CD28-CD80/86 and CD40-CD40L Interactions Promote Thymic Tolerance by Regulating Medullary Epithelial Cell and Thymocyte Development

Abstract

Development and central tolerance of T lymphocytes in the thymus requires both TCR signals and collaboration with signals generated through costimulatory molecule interactions. In this review, we discuss the importance of CD28-CD80/86 and CD40-CD40L costimulatory interactions in promoting normal thymic development. This discussion includes roles in the generation of a normal thymic medulla, in the development of specific T-cells subsets, including iNKT and T regulatory cells, and in the generation of a tolerant mature T-cell repertoire. We discuss recent contributions to the understanding of CD28-CD80/86 and CD40-CD40L costimulatory interactions in thymic development, and we highlight the ways in which the many important roles mediated by these interactions collaborate to promote normal thymic development.

I. INTRODUCTION

T-cell activation initiated by engagement of the T-cell receptor (TCR) is a highly regulated process. Much of that regulation is achieved by the participation of a number of different response-modifying molecules that serve either to increase or decrease T-cell activation following antigen encounter. Immune response modifiers that act to potentiate immune responses comprise a class of modifier known as costimulatory molecules. Costimulatory molecules do not initiate signaling; rather costimulatory signals interact with those mediated by antigen-specific receptors to promote responsiveness to antigens. Two prominent receptor/ligand members of the costimulatory molecule family that function as critical players in the priming and functioning of peripheral T cells are CD28-CD80/86 and CD40-CD40L. The function of these two costimulatory pairs in the peripheral immune system has been the topic of many excellent reviews.^1,2 In this review, we focus on the important roles that CD28-CD80/86 and CD40-CD40L interactions assume during thymic development and the generation in the thymus of a functioning, self-tolerant, T-cell repertoire.

Both the CD28-CD80/86 and CD40-CD40L costimulatory pairs were first described in the 1980s^3,4 and fit well into the signal 1/signal 2 paradigm put forth during the same time period.^5–7 This model applied to T cells essentially stipulated that productive activation of T cells in response to antigen required not only a signal 1 delivered through the TCR but also a collaborating Signal 2. If a TCR signal (signal 1) occurred in the absence of signal 2, the T cell would fail to activate and could be rendered anergic (unresponsive to subsequent stimulation by what is normally a productive antigenic stimulation). It became apparent that T-cell–expressed CD28, when engaged by its known ligands, CD80 (B7–1) or CD86 (B7–2) expressed on antigen-presenting cells (APC), could supply signal 2.

Similar to the CD28-CD80/86 receptor ligand pair, CD40-CD40L interactions are also a means by which T cells and APCs communicate. In this case, APC-expressed CD40 interacts with CD40L (CD154) expressed on TCR-signaled T cells. Engagement of CD40 results in various outcomes depending on the particular APC. In B cells, CD40 signaling interacts with B-cell receptor (BCR) signal to promote proliferation, differentiation, and isotype switching.⁸ Engagement of dendritic cell (DC) CD40 results in DC maturation along with increased stability of antigen/MHC complexes and is also associated with increased DC survival.⁹

The roles of costimulatory molecules in lymphocyte activation are multifaceted and include lowering the threshold for antigen stimulation, tailoring the immune response to be effective to particular pathogens, and helping to ensure that immune responses are focused on foreign rather than self-antigens. With regard to this latter point, CD28 and CD40 are usually constitutively expressed on T cells and APCs, respectively, whereas CD40L on T cells and CD80/86 on APCs are upregulated following activation.^8,10 This results in a dialogue between T cell and APC that, for optimal costimulation, requires both cell types to have been stimulated: the T cell upon recognition of antigen via the TCR, and the APC through recognition of antigen by the BCR (B cells) and/or recognition of pathogens through PAMP (pathogen-associated molecular pattern) receptors (i.e., B cells and DCs). If the T-cell population is tolerant to self-antigens, then T-cell activation should only occur in response to foreign or aberrant self-antigens and not to normal self-antigens.

Development of a tolerant peripheral T-cell repertoire begins in the thymus. In addition to providing the requisite signals and growth factors for T-lineage commitment and maturation, the thymus provides an environment for evaluating the potential of a developing thymocyte for self-reactivity. Thymocytes that express a high-affinity self-reactive TCR can have several possible fates, including death by negative selection, selection into the T-regulatory (Treg) cell compartment, and anergy induction.¹¹ How thymic T-cell tolerance is achieved has been a topic of intensive investigation for some time, but evidence to support critical roles for CD28-CD80/86 and CD40-CD40L costimulatory interactions has accrued over the past two decades. It is perhaps not surprising that the costimulatory signals encountered when T cells are presented with antigen in the periphery are present in the thymus. This ensures that the nascent T-cell repertoire has been tested for self-reactivity in the presence of the same immunologic milieu that will be encountered in the periphery. In this review, we elaborate on the critical functions of both CD28-CD80/86 and CD40-CD40L costimulatory interactions in promoting a tolerogenic thymic environment and facilitating the development of a self-tolerant T-cell repertoire.

