In vivo maintenance of T-lymphocyte unresponsiveness induced by thymic medullary epithelium requires antigen presentation by radioresistant cells

Abstract

The T-cell repertoire developing in the thymus is rid of autospecific cells by the process of thymic negative selection. Recognition of major histocompatibility complex (MHC)/self-peptide complexes expressed by thymic antigen-presenting cells (APC) of bone marrow origin leads to induction of apoptotic death of autospecific thymocytes. Induction of tolerance to self-antigens not presented by thymic APC is mediated by medullary thymic epithelial cells (mTEC) which express a very wide range of proteins, e.g. inducible and tissue-specific proteins. The main type of tolerance induced by mTEC is non-deletional and the issue of how it is maintained outside the thymus is therefore of crucial interest. We have previously shown that the non-T-cell receptor (TCR) -transgenic T-cell repertoire developing in conditions in which tolerance to self-MHC/peptide ligands is exclusively induced by mTEC is tolerant to syngeneic targets in vivo but lyses such targets in vitro. Here we report that this non-deletional in vivo self-tolerance is not due to active tolerance assured by known naturally occurring regulatory or immune-modulating T lymphocytes. Importantly, we show that in vivo maintenance of this therefore probably anergic state requires continued interaction of autospecific T cells with self-MHC/peptide ligands expressed by radioresistant cells while APC are incapable of maintaining the tolerant state. Therefore, maintenance of non-deletional T-lymphocyte tolerance to the wide range of self-antigens expressed by mTEC depends on continued interaction with radioresistant cells that very probably express a much more limited repertoire of antigens. Our data may therefore have important consequences for tolerance to tissue-specific and inducible self-antigens.

Introduction

Thymocytes that have successfully rearranged Tcra and Tcrb genes are submitted to positive and negative selection processes. Positive selection allows useful thymocytes [i.e. those expressing a self-major histocompatibility complex (MHC) restricted T-cell receptor (TCR)] to survive,¹ while negative selection eliminates or functionally inactivates potentially auto-reactive, and therefore dangerous, thymocytes.^2,3 Positive selection requires interaction between the thymocyte's TCR and MHC–peptide complexes expressed on cortical thymic epithelial cells (cTEC) which appear incapable of tolerance induction in vivo.^4–9 Establishment of ‘passive’ tolerance in the thymus is achieved by clonal deletion (i.e. induction of apoptotic cell death) or anergy induction.^2,3 Antigen-presenting cells (APC) of haematopoietic origin have been shown to mediate clonal deletion of potentially autoreactive thymocytes^10,11 whereas medullary thymic epithelial cells (mTEC) can induce anergy and possibly deletion.^12–21 In addition, ‘active’ T-cell tolerance assured by regulatory CD4⁺ CD25⁺ T cells develops in the thymus:²² their positive selection requires MHC class II expression on cTEC (ref. ²³ and our unpublished results), deletion can be mediated by APC,²⁴ while the potential role of mTEC in their differentiation remains unknown.

T-cell tolerance induction by APC and mTEC not only differs in the type of tolerance induced (deletion versus anergy) but also (and very importantly) in the self-antigens involved. APC present endogenous as well as serum antigens but have limited capacity to mediate tolerance to inducible or tissue-specific antigens. While non-responsiveness to the latter type of antigens was thought to be assured by peripheral mechanisms it has recently become clear that mTEC express and present a very large range of proteins and therefore also play an important role in the induction of tolerance to ‘peripheral’ and inducible antigens.^{19,21,25–29}

