Mode of Tolerance Induction and Requirement for Aire Are Governed by the Cell Types That Express Self-Antigen and Those That Present Antigen

Clonal deletion of autoreactive thymocytes and the production of regulatory T cells (Tregs) are the two major components required for establishment of self-tolerance in the thymus (1). Since the discovery of Aire as an essential gene responsible for tolerance induction, the role of medullary thymic epithelial cells (mTECs) has become the focus of intense research in the field (2–5).

Aire is one of the best-characterized transcriptional regulators, controlling the tissue-restricted self-antigen (TRA) expression of mTECs through several molecular pathways (3, 5–7). However, mTECs are not the only cell type contributing to TRA expression through the action of Aire within the thymus. Recent studies have demonstrated that Aire functions in thymic B cells for control of TRAs that are nonoverlapping with those from mTECs (8). Furthermore, Aire has been suggested to contribute to self-tolerance outside the thymus (9): a unique subset of EpCAM⁺ cells of hematopoietic-cell origin in the spleen and lymph nodes (i.e., extrathymic Aire-expressing cells [eTACs]) also express Aire, together with many TRAs whose spectrum also differs from that of mTECs for inactivation of autoreactive CD4⁺ T cells (10).

Regarding the tolerogenic action of Aire at the cellular level, it has been established that Aire plays an essential role in both clonal deletion of autoreactive T cells (11–13) and production of Tregs (14, 15). Interestingly, a recent study has also suggested that Aire enforces immune tolerance by directing autoreactive T cells toward the Treg lineage (16). Another action of Aire in inhibiting the production of autoimmune pathogenic IL-17–producing γδ T cells has also emerged (17). Given that Aire controls multitolerogenic events by acting within multiple cell types, understanding the precise function of Aire in each cell type is essential to gain an overall picture of Aire’s role in immune tolerance.

Because of the complexity of Aire-dependent multitolerogenic events involving different cell types that express TRAs (i.e., self-antigens) and several Aire-expressing APC types for tolerance induction both within and outside the thymus, it would be important to determine the relative contribution of TRAs from mTECs and bone marrow (BM)–derived cells to tolerance induction. Furthermore, there is a need to establish the precise identity of the thymic APCs responsible for the two major tolerogenic events (i.e., clonal deletion and Treg production), especially in relation to the requirement for Aire.

To analyze these Aire-dependent tolerogenic events, we have employed two different types of transgenic mouse (Tg) model in which different cell types express a prefixed model self-antigen of OVA and different Aire-expressing APCs are involved in the tolerance induction. In the first Tg model, expression of OVA was controlled by the rat insulin promoter (RIP–OVA Tg) (18). Because the insulin promoter is active in mTECs, and clonal deletion of OVA-specific T cells is Aire dependent, this Tg model has provided many insights into Aire-mediated immune tolerance (12, 13). However, the precise identity of the cell types that express self-antigens and details of the mode of tolerance induction by each thymic APC have not been fully addressed: in a previous study employing a Tg model in which hen egg lysozyme was expressed under the same rat insulin promoter, Aire-dependent clonal deletion of clonotypic T cells was also demonstrated (11).

The second model we employed was a Tg in which the model self-antigen of OVA was driven by the Aire promoter itself. This was achieved by insertion of the OVA gene into the Aire locus by homologous recombination in embryonic stem cells. Using this approach, we were able to strictly limit the expression of OVA by the mTECs and BM-derived APCs that expressed Aire. Of note, in both Tg models, OVA was tethered to the cell membrane using the transferrin receptor transmembrane domain (18) to minimize the effect of blood-borne soluble Ag. Our results suggested that mTECs and BM-derived APCs play a differential role in tolerance induction (i.e., clonal deletion versus production of Tregs) depending on the cell types that express self-antigen. Furthermore, the requirement for Aire was dependent on the cell types expressing self-antigens and/or the types of APCs responsible for the tolerogenic events. These results suggested an essential but nonuniversal role of Aire in tolerance induction.

