B7/CD28 in Central Tolerance: Costimulation Promotes Maturation of Regulatory T Cell Precursors and Prevents Their Clonal Deletion

According to the “two-step model,” the intrathymic generation of CD4+ regulatory T (Treg) cells segregates into a first, T cell receptor (TCR)-driven phase and a second, cytokine-dependent phase. The initial TCR stimulus gives rise to a CD25+Foxp3− developmental intermediate. These precursors subsequently require cytokine signaling to establish the mature CD25+Foxp3+ Treg cell phenotype. In addition, costimulation via CD28/B7 (CD80/86) axis is important for the generation of a Treg cell repertoire of normal size. Recent data suggest that CD28 or B7 deficient mice lack CD25+Foxp3− Treg cell progenitors. However, these data leave open whether costimulation is also required at subsequent stages of Treg differentiation. Also, the fate of “presumptive” Treg cells carrying a permissive TCR specificity in the absence of costimulation remains to be established. Here, we have used a previously described TCR transgenic model of agonist-driven Treg differentiation in order to address these issues. Intrathymic adoptive transfer of Treg precursors indicated that costimulation is dispensable once the intermediate CD25+Foxp3− stage has been reached. Furthermore, lack of costimulation led to the physical loss of presumptive Treg cells rather than their escape from central tolerance and differentiation into the conventional CD4+ T cell lineage. Our findings suggest that CD28 signaling does not primarily operate through enhancing the TCR signal strength in order to pass the threshold intensity required to initiate Treg cell specification. Instead, costimulation seems to deliver unique and qualitatively distinct signals that coordinately foster the developmental progression of Treg precursors and prevent their negative selection.


INTRODUCTION
CD4 + regulatory T (T reg ) cells expressing the transcription factor Foxp3 exert an essential function for the maintenance of selftolerance and immune homeostasis (Sakaguchi, 2004). There is good evidence that a substantial fraction of the T reg cell repertoire originates from the thymus; for instance, there is a large degree of sequence-overlap between the T cell receptor (TCR) repertoires of thymic and peripheral Foxp3 + cells (Hsieh et al., 2006;Pacholczyk et al., 2006;Lio and Hsieh, 2011).
Entry into the T reg cell lineage during thymocyte development is believed to depend upon instructive processes ensuing from selfantigen recognition (Wirnsberger et al., 2011). Evidence for this has been obtained in TCR/neo-self-antigen double transgenic systems (Jordan et al., 2001;Apostolou et al., 2002;Kawahata et al., 2002;Aschenbrenner et al., 2007) and also stems from observations that polyclonal thymocytes bearing superantigen-reactive TCRs are substantially enriched in Foxp3 + cells (Papiernik et al., 1998;Ribot et al., 2006). The exact parameters and modalities of antigen recognition that specify whether an autoreactive MHC II-restricted thymocyte enters the T reg lineage or is subject to negative selection remain to be established; however, there is some consensus that interactions of intermediate avidity may favor T reg cell differentiation over clonal deletion Atibalentja et al., 2009;Picca et al., 2009;Hinterberger et al., 2010). Furthermore, co-signals provided by common γ-chain cytokines [interleukin (IL)-2 in particular, but also IL-7 and -15;Fontenot et al., 2005a;Mayack and Berg, 2006;Yao et al., 2007;Bayer et al., 2008;Vang et al., 2008] as well as costimulation through CD28/B7 interactions are required for efficient intrathymic differentiation of T reg cells.
Mice deficient in CD28 or its ligands CD80 and CD86 (B7.1 and B7.2, respectively) display a significant decrease in the number of thymic and peripheral T reg cells (Salomon et al., 2000;Tang et al., 2003;Lohr et al., 2004;Tai et al., 2005). Although costimulation has been implicated in IL-2 production (Lindstein et al., 1989;Fraser et al., 1991;Jenkins et al., 1991), the failure of Cd28 −/− or Cd80/Cd86 −/− mice to generate a T reg cell pool of normal size is not directly linked to cytokine deprivation. Thus, the inefficient entry of Cd28 −/− thymocytes into the T reg lineage is not "rescued" by the presence of bystander Cd28 +/+ cells in mixed bone marrow chimeras, indicating that the paucity of thymic T reg cells in costimulation deficient mice primarily reflects a T cell-intrinsic function of the CD28 signaling axis (Tai et al., 2005).

