Interactions between double positive thymocytes and high affinity ligands presented by cortical epithelial cells generate double negative thymocytes with T cell regulatory activity

Abstract

Previous studies on thymocyte differentiation by using reaggregate cultures (RC) of double positive T cell receptor (TCR) transgenic thymocytes and the thymic epithelial cell line ANV indicated that low concentrations of high affinity ligands for the TCR were efficient inducers of thymocyte maturation to CD4 single positive (SP) functional cells. In this study, it is demonstrated that, when high concentrations of high affinity ligands are used in this RC system, double positive (DP) cells down-modulate expression of both coreceptors and that, as a result, large numbers of double negative (DN) cells are generated. These DN cells proliferated modestly in response to stimulation by antigen, and this response was considerably augmented by the addition of IL-2 to the cultures. Notably, these antigen-stimulated DN cells produced large amounts of IL-10. When the DN cells generated in RC were cocultured with naive TCR transgenic T cells in the presence of antigen, they suppressed the proliferative response of the naive T cells. Thus, high affinity ligands, when presented to DP thymocytes by cortical thymic epithelial cells in reaggregate cultures, rather than causing deletion of the immature thymocytes, induce their differentiation into immunoregulatory DN cells, suggesting a distinct mechanism by which self tolerance may be maintained.

The acquisition of T cell tolerance to self antigens and reactivity to foreign antigens are largely the result of positive and negative selection events within the thymus. From the study of T cell receptor (TCR) transgenic systems, strong evidence for a primary role of ligand affinity in dictating the outcome of thymic selection has been obtained (1–4). Although some exceptions have been reported, in general, exposure of developing thymocytes to high affinity ligands results in negative selection whereas lower affinity ligands induce positive selection and, if the affinity is too low, immature double positive (DP) thymocytes die of “neglect.” The cells that present the ligands to the TCR within the thymus are another critical factor that determines the fate of developing thymocytes. There is compelling evidence to support the concept that cortical epithelial cells are the major and perhaps sole cell type capable of inducing positive selection (5–9). There is also strong evidence that hematopoietic-derived antigen-presenting cells (APCs) are potent inducers of negative selection (7, 10) although there may also be a role for medullary and even cortical epithelial cells in negative selection (7, 11). Most of the studies examining the correlation between ligand affinity and positive and negative selection used in vivo systems in which TCR ligands of varying affinity were present or fetal thymic organ cultures to which ligands of varying affinity for a transgenic TCR were added (1–3). In both types of studies, cells capable of positive and negative selection were present so that it could not be ascertained whether the affinity window that results in positive selection was due to some inherent property of low affinity ligands presented by cortical epithelial cells or whether the affinity window was imposed by the cells capable of causing negative selection, which require presentation of a high affinity ligand to activate a deletional pathway. Experiments using reaggregate thymic cultures that contained a cortical epithelial cell line as the sole class II MHC-expressing cell capable of presenting ligands to class II MHC-restricted thymocytes indicated that the induction of positive selection was not restricted to ligands of low affinity and that in fact there was a direct correlation between ligand affinity and the efficiency of induction of CD4 single positive (SP) thymocytes (12). Thus, these experiments indicated that high affinity ligands are capable of inducing positive selection when presented by cortical epithelial cells and that the affinity window is imposed by the deletion of cells interacting with similar high affinity ligands presented by hematopoietic and/or medullary epithelial APCs.

The present study further explores the consequence of high affinity ligand presentation by cortical epithelial cells to immature thymocytes. It was found that, in contrast to low concentrations of high affinity ligand that induced maturation to CD4 SP cells, high concentrations of high affinity ligand induced the down-modulation of CD4 and CD8 coreceptors on DP thymocytes and that the resultant DN cells produced significant amounts of IL-10 after antigen stimulation. These double negative (DN) cells also expressed immunoregulatory activity when cocultured with bystander naive T cells and antigen.

Materials and Methods

Mice.