Since thymic positive and negative selection depends on the recognition of MHC–peptide complexes expressed by distinct cellular compartments these processes can be dissected. In bone marrow chimeras generated by reconstituting irradiated wild-type (wt) mice with MHC-deficient bone marrow (‘MHC I°II° → wt’) (and that are therefore deficient in thymic deletion by APC but not in tolerance induction by TEC) two- to three-fold greater numbers of T cells develop than in wt → wt controls, indicating that normally half to two-thirds of positively selectable thymocytes are deleted before reaching full maturity.¹¹ T cells developing in these chimeras can be activated by syngeneic MHC-expressing APC in vitro while in vivo (in syngeneic hosts) they are incapable of eliminating syngeneic haematopoietic targets.²⁰ A similar state of in vitro versus in vivo ‘split tolerance’ has been observed in P → F₁ chimeras.³⁰ In contrast, T cells generated in transgenic animals expressing MHC class I⁹ or MHC class II molecules^6,8 selectively on thymic cortical epithelium (and therefore deficient in tolerance induction by APC and mTEC) are strongly reactive to syngeneic APC in vitro and in vivo. These data indicated that the in vivo tolerance is due to a non-deletional thymic mechanism critically dependent upon the interaction of auto-specific thymocytes with mTEC.

The tolerogenic capacity of thymic epithelium is known from a variety of transgenic and chimeric systems, and several tolerance mechanisms have been implicated. In TCR-transgenic mice induction of deletional tolerance by thymic epithelium has been observed,^{14,16,17,19,21} but other interpretations of these results have been proposed.^31,32 Superantigen-expression restricted to radioresistant thymic tissue leads to induction of a profound anergic state of specific thymocytes.¹² T cells developing in bone marrow chimeras were tolerant to host-type MHC in vivo while they could be activated with syngeneic APC in vitro, demonstrating that a non-deletional mechanism is involved.^20,30 In the chick–quail chimera system tolerance to antigens expressed by only part of the thymic tissue was observed. In mouse models this tolerance was subsequently confirmed to be active, i.e. mediated by regulatory T cells.³³ Transgenic expression of MHC or antigen specifically on thymic medullary epithelium also induces non-deletional tolerance.^15,34–36

Therefore, most of the current information on tolerance induction by mTEC indicates that non-deletional mechanisms are involved. Importantly, a state of split-tolerance is generally induced by medullary epithelium: T cells are tolerant in vivo but not in vitro. Several reasons for the lack of in vivo reactivity can be envisaged. It has been shown that in certain cases active tolerance assured by regulatory T cells is involved.^2,33 Others have suggested a role for T-cell anergy possibly reversible in vitro but not in vivo.³⁰ It has also been reported that recognition of self-ligands at the surface of mTEC deletes high-affinity cells. The resulting T-cell repertoire would therefore have low-affinity receptors for self-MHC–peptide complexes and be helper T-cell-dependent.¹⁸ Importantly, in some systems injected T cells may be eliminated by natural killer (NK) cells (because of the lack of expression of host-type MHC class I).^11,20,30 In this report we examined the mechanisms responsible for in vivo versus in vitro T-lymphocyte split tolerance to self-MHC–peptide ligands induced by thymic epithelium in MHC I°II° → wt bone marrow chimeras.

Materials and methods

Mice

Wild-type C57BL/6 (B6) and B10.Q mice were purchased from CER Janvier, Le Genest, France. B6 mice deficient in MHC class I and MHC class II expression (MHC I°II°) because of targeted disruptions of the β₂-microglobulin and I-Aβ^b genes were obtained from CDTA, Orléans, France.³⁷

Primary haematopoietic chimeras

Irradiation bone marrow chimeras were prepared as previously described.¹¹ Briefly, anti-NK1.1 antibody-treated hosts (100 μg PK136³⁸, intraperitoneal on days −1 and 0 of reconstitution) were lethally irradiated (850 rads γ) using a ¹³⁷Cs source (700 rads/min) and next day were reconstituted by retro-orbital intravenous injection of 8 × 10⁶−15 × 10⁶ bone marrow cells that had been depleted of T cells and NK1.1⁺ cells using anti-Thy-1 antibody AT83³⁹ and PK136 plus complement (C; Saxon Europe, Cambridgeshire, UK). Chimeras were kept on antibiotic-containing drinking water (0·2% Bactrim, Roche, Basel, Switzerland) for the complete duration of the experiment, usually 6 weeks.