www.frontiersin.org
The "two-step model" of intrathymic T reg differentiation suggests a sub-division into an antigen-driven instruction phase and a cytokine-dependent (but largely antigen independent) consolidation phase. Accordingly, CD25 + Foxp3 − CD4 single-positive (SP) cells represent TCR-instructed, T reg lineage committed intermediates that require continual cytokine  signaling, but are largely independent of TCR stimulation, for their differentiation into "mature" CD25 + Foxp3 + T reg cells Lio and Hsieh, 2008). Recent data support the idea that CD28 costimulation and common γ-chain cytokine signaling operate at distinct stages of intrathymic T reg differentiation. Specifically, polyclonal CD25 + Foxp3 − cells, which are believed to contain T reg precursors that arise through TCR-mediated instruction ("step one") are strongly diminished in the thymus of Cd28 −/− mice (Lio et al., 2010;Vang et al., 2010).
The principle requirement for costimulation during intrathymic generation of the T reg cell pool has been well documented in polyclonal systems. However, assessing the number of Foxp3 + cells in a diverse TCR repertoire does not reveal insights into the "alternative" fate of presumptive T reg cells in the absence of costimulation. Thus, it is as yet unclear whether the respective TCR specificities are physically lost from the repertoire, i.e., negatively selected, or whether these cells instead escape from central tolerance induction and enter the pool of mainstream CD4 T cells. To address this issue, we have made use of a previously described TCR transgenic model of agonist antigen-driven T reg differentiation.

INTRATHYMIC TRANSFER
About 5 × 10 5 CD4 SP thymocytes or 4 × 10 5 cells of sorted subpopulations from TCR-HA × AIRE-HA donors (CD45.1) were injected in 3 μl PBS into one thymic lobe of CD45.2 recipients of the indicated genotype. The analysis of injected thymi was carried out by depletion of CD8 + cells, staining for the indicated surface markers and analysis of the entire thymus by flow cytometry.
Phycoerythrin-conjugated mAb to Foxp3 (FJK-16s) was from eBiosciences. The mAb to the TCR-HA was purified from hybridoma (6.5) supernatants and conjugated to phycoerythrin or Alexa Fluor 647 in our lab.

BRDU LABELING
One milligram of BrdU (Becton Dickinson) in 200 μl PBS was intraperitoneally injected into recipient mice. 24 h after injection mice were sacrificed and thymocytes were stained with the indicated surface markers. Subsequently cells were fixed, permeabilized, treated with DNase I, and stained with a mAb specific to BrdU according to the manufacturers protocol (BrdU Flow Kit, Becton Dickinson).

BONE MARROW CHIMERAS
Bone marrow was depleted of T cells with biotinylated mAbs to CD8 and CD4 followed by depletion with streptavidin MACS beads (Miltenyi Biotec) according to standard procedures. BALB/c recipient mice were irradiated with two split doses of 450 rad and were reconstituted with 8 × 10 6 bone marrow cells.

PURIFICATION OF CD4 SP CELLS OR T REG PRECURSORS
CD4 SP cells or subpopulations of CD4 SP cells (T reg precursors) were purified by CD8 depletion, staining for the indicated surface markers, and sorting with a FACSAria cell sorter (Becton Dickinson).

STATISTICAL ANALYSIS
Statistical significance was assessed by the two-tailed Student's t -test with unequal variance.