II. EXPRESSION PATTERNS OF CD28-CD80/86 AND CD40-CD40L IN THE THYMUS

Thymocytes proceed through a sequence of developmental stages in the thymus defined by the absence or presence of CD4 and CD8 coreceptors. Commitment to the T-cell lineage is complete at the CD4^negCD8^neg double-negative (DN) stage. At this stage, a small cohort of DN thymocytes rearrange and successfully express TCR γ and δ genes constituting a γδ TCR, and continue intrathymic development as γδ T cells.^12,13 Of the remaining DNs, those that successfully rearrange a TCRβ chain express a pre-TCR consisting of the rearranged TCRβ and an invariant preTα chain. Expression of a pre-TCR allows DNs to traverse the β-checkpoint and become CD4⁺CD8⁺ double-positive (DP) thymocytes (reviewed in Shah and Zuniga-Pflucker¹⁴). During the DP stage, thymocytes attempt to successfully rearrange the TCRα chain; those that do then test their nascent TCR for recognition of self-peptide/MHC complexes. DP thymocytes with TCRs that recognize self-MHC/peptide complexes with adequate affinity in the thymus are positively selected to become either CD4 or CD8 single-positive (SP) cells (reviewed in Hogquist and Jameson¹¹).

Thymocyte expression of CD28 and CD40L costimulatory molecules differs across developmental stages, with expression of CD28 preceding that of CD40L. CD28 expression levels vary with thymocyte maturation stage.¹⁵ TCR^negDN cells express low levels of CD28, which increase as cells progress to the pre-TCR⁺DN stage, and CD28 levels are highest on DP thymocytes. Direct evidence that stimulation of DNs regulates CD28 expression was provided by injecting Rag KO mice with anti-CD3; CD28 levels on DNs in anti-CD3-injected Rag KOs increased relative to untreated Rag KO mice and were comparable to the levels seen on TCR⁺DNs in WT mice.¹⁶ Thus, similar to peripheral T cells, signaling through the pre-TCR of DN thymocytes results in increased expression of cell-surface CD28.¹⁵ Therefore, it is surprising that DP thymocytes, which express either no TCRαβ or low levels of TCRαβ relative to SP cells, have the highest levels of cell-surface CD28 in the thymus. An explanation has been provided by examining CD28 levels on thymocytes in mice deficient for CD80/86.¹⁷ CD28 levels on all subsets of thymocytes in CD80/86 KO mice were elevated relative to WT mice. The most profound increase occurred on CD4 SP thymocytes, which, in the CD80/86 KO mice, displayed higher CD28 levels than DP thymocytes. This result suggests that normal expression of intrathymic CD80/86 downregulates surface expression of CD28 on developing thymocytes. Further support for the effect of CD28-CD80/86 interactions resulting in decreased surface CD28 expression levels comes from the observation that expression of either a CD80 or CD86 transgene, both of which resulted in high levels of expression on thymocytes, led to markedly decreased CD28 expression on total thymocytes.¹⁷ Thus, like peripheral T cells, where engagement of CD28 either with anti-CD28 antibody or CD80 results in diminished transcription of CD28 as well as cell surface expression,¹⁸ engagement of CD28 on thymocytes also results in reduced expression. Regulation of CD28 cell surface expression by CD80/86 engagement allows differences in thymocyte CD28 levels between WT and CD80/86 KO mice to be considered an indication of when and where developing thymocytes encounter CD80/86 ligands. Levels of CD28 on CD4 and CD8 SP thymocytes are increased 8- and 5-fold, respectively, in CD80/86 KO relative to WT mice, whereas CD28 levels on DP thymocytes are increased only 0.5-fold, suggesting that SP thymocytes are encountering more CD80/86 than DP thymocytes. SP thymocytes are localized to the thymic medulla, and immunohistochemical staining has revealed that the CD28 ligands, CD80/86, are predominantly expressed in the medulla with only isolated expression in the cortex, likely associated with rare cortical DC or macrophages.^19,20 Therefore, the concentrated expression of CD80/86 in the thymic medulla of WT mice represses the levels of CD28 on SP cells; this repression of CD28 is relieved in the CD80/86 KO mice, where SP cell CD28 expression levels are now higher than any other thymocyte subset.

In contrast to CD28, which is expressed on the majority of thymocytes, expression of CD40L is confined to CD4 SP thymocytes and is highest in the most mature CD4 SP.^21,22 Within the CD4 SP thymocyte population, CD40L is expressed in FoxP3^negCD4 SP but is not detectable in FoxP3⁺CD4 SP thymocytes.^22,23

Similar to CD28 and CD40L, expression of CD80, CD86, and CD40 has been most closely studied on bone-marrow–derived cells on which costimulatory molecule expression is generally detectable on freshly isolated B cells, conventional DCs, and macrophages.^24–28 Immunohistochemical staining of the thymus demonstrates that the thymic medulla stains intensely with anti-CD80, -CD86 and -CD40, while only isolated patches of staining are seen in the cortex.^19,29 In the thymus, most APCs (thymic dendritic cells, B cells and macrophages) and much of the epithelial-derived thymic stoma express high levels of cell surface CD40, CD80, or CD86.²⁴ Medullary thymic epithelium, but not cortical epithelium, expresses CD80, whereas most medullary and a fraction of cortical epithelial cells express CD40.^30,31 Thus, the intense thymic medullary staining observed with anti-CD80, -CD86, and -CD40 antibodies is not surprising. mTEC expresses high levels of both CD40 and CD80, and the most prevalent BM-derived APCs in the thymus, DCs and B cells, are localized in or around the medulla.^32,33