The AND transgenic mice bred onto a B10.A(4R) or a B10.A background were obtained from S. Hedrick (University of California, San Diego, CA). The DO11.10 transgenic mice were obtained from K. Murphy (Washington University, St. Louis, MO). B10.A and B10.A(4R) mice were purchased from The Jackson Laboratory. The neonatal mice (d1 to d5) were obtained from the breeding colony of the La Jolla Institute for Allergy and Immunology.

Antibodies and Cell Surface Staining.

Phycoerythrin (PE) anti-CD4 (Becton Dickinson), RED613 anti-CD8 and FITC anti-CD8 (GIBCO/BRL), FITC anti-Vα11 TCR, anti-CD3-ɛ, anti-CD69, anti-I-E^k, anti-I-A^k, biotinylated anti-CD62L, anti-CD45RB, and PE anti-Vβ3 (PharMingen) were used for cell surface analysis. Stained cells were analyzed on a FACScan flow cytometer (Becton Dickinson). Dead cells were gated out on the basis of forward and side scatter.

Cells.

DP thymocytes were purified from newborn B10 A(4R) TCR transgenic thymi by positive selection by using Ly-2 Microbeads (Miltenyi Biotec, Sunnyvale, CA). The purity of DP thymocytes was routinely >98%. DN thymocytes were purified to >98% by removing cells that express CD4 and/or CD8 by using microbeads. The cortical epithelial cell line ANV-41-2 was provided by A. Farr (University of Washington, Seattle, WA). A stable transfectant of ANV-41-2 that expressed moth cytochrome c peptide tethered to the I-E^kβ chain and I-Eα chain (W5b) was made. The peptide-tethered β-chain cDNA construct was provided by J. Kappler (National Jewish Center, Denver, CO). The I-Eα^k and I-Eβ^k cDNA in the expression vector pcEXV-3 were obtained from J. Miller (University of Chicago). Transfection of I-E^k genes into ANV cells for stable expression of IE^K has been described (12). The B cell lymphomas CH27 and A20 were used as APCs in some experiments. I-E^K-transfected fibroblasts originally generated by R. Germain (National Institutes of Health, Bethesda, MD) were also used as APCs after they were transfected with genes encoding intercellular adhesion molecule-1 (ICAM-1) and B7-1 (DCEK.ICAM) (13).

Thymic epithelial cells (TEC) were prepared from deoxyguanosine-treated d14–d16 B10.A fetal thymi (5). After trypsin digestion, the TEC cell suspension was depleted of CD45⁺ cells by using biotinylated anti-CD45 mAbs and Avidin micromagnetic beads, followed by passage through MiniMACS (Miltenyi Biotec) magnetic columns.

Naive CD4⁺ T cells were purified from spleen and lymph nodes of AND or DO11.10 TCR transgenic mice by collecting nylon wool nonadherent cells, and purifying CD4⁺ cells by negative selection by using biotinylated antibodies to CD8, class II MHC, B220, CD11b, and kappa light chains and magnetic beads coated with streptavidin.

Preparation of Reaggregate Cultures (RC).

Thymic RC were established by centrifuging a mixture of 0.5 × 10⁶ transfected (ANV-IE^k) or parental ANV-41-2 (pretreated for 2 days with IFN-γ; ref. 12) and 1 × 10⁶ DP thymocytes. The resultant pellet was placed onto a nucleopore filter set on a foam sponge in RPMI 1640 complete medium (10% FCS/5 × 10⁻⁵ M 2mercaptoethanol (2-ME)/l-glutamine/nonessential amino acid/sodium pyruvate/penicillin/streptomycin; GIBCO/BRL). The peptide antigen MCC_88–103 (MCCp) was added at various concentrations. The cell slurry aggregated to form a “lobe” by 18 h of culture. Cell suspensions from reaggregated lobes were made after varying times in culture, followed by nylon mesh filtration. Cell yields were determined by counting trypan blue-excluding lymphocytes. In some experiments, CH27 or DCEK.ICAM cells were used instead of ANV-IE^k or a 2:1 mixture of ANV-IE^k and the other APCs was used to establish RCs. Routinely, RCs were harvested after 4 days in culture, and the cells were analyzed for CD4, CD8, and other surface markers and/or placed into suspension culture. Before culture, contaminating epithelial cells were removed by collecting nonadherent cells and depleting I-E^k-expressing cells by panning on anti-I-E^k-coated plates. Dead cells were removed by Ficoll (Pharmacia) centrifugation, and the viable DN cells were placed into culture.

