Major Histocompatibility Complex Class II–Positive Cortical Epithelium Mediates the Selection of Cd4⁺25⁺ Immunoregulatory T Cells

The thymic development of T cells results in the acquisition of both immunologic tolerance to self and reactivity to foreign antigenic stimuli. The developmental processes of positive selection or self-restriction and negative selection or self-tolerance are compartmentalized within the thymus. The critical and unique function of thymic cortical epithelium in supporting positive selection of CD4⁺ T cells is well accepted 1,2. Conversely, multiple tissue types can mediate the opposing process of negative selection. Interactions between immature T cells and bone marrow–derived APCs, particularly dendritic cells, lead to tolerance via clonal deletion 3,4,5. In addition to APCs, thymic medullary epithelium can mediate incomplete deletion of high affinity thymocytes reactive to both endogenous mouse mammary tumor virus superantigens and MHC–peptide ligands 6,7. The question of whether clonal deletion or tolerance induction can occur in the cortex remains controversial and the accepted view is that the thymic medulla is the primary site of negative selection.

Interactions between immature T cells and thymic epithelium can also result in tolerance via nondeletional mechanisms. Early studies of murine radiation chimeras showed that functional unresponsiveness, rather than clonal deletion, was induced to self-antigens expressed by radioresistant thymic epithelium 8. Similarly, nude mice grafted with allogeneic fetal thymic epithelium devoid of hematopoietic precursors acquired donor-specific tolerance to heart and skin transplants but rejected both third party grafts and donor-specific hematopoietic cells 9,10,11. In this system, transplant tolerance could be transferred by CD4⁺ cells to naive nude mice and was “dominant” over the alloreactive effector cells. These experiments suggested that donor-specific tolerance to allografts generated by thymic epithelium was mediated by “regulatory” CD4⁺ T cells.

Separate experiments have implicated CD4⁺ regulatory cells generated in the thymus in the prevention of autoimmunity. Studies in two different systems have identified a population of CD4⁺ T cells constitutively expressing the IL-2 receptor α chain, CD25, which function as immunoregulatory cells. In the first model, peripheral mature CD4⁺ T cells depleted of the CD4⁺CD25⁺ population induced organ-specific autoimmunity when inoculated into histocompatible T cell–deficient nu/nu mice 12,13. In the second model, genetically susceptible mice thymectomized on day 3 of life (3dTx) developed similar organ-specific autoimmune disease 14,15. The disease process in 3dTx mice was induced by CD4⁺ cells but could be inhibited by normal adult CD4⁺ cells; the inhibitory activity was again contained in the subset of CD4⁺ cells coexpressing CD25 16.

There is strong evidence that regulatory CD4⁺CD25⁺ T cells develop and acquire their suppressive phenotype in the thymus 17,18,19. Clearly, the development of autoimmunity in 3dTx mice suggests a requirement for an intact thymus in the preservation of peripheral tolerance. Closer examination of CD4⁺CD25⁺ thymocytes demonstrate that they comprise ∼5% of the CD4 single positive (SP) pool and have a regulatory phenotype similar to their peripheral counterparts 18,19. They can prevent autoimmunity when adoptively transferred into 3dTx mice or, with CD4⁺CD25⁻ effector cells, into nu/nu hosts 18. Intrathymic transfer experiments suggest that CD4⁺25⁺8⁻ thymocytes arise from double negative thymocyte precursors, and fluorochrome labeling has demonstrated that CD4⁺CD25⁺ cells emigrate from the thymus to populate the periphery 17,18.

In several experimental systems, the expression of a regulatory cell phenotype does not occur until cells have entered the CD4 SP thymocyte pool 17,18,19. Recently, two of us (M.S. Jordan and A.J. Caton, Wistar Institute) have shown that an agonist peptide reexpressed as self on thymic epithelium is sufficient for the generation of functionally suppressive CD4⁺CD25⁺ T cells 19. Given that acquisition of a cell surface phenotype characteristic of regulatory CD4⁺ T cells (such as CD25) occurs “relatively late” during CD4-SP development, it has been suggested that the generation of regulatory CD4⁺CD25⁺ T cells is the result of high affinity interactions with medullary epithelium 19,20. While this model system demonstrated that thymic epithelium is sufficient for the development of these cells, it was unable to place the generation of CD4⁺CD25⁺ T cells in the context of positive and negative selection.