In vivo reactivity of splenic T cells

In vivo reactivity was tested using secondary bone marrow chimeras prepared as previously described.⁹ Briefly, NK-depleted splenocytes (15 × 10⁶−25 × 10⁶) from primary haematopoietic chimeras were injected intravenously into lethally irradiated hosts reconstituted as above except that a 50%/50% mixture of bone marrow cells form B6 and MHC I°II° mice previously treated with PK136 and AT83 plus C was used. In some experiments a 33%/33%/33% mixture of bone marrow cells from B6, B10.Q and MHC I°II° mice was used. Where indicated splenocytes to be transferred were depleted of CD8⁺, CD4⁺, or CD25⁺ cells using 31M⁴⁰, RL172.4⁴¹ or 7D4⁴² monoclonal antibodies (mAb) plus C, respectively. Two weeks after reconstitution, mice were killed and a suspension of their bone marrow cells was prepared. Cells were incubated with saturating concentrations of 2.4.G.2 mAb⁴³ to block Fcγ receptors, then stained with fluorescein isothiocyanate (FITC)-conjugated anti-H-2K^b mAb AF6-88.5 or in some cases FITC-conjugated anti-H-2K^q mAb KH114 (Pharmingen, San Diego, CA). After two washing steps in phosphate-buffered saline containing 2% fetal calf serum and 0·02% NaN₃, cells were analysed by flow cytometry using a Coulter Epics XL 4C flow-cytometer (Coulter, Miami, FL).

Results

T cells that had exclusively been submitted to tolerance induction by TEC lyse allogeneic but not syngeneic targets in vivo

We have previously shown that thymocytes and splenocytes from MHC I°II° → B6 bone marrow chimeras (that therefore had been submitted to thymic tolerance induction by TEC but not by APC) exhibited strong in vitro reactivity toward syngeneic haematopoietic targets while in vivo (in syngeneic recipients) they were tolerant to syngeneic bone marrow grafts.²⁰ Since MHC I°II° → B6 chimera-derived splenocytes do not express MHC class I molecules and are therefore potential targets for host or donor NK cells, we wished to investigate if injected T cells survived functionally. We have developed an experimental system in which in vivo reactivity towards transplanted syngeneic and allogeneic bone marrow cells could be tested in the same mouse. Irradiated B6 hosts were reconstituted with an equilibrated mixture of (T- and NK-depleted) MHC I°II° (to allow survival of hosts), B6 (test) and B10.Q (positive control) bone marrow. The mice were simultaneously injected with MHC I°II° → B6 or B6 → B6 chimera-derived splenocytes. In mice injected with bone marrow cells only B6 and B10.Q cells had reconstituted the host (Fig. 1). When B6 → B6 chimera-derived splenocytes were co-injected, 2 weeks later H-2^q-expressing cells were absent from the bone marrow (Fig. 1), indicating that as expected B6 → B6 chimera-derived T cells had lysed allogeneic B10.Q bone marrow cells. Injection of MHC I°II° → B6 splenocytes led to elimination of B10.Q but not B6 bone marrow. These data confirm the lack of in vivo syngeneic reactivity of MHC I°II° → B6 chimera derived T cells in a novel experimental system and, importantly, indicate that injected T cells functionally survive and had not simply been eliminated by NK cells.

Figure 1.

MHC I°II → B6 chimera-derived T cells reject allogeneic but not syngeneic bone marrow grafts in vivo. Lethally irradiated B6 hosts were reconstituted with a mixture of B6, B10.Q and MHC I°II° cells alone (upper panels) or in combination with either B6 → B6 (middle panels) or MHC I°II° → B6 (lower panels) chimera-derived splenocytes. Reconstitution by B6 and B10.Q bone marrow was assessed 2 weeks later by flow cytometry of bone marrow using anti-H-2K^b (detecting B6 precursors, left panels) and anti-H-2K^q (detecting B10.Q precursors, right panels) antibody.