Proliferation Assay.

To measure the functional capacity of DN cells, 10⁵ cells were cultured with 10⁵ mitomycin C-treated CH27 cells in U-bottom wells together with serial dilutions of MCCp. To some cultures, IL-2 (100 units/ml) was added at the time of stimulation. After 30 h, cultures were pulsed with 1.0 μCi (1 Ci = 37 GBq) [³H]thymidine and harvested at 48 h.

Cytokine Assays.

For IL-2, IL-4, IL-10, and IFN-γ analyses, aliquots of supernatants from the cultures that had been established to measure the proliferative response were removed at 24 h or 48 h and assayed by ELISA.

Carboxyfluorescein Diacetate Succinimidyl Ester (CFSE) Labeling.

CFSE (Molecular Probes) labeling was performed as previously described (14). Cells were washed with PBS containing 0.1% BSA (ICN), resuspended at 10⁷ cells/ml, and incubated with 2 μl of CFSE stock for 10 min at 37°C. Cells were then washed and resuspended in culture medium.

Apoptosis Assays.

To detect apoptotic cells, the terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) assay was used (11). After labeling, the cells were analyzed by flow cytometry. Propidium iodide (PI) staining was also used to detect apoptotic cells. Cells were stained with 0.05 μg/ml PI (Sigma) for 1 min and analyzed by flow cytometry.

Peptide Synthesis.

Peptides were synthesized on a Symphony peptide synthesizer (Rainin Peptide Technologies, Washington, DC) as previously described (15). The identity of peptides was substantiated by amino acid sequence and/or composition analysis. They were routinely >95% pure after HPLC.

Results

High Concentrations of Antigen Lead to the Generation of DN Cells in RCs of DP Thymocytes and the TEC Line ANV-41-2.

The thymic cortical epithelial cell line ANV-41-2 transfected with I-E^k (ANV-IE^k) has been used in the RC system previously to study the role of ligand affinity in the positive selection of TCR transgenic thymocytes. Positive selection was observed in a dose-dependent manner when the cognate antigen was added to the cultures (12). In further investigation of the antigen dose dependency, a broad peptide dose range was tested (Fig. 1). At the optimal dose of antigen for the induction of positive selection (0.04 μM), approximately 50% of recovered cells were CD4 SP (Fig. 1B). However, with higher antigen doses, the percentage of CD4 SP cells decreased, and large numbers of DN cells appeared. At 40 μM MCCp, there was less than 10% CD4 SP cells, with the rest being DN cells. These results are similar to those published in a previous report (16). The increase in the DN population was also observed when normal thymus cortical epithelial cells were used in the RC in the presence of high concentrations of MCCp (Fig. 1C) or when ANV cells transfected with IE^k whose β chain contained MCCp tethered to the N terminus (data not shown).

Figure 1.

High concentrations of cognate antigen lead to the generation of DN thymocytes in RCs. RCs were established with TCR transgenic DP thymocytes and the thymic epithelial cell line ANV-IE^k together with varying concentrations of the antigen MCCp. After 4 days in culture, cells were analyzed for CD4 and CD8 expression. (A) Starting population of DP thymocytes. (B) Effect of varying concentrations of MCCp on the maturation of DP thymocytes. (Right) PI staining of the DN populations obtained from the RC incubated without antigen (Upper) or with 40 μM antigen (Lower). (C) Large numbers of DN cells were also observed when normal TECs were used as APCs in RC.

We considered three possible sources of the DN cells generated after high dose antigen exposure. First, it was possible that the small number of DN cells that contaminated the DP cells placed into culture (Fig. 1A) underwent massive expansion during the 4-day culture period. Second, in the presence of high concentration of antigen, apoptosis of DP cells was induced, and, in the process of dying, these cells converted to DN. Third, the DP cells were viable and had down-modulated their coreceptors in response to the high dose of antigen.