LOSS OF PRESUMPTIVE T REG CELLS IN THE ABSENCE OF COSTIMULATION
Studies in polyclonal systems have clearly indicated a substantial reduction in the thymic production of T reg cells in the absence of costimulation (Bour-Jordan et al., 2011). However, these analyses did not reveal the fate of presumptive T reg cells under these circumstances, that is, whether the respective TCR specificities are physically lost from the repertoire or instead enter the naïve pool of CD4 T cells. To address this issue, we used TCR-HA × AIRE-HA double transgenic mice. In these animals, expression and presentation of cognate antigen by medullary thymic epithelial cells (mTECs) promotes the negative selection of the majority of influenza HA specific CD4 SP thymocytes, while at the same time a distinct and traceable cohort of HA-specific CD4 SP cells differentiate into T reg cells (Aschenbrenner et al., 2007;Hinterberger et al., 2010). T cells expressing the HA-specific TCR (TCR-HA) can conveniently be traced using the anticlonotypic antibody 6.5.
thymocytes when compared to costimulation competent TCR-HA × AIRE-HA controls ( Figure 1A). These somewhat surprising initial findings indicated that lack of costimulation augmented the antigen-driven loss of HA-specific CD4 SP cells.
The CD25 − Foxp3 − surface phenotype of the majority of TCR-HA + CD4 SP cells in costimulation deficient mice might have indicated that these cells are naive cells that have not received a "T reg instructing" TCR signal of appropriate strength. Potentially, such cells might escape from central tolerance induction and seed peripheral lymphoid organs. If this were the case, one might expect to find TCR-HA + non-T reg CD4 + T cells in the periphery of costimulation deficient TCR-HA × AIRE-HA mice. However, inspection of peripheral CD4 T cell compartments revealed the complete absence of TCR-HA + cells in costimulation deficient mice ( Figure 1D). Specifically, not only was the distinct population of TCR-HA + CD25 + T reg cells that is seen in costimulation sufficient mice absent, but there was also no discernible emergence of TCR-HA + CD25 − cells in peripheral lymphoid organs ( Figure 1D).
In order to address in how far either CD80 or CD86 provided the essential signals for T reg cell differentiation, we bred the TCR-HA × AIRE-HA system onto the respective single knockout background. This revealed a degree of redundancy of the two B7 family members in that both Cd80 −/− and Cd86 −/− mice only showed a relatively mild reduction of CD25 + Foxp3 − precursors and their "mature" CD25 + Foxp3 + progeny among TCR-HA + CD4 SP thymocytes (Figures 2A,B).
In sum, these observations are consistent with a role of costimulation in the TCR-driven development of early intermediates of thymic T reg development. A similar conclusion has recently been drawn from the absence of CD25 + GITR + CD122 + cells among polyclonal CD4 SP cells of Cd28 −/− mice (Lio et al., 2010;Vang et al., 2010). Importantly, our data suggest that lack of costimulation, rather than allowing these presumptive T reg cells to escape from clonal deviation and to enter the naïve repertoire, leads to physical loss of the respective specificities. In other words, under conditions that are otherwise permissive for T reg cell differentiation (i.e., appropriate strength of TCR stimulus), lack of costimulation results in the conversion of T reg differentiation into negative selection.

THE FUNCTION OF COSTIMULATION EXTENDS BEYOND IL-2 SIGNALING AND IS CELL-INTRINSIC
CD28 costimulation has been implicated in IL-2 production (Lindstein et al., 1989;Fraser et al., 1991;Jenkins et al., 1991). Hence, its abrogation may impinge on T reg cell differentiation through lack of IL-2 mediated cell extrinsic survival and/or differentiation signals that orchestrate the cytokine-dependent"second" phase of T reg cell differentiation Lio and Hsieh, 2008;Wirnsberger et al., 2009). However, upon breeding onto an Il2 −/− background, thymi of TCR-HA × AIRE-HA mice -in contrast to what was observed in Cd28 −/− or Cd80/86 −/− mice -did not show a reduction of TCR-HA + CD4 SP cells and of mature CD25 + cells within this population (Figure 3). This is consistent with earlier observations that IL-2 acts on thymic T reg cell differentiation in an at least partly redundant manner with other common γ-chain cytokines such IL-7 or IL-15 (D'Cruz and Klein, 2005;Fontenot et al., 2005a;Vang et al., 2008) and indicates that the apparent developmental blockade and loss of TCR-HA + T reg cells in CD28 or CD80/86 deficient TCR-HA × AIRE-HA mice cannot be explained by an eventual requirement of CD28/B7 costimulation solely for IL-2 production.