The expression patterns of CD28, CD80/86, CD40, and CD40L in the thymus suggest that interactions between these pairs of costimulatory molecules are most likely to occur in the thymic medulla. Positive selection of developing thymocytes not only results in commitment of DP thymocytes to either the CD4 or CD8 SP lineage but also establishes a switch in chemokine responsiveness such that SP thymocytes begin to express CCR7 and lose CCR9 expression. This selection promotes a loss of responsiveness to cortically expressed CCR9 ligands (CCL25) and a gain of responsiveness to medullary-expressed CCL19/21 and, hence, in migration from cortex to medulla.^34,35 In the medulla, SP thymocytes experience a much denser landscape of CD80/86 and CD40 expression than in the cortex. In addition to the continued expression of CD28 on both CD8 and CD4 SP thymocytes as they transition from DPs to SPs, CD40L, which is not expressed on less mature thymocytes, is expressed by CD4 SPs.^36,37 Thus, the importance of CD28-CD80/86 and CD40-CD40L interactions in the thymus, while not exclusively limited to the medulla, is most clearly manifest during developmental steps that occur in the medulla. Interestingly, as discussed in the section below, normal development of the thymic medulla is itself dependent on CD28-CD80/86 and CD40-CD40L interactions.

III. DEVELOPMENT OF NORMAL THYMIC MEDULLARY COMPARTMENT REQUIRES CD28-CD80/86 AND CD40/CD40L COSTIMULATORY PATHWAYS

The stromal cells of the medullary region of the thymus play a critical role in ensuring that the mature T-cell population that emerges from the thymus is functionally self-tolerant. Among the stromal cells of the thymic medulla that are known to participate in this tolerization process are thymic medullary epithelium (mTEC), thymic dendritic cells (tDCs), and thymic B cells (reviewed in Klein et al.³⁸). mTECs, due at least in part to expression of the mTEC-specific transcription-associated factor, AIRE, are unique in expressing a wide array of tissue-associated antigens (TSA). These antigens can either be directly presented in association with MHC by mTEC or can be transferred to tDC for presentation to thymocytes. Thus, mTEC and dendritic cells work in concert to tolerize the developing thymocyte population. It has been appreciated for some time that maintenance of a normal mTEC compartment requires the presence of bone-marrow–derived cells.^39,40 In embryonic and early neonatal life, lymphoid tissue inducer (LTi) cells and Vγ5⁺ dendritic epidermal T cell (DETC) progenitors are of particular importance for mTEC development.^41,42 Studies in adult mice demonstrated that TCR⁺ thymocytes are critical for normal mTEC compartment development,^21,39 and it now appears that CD4 SP cells are especially important for maintaining this population.⁴³ LTi, Vγ5⁺ DETC progenitors and CD4 SP thymocytes express a number of TNF family ligands, including RANKL, LTαβ, and CD40L (CD4 SPs only), which have been shown to be critical regulators of mTEC development, including the subpopulation of mTECs expressing AIRE (reviewed in Anderson and Takahama⁴⁴).^{21,25,43,45,46}

With the first demonstration that TCR⁺ thymocytes were required for maintenance of the thymic medulla in adult mice, the term “thymic crosstalk” was introduced to describe the interrelationship between thymocytes and the mTEC compartment, in which each cell type influences the development of the other.⁴⁷ The importance of mTEC in achieving normal thymic central tolerance is illustrated by the finding that the absence or reduction of mTEC is associated with a breakdown of self-tolerance in the T-cell compartment.^45,48–50 In turn, thymocytes expressing TCRs with high affinity for self (i.e., autoreactive) have been shown to be critical for maintaining a normal mTEC compartment.⁴³ The importance of high-affinity TCR–MHC interactions in promoting mTEC suggests that costimulatory molecules, known to augment TCR signaling, might have a role in thymic crosstalk. Indeed, recent studies from our lab have described a cooperative role for CD40L-CD40 and CD28-CD80/86 interactions in promoting mTEC development/maintenance as we find a profound mTEC defect occurs in the combined absence of CD40-CD40L and CD28-CD80/86 interactions.⁵¹ Earlier studies described relatively mild mTEC defects in both CD40L and CD40 KO mice,^22,25,45,52 which we also observed. However, mTEC numbers in CD40/CD80/86 KOs are far more severely decreased than in the CD40 KO alone, and are comparable to the reduced numbers of mTECs observed in complete absence of SP thymocytes. CD4 SP thymocyte expression of LTα and LTβ has been demonstrated to be compromised in CD28 KO and in CD80/86 KO mice ^51,53 as well as in CD40/CD80/86 KO mice.⁵¹ Because mTEC numbers are significantly decreased in the absence of LTαβ-LTβR interactions, it was proposed that CD28-CD80/86 interactions affect mTECs by regulation of LTαβ expression.^51,53 Indeed, similar to the CD40/CD80/86 KO, CD40/LTβR doubleKO mice displayed an mTEC defect more profound than either single KO, while the decrease in mTEC observed in LTβR KO mice was not further compromised in CD28/LTβR KOs.⁵¹ These findings support a model in which CD28-CD80/86 interactions function to support mTECs primarily through regulation of LTαβ expression and in which LTβR signaling, together with CD40 signals, are required for a normal adult mTEC compartment (Figure 1).

FIG. 1: Development of Aire-expressing medullary thymic epithelial cells (mTEC).