Neither regulatory CD4⁺ CD25⁺ nor immune deviating CD4⁺ T cells are involved in the non-deletional tolerance induced by mTEC

CD4⁺ CD25⁺ regulatory T cells have been shown to negatively control development of auto-immune disorders and to limit the induction of anti-tumour immunity.^44–46 These cells are known to inhibit CD4⁺ or CD8⁺ T lymphocytes as well as NK cells.^44–48 It has also been suggested that in vivo tolerance can be established by immune deviation by CD4⁺ T lymphocytes.^49,50 Perhaps the most obvious explanation for the observation that (MHC I°II° → wt chimera-derived) T cells that had undergone thymic tolerance induction exclusively by TEC are self-tolerant in vivo but self-reactive in vitro is a potential activity of regulatory or immune-deviating T cells. To test if the in vivo tolerance was due to CD4⁺ CD25⁺ regulatory or CD4⁺ immune deviating T cells we reconstituted lethally irradiated B6 hosts with a 50%/50% mixture of MHC I°II° and B6 bone marrow cells and co-injected MHC I°II° → B6 chimera-derived T cells depleted or not of CD25⁺ or CD4⁺ cells. Even in absence of CD25⁺ or of CD4⁺ cells MHC I°II° → B6 chimera-derived T cells failed to lyse syngeneic haematopoietic targets in vivo (Fig. 2a). In the several experiments performed no significant difference between the percentage of H-2K^b cells surviving in presence of total or CD25- or CD4-depleted MHC I°II° → B6 T cells was observed (Fig. 2b). Similar CD25-depleted CD4⁺ T-lymphocyte preparations induced inflammatory bowel disease in RAG2-deficient mice, demonstrating the functional efficiency of the regulatory T-cell depletion (our unpublished data). Therefore, TEC-induced in vivo T-lymphocyte tolerance is definitively not mediated by the only currently known naturally occurring regulatory T-cell population CD4⁺ CD25⁺ neither by immune deviating CD4⁺ T cells. Combined with the fact that these MHC I°II° → wt chimera-derived cells lyse syngeneic targets in vitro and can therefore not have been deleted these data indicate that the tolerance mechanism responsible is induction of anergy.

Figure 2.

Known regulatory T-lymphocyte populations are not responsible for in vivo tolerance of MHC I°II° → B6 T cells. Lethally irradiated B6 hosts were reconstituted with a mixture of B6 and MHC I°II° B6 bone marrow cells that were co-transferred with MHC I°II → B6 splenocytes. Where indicated the MHC I°II° → B6 splenocytes were depleted of either CD4⁺ (ΔCD4) or CD25⁺ (ΔCD25) cells (by antibody plus C treatment) before transfer. Reconstitution by B6 bone marrow was assessed 2 weeks later by flow cytometry as described in the legend of Fig. 1. (a) A representative experiment is shown. (b) The percentage of H-2K^b bone marrow in the different experimental conditions is indicated for all mice tested. Bars indicate mean values ± standard deviation.

The anergic state of T cells rendered tolerant by TEC is lost in the absence of MHC expression by radioresistant cells

In superantigen and TCR-transgenic systems it has been shown that maintenance of T-lymphocyte anergy requires continuous expression of recognized ligands.^51–53 We therefore investigated if maintenance of anergy in our system (in which a normal repertoire of T cells and MHC–peptide ligands is involved) also required persistence of antigen-presentation. We have previously shown that in vitro culture of MHC I°II° → wt chimera-derived T cells in the presence of syngeneic splenic APC resulted in the activation and induction of cytolytic T-lymphocyte effectors.⁹ This result indicated that despite the presence of recognized ligands the T-cell anergy was not maintained in vivo. We therefore wished to test if in vivo maintenance of TEC-induced anergy required interaction with radioresistant host cells. MHC I°II° → B6 or control B6 → B6 chimera-derived T cells were injected into lethally irradiated B6 or MHC I°II° recipients reconstituted with a 50%/50% mixture of MHC I°II° and B6 bone marrow. In MHC I°II° hosts injected T cells do not encounter self antigens presented by radioresistant cells but only at the surface of bone marrow target cells which we know are incapable of maintaining the anergic state in vitro. As shown in Fig. 3 when transferred into B6 animals the percentages of H-2K^b bone marrow cells reconstituting the host in presence of MHC I°II° → B6 or B6 → B6 splenocytes were similar. Importantly, when MHC I°II° → B6 (but not B6 → B6) chimera-derived T cells were transferred into irradiated MHC I°II° mice (co-injected with B6 plus MHC I°II° bone marrow) no H-2K^b cells reconstituted the host, indicating that B6 cells had been lysed by the T lymphocytes. Therefore, maintenance of the anergic state of injected T lymphocytes required antigen presentation by radioresistant host cells. In contrast, recognition of MHC–peptide ligands at the surface of bone marrow-derived cells did not allow for the maintenance of T-cell anergy in vitro or in vivo.