Frontiers in Immunology | Immunological Tolerance
Together, these findings clearly indicated that costimulation sufficient bystander cells do not rescue the progression of CD28 deficient cells toward a mature T reg cell phenotype, for instance through provision of IL-2 or other factors in trans. Instead, there is a cell-intrinsic requirement for CD28 signaling at the earliest stages of T reg cell differentiation that is unrelated to the presumed role of IL-2 at a subsequent stage of this process.

CD28 DEFICIENT HA-SPECIFIC CD25 − FOXP3 − CELLS ARE NOT NAIVE
Our results so far revealed that in the presence of cognate antigen, HA-specific CD4 SP cells with a CD25 − Foxp3 − phenotype could be found in similar proportions irrespective of whether or not CD28/B7 costimulation was available, whereas CD25 + Foxp3 − and CD25 + Foxp3 + cells were strongly reduced in the absence of costimulation. This suggested a developmental blockade at the transition to a CD25 + Foxp3 − phenotype, i.e., at "step one" of T reg cell differentiation. Alternatively, it was possible that CD25 − Foxp3 − CD4 SP cells only in a costimulation sufficient environment represented a true T reg intermediate downstream of the initiating TCR stimulus, whereas in the absence of costimulation, CD25 − Foxp3 − CD4 SP cells may instead actually be naïve cells.
In order to distinguish these two possibilities, we performed a more detailed surface marker analysis of Cd28 +/+ and Cd28 −/− CD25 − Foxp3 − CD4 SP thymocytes in the mixed bone marrow chimeras depicted in Figure 4A and compared their phenotype to bona fide "naïve" CD25 -Foxp3 -CD4 SP thymocytes from TCR-HA single-transgenic mice ( Figure 4D). Both Cd28 +/+ and Cd28 −/− CD25 − Foxp3 − CD4 SP thymocytes displayed a similar up-regulation of the surface molecules PD-1 and GITR, whereas truly naïve CD4 SP cells were PD-1 negative and GITR low . In further support that Cd28 +/+ and Cd28 −/− CD25 − Foxp3 − CD4 SP thymocytes had received a similar TCR stimulus, expression of the TCR was similarly down-regulated on either population, presumably as a result of cognate antigen encounter ( Figure 4D).
In sum, these findings provided further evidence that in the absence of costimulation, HA-specific cells do not escape as naïve T cells. Instead, our observations support the idea that irrespective of whether or not costimulation is provided, TCR-HA + progenitors receive a TCR signal that is sufficient to mediate the acquisition of an"early"T reg progenitor phenotype. However, in the absence of CD28 signals, these cells only very inefficiently progress toward the subsequent CD25 + Foxp3 − stage and the mature CD25 + Foxp3 + T reg phenotype.