Aire/OVA–knock-in (KI) mice were generated by homologous recombination in embryonic stem cells established from C57BL/6 (B6) mice using the same design as the generation of Aire/GFP-KI mice (19), with the exception of the KI cassette: OVA was tethered to cell membrane using the transmembrane domain of transferrin receptor (18). After the targeted cells had been injected into morula-stage embryos, the resulting chimeric male mice were mated with B6 females (CLEA Japan) to establish germline transmission. The mice were then crossed with a transgenic line expressing the general deleter Cre recombinase to remove the neo^r gene cassette. TCR Tg mice specific for OVA in the context of I-Ab (OT-2 Tg) mice (20) and mice deficient for B cells (21) were from The Jackson Laboratory. Mice deficient for MHC class II (MHC-II) (22) are from Taconic. Aire-deficient mice on a B6 background (23), Aire/GFP-KI mice (19), and CD11c-diphtheria toxin A (DTA) Tg mice (24) were generated as described previously. Transgenic mice expressing OVA under control of the rat insulin promoter (RIP-OVA Tg) were kindly provided by Dr. M.J. Bevan (25). The mice were maintained under pathogen-free conditions. The protocols used in this study were in accordance with the Guidelines for Animal Experimentation of Tokushima University School of Medicine.

Preparation of TECs and flow-cytometric analysis with a FACSCalibur (BD), a FACSAria II (BD), and a Gallios (Beckman Coulter) were performed as described previously (19, 26). The mAbs used were anti-CD4, anti-CD8α, anti-CD11c, anti-CD24, anti-CD45, anti-CD62L, anti-CD103, and anti-Vα2 mAb, all purchased from BioLegend. Anti-CD19, anti-CD25, anti-CD80, anti-B220, anti-Foxp3, and anti-Qa2 mAb were from eBioscience. Anti-Vβ5 and anti-Sirpα mAb were from BD, and Ulex europaeus agglutinin 1 was from Vector Laboratories.

Tissues were harvested after perfusion of mice and fixation with tissue fixative. Paraffin-embedded blocks were cut into sections 6 μm thick. Before in situ hybridization, the tissue sections were de-waxed with xylene and rehydrated by repeated rinses in ethanol and PBS. They were then fixed with 4% paraformaldehyde in PBS for 15 min, and washed with PBS. The sections were treated with 15 μg/ml Proteinase K in PBS for 30 min at 37°C, followed by washing with PBS. After refixing with 4% paraformaldehyde in PBS, they were placed in 0.2 N HCl for 10 min. After washing with PBS, the sections were acetylated by incubation in 0.1 M triethanolamine-HCl (pH 8.0) containing 0.25% acetic anhydride for 10 min. After washing with PBS, the sections were dehydrated through a series of ethanol washes. Hybridization was performed with a probe spanning a 664 bp stretch of OVA cDNA at a concentration of 300 ng/ml at 60°C for 16 h. After hybridization, the sections were washed in 5 × SSC at 60°C for 20 min and then in 50% formamide, 2 × SSC at 60°C for 20 min, followed by RNase treatment with 50 μg/ml RNase A in 10 mM Tris-HCl (pH 8.0) at 37°C for 30 min. The sections were then washed with 2 × SSC at 60°C for 20 min, 0.2 × SSC at 60°C for 20 min, and with TBST (0.1% Tween 20 in TBS). After treatment with 0.5% blocking reagent (Roche) in TBST for 30 min, the sections were incubated with anti–digoxigenin-alkaline phosphatase, Fab fragment (Roche) diluted 1:1000 with TBST for 2 h at room temperature. The sections were washed with TBST, and then incubated in 100 mM Tris-HCl (pH 9.5) containing 0.1% Tween 20. Coloring reactions were performed with NBT/BCIP solution (Sigma-Aldrich) overnight, and then washed with PBS. The sections were counterstained with Kernechtrot stain solution (Mutoh), and mounted with CC/Mount (Diagnostic BioSystems).

BM transfer and thymus grafting onto BALB/c nude mice (CLEA Japan) or other recipient mice were performed as previously described (23, 27). In brief, thymic lobes were isolated from embryos at 14.5 d postcoitus. The lobes were then cultured for 4 d on Nucleopore filters (Whatman) placed on culture media containing 1.35 mM 2′-deoxyguanosine (Sigma-Aldrich) to eliminate BM-derived cells within the thymus prior to the graft. The recipient mice were used for analyses 6–10 and 5 wk after BM transfer and thymus grafting, respectively.