In the embryonic thymus, RANKL⁺ lymphoid tissue inducer (LTi) cells and/or Vγ5⁺ thymocytes interact with Aire⁻ mTEC progenitors and induce their maturation into Aire⁺ mature mTECs. In the post natal and adult thymus, positively selected CD4SP thymocytes provide CD40L/RANKL for continued medulla maturation. In addition to RANKL/RANK and CD40L/CD40 interactions, signaling mediated through LTβR also plays crucial roles in medullary development. CD28/CD80 interactions enhance LTαβ expression to potentiate LTβR signaling and together with CD40L/CD40 promote mTEC development/maintenance.

So how might the function of costimulatory signals be integrated with a requirement for high-affinity TCR-MHC/peptide recognition in driving mTEC development/maintenance? Insight into how this may be accomplished comes from studies by Irla et al., in which the expression of CD40L, LTαβ, and RANKL was investigated in thymocytes from TCR transgenic OTII/Rag KO (model for positive selection unaccompanied by negative selection) and RIP-mOVA OTII/Rag KO mice (model for negative selection).⁵⁴ As expected, given the importance of high-affinity TCR interactions in driving mTEC development, the number of mTECs was increased 4-fold in RIP-mOVA OTII/Rag KO compared with OTII/Rag KO mice. Of the three TNF family members examined, only LTα was differentially regulated in OTII/Rag KO and RIP-mOVA OTII/Rag KO mice. LTα mRNA levels were comparable in OTII/Rag KO and RIP-mOVA OTII/Rag KO DP thymocytes but were substantially increased in RIP-mOVA OTII/Rag KO CD4 SP relative to OTII/Rag KO CD4 SP thymocytes. Thus, in this system, only LTα expression was influenced by interactions between CD4 SP thymocytes and antigen-expressing mTEC, thereby providing a potential mechanism by which high-affinity, autoreactive CD4 SP thymocytes participate in mTEC development.⁵⁴ Taken together with the finding that engagement of CD28 upregulates LTαβ expression, these results suggest that TCR and CD28 signals, delivered when autoreactive CD4 SP thymocytes encounter antigen, help expand the mTEC compartment, thus ensuring that the medulla will be sufficient to tolerize the developing thymocyte population (Figure 1).

Thymic DCs are another important component for ensuring the self-tolerance of developing thymocytes (reviewed in Klein et al.⁵⁵). As described earlier, tDCs are found predominantly in the medulla and can present both peripheral and thymic-derived antigens to SP thymocytes. Recently, a defect in maturational status of tDC was found in mice that lacked TCR⁺ thymocytes and in TCR transgenic animals lacking cognate antigen expression.⁹ The defect in tDC maturation was determined to be largely due to decreased CD40-CD40L interactions, as the increased tDC maturation observed in CD4⁺ TCR transgenics exposed to cognate antigen was prevented by treatment with CD40L-blocking antibody.⁹ This finding suggests that upregulation of CD40L on developing CD4 SP thymocytes is of particular importance in driving tDC maturation and hence, more effective intrathymic antigen presentation. Similar findings have recently been published regarding the role of TCR⁺ thymocytes and thymic B cells (tB cells); tB cells were much reduced in TCRα KO mice (which lack TCR⁺ thymocytes and thus lack TCR-MHC/peptide interactions) and in CD40 KO or CD40L KO mice. These findings suggest that CD40L expressed by CD4 SP thymocytes is required for a normal frequency of tB cells.⁵⁶ This study also provides evidence to support the unique and crucial role of B cells in the negative selection of superantigen-reactive TCR-expressing thymocytes.

Thus, three thymic medullary cell populations critical for thymocyte self-tolerization, mTEC, tDC, and tB cells, depend on crosstalk with the developing thymocytes they affect. This process of crosstalk is, in turn, dependent on intact CD28-CD80/86 and CD40-CD40L costimulatory interactions.

IV. ABERRANT DEVELOPMENT OF SPECIFIC THYMOCYTE SUBSETS IN THE ABSENCE OF COSTIMULATORY INTERACTIONS

Recent advances have highlighted the complexity of the various cell subpopulations that emerge from the thymus to populate the periphery. In addition to conventional TCRαβ⁺ CD4 and CD8 SP cells, the list includes γδ T cells, iNKT cells, Treg cells, and innate-like CD8 SP cells. While it is beyond the scope of this review to describe in detail what is known regarding the functional significance of these various T cell subpopulations, excellent reviews are available on γδ T cells,⁵⁷ iNKT cells,⁵⁸ Treg,⁵⁹ and innate-like CD8T cells.⁶⁰ CD28-CD80/86 and CD40-CD40L interactions impact the development of both conventional TCRαβ⁺ SP T cells and the rarer thymocyte subpopulations. The influence of costimulatory interactions on conventional TCRαβ⁺ SP T cells are covered in later sections of this review. Here, the impact of costimulation, particularly of CD28-CD80/86 interactions, on development of the rarer thymocyte subpopulations is discussed. It should be mentioned at the outset that the effects of costimulatory interactions on development of the thymocyte subpopulations discussed can be direct (i.e., intrinsically influence the subpopulations’ development) or indirect (e.g., by impacting normal development of the thymus medulla, the region in which many of these subpopulations develop/mature).