COSTIMULATION DOES NOT ACT VIA PROLIFERATIVE EXPANSION OF T REG CELL PRECURSORS
So far, we have considered that in the absence of costimulation, the earliest phase of T reg differentiation represents a developmental dead end. An alternative explanation for the paucity of CD25 + Foxp3 − cells and their CD25 + Foxp3 + progeny in CD28 or CD80/86 deficient mice would be that costimulation would orchestrate the entry of T reg cell precursors into cell cycling, thereby mediating the proliferative expansion of intermediate T reg precursors rather than their actual developmental progression. Of note, despite a certain consensus that cycling of "mature" Foxp3 + thymocytes is barely detectable, it is as yet unclear whether T reg cell differentiation involves an early expansion phase prior to Foxp3 expression. This is particularly relevant for the earliest CD25 − Foxp3 − progenitor stage, because in a polyclonal repertoire these early T reg precursors are essentially impossible to distinguish from the bulk of "naïve" non-T reg cell precursors.
In order to address this question, we performed BrdU labeling experiments. 24 h after a single injection of BrdU into Cd28 +/+ TCR-HA × AIRE-HA mice, a substantial fraction of TCR-HA + CD25 − Foxp3 − cells and to a lesser extent also of CD25 + Foxp3 − "intermediate"precursors had incorporated BrdU, whereas BrdU + cells were very rare among mature Foxp3 + cells ( Figure 5A). In the absence of costimulation (in Cd28 −/− TCR-HA × AIRE-HA mice), TCR-HA + CD25 − Foxp3 − cells incorporated similar amounts of BrdU when compared to their counterparts in Cd28 +/+ mice, indicating that entry into the cell cycle of this early T reg cell precursor-population is independent of CD28/B7mediated costimulatory signals ( Figure 5A). Somewhat surprisingly, the incorporation of BrdU by CD25 + Foxp3cells and also by "mature" CD25 + Foxp3 + thymocytes was even increased rather than diminished in the absence of CD28 co-signals (Figures 5A,B).
In order to address whether these observations similarly applied to non-transgenic polyclonal TCR specificities, we also compared the BrdU incorporation by TCR-HA − CD4 SP thymocytes of Cd28 +/+ and Cd28 −/− TCR-HA × AIRE-HA mice. These cells express endogenously rearranged TCRs, and their eventual entry into the T reg lineage reflects polyclonal T reg development. Indeed, a clear tendency toward more proliferation in the absence of costimulation was also observed for CD25 + Foxp3 − and www.frontiersin.org "mature" CD25 + Foxp3 + cells among TCR-HA − CD4 SP thymocytes, emphasizing that our observations for TCR transgenic T reg cells and their precursors faithfully recapitulated the behavior of polyclonal T cells (Figure 5B).
Taken together, our findings suggest that the early specification into the T reg cell lineage indeed coincides with entry of "pre-Foxp3" T reg precursors into cell cycling. However, our data strongly argue against a requirement for CD28/B7 costimulation for proliferative expansion of a minute "TCR-primed" precursor-population.

THE TCR-DRIVEN INSTRUCTIVE BUT NOT THE CYTOKINE-DEPENDENT CONSOLIDATION PHASE OF T REG DIFFERENTIATION REQUIRES COSTIMULATION
A precise assessment of where and when costimulation is required during intrathymic T reg cell development is difficult to achieve when studying steady state thymocyte differentiation. For instance, it is possible that the requirement for costimulation even precedes the TCR stimulus, whereby costimulation may somehow prime cells for a subsequent instructive signal. Similarly, an early bottleneck in T reg differentiation may mask a continual requirement for costimulation also at a subsequent stage of T reg differentiation.