To evaluate whether the small population of DN cells present in the starting population of DP cells (≤1%) was the source of the DN cells found at the end of the culture period, we purified DN and DP cells from newborn thymi, labeled them with CFSE, and placed them in RC in the presence of high dose antigen (Fig. 2). Neither DN nor DP thymocytes divided significantly until day 3, and only about half of the DN cells underwent one or two rounds of division by the end of the 4-day culture period (Fig. 2B). In contrast, most of the DP cells underwent between 1–5 divisions, with most of the cells having undergone 3–4 divisions by day 4 (Fig. 2A). Thus, the large increase in absolute number of DN cells could not have resulted from preferential expansion of preexisting DN cells.

Figure 2.

The capacity of DN and DP cells to proliferate in RCs. DP and DN thymocytes from the thymus of TCR transgenic B10A (4R) mice were purified, stained with CFSE, and placed into RC with 40 μM MCCp. Cultures were harvested daily over a 4-day period and analyzed by flow cytometry for CFSE content. (A) DP thymocytes. (B) DN thymocytes.

Because it has been shown that down-modulation of CD4 and CD8 coreceptors occurs when DP thymocytes undergo apoptosis, we investigated whether the DN cells generated in response to high dose antigen exposure in our RC system were viable, as measured by PI uptake or TUNEL assay. As shown in Fig. 1 Right, only 21% of the DN cells derived from RC with high dose antigen were dead as measured by PI staining. This result was confirmed by the more sensitive TUNEL assay, which indicated that most of the DN cells were viable (65%) after the RC (data not shown). Thus, cells undergoing apoptosis were not a major cause of the DN cells generated in the RC system. In contrast, when DN thymocytes were isolated from RC cultured in the absence of antigen, they were predominantly PI positive (approximately 65%), with relatively few viable DN recovered after the 4-day culture period (0.85 × 10⁴). In comparison, the recovery of live DN cells after culture of DP thymocytes with 40 μM antigen was much higher (29.2 × 10⁴; Table 1). This recovery is similar to that of SP cells from RC in the presence of concentrations of MCCp that are optimal for positive selection (5–20 × 10⁴; ref. 12). Taken together, these results indicate that DP thymocytes develop into viable DN cells during the course of the RC with a thymic cortical epithelial cell line in the presence of high concentrations of the cognate antigen.

Table 1.

Viability and yield of DN cells recovered following RC

Ag, μM	Total cells recovered (× 10−4)	% DN	% Viable DN	Viable DN cells recovered (× 10−4)
0	4.0	62	35	0.9
40	45.4	87	74	29.2

Reaggregate cultures established with DP AND thymocytes and the ANV-IE^k thymic epithelial cell line either in the absence or presence of 40 μM MCCp were harvested after 4 days of culture, and the cells were analyzed for the yield and viability of DN cells. 

Kinetics of Phenotypic Changes of DP Thymocytes Cultured in the Presence of High Dose Antigen.

To further investigate the mechanism by which DN cells were generated in the RC of DP thymocytes in the presence of 40 μM antigen, we studied the appearance of DN cells throughout the 4-day culture period. Purified DP TCR transgenic thymocytes were cultured with ANV-IE^k cells in the presence of 40 μM MCCp, and samples were analyzed daily by flow cytometry (Fig. 3). The starting population expressed CD3 at relatively low levels and were CD69 negative (day 0, Fig. 3 B and C). By day 1, DP thymocytes had started to down-modulate both coreceptors. This result was associated with TCR down-modulation and up-regulation of CD69 expression. At this time, there were also some CD4⁺ CD8^low cells and very few DN cells. By day 3, most of the cells were DN cells with low CD69 expression. Cells at this stage still had somewhat lower CD3 expression than at day 0. At day 4, CD3 expression had fully returned to day 0 levels, whereas the expression of the other markers remained similar to that of day 3. From these kinetics, it appears that the majority of DN cells are directly derived from DP thymocytes, although some cells appeared to transit through a CD4⁺ CD8^low stage before becoming DN cells. The pattern of modulation of CD69 expression during the culture mimicked the expression of this marker during SP thymocyte differentiation; however, the changes of TCR, CD4, and CD8 expression did not follow any conventional differentiation pathway.