A. γδ Thymocytes

While both CD40L and CD28 can be expressed on subsets of peripheral γδ cells,^61,62 only CD28 expression has been closely examined on developing γδ thymocytes. Expression of CD28 is readily detectable on most γδ thymocytes at levels similar to or higher than peripheral γδ T cells.⁶² In addition, numbers of thymic γδ T cells are significantly lower in CD28 KO relative to CD28 WT mice. This decrease in numbers does not appear to be due to either increased cell death or decreased proliferation of γδ thymocytes but rather has been suggested to be indirectly affected by the numbers of DN thymocyte precursors, as DN1 were found to have decreased percentages of proliferating cells in the CD28 KO.⁶²

B. iNKT thymocytes

As has been known for a number of years, thymic invariant NKT (iNKT) cells have unique development requirements. In addition to being positively selected on CD1d expressed on DP thymocytes, they also depend on activation of the NF-κB family member, RelB, in thymic stromal cells; both of these features distinguish their development from that of conventional TCRαβ⁺ thymocyte development,^63–65 which generally is positively selected on nonhematopoietic cells and can occur in the absence of RelB. Interestingly, mTEC development is severely compromised in the absence of RelB, suggesting that developing iNKT cells are directly or indirectly dependent on this epithelial cell population.⁴⁷ Recent work from White et al. demonstrated that the requirement for mTEC is due to the capacity to produce and trans-present IL-15, a cytokine required for later stages of iNKT cell development.⁶⁶ In this context, it is of interest to consider how development of thymic iNKT cells, which depend on an intact mTEC population, is affected by the absence of CD40L-CD40 and CD28-CD80/86 interactions. We reported that thymic iNKT cells are almost completely absent in the combined absence of CD28-CD80/86 and CD40L-CD40 interactions.⁵¹ The reason for this decrease may be multifaceted. Although we did not measure it directly, the profound decrease in mTEC observed in CD40/CD80/86 KO mice would be expected to result in a decrease in available IL-15. The absence of either CD40L-CD40 or CD28-CD80/86 interactions results in a loss of thymic NKT cells, while not as severe as that observed in CD40/CD80/86 KO mice, is profound. CD80/86 KO mice, which do not have a significant loss in mTEC numbers, do display a significant loss of thymic NKT cells. The absence of thymic CD28-CD80/86 interactions does not appear to compromise the early stages (NK1.1^neg) of iNTK development,^67,68 but it is associated with a 4-fold decrease in proliferation of thymic NK1.1⁺ iNKT cells.⁶⁸ Precisely how the absence of CD28-CD80/86 interactions impacts NKT expansion/differentiation has not been determined, although CD28 signaling has been shown to be important for optimal expression of IL-15Rβ (CD122) in CD8 memory T cells.⁶⁹ We also reported a decrease in thymic NKT cells in CD40-deficient mice, comparable to that observed in the CD28 KO and CD80/86 KOs. CD40 KO mice do have a decrease in the mTEC MHC II^lo population, which White et al. identified as an important trans-presenter of IL-15.⁶⁶ In addition, CD40 signals have been directly linked in peripheral and BM-derived DCs to upregulation of IL15Rα and IL-15,⁷⁰ a feature which may also be shared by tDC and mTECs. In fact, Cuss and Green found that intrathymic levels of IL-15 are reduced in CD40L KO mice.⁷¹ Thus, in CD40/CD80/86 KO mice, a decrease in IL-15Rβ (due to the absence of CD28 signaling) as well as a reduction in the generation of IL15-α/IL-15 complexes (due to the absence of CD40 signals) could be responsible for the dramatic decrease in iNKT observed in this combined KO strain (Figure 2).

FIG. 2: CD40L/CD40 and CD28/CD80 costimulation affects development of iNKT and innate-like CD8⁺ thymocytes.

In the thymic medulla, the interaction of CD40L/CD40 between TCRαβ⁺CD4SP thymocytes and mTECs/DCs enhances the expression of IL-15/IL-15Rα. The interaction of CD80/CD28 induces optimal expression of IL-15Rβ in immature iNKT and promotes their expansion and maturation into mature iNKT. Thymic iNKT cells produce IL-4 and regulate innate-like CD8⁺ T-cell generation.

C. Innate-like CD8 SP T cells

Thymic NKT cells have been linked to IL-4 dependent development of innate-like CD8 SP in the thymus. Innate-like CD8 T cells differ from conventional CD8 SP thymocytes and from memory CD8 T cells. Unlike conventional CD8 SP thymocytes, innate-like CD8 thymocytes have a memory phenotype; they are CD44^hi and express the T-box transcription factor Eomes, and they can rapidly produce cytokines in response to stimulation. Innate-like CD8 T cells differ from memory CD8 cells in that they do not require antigen experience to assume a memory phenotype, and while both cell types are Eomes⁺, innate-like CD8 T cells do not express T-bet.⁷² The number of thymic innate-like CD8 SPs is significantly reduced in both CD1d (lacking NKT cells) and IL-4 deficient mice.⁷³ While innate-like CD8 thymocytes can be found in normal mice,⁷³ they are dramatically increased in several gene-deficient mouse models including mice deficient for Itk, a Tec family kinase.⁷⁴ Although the increased frequency of CD8 SPs in Itk KO mice is not dependent on the presence of CD28, the “innate” characteristics (CD44^hi, Eomes⁺) of these CD8 cells were significantly reduced (i.e., the CD8 SP cells present were CD44^lo, Eomes⁻) in Itk/CD28 double-deficient animals.⁷⁴ Because CD28-CD80/86 interactions impact iNKT cell differentiation, CD28-CD80/86 interactions likely contribute to development of thymic innate-like CD8 T cells by enabling thymic NKT cell development as well as by supporting the development of innate-like characteristics in a subset of CD8 SPs (Figure 2).