We have delineated the thymic development of CD4⁺ CD25⁺ cells in two different models of attenuated negative selection. In K14-A_β^b mice, an MHC class II I-A^b transgene is expressed only on thymic cortical epithelium; both medullary epithelium and bone marrow–derived APCs are class II-negative 21. Clonal deletion of CD4⁺ T cells cannot be documented, and CD4⁺ T cells react in MLRs to I-A^b–positive hematopoietic APCs. H2-DMα–deficient mice have H-2A^b molecules loaded almost exclusively with the class II–associated invariant chain peptide 22,23,24. The CD4⁺ T cells selected in H2-DMα–deficient mice are not tolerant to the broad array of peptides expressed by wild-type APCs and these cells also react vigorously in MLRs against B6 hematopoietic APCs 22,23. In both K14-A_β^b and H2-DMα–deficient mice, the introduction of I-A^b-positive hematopoietic APCs eliminates these autoreactive cells 25,26. Thus, clonal deletion is deficient in both K14-A_β^b and H2-DMα–deficient mice. The nondeletional mechanisms of T cell tolerance have not been examined in these models.

The data presented herein demonstrate that CD4⁺25⁺ T cells are selected on cortical epithelium and that these cognate interactions are sufficient for the phenotypic and functional development of CD4⁺CD25⁺ immunoregulatory T cells. Furthermore, analysis of the specificities of CD4⁺25⁺ T cells in K14-A_β^b and H2-DMα–deficient mice suggests that a subset of CD4⁺25⁺ T cells is subject to negative selection on hematopoietic APCs.

C57BL/6 (designated WT or B6 in text) and C57Bl/6 recombination activating gene 2-deficient (Rag-2°) were purchased from The Jackson Laboratory or Centre de Dévelopment des Techniques Avancées pour l'experimentation animale (Orleans, France). The MHC class II–deficient mice 27, K14-A_β^b mice 21, and H2-DMα–deficient mice 24 have been described previously. H2-DMα–deficient mice were a gift from Luc van Kaer (Vanderbilt University, Nashville, TN) and are seven generations backcrossed onto a C57BL/6 background. Both K14-A_β^b and MHC class II–deficient mice are >20 generations. All mice were maintained under SPF conditions in the animal facilities of the University of Pennsylvania or the Institut Pasteur.

All cells were grown in RPMI 1640 (Mediatech) supplemented with 10% heat-inactivated FCS, 100 U/mL penicillin, 100 μg/ml streptomycin, 2 mm l-glutamine, 10 mm Hepes (all from GIBCO BRL), and 50 μm 2-ME (Sigma-Aldrich). FITC and biotin–anti-CD25 (7D4), PE, and APC–anti-CD4 (RM4-5), and streptavidin–PE were purchased from BD PharMingen. Purified anti-CD3ε (2C11) was purchased from Accurate. Murine rIL-2 was purchased from Peprotech. Cells were analyzed on a Becton Dickinson FACSCalibur™ using CELLQuest™ software (Becton Dickinson).

Thymocytes, LN, and spleen cells were initially prepared by lysing erythrocytes with ACK lysis buffer. Cell preps were then enriched for CD4⁺ T cells by negative selection. Single cell suspensions were incubated with anti-CD8 (2.43), anti-B220 (RA3), anti-MAC-1(M1/70.15), F4/80, and anti-MHC class II (M/5114) culture supernatants for 30 min at 4°C followed by incubation with microbead-conjugated goat anti–rat IgG (PolySciences, Inc.) for 30 min at 4°C. Cell preps were placed on a Bio-Mag magnetic stand (PolySciences, Inc.) and the negative fraction was harvested. CD4⁺25⁺ T cells were subsequently labeled with biotin–anti-CD25 (7D4), APC-anti-CD4 (RM4-5), and streptavidin–PE and purified by flow cytometry on a FACSVantage™ Cell Sorter (Becton Dickinson). The purity of CD4⁺25⁺ T cells was >95%.

In anti-CD3ε stimulated proliferation assays, CD4⁺25⁻ cells (5 × 10⁴) were cultured in 96-well plates with MHC class II–deficient APCs (10⁵) or B6 APCs as indicated, 1 μg/ml anti-CD3ε, and the indicated numbers of CD4⁺25⁺ cells for 72 h at 37°C/5% CO₂. In the K14-A_β^b anti-B6 MLR or H2-DMα–deficient anti-B6 MLR, either K14-A_β^b or H2-DMα–deficient CD4⁺25⁻ cells (5 × 10⁴) were cultured in 96-well plates with 1–2 × 10⁵ irradiated, T cell–depleted B6 or H2-DMα–deficient spleen cells as indicated. B6, MHC class II–deficient, or H2-DMα–deficient stimulators (APCs) were initially prepared by erythrocyte lysis with ACK buffer. APCs were incubated with anti-Thy 1.2 (MMT1), anti-CD4 (172.4), and anti-CD8 (TIB-211) for 45 min on ice. APCs were subsequently incubated with rabbit complement (Accurate Antibodies) for 45 min at 37°C. APCs were then irradiated at 2,500 rad of γ-irradiation. MLRs were cultured for 96 h at 37°C/5% CO₂, pulsed with ³[H]thymidine for the final 16 h, and harvested on an automatic harvester (Wallac). All proliferation experiments were set up in triplicate and results are expressed as a mean of triplicates.