Our observations so far did not reveal whether the costimulatory interactions that support T reg differentiation occur before the CD4 SP T cell stage, for instance concomitant to positive selection. We have shown previously that T reg differentiation in the TCR-HA × AIRE-HA thymus can be dissociated from positive selection and CD4 lineage commitment. Specifically, injection of CD4 SP cells from TCR-HA Rag2 −/− mice, i.e., truly naïve, monoclonal cells that did not contain any pre-existing Foxp3 + cells, into AIRE-HA thymi resulted in a substantial fraction of cells entering Frontiers in Immunology | Immunological Tolerance the CD25 + Foxp3 + T reg cell lineage ; see also Figure 6). These finding indicated that self-antigen-driven intrathymic T reg differentiation can be initiated in the absence of "nominal" antigen encounter prior to the CD4 SP stage.
In order to dissociate positive selection in the absence or presence of CD28/B7 costimulatory interactions from cognate antigen encounter at the CD4 SP stage in the absence or presence of costimulation, we intrathymically (i.t.) injected CD28 deficient Rag 2 −/− TCR-HA SP thymocytes into AIRE-HA recipients. In a "reciprocal" setting, we injected Rag 2 −/− TCR-HA SP cells from costimulation sufficient animals into Cd80/86 −/− recipients (Figure 6). Both sets of experiments yielded essentially identical outcomes, namely an almost complete absence of T reg differentiation, suggesting that costimulation is necessary concomitant to or immediately subsequent to the instructing TCR stimulus (Figure 6).
Our data so far revealed an essential requirement for costimulation simultaneous to or in close temporal proximity to the instructing TCR stimulus. When analyzing steady state T reg cell development in the absence of costimulation, the early developmental arrest at the CD25 − Foxp3 − stage precludes the analysis of an eventually continual requirement for CD28/B7 interactions at subsequent stages of T reg differentiation. In order to address this issue, we isolated CD25 − Foxp3 − GITR + cells (i.e., the earliest distinct subset of TCR-triggered T reg cell precursors) and cells at the subsequent CD25 + Foxp3 − intermediate stage (i.e., cells that require common γ-chain cytokines -but not TCR stimulation -to mature into CD25 + Foxp3 + cells) from costimulation sufficient TCR-HA × AIRE-HA mice and injected them into Cd80/86 -/recipient thymi ( Figure 7A). This revealed that CD25 -Foxp3 -GITR + input cells were strongly dependent upon persistent costimulation to progress toward a mature T reg phenotype, whereas CD25 + Foxp3 − cells gave rise to mature T reg cells irrespective of whether or not continual costimulation was provided in the host microenvironment (although T reg occurred perhaps slightly less efficient in Cd80/86 −/− recipients; Figure 7B). Taken together, these data support a model whereby B7/CD28 costimulation www.frontiersin.org FIGURE 6 | Costimulation is required simultaneous to or in close temporal proximity to the T reg lineage-instructing TCR stimulus at the CD4 SP stage. 5 × 10 5 CD4 SP cells from TCR-HA Foxp3 gfp transgenic animals on a Rag2 −/− background (CD45.1) were intrathymically injected into Cd80/86 +/+ (left) or Cd80/86 −/− AIRE-HA (right) recipients (CD45.2). In a "reciprocal" setting, we injected CD4 SP cells from Cd28 −/− TCR-HA Foxp3 gfp Rag2 −/− mice into the thymus of costimulation sufficient (Cd80/86 +/+ ) AIRE-HA recipients (middle). 6 days after transfer, injected cells were analyzed for CD25 and Foxp3 expression. Numbers indicate the average frequency (±SD) of cells within gates (n = 4 for all groups).
is tightly linked to the TCR-driven first phase of T reg differentiation, but is dispensable at the cytokine-dependent second phase.