Figure 3.

MHC I°II → B6-derived T cells reject syngeneic grafts in MHC-deficient hosts. Lethally irradiated B6 or MHC I°II° hosts were reconstituted with a mixture of B6 and MHC I°II° bone marrow cells co-transferred with B6 → B6 or MHC I°II° → B6 chimera-derived splenocytes. Reconstitution by B6 bone marrow was assessed as described in the legend to Fig. 1. Numbers indicate mean values ± standard deviation.

Since we now had an in vivo system available in which autospecific T cells lysed syngeneic targets in vivo we wished to validate our interpretation of the results shown in Fig. 2 that CD4⁺ T cells were not responsible for the in vivo tolerance in B6 hosts. We therefore sought to identify the population of MHC I°II° → B6 chimera-derived T cells responsible for the syngeneic bone marrow rejection in MHC I°II° hosts. MHC I°II° → B6 chimera-derived T cells were depleted with anti-CD4 or anti-CD8 mAb plus complement before transfer into irradiated MHC-deficient hosts simultaneously reconstituted with B6 and MHC-deficient bone marrow. As shown in Fig. 4 the in vivo lysis of B6 bone marrow by MHC I°II° → B6 chimera-derived T lymphocytes was fully abolished when CD8⁺ but not when CD4⁺ T cells were removed. This result indicates that CD8⁺ cells are responsible for the rejection of syngeneic bone marrow grafts in our model and validates our conclusion that CD4⁺ cells do not inhibit in vivo reactivity of MHC I°II° → wt chimera-derived T cells. Importantly, MHC I°II° → B6 chimera-derived CD8⁺ T cells efficiently performed their effector functions independently of CD4-help.

Figure 4.

In MHC I°II° hosts CD8⁺ but not CD4⁺ T lymphocytes lyse syngeneic bone marrow. Lethally irradiated MHC I°II° hosts were reconstituted with a mixture of B6 and MHC I°II° B6 bone marrow co-transferred with MHC I°II° → B6 splenocytes. Where indicated the MHC I°II → B6 splenocytes were depleted of either CD4⁺ (ΔCD4) or CD8⁺ (ΔCD8) cells (by antibody plus C treatment) before transfer. Reconstitution by B6 bone marrow was assessed as described in the legend of Fig. 1.

Discussion

Thymocyte tolerance to extrathymic tissue-specific as well as to inducible antigens is mainly induced by mTEC that are known to induce an anergic state of auto-specific cells. In this report we have shown that peripheral maintenance of this therefore very important T-cell tolerance mechanism requires antigen presentation by radioresistant cells. Surprisingly, bone marrow-derived APC are incapable of maintaining the anergic state and in the absence of antigen presentation by radioresistant cells actually activate auto-specific T cells.

Studies using chimeric mice have indicated that thymic epithelium is capable of T-cell tolerance induction.^30,33 While results with TCR-transgenic models have suggested that thymic epithelial cells can induce deletion of potentially autoreactive thymocytes² this interpretation of the data is contested.^31,32 In contrast, it has convincingly been shown that anergy as well as active tolerance assured by regulatory T cells can be induced by thymic epithelial cells.^2,23 In our MHC I°II° → wt bone marrow chimeras (in which radioresistant thymic epithelial but not radiosensitive bone marrow-derived APC express MHC molecules) CD8⁺ cytotoxic thymocytes or peripheral T lymphocytes (CTL) develop that in vivo are self-tolerant, but that in vitro lyse syngeneic haematopoietic targets.²⁰ These data clearly indicated that reversible (non-deletional) tolerance is induced by thymic epithelial cells. Combined with the fact that in transgenic mice expressing MHC class I or II molecules exclusively on cTEC T cells develop that efficiently lyse syngeneic haematopoietic targets in vivo^6,8,9 our data indicated that medullary epithelial cells are responsible for the induction of the non-deletional in vivo tolerance to haematopoietic targets observed.

