF- signaling in the development of natural CD4+CD25+Foxp3+ regulatory Tcells. Nature Immunol. 9, 632640 (2008). 82. Ouyang, W., Beckett, O., Ma, Q. & Li, M.O. Transforming growth factor- signaling curbs thymic negative selection promoting regulatory Tcell development. Immunity 32, 642653 (2010). 83. Wirnsberger, G., Hinterberger, M. & Klein, L. Regulatory T-cell differentiation versus clonal deletion of autoreactive thymocytes. Immunol. Cell Biol. 89, 4553 (2011). 84. Bensinger, S.J., Bandeira, A., Jordan, M.S., Caton, A.J. & Laufer, T.M. Major histocompatibility complex class II-positive cortical epithelium mediates the selection of CD4+ CD25+ immunoregulatory Tcells. [Link]. Med. 194, 427438 (2001). 85. Fontenot, J.D. etal. Regulatory Tcell lineage specification by the forkhead transcription factor Foxp3. Immunity 22, 329341 (2005). 86. Liston, A. etal. Differentiation of regulatory Foxp3+ Tcells in the thymic cortex. Proc. Natl Acad. Sci. USA 105, 1190311908 (2008). 87. Wan, Y.Y. & Flavell, R.A. Identifying Foxp3-expressing suppressor Tcells with a bicistronic reporter. Proc. Natl Acad. Sci. USA 102, 51265131 (2005). 88. Cabarrocas, J. etal. Foxp3+ CD25+ regulatory Tcells specific for a neo-self-antigen develop at the doublepositive thymic stage. Proc. Natl Acad. Sci. USA 103, 84538458 (2006). 89. Lee, H.M. & Hsieh, C.S. Rare development of Foxp3+ thymocytes in the CD4+CD8+ subset. [Link]. 183, 22612266 (2009). 90. Aschenbrenner, K. etal. Selection of Foxp3+ regulatory Tcells specific for self antigen expressed and presented by Aire+ medullary thymic epithelial cells. Nature Immunol. 8, 351358 (2007). 91. Proietto, A.I. etal. Dendritic cells in the thymus contribute to T-regulatory cell induction. Proc. Natl Acad. Sci. USA 105, 1986919874 (2008). 92. Klein, L., Hinterberger, M., Wirnsberger, G. & Kyewski, B. Antigen presentation in the thymus for positive selection and central tolerance induction. Nature Rev. Immunol. 9, 833844 (2009). 93. Li, J., Park, J., Foss, D. & Goldschneider, I. Thymus-homing peripheral dendritic cells constitute two of the three major subsets of dendritic cells in the steady-state thymus. [Link]. Med. 206, 607622 (2009). 94. Mathis, D. & Benoist, C. Aire. Annu. Rev. Immunol. 27, 287312 (2009). 95. Nedjic, J., Aichinger, M., Emmerich, J., Mizushima, N. & Klein, L. Autophagy in thymic epithelium shapes the T-cell repertoire and is essential for tolerance. Nature 455, 396400 (2008). 96. Koble, C. & Kyewski, B. The thymic medulla: a unique microenvironment for intercellular self-antigen transfer. [Link]. Med. 206, 15051513 (2009). 97. Klein, L., Hinterberger, M., von Rohrscheidt, J. & Aichinger, M. Autonomous versus dendritic celldependent contributions of medullary thymic epithelial cells to central tolerance. Trends Immunol. 32, 188193 (2011). 98. Hubert, F.X. etal. Aire regulates the transfer of antigen from mTECs to dendritic cells for induction of thymic tolerance. Blood 118, 24622472 (2011). 99. Gallegos, A.M. & Bevan, M.J. Central tolerance to tissue-specific antigens mediated by direct and indirect antigen presentation. [Link]. Med. 200, 10391049 (2004). 100. Daniely, D., Kern, J., Cebula, A. & Ignatowicz, L. Diversity of TCRs on natural Foxp3+ Tcells in mice lacking Aire expression. [Link]. 184, 68656873 (2010).

Competing interests statement

The authors declare no competing financial interests.

NATURE REVIEWS | IMMUNOLOGY 2012 Macmillan Publishers Limited. All rights reserved

VOLUME 12 | MARCH 2012 | 167