Figure 3.

Kinetics of DN thymocyte generation in RC containing high concentration of antigen. RCs were established as described in Fig. 1, and samples were taken daily and analyzed for expression of CD4 and CD8 (A), CD3 (B), and CD69 (C).

Effects of other APCs on the Generation of DN Thymocytes.

The generation of large numbers of DN thymocytes in RC with the ANV-IE^k cell line in the presence of high concentrations of the cognate antigen raised the issue of whether this was a special feature of thymic epithelial cells or whether other class II MHC-expressing APCs would also induce the generation of this cell population. To explore this possibility, RCs were established with transfected fibroblasts that express I-E^k, B7 and ICAM-1 (DCEK.ICAM), and the I-E^k-expressing B cell lymphoma CH27. Both of these cell types are excellent APCs for peripheral T cells (17–19). When DP thymocytes were cultured with either CH27 or DCEK.ICAM and 40 μM MCCp, no discrete reaggregrated “lobes” were formed, and after the 4-day culture period, very few viable cells were recovered (Table 2). Furthermore, when CH27 or DCEK.ICAM cells were added along with ANV-IE^k and 40 μM MCCp, although the thymocytes aggregated into “lobes” as well as when mixed with ANV-IE^k cells alone, very few viable thymocytes were recovered after 4 days in culture. Thus, in this in vitro RC model of thymic differentiation and selection, noncortical epithelial cells presenting high concentrations of antigen to DP thymocytes resulted in deletion of DP thymocytes. In contrast, thymic cortical epithelial cells induced cell division of the thymocytes together with the down-modulation of CD4 and CD8 coreceptors, resulting in the appearance of large numbers of DN thymocytes.

Table 2.

The effect of different APCs on the generation of DN thymocytes in RC

APCs	No. of DN thymocytes (× 104)
ANV	14.7
DCEK.ICAM	1.7*
CH27	0.7*
ANV + DCEK.ICAM (2:1)	0.2
ANV + CH27 (2:1)	0.5

Reaggregate cultures were established utilizing as APCs ANV-IE^k, DCEK.ICAM, or CH27 cells. In some cultures, a 2:1 mixture of ANV-IE^k with either DCEK.ICAM or CH27 cells was used. After 4 days, the cultures were harvested, and the number of viable DN cells generated in the culture was determined. 

^*No lobes formed. 

Functional Capacity of DN Cells Generated in RC.

The functional capacity of the DN cells generated in RC in the presence of high concentrations of antigen was assessed by their capacity to proliferate and produce cytokines after antigen stimulation. DN cells obtained from a 4-day RC by using ANV cells expressing IE^k with tethered MCCp were cultured with CH27 cells and a dose range of MCCp. A modest early proliferative response, which peaked 1–2 days after antigen stimulation, was markedly enhanced when IL-2 was added at the start of the culture (Fig. 4). Table 3 shows the production of cytokines in the same cultures. No cytokines were detected in the cultures that were stimulated by antigen in the absence of added IL-2, and, by 48 h, most DN cells in these cultures were dead. In the cultures that were stimulated with antigen plus IL-2, a considerable quantity of IL-10 was produced along with a small amount of IL-4. The production of these cytokines was antigen dependent, because cells cultured in the presence of IL-2 but without antigen produced no detectable cytokines (data not shown). For comparative purposes, also shown in Table 3 is the cytokine production of naive, antigen-stimulated peripheral CD4⁺ cells. Substantial amounts of IL-2 were made by these cells, along with small quantities of IL-4 and IFN-γ, and no IL-10. All cultures were also tested for the production of TGFβ, but in no instance was there any detection of this cytokine (data not shown). From these data, we conclude that the DN cells generated in RC can respond in an antigen-specific manner but require an exogenous source of IL-2 to maintain their viability and their capacity to make cytokines, with the predominant cytokine produced being IL-10.