V. COSTIMULATION AND FOXP3⁺ TREG CELL DEVELOPMENT IN THE THYMUS

Ensuring that peripheral T cells will be functionally tolerant of self-antigens is brought about during thymic development primarily by two mechanisms: (1) deletion of thymocytes bearing TCRs with high affinity for self-antigens and (2) conversion of a subpopulation of self-reactive thymocytes to a T-regulatory fate. As discussed in this and the following section on negative selection, both CD28 and CD40 signals are required for these two tolerance-inducing mechanisms to operate normally. As the work described below illustrates, CD28 signals figure prominently in the fate of self-reactive T cells, particularly in determining whether self-reactive thymocytes die by negative selection or adopt a T-regulatory fate.

CD28-mediated costimulation is necessary for Treg development in the thymus; mice deficient in either CD28 or CD80/86 have dramatically reduced numbers of thymic and peripheral FoxP3⁺ Treg cells.^75,76 Although CD28 costimulation is required for induction of IL-2 secretion, and IL-2/IL-2R signaling is important for Treg generation and survival,⁷⁷ there is a T-cell intrinsic requirement for CD28 costimulation, indicating that the role for CD28 is not mediated solely via IL-2 production. The cell intrinsic requirement for CD28 is evidenced by the fact that bystander wild-type T cells, which are able to produce IL-2, cannot restore Treg development of CD28 KO thymocytes in mixed bone marrow chimera in which WT and CD28 KO thymocytes coexist.⁷⁶ CD28 signaling in the generation of Tregs has been further addressed using CD28 knock-in and transgenic mice expressing CD28 molecules in which different signaling motifs within the cytoplasmic tail have been mutated.^76,78 Results from these studies indicate that efficient Treg generation does not require an intact PI3-K or an intact Itk kinase-binding motif but does require an intact Lck-binding motif. The Lck-binding motif couples to NF-κB activation, which, as will be discussed below, is important in CD28-mediated induction of FoxP3 expression. The relative importance of cell-intrinsic CD28 signaling in allowing cytokine responsiveness (e.g., CD25 upregulation) to develop remains to be determined.

Two nonmutually exclusive models have been proposed to explain the role of thymocyte-intrinsic CD28 costimulation in the generation of Treg precursors. The earlier of the two models stipulated that the role of CD28 is to give rise to FoxP3⁻CD25⁺ Treg precursors that have an enhanced response to IL-2, which in turn, is required for Treg precursors to differentiate into functionally mature Tregs expressing FoxP3.^79,80 In the second model, CD28 costimulation is required to induce FoxP3⁺CD25⁻ precursors.⁸¹ In this model, CD28 signaling acts through the Lck-binding domain to activate c-Rel, which then binds to the CD28 response element in the CNS3 region of the FoxP3 gene, resulting in its expression. Consistent with this model, mice expressing a constitutively active form of κB kinase β (IKKβ) transgene, which results in increased NF-κB activity, have numerous CD4⁺FoxP3⁺CD25⁻ thymocytes.⁸² On the other hand, Farrar et al. showed that CD28 costimulation, acting via the Lck-binding motif and NF-κB activation, is important for Foxp3⁻CD25⁺ Treg precursor cell generation.⁸³ Therefore, CD28 costimulation depends on Lck-mediated c-Rel activation to induce both Foxp3⁺CD25⁻ and Foxp3⁻ CD25⁺ Treg precursors. CD28 costimulation also promotes IL-2 production, which is required for both Foxp3⁺CD25⁻ and Foxp3⁻CD25⁺ Treg precursors to differentiate into Foxp3⁺CD25⁺ mature Tregs.^80,81,84 The identical Lck-binding motif in the CD28 cytosolic tail is also required for IL-2 production.^76,85,86 Taken together, CD28 costimulation promotes Treg cell generation not only by inducing precursor cell generation but also by promoting IL-2 production. The Lck-binding motif in the cytosolic domain of CD28 is required for all these steps. Just how these steps function in conjunction with one another to generate thymic Tregs requires further investigation.

The requirement for CD28 signaling to promote FoxP3⁺ Treg cell development stands in contrast to the role of CD28 signaling in promoting negative selection of immature conventional TCRαβ⁺ thymocytes (discussed in section VI below). In addition, it has recently been demonstrated that FoxP3 is proapoptotic; its expression results in a distinctive proapoptotic protein signature, and it represses expression of the prosurvival protein Bcl-2.⁸¹ This finding raises the question of how developing Tregs survive in the context of high-affinity TCR signals, CD28 costimulation and FoxP3 expression. Treg survival in this signaling milieu is likely due to their expression of CD122. Unlike conventional CD4 T cells, which do not express CD122, FoxP3⁺CD25⁻Treg precursors express CD122 and thus can form the intermediate-affinity IL-2R (X. Tai, unpublished data). Both the intermediate- and high-affinity (CD122 plus CD25) IL-2R are capable of inducing IL-2 signaling and overcoming FoxP3-induced lethality via upregulation of Bcl-2 expression.⁸¹ Expression of a Bcl-2 transgene driven by the proximal Lck enhancer⁸⁷ has been shown to rescue both CD28 costimulation- and FoxP3-induced apoptosis in developing thymocytes.^81,88 Together, these findings point to IL-2 signaling as the mechanism by which FoxP3⁺ Tregs are protected from CD28-and FoxP3-induced apoptosis. Notably, two different pathways for generating FoxP3⁺CD25⁺ thymic Tregs described earlier likely differ in their dependence on IL-2 survival signals. FoxP3⁺CD25⁻ precursors are highly susceptible to FoxP3-induced apoptosis while FoxP3⁻CD25⁺ Treg precursors are not. It remains to be determined whether important functional differences exist in the mature FoxP3⁺CD25⁺ Tregs derived from these two different precursor populations.