Donor CD4⁺ T cells were purified and enriched from spleen and LNs by negative selection as indicated above. Single cell suspensions were labeled with CFSE (Molecular Probes) as described previously 28. A total of 1–3 × 10⁷ CFSE-labeled C57BL/6 CD4⁺ T cells were injected intravenously into indicated mice in a total volume of 0.2 mL sterile PBS. Recipients were killed at 72 h and spleen and LNs were harvested. Total spleen and LN cells were purified, labeled with APC–anti-CD4 (RM4-5) and analyzed by flow cytometry as above.

For the preparation of CD45RB^high CD4 T cells, C57Bl/6-Thy1.1 splenic single cell suspensions were first enriched for CD4⁺ by positive selection on midiMacs™ columns (Miltenyi Biotec) according to the manufacturer's instructions. In brief, cells were first incubated with anti-CD4–FITC antibodies for 20 min on ice in PBS supplemented with 0.5% FCS, then incubated in the same buffer with anti-FITC microbeads for 15 min. The magnetically labeled positive fraction was retained on a midiMacs column. After enrichment, the cells were then labeled with anti-CD45RB-PE antibodies for 20 min on ice and then sorted on a FACStar^Plus™ (Becton Dickinson). For CD45RB, the brightest 40–50% of CD4⁺ cells were sorted as “high.” The purity of the population was >96%. For the preparation of CD25⁺ CD4 T cells, splenic single cell suspensions were enriched by positive selection on midiMacs columns with anti-CD25–FITC or biotin-labeled antibodies followed by anti-FITC or streptavidin–PE magnetic microbeads. Given that class II° mice contain very low numbers of CD25⁺ CD4 T cells, instead of sorting, one or two additional rounds of magnetic sorting was performed. For all populations, >95% of the CD4 T cells were CD25⁺.

6–8-wk-old C57Bl/6 RAG-2° mice were injected intravenously with 3 × 10⁵ CD4⁺CD45RB^high T cells either alone or with 3 × 10⁵ purified CD4⁺25⁺ T cells. Mice were monitored and weighed every other day and killed if they lost 20% of their initial body weight, had extensive diarrhea, or appeared morbidly ill. Whole intestines were first flushed extensively to eliminate the lumen content, then longitudinally opened and cut into 0.5 cm pieces for fixation in PBS-buffered formalin before sectioning. Paraffin sections were stained with H&E.

To understand the thymic requirements for the development of CD4⁺25⁺ T cells, we compared the phenotype and function of these cells in K14-A_β^b and wild-type C57BL/6 (B6) mice. Flow cytometric analysis of LNs and spleen revealed that peripheral K14-A_β^b CD4⁺ T cells expressed CD25 (Fig. 1 A). Like B6 mice, K14-A_β^b CD4⁺CD25⁺ cells are also found in the mature CD4 SP thymocyte pool (Fig. 1 B).

To determine if these cells have immunoregulatory characteristic similar to their wild-type counterparts, purified CD4⁺CD25⁺ T cells from K14-A_β^b or B6 animals were cultured for 3 d with T cell–depleted, MHC class II–deficient APCs and soluble anti-CD3ε antibody, with or without exogenous rIL-2. In this and the following assay, MHC class II–deficient APCs were used as K14-A_β^b CD4⁺ cells respond to I-A^b-positive (B6) APCs 21. CD4⁺CD25⁺ T cells from both B6 mice and K14-A_β^b mice did not proliferate to anti-CD3ε stimulation unless provided with exogenous rIL-2 (Fig. 2 A). Thus, these data demonstrate that both K14-A_β^b and B6 CD4⁺CD25⁺ T cells are anergic in vitro.

To examine suppressor function 29,30, a fixed number (5 × 10⁴ cells) of K14-A_β^b or wild- type B6 CD4⁺CD25⁻ effector T cells were cultured with T cell–depleted, MHC class II–deficient APCs, and soluble anti-CD3ε in the presence of increasing numbers of CD4⁺CD25⁺ lymphocytes from the same strain. Proliferation of responder K14-A_β^b and B6 CD4⁺25⁻ T cells was inhibited by the addition of CD4⁺25⁺ T cells to cultures in a dose-dependent fashion (Fig. 2 B). The ratios of K14-A_β^b CD4⁺25⁺ cells required to suppress the proliferation of a fixed number of effector cells were equivalent to wild-type regulatory cells. For example, a final effector to responder ratio of 1:1 resulted in an 80% inhibition of CD4⁺25⁻ proliferation in both K14-A_β^b and B6 cultures. Thus, K14-A_β^b CD4⁺25⁺ T cells appear functionally indistinguishable from wild-type regulatory CD4⁺25⁺ T cells.