Figure 4.

Functional capacity of DN thymocytes generated in RC. After the 4-day culture period of DP thymocytes and MCCp-expressing ANV cells (W5b), RC were harvested, and the thymocytes were placed into culture with CH27 APCs and varying concentrations of MCCp. [³H]Thymidine incorporation was measured after 2 days in culture. Cultures without IL-2 added (♦); cultures with 100 units/ml IL-2 added at the initiation of culture (■).

Table 3.

Cytokine production of DN T cells generated in RC by high dose of antigen

T cell source (AND TCR Tg)	IL-2, units/ml	IL-4, ng/ml	IL-10, ng/ml	IFN-γ, ng/ml
Peripheral CD4+	91	0.41	—	0.27
DN (RC)*	—	—	—	—
DN (RC) + IL-2*	ND†	0.33	1.17	—

—, Not detectable. Limits of the assay were as follows: IL-4, 0.2 ng/ml; IFN-γ, 0.25 ng/ml; IL-10, 0.3 ng/ml; and IL-2, 3 units/ml. 

^*DN cells were generated after 4 days of reaggregate thymic organ culture of DP thymocytes with I-E^k-transfected ANV-41-2 thymic cortical epithelial cells in the presence of 40 μM MCCp. Those cells were stimulated with 10 μg/ml of MCCp and CH27 APCs for 1 to 2 days either in the absence or presence of exogenous IL-2 (100 units/ml). 

^† ND, not determined. 

Regulatory Capacity of RC-Derived DN Cells.

The cytokine profile of the RC-derived DN cells was similar to that described for some regulatory cells: namely, production of large amounts of IL-10 with little or no IL-2 (20). Because of this result, we conducted experiments to determine whether the RC-derived DN cells had regulatory effects on other T cells. Naive TCR transgenic CD4⁺ T cells were labeled with CFSE and placed in culture with RC-derived DN cells and APCs, and their proliferative response in the presence of antigen was monitored by flow cytometry. To determine whether a shared antigen specificity between the DN cells and the naive T cells was critical, naive T cells with the same (MCCp) or with a different [ovalbumin 323–339 (OVAp)] specificity were used. The data are shown in Fig. 5. In the control cultures of MCCp-specific AND T cells (Fig. 5A), most of the AND T cells divided at least once, with the greatest number of cells undergoing two divisions. In contrast, in the presence of an equal number of DN cells, over one half of the naive T cells did not divide at all, and most of those that did went through a single round of division. Similar results were obtained with naive T cells that had an antigen specificity different from that of the DN cells (Fig. 5B). In the absence of DN cells, the OVAp-specific DO11.10 cells divided on average two to three times, whereas, in the presence of MCCp-specific DN cells, about one half of the cells remained undivided, and most of the rest divided only once. Similar but less complete inhibition of proliferation was observed when DN cells were present in a 1:3 ratio with CD4⁺ cells (data not shown). Antigen stimulation of the DN cells was required, because DO11.10 cells cultured in the presence of DN cells and OVAp but without MCCp divided similarly to the DO11.10 cells cultured in the absence of DN cells (data not shown). Analysis of supernatants indicated that large quantities of IL-10 were produced in the cultures containing T cells and DN cells stimulated with their respective antigens, whereas the supernatants from the separate cultures of the antigen-stimulated naive T cells or antigen-stimulated DN cells contained no detectable IL-10 (Fig. 5C). This result suggests that the IL-2 synthesized by the antigen-stimulated CD4⁺ T cells was required to support the viability and IL-10 production of the antigen-stimulated DN cells.

Figure 5.