Interactions between CD40L and CD40 have also been demonstrated to have a role in the maintenance of normal numbers of thymic Treg cells.^26,89 A recently published study reported that the decrease in thymic Treg in CD40- or CD40L-deficient mice is the result of lower expression of thymic IL-2. The decrease in IL-2 does not affect the generation of FoxP3⁺ Treg cells in the thymus but rather results in a decreased homeostatic proliferation of resident thymic Treg cells; numbers of resident thymic Treg cells increased to normal levels when CD40L-deficient mice were given IL-2.⁷¹ Therefore, it appears that the level of intrathymic IL-2 required to support thymic resident Treg cells differs from that required for the thymic generation of Treg cells. A number of studies have attempted to define the thymic site of origin of Tregs. Although several reports have suggested that Treg commitment can occur in the cortex, FoxP3-GFP reporter mice have been used to document the localization of FoxP3⁺ cells almost exclusively to the thymic medulla.⁹⁰ The importance of CD28 signaling in generating Treg precursors suggests that the prevalence of CD80/86 expressing cells in the medulla is one feature that facilitates the development of Treg cells in this region of the thymus. Studies of Treg generation during ontogeny show that the appearance of thymic Tregs parallels development of a CD80/86-expressing medulla, ⁹⁰ and Treg generation is greatly impaired in a thymus that lacks a medullary region.⁹¹ Elegant work by Roman et al.⁹² has demonstrated that, similar to activation of mature T cells in the periphery, the TCR signal and CD28 costimulation in vivo must be presented in cis, i.e., on the same cell, to support Treg development. Likewise, Tai et al.⁷⁶ found that induction of FoxP3 in stimulated DP thymocytes in vitro required simultaneous stimulation with anti-TCR and anti-CD28; stimulation through the TCR followed by CD28 triggering did not upregulate FoxP3. As will be described below, the role of CD28 in promoting negative selection of autoreactive thymocytes also requires simultaneous engagement of the TCR and CD28. Thus, the signaling requirements for TCR and CD28 in negative selection and T-regulatory cell generation are similar; what remains to be determined is how the “decision” to purge the T-cell repertoire of a self-reactive thymocyte by deletion or to convert the thymocyte to a T-regulatory cell fate is made.

VI. CD28-CD80/86 AND CD40-CD40L INTERACTIONS IN NEGATIVE SELECTION

A number of in vitro and in vivo studies have demonstrated that CD28 signals play an important role in thymic negative selection. Early in vitro studies by Punt et al.^88,93 and later by others^94,95 showed that TCR signals alone were not sufficient to mediate cell death in DP thymocytes. When various costimuli were tested for the capacity to promote death in TCR-stimulated DP cells, only anti-CD28 was able to do so. As in the case of TCR plus CD28 stimulation of DPs to induce FoxP3 expression and Treg cell development, engagement of TCR and CD28 had to occur simultaneously to elicit a death response.⁸⁸ As striking as these in vitro demonstrations of the role of CD28 in negative selection are, assessment of negative selection in CD28 KO mice has generated conflicting results. Examination of in vivo negative selection in a number of different models has failed to show an effect on selection in the absence of CD28.^96,97 Other studies, however, have indicated that CD28 signals delivered in vivo do have a role in negative selection. Compelling data supporting a role for CD28 signaling in promoting death of immature autoreactive thymocytes have been presented in studies by Kishimoto and Sprent,⁹⁴ where they demonstrate that injection of neonatal mice with a low dose of a deleting antigen (SEB in H-2^d mice; OVA peptide in DO11 TCRtg mice) allows recovery of significantly more CD4⁺HSA^hi immature CD4 SPs in CD28 KO compared to WT mice. At high antigen doses, significant loss of both CD28 KO and WT CD4⁺HSA^hi immature CD4 SPs was observed.

Significant insight into the conflicting data regarding the role of CD28 signaling in negative selection came from studying the large TCR^hi DN thymocyte population that is present in mice lacking CD28 or CD80/86.⁹⁸ Staining with CD1d tetramer showed that the increase in DN TCR^hi thymocytes relative to WT mice was not the result of an increase in iNKTs. In fact, as expected from earlier studies, thymic iNKTs were reduced in CD28 KO and CD80/86 KO mice.^67,68 Instead, it was determined that the DN TCR^hi cells were enriched for self-reactive thymocytes that had differentiated to at least the DP stage and had then been developmentally diverted into the DN population. Though the DN TCR^hi thymocytes expressed self-reactive TCRs, they were shown to be anergic and thus functionally tolerant. Interestingly, these self-reactive DN TCR^hi thymocytes do not reside indefinitely in the thymus; rather, they migrate to the intestine where they re-express CD8α and become part of the cohort of CD8αα intraepithelial lymphocyte (IEL) population. Thus, it appears that even in vivo CD28 signals are uniquely able to promote death of immature thymocytes expressing a TCR with high affinity for self-antigen expressed intrathymically; when CD28 is absent, self-reactive thymocytes are developmentally diverted and become anergic DN TCR^hi thymocytes, which can migrate to the gut and become intestinal IELs, which, like their DN TCR^hi thymocyte precursors, do not respond to anti-TCR stimulation (Figure 3).⁹⁸

FIG. 3: TCR-mediated differentiation and thymic selection.