K14-A_β^b mice were generated on a MHC class II–deficient background and MHC class II–deficient mice contain CD4⁺ T cells which are restricted to classical and nonclassical MHC class I antigens 31. Thus, it was possible that the CD4⁺25⁺ T cells present in K14-A_β^b mice were selected on MHC class I antigens. We addressed this possibility by (i) examining the phenotype and function of CD4⁺25⁺ T cells in MHC class II–deficient mice and (ii) generating K14-A_β^b mice lacking β2-microglobulin (β₂m) and MHC class I antigens (designated K14-A_β^b/β₂m^−/− in this text).

A significant fraction of the residual CD4⁺ T cells present in LNs and the spleen of MHC class II–deficient mice express CD25 (Fig. 3 A, and data not shown). To examine the suppressor function of these cells, a fixed number of B6 CD4⁺CD25⁻ effector T cells were cultured with T cell–depleted, B6 APCs, and soluble anti-CD3ε. In addition, indicated cultures (Fig. 3 B) received an equal number of FACS^®-sorted B6 or MHC class II–deficient CD4⁺CD25⁺ T cells. CD4⁺CD25⁺ T cells from MHC class II–deficient mice did not suppress the proliferation of wild-type effector T cells. Rather, the addition of these cells to B6 CD4⁺CD25⁻ effector T cells resulted in a two to threefold increase in the proliferation of cultured cells (Fig. 3 B). Thus, MHC class II–deficient CD4⁺CD25⁺ T cells are functionally distinct from B6 CD4⁺CD25⁺ T cells as they do not have regulatory activity.

FACS^®-sorted CD4⁺25⁺ lymphocytes from K14-A_β^b/β₂m^−/− (MHC class I–deficient) mice cultured with responders at a final effector to responder ratio of 1:1 resulted in an 85% inhibition of proliferation (data not shown). Thus, these cells have the same suppressive capabilities as those from MHC class I–sufficient K14-A_β^b mice, suggesting again that selection of these cells is MHC class II dependent. Taken together, these data suggest that the immunoregulatory CD4⁺25⁺ T cells which are present in K14-A_β^b mice require MHC class II expression on cortical epithelium for their selection.

The suppressive function of immunoregulatory CD4⁺25⁺ T cells is acquired in the thymus 18,19 and purified K14-A_β^b CD4⁺25⁺ SP thymocytes did not proliferate to TCR signals unless exogenous rIL-2 was added to cultures (Fig. 4 A). Furthermore, K14-A_β^b CD4⁺25⁺ SP thymocytes suppressed proliferation of anti–CD3ε-treated K14-A_β^b CD4⁺25⁻ effector cells similarly to peripheral K14-A_β^b CD4⁺25⁺ T cells examined in parallel (Fig. 4 A). Thus, K14-A_β^b CD4⁺25⁺ thymocytes are anergic and suppressive. These data confirm that selection on MHC class II-positive cortical epithelium is sufficient for the complete development of CD4⁺25⁺ immunoregulatory T cells.

The suppressive ability of K14-A_β^b CD4⁺25⁺ cells suggests that peripheral expression of MHC class II molecules is not required for the functional and phenotypic development of these cells. However, it remained possible that these immunoregulatory thymocytes could derive from activated recirculating cells 32. Therefore, it was important to establish that K14-A_β^b mice do not express functional MHC class II antigens in the periphery.

The division of transferred CFSE-labeled alloreactive CD4⁺ T cells is a very sensitive measure of functional MHC class II expression in host mice 33,34. We transferred alloreactive CFSE-labeled H-2^bm12 CD4⁺ T cells into K14-A_β^b, MHC class II–deficient, or B6 mice and examined their proliferation after 72 h. Alloreactive H-2^bm12 CD4⁺ T cells extensively divide 72 h after transfer into B6 hosts and localize in the spleen (Fig. 4 B) and LNs (data not shown). In contrast, no cells undergoing division were identified 72 h after H-2^bm12 CD4⁺ T cells were transferred into either K14-A_β^b or class II–deficient mice. Similar results were obtained when host mice were sublethally irradiated with 450 rads of whole body γ-irradiation 48 h before adoptive transfer (data not shown). These results do not directly address the requirement for peripheral antigens in the development of regulatory T cells which can suppress organ-specific autoimmune disease. However, they do supply functional corollaries to our earlier immunohistochemical and FACS^® data that MHC class II antigens are not available in the periphery of K14-A_β^b mice 21, and suggest that peripheral cognate T cell–MHC interactions are not required for the development of CD4⁺25⁺ regulatory cells.

To examine their in vivo function, CD4⁺25⁺ fractions were sorted from B6, K14-A_β^b, or MHC class II–deficient mice and tested for their ability to inhibit the colitis induced by transfer of CD4⁺45RB^high cells into Rag-2° recipients. It has been reported that severe combined immunodeficient mice reconstituted with 3 × 10⁵ CD4⁺45RB^hi B6 effector T cells develop an inflammatory bowel syndrome (IBD) which results in severe wasting and histologically resembles Crohn's disease 35,36. In this model, the wasting and IBD can be controlled by the cotransfer of CD4⁺25⁺ T cells 37,38.