DN cells derived from RCs in the presence of 40 μM MCCp have immunoregulatory activity. (A) Naive TCR transgenic CD4⁺ T cells from MCCp-specific (AND) TCR transgenic mice or (B) ovalbumin-specific (D011.10) TCR transgenic mice were labeled with CFSE and placed in culture with or without an equal number (5 × 10⁴) DN AND thymocytes generated in RC in the presence of high concentrations of MCCp. Mitomycin-treated CH27 (A) or CH27 and A20 cells (B) were added as APCs together with varying concentrations of MCCp (A) or OVAp plus 1 μg/ml MCCp (B). After 3 days, the cultures were harvested and analyzed for cell division of the naive T cells by flow cytometry analysis of CFSE content. (C) Culture supernatants were obtained at 48 h and assayed for IL-10 content.

Of the other cytokines assayed (IL-2, IFN-γ, IL-4, and TGFβ), IL-2 was produced in significant quantities after antigen stimulation of cultures that contained both CD4⁺ T cells and DN cells or in the cultures that contained only CD4⁺ T cells, and small amounts of IL-4 were produced in the cultures containing both CD4⁺ T cells and DN cells (data not shown).

Discussion

We previously reported that, when the only MHC class II⁺ APCs present were cortical epithelial cells, there was a direct correlation between the affinity of a peptide for the TCR and its ability to drive positive selection of DP thymocytes in a RC system (12). In other systems, such as fetal thymic organ culture, that contain bone marrow-derived APCs and medullary epithelial cells as well as cortical thymic epithelium, addition of high affinity peptides results in deletion of the DP thymocyte population and failure to generate CD4 SP thymocytes. These observations support the idea that cortical epithelial cells are uniquely capable of driving positive selection and are either incapable of, or are deficient in, causing deletion of developing thymocytes.

In the present study, the unique function of this cell type was further demonstrated by the finding that high concentrations of high affinity ligands presented to DP thymocytes by cortical epithelial cells did not result in deletion but, instead, caused their differentiation to DN cells that, on antigen stimulation in the presence of IL-2, produced a considerable quantity of IL-10. Furthermore, these DN cells suppressed the proliferative response of bystander CD4⁺ T cells present in the same culture. Whether IL-10 is the sole mediator of this suppression or whether other mechanisms may also be involved remains to be determined.

By examining the kinetics of cell division and the appearance of DN cells in the RC, the following conclusions were made: (i) the vast majority of DN cells that appeared after 3–4 days in culture were derived from the DP input population as a result of the synchronized down-modulation of both coreceptors, although some cells appeared to transit through a CD4⁺, CD8^low stage before their conversion to DN cells; (ii) the majority of DN cells that were generated were viable and could be maintained in culture as DN cells if a source of IL-2 was provided; and (iii) the generation of DN cells after culture with high concentrations of high affinity ligand required that cortical epithelial cells were the only APCs available; addition of B cell or class II transfected fibroblast APCs to the RC resulted in deletion of the DP cells and recovery of very few viable cells.

Although the literature contains conflicting reports, there is a general consensus on certain aspects of the function of the three class II MHC-positive cell types (medullary epithelial cells, cortical epithelial cells, and bone marrow-derived cells) in the process of thymocyte differentiation and selection. Bone marrow chimera studies have convincingly demonstrated the capacity of bone marrow-derived APCs to cause negative selection and their inability to induce positive selection (21, 22). Medullary epithelial cells are also capable of negative selection although less efficient in this regard than bone marrow-derived cells (23). In contrast, cortical epithelial cells appear to be uniquely capable of driving positive selection (5, 9, 24, 25). This result has been demonstrated perhaps most clearly by using transgenic mice in which class II MHC expression was limited to cortical epithelial cells (9). Furthermore, when antigen presentation was limited to cortical epithelial cells, persistence of certain autoreactive T cells was observed (9), indicating a limitation in the capacity of these cells to cause negative selection. Similarly, in an OT-1 TCR transgenic system in which the cognate ovalbumin peptide antigen was expressed only on cortical epithelial cells, deletion of double positive cells did not occur but rather partial down-modulation of CD4 and CD8 coreceptors and TCR was observed (26), reminiscent of our in vitro data at early stages (d1–d2) of the RCs.