TCR signals induce the differentiation of DP thymocytes into CD4⁺CD8⁻ intermediate thymocytes. At the intermediate stage, TCR-signaled thymocytes are induced to differentiate into different lineage fates. In the absence of CD28 costimulation, strong persistent TCR signals stimulated by high affinity ligands induce intermediate thymocytes to undergo differentiation into TCRαβ⁺DN thymocytes which then receive IL-15 signal differentiate into TCRαβ⁺CD8αα thymocytes. The presence of CD28 costimulation in signaled intermediate thymocytes results in Treg differentiation or cell death (i.e. clonal deletion). While CD40L costimulation is crucial for clonal deletion, it is not necessary for the initiation of Treg differentiation but plays an important role in the maintenance of thymic Tregs.

The role of CD40-CD40L interactions in thymic negative selection has not been as extensively examined as that of CD28-CD80/86. It has been established that deletion of developing thymocytes expressing TCRs recognizing endogenous superantigens (SAg) does not occur in either CD40 KO or CD40L KO mice.⁹⁹ Bone marrow chimera experiments demonstrated that the role of CD40L in SAg-mediated deletion is non-cell autonomous, i.e., thymocytes of CD40L deficient origin were as efficiently deleted as those expressing CD40L when the two populations coexisted in WT host mice.¹⁰⁰ Thus, the role of CD40L during negative selection does not require that CD40L directly signal to a thymocyte to promote negative selection of that thymocyte. Whether the requirement for CD40-CD40L interactions in thymic negative selection extends beyond that of deletion in response to endogenous SAg has not been extensively examined. Foy et al. did find that blocking CD40-CD40L interactions with anti-CD40L in neonatal AND TCR tg mice expressing PCC, the agonist ligand for the AND TCR, inhibited thymic negative selection in this TCR tg model. ⁹⁹ In a separate set of studies by Ozaki et al. ¹⁰¹ the authors conclude that the importance of CD40-CD40L interactions during negative selection applies only when B cells serve as APCs. This is an interesting observation that requires further investigation.

VII. THE PARADOX OF CD28 SIGNALING IN THE THYMUS

A question that emerges when considering the role of CD28 signaling in the thymus is how it is possible for CD28 signals to be a critical component of cell death during negative selection but to provide pro-survival signals for TCR-signaled peripheral T cells. A series of studies examining the signaling pathways activated downstream of TCR plus CD28 engagement in DP thymocytes provided clear evidence that in DP cells, these signaling pathways converge to push the cell toward death. In DP thymocytes, TCR plus CD28 signals act to increase expression of apoptosis activators, Bim and Nur77, as well as to simultaneously inhibit ERK/MAPK Bcl-2 up-regulation.^102–104 This activity contrasts with that in peripheral T cells, where CD28 engagement cooperates with TCR signaling to promote survival in part via increased expression of Bcl_xL.¹⁰⁵ The mechanism responsible for changing how CD28-derived signals interact with and modify those of the TCR as thymocytes mature from the DP stage to mature T cells is not yet understood.

Another intriguing aspect of CD28 signaling in the thymus is its critical role in the generation of Treg cells. Whether CD28 signals function to generate a FoxP3⁻CD25⁺ (which then can respond to IL-2 to upregulate FoxP3) or a FoxP3⁺CD25⁻ Treg precursor (which then requires IL-2 to upregulate CD25), it is a necessary component in allowing developing thymocytes with high affinity for self-antigens to survive and develop into mature FoxP3⁺CD25⁺ Treg cells. As previously pointed out in this review, TCR and CD28 appear to require engagement in cis to promote negative selection as well as Treg development. Just when and how the decision is made to adopt one fate over the other is a current topic of investigation (Figure 3).

VIII. CONCLUSION

The thymus is a microcosm in which an immunologic version of much of the peripheral self is recreated. This allows developing thymocytes to be presented with the MHC/self-antigens and costimulatory molecules they will encounter in the periphery and to become tolerant to these immunologic stimuli. Two important means by which tolerization is achieved in the thymus are (1) a purging of autoreactive thymocytes (negative selection) and (2) induction into a T-regulatory fate. CD28-CD80/86 and CD40-CD40L costimulatory interactions are intimately involved in both of these processes. Interestingly, CD28-CD80/86 and CD40-CD40L costimulatory interactions are required both for negative selection of autoreactive thymocytes as well as for normal development of the thymic medulla critical for imposing tolerance on developing thymocytes. Evidence suggests that crosstalk between autoreactive thymocytes and mTEC, mediated at least in part by costimulatory molecule interactions, leads to expansion of the mTEC compartment. The result is a dynamic thymic tolerizing compartment that responds to meet the requirements of the developing T-cell repertoire, allowing both negative selection and T-regulatory cell development to occur appropriately.

ABBREVIATIONS:

TCR

T-cell receptor

APC

antigen presenting cells

BCR

B-cell receptor

DC

dendritic cell

Treg

T regulatory

DN

double negative

SP

single positive

mTEC

thymic medullary epithelium

TSA

tissue associated antigens

LTi

lymphoid tissue inducer

DETC

dendritic epidermal T cell

iNKT

invariant NKT

IEL

intraepithelial lymphocyte