Preparation of thymic slices and imaging using a two-photon laser microscope were performed as described previously (28). In brief, a thymic lobe was embedded on a handmade plastic pedestal with low-gelling agarose (Sigma-Aldrich). The thymic slice was cut every 400 μm using a vibratome (DSK, Kyoto, Japan), and incubated with culture medium (IMDM containing 4% FCS, 50 μM 2-ME, and antibiotics) for a few minutes at room temperature. The slices were placed onto a Millicell insert (30 mm organotypic PTFE; Millipore) in a 3.5 mm plastic dish filled with 1 ml of culture medium, and then enclosed using silicone grease. The thymocytes were labeled with 1 μM CMTMR (Invitrogen) in 1 ml per 1 × 10⁷ of RPMI 1640 containing 1% FCS for 25 min at 37°C. The labeled CD4⁺ OT-2 T cells (1 × 10⁵ cells) were suspended in culture medium and loaded onto slices, and the cells on the slice were incubated for more than 3 h at 37°C/5% CO₂ to allow cells to enter the tissue. Two-photon imaging was performed using a FV1000 upright microscope (Olympus) with a Ti:sapphire laser (MaiTai HP DeepSee-OL, Spectra-Physics) tuned to 890–900 nm and a 25×/1.05 NA (XLPLN25XWMP; Olympus) objective. The emitted light was passed through 495−540 nm and 575−630 nm bandpass filters to PMT for detection of CFSE (EGFP) and CMTMR fluorescence. Acquisition was controlled with FV10-ASW software (Olympus), and Z stacks were typically acquired every 15−20 s, with Z steps spaced 2−2.5 μm apart. Imaging was performed 10−100 μm beneath the cut surface of the slice under a flow (1–2 ml/min) of RPMI 1640 medium bubbled with 95% O₂/5% CO₂ at 37°C. Volocity (Improvison) was used for analyzing cell positions over time in three dimensions. Image stack sequences were transformed into volume-rendered three-dimensional images or movies.

RNA was extracted from the total thymus with RNeasy Mini Kits (Qiagen) and made into cDNA with SuperScript III RT Kits (Invitrogen) in accordance with the manufacturer’s instructions. Real-time PCR for quantification of the OVA and Hprt gene was performed as described previously (29, 30).

All results are expressed as mean ± SEM. Statistical analysis was performed using Student two-tailed unpaired t test for comparisons between two groups. Differences were considered significant for p values ≤0.05.

The first model we employed for assessing the fate of autoreactive T cells was MHC-II–restricted/OVA-specific OT-2 TCR-Tg (OT-2 Tg) in combination with RIP-OVA Tg (18, 20, 25). In this model, clonal deletion of OT-2 T cells specific for OVA [i.e., clonotypic CD4-single-positive (SP) thymocytes] in the thymus occurred in an Aire-dependent manner when the deletion was induced either by crossing the two strains (Fig. 1A) or by transferring the BM cells from OT-2 Tg into irradiated RIP-OVA Tg recipients (Fig. 1B).

Production of Ag-specific Tregs depends on the recognition of self-antigens with a rather high affinity (31, 32). Indeed, we observed the production of clonotypic Tregs in OT-2 Tg only when they were crossed with RIP-OVA Tg, and this process was also Aire dependent (Fig. 1C). Aire-dependent production of clonotypic Tregs was also observed in BM chimeras (Fig. 1D), clearly indicating that Aire controls the process of both clonal deletion and production of Tregs in OT-2 Tg when the self-antigen was expressed under the control of the insulin promoter, which is active in the thymic stroma (33).

One caveat for such Aire-dependent tolerogenic events using RIP-OVA Tg is that the cell types present among TECs and/or the BM-derived cells expressing the model self-antigen of OVA are not precisely known. It is also unknown which types of APC are involved in the cognate interaction between OT-2 T cells and OVA/MHC-II complex. In this regard, it also remains controversial whether the level of OVA expression in mTECs is reduced (12) or preserved (13) by the lack of Aire in RIP-OVA Tg. This issue is particularly important when attempting to clarify whether Aire dictates central tolerance primarily by controlling the expression level of self-antigens, or by some additional means. To examine these issues, we first evaluated OVA expression in thymi from Aire-sufficient and Aire-deficient RIP-OVA Tg using in situ hybridization. Detection of sense OVA mRNA from the thymi using an anti-sense probe showed similarly scattered signals from the medulla in both Aire-sufficient and Aire-deficient RIP-OVA Tg (Supplemental Fig. 1A). Unexpectedly, however, we observed much more abundant signals from the thymi of both sets of mice using a sense OVA probe (Supplemental Fig. 1A), probably due to the transgenic nonphysiological expression of anti-sense OVA mRNA. Although the significance of this finding is unclear, these aberrant mRNAs would be irrelevant to the OVA-specific immune response unless they complementarily bind to sense OVA transcripts and regulate their activity (34). Aberrant expression of the anti-sense OVA transcript was not observed in pancreatic β-islets (Supplemental Fig. 1A).