Although our in vitro data do not provide direct evidence regarding the possible in vivo physiologic relevance of the DN regulatory cells induced by the interaction of developing thymocytes with high affinity ligand presented by cortical epithelial cells, the phenotype of the cells we described is similar in some respects to cells that have been reported to occur in vivo. Perhaps the most relevant in vivo studies have been experiments using TCR transgenic mice that develop in the presence of a high affinity ligand for the TCR. Studies with HY-specific TCR transgenic mice demonstrated that a large population of DN TCR⁺ T cells were generated in male mice that was not seen in female mice (27). Similarly, in a class II-restricted TCR transgenic system using D011TCR transgenic mice bred on a H2^b background, large numbers of DN T cells were found in the periphery (28). Because the IA^d-restricted D011 TCR shows alloreactivity to IA^b-expressing APCs (29), it was suggested that DO11 has a higher affinity for IA^b/self than IA^d/self and that this higher affinity interaction led to the appearance of DN T cells in the periphery. Although the authors argued that these DN T cells did not pass through a DP stage, the experiments that reported on the state of methylation of the CD8 gene in the DN T cells were compatible with the interpretation that a significant portion of the DN cells did at one time express CD8.

A third study that used a class II-restricted TCR specific for an antigen (IEα_52–67) expressed in the TCR transgenic mice demonstrated that, along with the expected extensive deletion of DP thymocytes, a significant population of TCR positive DN T cells were generated in both thymus and spleen (30).

In two other situations in which peripheral DN T cells have been described, the derivation of the DN cells from DP thymocytes has been studied. In one, the large population of DN T cells found in the periphery of the MRL^lpr/lpr autoimmune mouse strain were shown to have a hypomethylated CD8 gene indicative of previous expression of CD8 (31). Similarly, DN natural killer (NK) T cells not only have been shown to have a hypomethylated CD8 gene but, in adoptive transfer experiments, were shown to be derived from DP precursor thymocytes (32).

There have been several reports that have described negative regulatory activity to DN peripheral T cells. Recent attention has been directed toward the NK1.1⁺ DN T cell as a regulatory cell (33). These cells are unique in that they express a very restricted TCR repertoire that recognizes glycolipid antigens presented by the class I molecule CD1d (34). They also are capable of secreting large amounts of IL-4 and/or IFN-γ. Although NK T cells drastically differ from the DN T cells described in this paper (i.e., NK1.1⁻, class II MHC-restricted, lack of production of IL-4 and IFNγ), it is possible that the mechanism that generates NK T cells in the thymus is similar to that which we propose for the generation of DN T cells in the RCs: namely, the presentation of high affinity ligand by APCs that are inefficient at causing deletion. First, NK T cells have been shown to be autoreactive, in that they can be stimulated in vitro by syngeneic thymus cells, suggesting the presence of a high affinity autoantigen. Second, CD1d-expressing thymocytes have been implicated as the APCs responsible for positive selection of NK T cells (35), a population that may not be capable of inducing deletion. Thus, it may be that high affinity glycolipid self antigen presentation by CD1d⁺ thymocytes to DP thymocytes leads to the down-modulation of CD4 and CD8, resulting in the generation of DN regulatory NK1.1⁺ T cells.

Finally, there have also been reports of NK1.1⁻ DN regulatory T cells capable of suppression of allograft rejection (36) and graft-vs.-host disease (37). Although these cells have characteristics that appear to distinguish them from NK T cells, whether they represent a distinct population has yet to be established.

In conclusion, we present data that suggest that thymic cortical epithelial cells may contribute to the process of self tolerance, not by causing deletion of autoreactive thymocytes but, rather, when the concentration and affinity of antigen is high, by the induction of DN T cells that could help maintain peripheral tolerance to self antigens in part by the elaboration of IL-10.

Abbreviations

TCR

T cell receptor

RC

reaggregate culture

DN

double negative

DP

double positive

APC

antigen-presenting cell

MCCp

peptide antigen MCC_88–103

CFSE

carboxyfluorescein diacetate succinimidyl ester

ICAM

intercellular adhesion molecule

TEC

thymic epithelial cell

SP

single positive

TUNEL

deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling

PI

propidium iodide

NK

natural killer

OVAp

ovalbumin peptide 323–339