DISCUSSION
Our findings suggest that the critical function of B7/CD28 costimulation is to support the development and survival of the CD25 + Foxp3 − intermediate stage of T reg differentiation. Furthermore, using adoptive transfer of T reg precursors, we could show that costimulation is largely dispensable once the CD25 + Foxp3 − intermediate stage of T reg differentiation has been reached. Hence, the B7 co-stimulus is mainly required simultaneous to or in close temporal proximity to the instructive TCR signal, i.e., at "step one" of T reg differentiation. These findings are consistent with two recent reports indicating that there is a substantial diminution of polyclonal CD25 + Foxp3 − T reg precursor cells in CD28 deficient mice (Lio et al., 2010;Vang et al., 2010). Importantly, these analyses of polyclonal T reg development did not identify the actual fate of "presumptive" T reg cells in the absence of B7/CD28 costimulation. Here, the use of a TCR transgenic model of cognate antigen-driven T reg differentiation allowed us to reveal that lack of costimulation leads to the physical loss of T reg precursors from the T cell repertoire. As a net effect, it thus appears that CD28 signaling protects T reg precursors from clonal deletion and thereby promotes the emergence of a T reg repertoire of normal size.
Our findings have obvious implications for the observation that autoimmune prone NOD mice on a CD28 or B7 deficient background develop a more severe and accelerated form of diabetes (Salomon et al., 2000). Thus, it appears that the aggressive form of diabetes in this setting is caused by a deficiency in T reg cells rather than by escape of otherwise "vetoed" T cells specificities from central tolerance. Consistent with this, adoptive transfer of polyclonal or islet antigen specific T reg cells prevented diabetes in NOD Cd28 −/− mice (Salomon et al., 2000;Tang et al., 2004).
The avidity model of T reg differentiation posits that T reg differentiation ensues from cognate antigen interactions whose strength lies in between the signaling intensity required for positive selection on the one hand and clonal deletion on the other hand Atibalentja et al., 2009;Picca et al., 2009;Simons et al., 2010). We have recently obtained further evidence for this hypothesis by attenuating antigen presentation in the TCR-HA × AIRE-HA model through "designer micro-RNA" mediated knock-down of MHC class II on mTECs. This resulted in a diminished extent of negative selection and an increased emergence of T reg cells, which is consistent with the notion that intermediate avidity-interactions favor T reg differentiation over negative selection (Hinterberger et al., 2010). Considering the predictions of the avidity hypothesis, one may have expected TCR-HA + cells to escape from negative selection and T reg induction and to eventually enter the naïve CD4 T cell pool, if B7/CD28 costimulation merely were to amplify the strength of an integrated signal downstream of the TCR and CD28. However, this is clearly not the case. Instead, lack of costimulation increases the antigen-driven net loss of TCR-HA + cells. Hence, our findings indicate that CD28 signaling does not operate primarily through amplifying the TCR signal, but through qualitatively changing the interpretation of the TCR signal and thereby initiating a distinct genetic program. Consistent with this, we found that in the presence of the AIRE-HA transgene, TCR-HA + CD25 − Foxp3 − cells displayed identical signs of early activation (up-regulation of PD-1 and GITR and downregulation of the TCR) irrespective of whether they were Cd28 +/+ or Cd28 −/− . Parallel signals emanating from CD28/B7 costimulation may then support the progression toward the cytokinedependent "step two" of T reg differentiation. It remains possible that the early events associated with entry into the T reg lineage can even be set off by a TCR signal of matching strength independent of costimulation.
Generally, CD28 co-signals are thought to stabilize mRNAs and amplify the activation of nuclear factor of activated T cells (NFAT) and nuclear factor-κB (NF-κB), thereby supporting T cell cytokine production, proliferation, survival, and differentiation (Rudd et al., 2009). Concerning a potential role of CD28 signaling in cytokine production, it is hard to see how this should Frontiers in Immunology | Immunological Tolerance (B) Four days after injection transferred cells were analyzed for CD25 and Foxp3 expression. Numbers indicate the average frequency (±SD) of cells within gates (n = 3 for transfer of CD25 − Foxp3 − GITR + cells and n = 5 for transfer of CD25 + Foxp3 − cells; P = 0.0065 for CD25 − Foxp3 − GITR + input cells and P = 0.15 for CD25 + Foxp3 − input cells). account for the block of thymic T reg development at "step one," which is believed to be TCR-driven but cytokine independent. Along these lines, we and others found that the bottleneck in T reg development caused by CD28 deficiency affects a stage of T reg differentiation considerably upstream of the perturbations that are caused by IL-2 deficiency (Bayer et al., 2005;D'Cruz and Klein, 2005;Fontenot et al., 2005a;Setoguchi et al., 2005;Vang et al., 2008). As already discussed above, it also appears highly unlikely that CD28 functions to merely amplify the TCR signal. Sequence analyzes of polyclonal T reg cells generated in the absence or presence of costimulation also argue against this scenario (Lio et al., 2010). Thus, it was found that the residual T reg cell repertoire generated in the absence of CD28 was not dramatically altered at the level of TCR specificities. Instead, the relative abundance of individual TCR specificities within the contracted T reg pool of Cd28 −/− mice resembled that of the WT T reg repertoire, at least with regard to abundant specificities (Lio et al., 2010). On this basis, it was suggested that CD28 signaling provides signals (parallel to TCR stimulation) that facilitate T reg development, but by themselves are not truly essential (Lio et al., 2010).