In this report we investigated if the reversible non-deletional in vivo tolerance induced by mTEC is active (i.e. mediated by regulatory T cells) or passive (i.e. due to induction of anergy). In vivo active T-cell tolerance has been reported to be assured by regulatory CD4⁺ CD25⁺ as well as by immune-deviating CD4⁺ T lymphocytes. While elimination of CD25⁺ regulatory cells can lead to auto-immunity and immunopathology^45,46 it did not allow the MHC I°II° → wt chimera-derived CD8⁺ T cells to lyse syngeneic targets in vivo in our experimental system indicating that CD4⁺ CD25⁺ regulatory T cells are not responsible for the in vivo tolerance. Depletion of CD4⁺ T cells did not reverse the in vivo self-tolerant state of the auto-specific CD8⁺ CTL neither, eliminating a potential role for immune-deviating CD4⁺ T cells. However, a role for CD25^– or CD4^– regulatory T cells, such as certain NKT⁵⁴ or CD8⁺⁵⁵ cells, cannot be excluded. Since the in vivo self-tolerance of MHC I°II° → wt chimera-derived T cells is reversible in vitro it must be due either to anergy or to dominant in vivo tolerance. Our data indicating that the only naturally occurring regulatory T-cell subsets known to date are not involved strongly suggest a role for anergy. Since in our system tolerance of a normal T-cell repertoire to physiological MHC/peptide ligands is assessed these data substantially extend previous data on thymic epithelium-induced tolerance to superantigens¹² and of TCR-transgenic T cells.^14,34

It has been suggested that MHC–peptide ligand expression by thymic epithelium can cause deletion of (high-affinity/avidity) helper-independent T-cells.^13,18 The remaining helper-dependent (low-affinity/avidity) cells could potentially be activated if T-cell help were delivered in trans. However, in thymus-transplanted mice grafting allogeneic skin did not allow for rejection of skin grafts expressing the MHC haplotype of the thymic epithelium.¹⁸ When we grafted a mixture of fully allogeneic, autologous, and MHC-deficient bone marrow in irradiated mice that had received MHC I°II° → wt chimera-derived T cells only the allogeneic bone marrow was rejected. However, the vigorous alloreactivity did not affect the survival of host-type MHC expressing bone marrow (and therefore the in vivo activity of autospecific CTL). Moreover, as discussed below, in MHC-deficient hosts MHC I°II° → wt derived CTL lyse syngeneic targets in the absence of CD4⁺ T lymphocytes, confirming their helper-independence. Therefore, our data and those of Hoffmann¹⁸ are inconsistent with the hypothesis that helper-dependent T cells are specifically deleted in mice expressing tolerizing ligands exclusively by mTEC. These data also convincingly demonstrate that the injected T cells functionally survive. This is an important issue since the injected T cells do not express MHC class I molecules that inhibit lysis by persisting host NK cells.