B6 CD4⁺45RB^high effector CD4 cells were transferred into B6 Rag-2° mice either alone or with CD4⁺CD25⁺ cells purified from K14-A_β^b, class II°, or B6 mice. Animal weight was monitored every other day and animals were killed either when they had lost 20% of their body weight, had significant diarrhea, or between days 40 and 44.

Effector CD4⁺45RB^high cells alone induced an overt wasting disease with diarrhea and weight loss (Fig. 3 C) associated with colonic thickening and inflammatory infiltration of the colon (data not shown). Cotransfer of either B6 or K14-A_β^b CD4⁺CD25⁺ cells prevented intestinal inflammation; recipient mice gained weight over the course of the experiment and had normal colonic histology (Fig. 3c and Fig. d). In contrast, recipients of cotransfered CD4⁺45RB^high effector T cells and MHC class II–deficient CD4⁺CD25⁺ T cells developed severe wasting diarrhea and weight loss. These symptoms were associated with chronic colitis with hyperplasia of the glandular mucosae, increased numbers of crypts, and diffuse mononuclear cell infiltration. These results support the in vitro observations that K14-A_β^b CD4⁺25⁺ regulatory T cells are functionally similar to the regulatory T cells present in wild-type mice. Secondly, the population of CD4⁺ cells present in MHC class II–deficient mice which expresses CD25 (presumably as an activation marker) do not have this regulatory function.

The experiments presented above show that regulatory cells which developed in either B6 or K14-A_β^b mice could inhibit an inflammatory process mediated by B6 effector cells. However, the effector T cell responses in this model of IBD are promoted by bacterial antigens, rather than self-antigens 37. Therefore, we asked if CD4⁺25⁺ T cells could control the activation of a polyclonal self-reactive repertoire. K14-A_β^b CD4⁺ T cells proliferate vigorously in vitro to B6 APCs 21. These in vitro MLRs were performed using unfractionated K14-A_β^b CD4⁺ T cells in which the ratio of CD25⁻ to CD25⁺ cells is ∼10–15:1. Ratios of K14-A_β^b CD25⁺ regulatory to CD25⁻ effector cells of 1:1 are required to inhibit 80% of anti-CD3ε-induced proliferation (Fig. 2 B). Therefore, we examined the regulatory function of CD4⁺25⁺ cells in an MLR at a responder to suppressor ratio of 1:1. Purified K14-A_β^b CD4⁺25⁻ effector proliferated significantly to γ-irradiated, T cell–depleted B6 hematopoietic APCs in an MLR. The addition of an equal number of K14-A_β^b CD4⁺25⁺ T cells to this MLR suppressed the proliferation of B6-reactive K14-A_β^b responder cells in this MLR to 10% of the unsuppressed level (Fig. 5 A). Strikingly, CD4⁺25⁺ cells purified from B6 mice were unable to significantly suppress the anti-B6 proliferation of K14-A_β^b CD4⁺25⁻ responders at similar ratios (Fig. 5 A).

It was possible that B6 CD4⁺25⁺ T cells failed to inhibit the K14-A_β^b anti-B6 MLR because regulatory T cells only suppressed the activity of syngeneic cells. However, after stimulation via anti-CD3ε added to the culture, B6 CD4⁺25⁺ T cells could suppress the K14-A_β^b CD4⁺25⁻ anti-B6 MLR (Fig. 5 A). CD4⁺25⁺ T cells require TCR signals to become active suppressors 29,30 and these results suggest that B6 CD4⁺25⁺ T cells were not activated by B6 APCs, and therefore could not suppress a K14-A_β^b anti-B6 MLR. In contrast, K14-A_β^b CD4⁺25⁺ cells received activation signals from B6 hematopoietic APCs to block an I-A^b reactive MLR.

This pattern of restriction of the CD25⁺ fraction of CD4⁺ cells mirrors total CD4⁺ T cells from these respective mice: K14-A_β^b CD4⁺ T cells proliferate in MLR to syngeneic B6 APCs; B6 CD4⁺ T cells do not 21. To directly test if K14-A_β^b CD4⁺25⁺ T cells were reactive to I-A^b-positive APCs, we cultured FACS^®-purified K14-A_β^b or B6 CD4⁺25⁺ lymphocytes (from LNs and spleen) with γ-irradiated, T cell–depleted B6 splenocytes and measured the proliferative response. We previously observed that CD4⁺25⁺ T cells were unresponsive to stimulation with anti-CD3ε Ab. Consistent with these results, neither B6 nor K14-A_β^b CD4⁺25⁺ cells proliferated in the presence of B6 APCs. However, when exogenous rIL-2 was added to cultures, K14-A_β^b CD4⁺25⁺ T cells proliferated to the same extent as K14-A_β^b CD4⁺25⁻ when stimulated by B6 APCs. In contrast, B6 CD4⁺25⁺ T cells proliferated only minimally (Fig. 5 B). Furthermore, the addition of Y-3P 39, an anti-I-A^b–blocking antibody, to cultures significantly inhibited the proliferation of K14-A_β^b CD4⁺25⁺ T cells (data not shown) confirming that the reactivity of these cells is I-A^b dependent. Thus, the specificity of K14-A_β^b CD4⁺25⁺ cells is significantly different from B6 mice but similar to the CD4⁺25⁻ effector cells from K14 mice.