An alternative explanation why the polyclonal T reg compartment is reduced by about 80% in Cd28 −/− mice would be that some, but not other TCRs depend upon CD28 co-signals to segregate into the T reg compartment. However, our observations in a TCR transgenic system are more consistent with the "facilitator" scenario, as the differentiation of quasi-monoclonal TCR-HA + www.frontiersin.org T reg cells is diminished by a factor of about five-fold rather than being fully abolished (or not being affected at all).
In order to explain why the defect in CD28 or B7 deficient mice is quantitative rather than qualitative, we considered the hypothesis that costimulation might foster T reg generation through promoting the proliferative expansion of T reg precursors rather than actually instructing their differentiation per se. However, we could not find any evidence that this was the case. In fact, the proliferation of T reg precursors was even increased in the absence of costimulation, perhaps suggesting a compensatory mechanism. On the basis of this finding, the most plausible scenario is that CD28 signaling serves a dual, partly instructive (as bona fide differentiation factor) and partly permissive (as survival factor) function during T reg differentiation. Of note, neither function appears to be truly essential, so that the role of costimulation is indeed perhaps better described as that of a "catalyst." The full spectrum of molecular events downstream of CD28 signaling during T reg differentiation remains to be established. However, recent work has shed light on how costimulation may support the differentiation of T reg precursors through qualitatively modulating signaling events downstream of the TCR. CD28 communicates with several downstream signaling cascades through distinct motifs in its cytoplasmic tail that mediate interactions with Lck and the PI3K pathway, respectively. Several groups have reported that efficient T reg cell generation does not require CD28's PI3K-binding motif, whereas the Lck-interacting P 187 YAPP motif seems to be crucial for T reg differentiation (Tai et al., 2005;Lio et al., 2010;Vang et al., 2010). Mutations in the CD28 P 187 YAPP motif strongly diminish TCR/CD28 mediated NF-κB activation (Sanchez-Lockhart et al., 2008), and the ablation of genes involved in NF-κB activation (PKC-θ, CARMA-1, Bcl-10, IKK-2) impairs thymic T reg differentiation (Schmidt-Supprian et al., 2004;Barnes et al., 2009;Medoff et al., 2009). The recent identification of c-Rel as essential NF-κB family transcription factor in T reg differentiation may provide important clues as to how integrated TCR/CD28 signaling activates the transcriptional program that controls T reg differentiation (Isomura et al., 2009;Long et al., 2009;Ruan et al., 2009;Deenick et al., 2010;Visekruna et al., 2010). One aspect of c-Rel's function seems to be direct control of the Foxp3 gene through binding to a DNA motif resembling the CD28-response element in the IL-2 gene (Zheng et al., 2010). It has been suggested that through opening and remodeling of the Foxp3 locus, c-Rel activation downstream of TCR/CD28 signaling may serve a bona fide lineage instructing function (Josefowicz and Rudensky, 2009). However, considering that T reg differentiation can proceed surprisingly well in the absence of a functional Foxp3 gene (Gavin et al., 2007;Hill et al., 2007;Lin et al., 2007;Lahl et al., 2009), it appears reasonable to assume that NF-κBsignaling or other signaling pathways downstream of CD28 also initiate further -as yet unknown -instructive molecular events not related to Foxp3-induction. At the same time, it is likely that CD28 signaling in parallel elicits a transcriptional program that is of rather permissive nature. It may thereby set the stage for "step two" of intrathymic T reg differentiation, for instance by up-regulating components of the IL-2 receptor (Lio et al., 2010;Vang et al., 2010). Unraveling these functions will be challenging, since the presumed lineage instructing function of IL-2 signaling in T reg cells is, on the one hand, not absolute and, on the other hand, inextricably intertwined with its pro-survival function (Malek et al., 2002;D'Cruz and Klein, 2005;Fontenot et al., 2005a). Furthermore, it also remains to be established in how far CD28 costimulation may directly influence the survival of T reg precursors through controlling pro-survival genes such as Bclx L , akin to its function in mature, "conventional" T cells (Boise et al., 1995;Shi et al., 1995;Noel et al., 1996;Radvanyi et al., 1996). However, given the evidence that the PI3-kinase pathway is important for induction of Bcl-x L by CD28 (Burr et al., 2001;Okkenhaug et al., 2001), yet that the PI3K interacting motif in CD28 is dispensable for efficient T reg induction (Tai et al., 2005;Lio et al., 2010;Vang et al., 2010), this scenario appears less likely.

ACKNOWLEDGMENTS
This work was supported by grants from the Deutsche Forschungsgemeinschaft (KL 1228/3-1 to Ludger Klein and Maria Hinterberger; SFB 571 to Ludger Klein and Gerald Wirnsberger).