The absence of a role for active tolerance mediated by CD4⁺ or CD25⁺ cells strongly suggests that anergy is responsible for the in vivo MHC I°II° → wt chimera-derived CTL tolerance to syngeneic targets. A role for thymic deletion is excluded by the fact that peripheral CTL lyse syngeneic haematopoietic targets in vitro (and in vivo in MHC-deficient mice, see below). The fact that in transgenic mice expressing MHC class I or II molecules in the thymus exclusively on cTEC no in vivo tolerance to syngeneic haematopoietic targets in syngeneic wt hosts has been observed^6,8,9 indicates that anergy is induced by mTEC and excludes a critical role for exclusively peripherally induced tolerance (e.g. deletion). However, given the slight differences in the experimental systems used in our present and previously published reports^9,20 we cannot rule out formally the possibility that thymic medulla and parenchymal cells collaborate in the induction phase of the tolerance studied. For superantigen specific CD4⁺ as well as for TCR-transgenic CD8⁺ T cells it has previously been shown that the maintenance of an anergic state of autospecific T lymphocytes requires continued interaction with antigen.^51–53 However, no data concerning anergy maintenance of CTL specific for the physiological wide range of self-MHC–peptide complexes exist. Neither is it known if radiosensitive APC and/or radioresistant compartments are capable of delivering the anergy-maintaining signals in vivo. In this report we have shown that MHC I°II° → wt bone marrow chimera-derived CTL efficiently and specifically lyse syngeneic bone marrow cells in MHC deficient hosts. Therefore maintenance of anergy of CTL specific for haematopoietic (radiosensitive) cells requires MHC/peptide expression by radioresistant cells. Moreover, in the absence of MHC expression by radioresistant cells the recognition of TCR ligands expressed by bone marrow-derived APC does not allow for maintenance of anergy but rather activates anergic autospecific CTL. Therefore, differences between physiological properties of radioresistant versus radiosensitive cells other than the peptides they present are responsible for the fact that the former cells (but not the latter) are capable of maintaining T-cell anergy. Interestingly, peripheral T-cell tolerance to parenchymal antigens cannot be induced by parenchymal cells alone but requires presentation by APC.⁵⁶ Therefore, the lack of expression of co-stimulatory molecules by most radioresistant cells is probably not solely responsible for their role in the maintenance of the T-cell tolerance induced by mTEC in our system.

The anergy-maintaining radioresistant cells may be intra- and/or extrathymic. Since in our experimental model the analysed autoreactivity is fully attributable to CD8⁺ CTL the (MHC class I/peptide) ligands required for anergy maintenance may be expressed by a wide variety of radioresistant cell types. If the requirement of antigen presentation by radioresistant cells is also valid for the maintenance of CD4⁺ T-cell anergy, the identity of peripheral radioresistant (MHC class II expressing) anergy-maintaining cells is less obvious.

In both cases only a small proportion of the very large repertoire of antigens expressed by mTEC²⁹ is probably expressed by extrathymic radioresistant cells to which naive T cells have access: in contrast to neonatal mice, in adult mice naive T cells do not circulate in tissues.^57,58 How anergy can be maintained for T cells specific for tissue-specific antigens therefore remains an important issue. It has been shown that recirculation to the thymus is limited to T cells with an activated phenotype^59–61 which would be consistent with the possibility that anergic T cells maintain their unresponsive state through recirculation into the thymus.^62,63 Given our results indicating that maintenance of mTEC-induced anergy requires ligand recognition at the surface of radioresistant cells, the physiological role of T-cell recirculation to the thymus merits further investigation (e.g. by using athymic, MHC transgenic, or thymus-transplanted secondary hosts).

T-cell tolerance is maintained by several mechanisms which are partly redundant. However, the fact that development of the distinct types of tolerance depends on different thymic compartments [differentiation of regulatory T cells by cTEC, anergy induction by mTEC, and deletion (mainly) by APC] which express and present an overlapping but definitively not identical antigen repertoire,^29,36,64,65 indicates the importance of each of the individual mechanisms. Our data indicating that maintenance of CTL-tolerance induced by mTEC (which express a wide range of antigens) critically depends upon continued recognition of antigen presented by radioresistant cells to which T cells have access (probably expressing a much more limited repertoire of antigens in adult mice) therefore may have important implications for autoimmunity.

Abbreviations

APC

antigen-presenting cells

B6 C57BL/ C

complement

cTEC

cortical thymic epithelial cells

CTL

cytotoxic T lymphocytes

MHC I°II°

major histocompatibility complex class I and class II deficient

mTEC

medullary thymic epithelial cells

NK

natural killer

TCR

T-cell receptor

TEC

thymic epithelial cells

wt

wild-type