Both CD4⁺25⁻ effector cells and CD4⁺25⁺ regulatory T cells in K14-A_β^b mice have significant reactivity to syngeneic I-A^b-positive B6 APCs. Based on our observation that K14-A_β^b CD4⁺25⁺ T cells contain an autoreactive repertoire we hypothesized that, similarly to CD4⁺ effector cells, CD4⁺25⁺ T cells which react to MHC class II-positive hematopoietic thymic APCs undergo clonal deletion. To extend this hypothesis, we examined the function of CD4⁺25⁺ T cells in another model of attenuated negative selection, the H2-DMα–deficient mouse 22,23,24.

Flow cytometric analysis demonstrated that CD4⁺25⁺ cells were present in the thymus and periphery of H-2DMα-deficient mice (Fig. 6 A, and data not shown). H2-DMα–deficient CD4⁺ cells react in MLRs against wild-type (B6) APCs loaded with a broad array of I-A^b–associated peptides 22,23,24. Therefore, we asked if, similar to K14-A_β^b regulatory cells, H-2DMα–deficient CD4⁺25⁺ T cells could be activated by B6 hematopoietic APCs to suppress this proliferation, confirming a role for negative selection in shaping the CD4⁺25⁺ repertoire. The addition of purified H-2DMα–deficient CD4⁺25⁺ suppressor cells to H-2DMα–deficient CD4⁺25⁻ T cells culture at a ratio of 1:1 inhibited this proliferation to <1% of effector T cell proliferation. Again, FACS^®-sorted CD4⁺25⁺ cells from B6 mice had no significant effect on this anti–I-A^b proliferation, unless TCR signals were provided by the addition of antibodies to CD3ε (Fig. 6 B).

To determine if, similar to K14-A_β^b regulatory cells, H2-DMα–deficient CD4⁺25⁺ T cells were activated by B6 I-A^b molecules expressing a broad array of peptides, CD4⁺25⁺ cells were cocultured with either self H2-DMα–deficient APCs or B6 APCs with or without exogenous rIL-2. As shown in Fig. 6 C, H2-DMα–deficient regulatory cells did not proliferate significantly to self H2-DMα–deficient APCs in the presence of rIL-2. In contrast, H2-DMα–deficient CD4⁺25⁺ T cells proliferated vigorously when stimulated with B6 APCs in the presence of exogenous rIL-2. Thus, CD4⁺25⁺ T cells are tolerant to the repertoire of self-peptides on hematopoietic APCs and reactive to foreign antigen suggesting that the repertoire of these cells grossly mirrors that of the effector CD4⁺ T cell populations which develop in the same thymus.

It is generally accepted that CD4⁺25⁺ T cells are generated in the thymus, however, the selection events involved in the development of these cells remain poorly defined. In this report, we have taken a reductionist approach to determine the cognate interactions which are sufficient for the complete development of CD4⁺25⁺ immunoregulatory T cells. We find that CD4⁺25⁺ T cells are present in the thymus and periphery of K14-A_β^b and DMα-deficient mice. Like their wild-type counterparts, these cells do not proliferate to TCR-mediated stimulation unless exogenous IL-2 is provided, and their activation dominantly suppresses the proliferation of cocultured responder CD4⁺ T cells. We also demonstrate that the repertoire of CD4⁺25⁺ T cells from two distinct models of attenuated negative selection, K14-A_β^b and DMα-deficient mice, contain I-A^b reactive specificities, whereas B6 mice do not. Taken together, these data indicate that CD4⁺25⁺ T cells are selected on thymic cortical epithelium and that cognate interactions with MHC class II on cortical epithelium are sufficient for the development of CD4⁺25⁺ regulatory T cells. Our data also suggest that a subset of developing CD4⁺25⁺ T cell population are subject to negative selection on hematopoietic APCs.

To prove that the regulatory population of CD4⁺25⁺ cells was class II dependent, we examined the phenotype and function of CD4⁺25⁺ T cells in MHC class II-negative mice. As we noted above, a significant percentage of the CD4⁺ T cells remaining in MHC class II–deficient mice are CD25-positive (Fig. 3 A). However, we found that CD4⁺25⁺ T cells purified from MHC class II-negative mice do not suppress the proliferation of cocultured effector CD4⁺ T cells in vitro (Fig. 3 B). Hence, CD4⁺25⁺ T cells in MHC class II-negative mice are functionally distinct from CD4⁺25⁺ immunoregulatory T cells. Therefore, the development of regulatory CD4⁺25⁺ T cells requires the thymic expression of MHC class II molecules.

Previous examination of CD4⁺25⁺ T cell development suggested that thymic differentiation of these immunoregulatory T cells occurred at a late stage of CD4⁺ T cell development. These conclusions were based on BrdU labeling studies which demonstrated that these cells were not derived from CD25⁺ CD4⁺8⁺ double positive (DP) thymocytes and, separately, that expression of the CD25 antigen was not apparent until the CD4⁺ SP stage 17,19. Early models proposed that the differentiation of immunoregulatory CD4 ⁺T cells resulted from cognate interactions with self-antigens on hematopoietic APCs of sufficient affinity for the induction of anergy and the acquisition of suppressive characteristics but of insufficient affinity to induce clonal deletion 18,40. In contrast, a recent report has used transgenic/bone marrow chimeras to show that immunoregulatory CD4⁺25⁺ T cells are generated by high affinity interactions with self-peptide on thymic epithelium rather than hematopoietic APCs 19. This developmental pathway is reminiscent of the tissue requirements for the induction of nondeletional transplant tolerance to epithelial antigens which is also mediated by regulatory CD4⁺ cells 10. Indeed, Coutinho and his colleagues first proposed that regulatory cells developed after high avidity interactions with thymic epithelium. However, in both of these model systems the individual role of cortical and medullary epithelium could not be distinguished. Our observation that immunoregulatory CD4⁺25⁺ T cells are present in K14-A_β^b (Fig. 1 A) suggests that, similar to other CD4⁺ T cells, positive selection on cortical epithelium is necessary and sufficient for their differentiation from DP precursors.

Our experiments do not rule out the possibility that medullary epithelium and bone marrow–derived APCs play an inductive role that does not require MHC–TCR interactions. Punt, Singer, and colleagues have suggested that such a two-step strategy of clonal deletion operates to purge autoreactive DP thymocytes which encountered cognate self-antigen on cortical epithelium and the development of regulatory T cells could have a similar requirement 41. The current systems do not permit us to distinguish these possibilities.

It has been suggested in an animal model of thyroiditis that antigen-specific regulatory T cells require the relevant autoantigen in the periphery for complete function 42. However, this study monitored the onset of autoimmune disease as a marker of the presence of functional regulatory cells and did not distinguish development of suppressor cells from either persistence of these cells in the peripheral T cell pool or activation to suppress autoreactivity. Additionally, recent examination by Tung and his colleagues of autoimmune oophoritis showed that persistence of a relevant self-antigen in the periphery is required for the maintenance of T cell tolerance 43. Our results indicate that cognate recognition of MHC class II molecules in the periphery is not an absolute requirement for development of the regulatory population of CD4⁺ cells in that we have clearly shown that there is a significant number of functional CD4⁺25⁺ T cells in the periphery of K14-A_β^b mice (Fig. 1 A and 2 B). While it remains possible that K14-A_β^b mice express MHC class II on squamous epithelium (expression can be induced on keratinocytes during inflammation, unpublished data), the expression levels in disease-free mice are below our ability to detect by immunohistochemistry or FACS^® 21 and insufficient to drive allogeneic CD4⁺ T cell proliferation (Fig. 3 C). As K14-A_β^b CD4⁺25⁺ T cells can be activated by B6 APCs to suppress a K14-A_β^b anti-B6 MLR, these cells recognize self-antigens on I-A^b-positive hematopoietic APCs which are not present in either the thymus or the periphery of K14-A_β^b mice. We have not directly examined suppression of organ-specific autoimmunity and it remains possible that activation of suppressor cells in the periphery requires the presence of the inciting autoantigen. Nevertheless, our data indicate that peripheral MHC class II is not required for the development of functional CD4⁺25⁺ immunoregulatory T cells.

In K14-A_β^b mice, the thymic cell types which can mediate clonal deletion (hematopoietic APCs and medullary epithelium) are MHC class II-negative. In this setting, we have previously shown that ∼15% of the K14-A_β^b CD4⁺ T cell repertoire responds to B6 hematopoietic APCs 21. Therefore, it is striking that the MHC class II antigen reactivity of K14-A_β^b CD4⁺25⁺ T subset mirrors this autoreactivity and is significantly different from that of B6 CD4⁺25⁺ T cells.

Activation of regulatory CD4⁺25⁺ T cells by self-antigens suppresses autoimmunity. It is, thus, quite surprising that the peripheral CD4⁺25⁺ T cell population of B6 mice neither proliferates to nor suppresses an anti-B6 MLR response. These data indicate that the repertoire of B6 CD4⁺25⁺ T cells differs significantly from K14-A_β^b CD4⁺25⁺ T cells in that it has been purged of cells specific for self-hematopoietic APCs. The restriction pattern of CD4⁺25⁺ T cells seems to mirror the repertoire of the effector CD4⁺ T cell populations selected in the same thymus. It could be argued that the CD4⁺ T cell specificities generated in K14-A_β^b mice represent the positively selected CD4⁺ T cell repertoire before efficient clonal deletion. Thus, our observation that K14-A_β^b CD4⁺25⁺ T cells, but not B6 CD4⁺25⁺ T cells, are activated by syngeneic APCs suggests that expression of MHC class II on hematopoietic APCs results in the deletion of a subset of CD4⁺25⁺ thymocytes that are reactive to self-antigens expressed on hematopoietic APCs.

We turned to H2-DMα–deficient mice to extend these observations about the repertoire of CD4⁺25⁺ T cells 22,23,24. Similar to K14-A_β^b mice, these mice also have attenuated negative selection. Similar to K14-A_β^b CD4⁺25⁺ T cells, the repertoire of H2-DMα–deficient CD4⁺25⁺ T cells also parallels that of the effector CD4⁺ T cell population. A simple explanation for these data would be that restriction of CD4⁺25⁺ regulatory T cells occurs at the level of the T cell, that regulatory cells suppress autoreactive T cells which arise in the same thymus. However, wild-type (B6) CD4⁺25⁺ T cells can be activated by TCR signals supplied by anti-CD3ε to suppress both K14-A_β^b and H2-DMα–deficient effector cells, arguing that there is a deficiency in the B6 regulatory T cell repertoire rather than its function. These data argue that the repertoire differences noted for K14-A_β^b CD4⁺25⁺ T cells are not unique to the K14-A_β^b system and that our conclusions that CD4⁺25⁺ T cells are subject to clonal deletion on APCs may be generalized. These results also suggest that regulatory CD4⁺25⁺ T cells don't function in vivo in the periphery to prevent autoreactivity to self dendritic cells, presumably because tolerance to self-peptides present on hematopoietic APCs is mediated by clonal deletion. Rather, we would propose that CD4⁺25⁺ T cells routinely function to suppress reactivity to peripheral antigens which cannot induce deletion.

Recently, two of us have reported that development of regulatory CD4⁺25⁺ cells in a TCR transgenic system occurs after high affinity interactions with agonist self-peptide on thymic epithelium 19. In this model, it was proposed that generation of these cells was a high avidity event on medullary epithelium 19,20. Analysis of the K14-A_β^b system demonstrates that cognate interactions with MHC class II-positive cortical epithelium results in the development of functional CD4⁺25⁺ regulatory T cells. We also hypothesize that positively selected CD4⁺25⁺ thymocytes which recognize ubiquitous self-antigens with high affinity are subject to negative selection on hematopoietic APCs. Therefore, a subset of CD4⁺25⁺ T cells which are negatively selected in wild-type mice remain present in K14-A_β^b and H2-DMα–deficient thymi and can function to suppress an anti-B6 MLR. These selection events would restrict the repertoire of developing CD4⁺25⁺ T cells to tissue (epithelial)-specific antigens. This scenario is consistent with the previous observations that depletion and or manipulation of CD4⁺25⁺ T cells in wild-type animals results in organ-specific autoimmune disease rather than systemic autoimmunity.

We suggest that high affinity cognate interactions between developing CD4⁺ T cells and self-peptides on thymic cortical epithelium leads to self-tolerance through the induction of regulatory T cells. Recently, Hogquist and colleagues used the K14 promoter to target an MHC class I–restricted high affinity peptide to thymic cortical epithelium 44. In this system, they noted the reexpression of RAG proteins and receptor editing, rather than clonal deletion of self-reactive cells. Taken together with the data presented in this report, these results suggest that thymic cortical epithelium induces tolerance to self-antigens through nondeletional mechanisms.

In conclusion, our results indicate that anergic/suppressive CD4⁺25⁺ T cells are selected through cognate interactions on thymic cortical epithelium and are subject to clonal deletion on hematopoietic APCs.

We thank J. Salaun and P. Belo (Institute d'Embryologies, Norgent-sur-Marne, France) for the histology and M. Huerre (Unité d'Histopathologies, Institut Pasteur) for examination and interpretation of the histological sections. We thank T.Q. Lifsted, O. Burlen-Defranoux, and the Flow Cytometry Core Facility at the University of Pennsylvania for assistance and members of the Laufer laboratory and D. Izon for helpful discussion.

This work was funded by the generous support of the Division of Rheumatology of the University of Pennsylvania, National Institutes of Health (NIH) grant AI24541 to A. Caton, and NIH training grant T32AR07442 and 5T32EY07131. A. Bandeira was supported by a grant from the Association pour la Recherche sur le Cancer (ARC 5804).