We then measured the levels of OVA expression in several thymic cell populations including mTECs and BM-derived thymic APCs [i.e., plasmacytoid dendritic cells (DCs), conventional DCs (cDCs), and B cells] by real-time PCR in either the presence or absence of Aire (Fig. 1E). Because of the presence of a large amount of anti-sense OVA mRNA species as described above, we focused on the sense OVA product by performing the reverse transcription using an anti-sense primer set within the OVA sequence: both the sense and anti-sense OVA transcripts had been measured at least in one previous study (12) because random primers were used for the reverse transcription. The mTEC^low population from Aire-sufficient RIP-OVA Tg showed the highest expression of the sense OVA transcript, followed by the mTEC^high population, whose expression level was one third of that of the mTEC^low population. Because Aire expression is confined to the mature mTEC^high population (19, 35), there was no clear correlation between the levels of OVA expression and Aire expression from each mTEC subpopulation in RIP-OVA Tg. Additionally, OVA expression in thymic plasmacytoid DCs, thymic cDCs, and thymic B cells was close to the levels observed in the mTEC^high population. CD4SP thymocytes from RIP-OVA Tg, but not from non-Tg littermates, also expressed OVA at levels comparable to that in thymic APCs, probably due to leaky promoter activity (Supplemental Fig. 1B). Lack of Aire had no significant impact on the expression of sense OVA mRNA in all cell types (Fig. 1E). Thus, Aire-dependent clonal deletion and production of Tregs may not be solely governed by the transcriptional level of OVA. Although these results are consistent with one previous report (13), it was rather unexpected in view of the fact that Aire controls the expression levels of many TRAs from mTECs including the insulin 2 gene (36).

In addition to the cell types that express OVA, it is also necessary to clarify which APCs (i.e., mTECs and/or BM-derived APCs) are responsible for inducing clonal deletion of OT-2 T cells in RIP-OVA Tg. Because our real-time PCR analysis demonstrated a rather abundant OVA sense transcript in thymic DCs and thymic B cells (Fig. 1E), and these BM-derived APCs express high levels of the MHC-II molecule (Y. Mouri and Mitsuru Matsumoto, unpublished observations), the functional significance of this finding needs to be determined. Therefore, we investigated whether these BM-derived APCs can directly delete OT-2 T cells using OVA produced themselves. To do this, we transferred BM cells from double Tg (OT-2 Tg crossed with RIP-OVA Tg) into wild-type recipient mice. In this experimental setting, BM-derived cells are the only source of OVA in the thymus. We observed neither clonal deletion of OT-2 T cells nor production of clonotypic Tregs in thymi from the recipient mice (Fig. 1F); this situation was in marked contrast to mice expressing OVA under the control of the Aire promoter (i.e., Aire/OVA-KI mice) (see below). The results suggested that self-antigen (OVA) expressed by BM-derived APCs in the thymus does not contribute to central tolerance in a cell-autonomous manner in RIP-OVA Tg. Instead, OVA produced by mTECs is the most plausible source of the tolerance induction in this model.

If OVA expressed by BM-derived APCs is not directly involved in the tolerogenic events for OT-2 T cells in RIP-OVA Tg, it is important to know exactly how OVA produced by the mTECs mediates tolerance induction. One study has suggested that mTECs directly present OVA as APCs to induce clonal deletion of OT-2 T cells (12). However, another study has concluded that mTECs expressing OVA do not directly contribute to the clonal deletion of OT-2 T cells (25). In the latter case, BM-derived APCs (e.g., thymic DCs and/or thymic B cells) are responsible for the presentation of OVA transferred from mTECs (37). Because both studies transferred MHC-II–deficient OT-2 BM cells into irradiated RIP-OVA Tg recipients to eliminate the contribution of BM-derived APCs to the tolerogenic events in BM-chimeras, one possible reason for the discrepancy in the results of the two studies might have been the presence of radio-resistant MHC-II–sufficient BM cells derived from the recipient animals. These residual MHC-II–sufficient BM cells may pick up OVA and present it to OT-2 T cells for clonal deletion.

To eliminate this possibility, we performed a thymus graft experiment. We grafted fetal thymi (FT) from RIP-OVA Tg (MHC-II sufficient) into MHC-II–deficient OT-2 Tg. Because of the lack of MHC-II expression on any APCs (i.e., BM-derived APCs together with TECs from endogenous thymi) in the recipient mice, the stromal components (i.e., TECs) of the grafted thymi are the only cell types that possess MHC-II expression, thereby contributing to the positive selection of OT-2 T cells that is followed by negative selection.

To confirm the Ag specificity of this experimental system, we first grafted thymi from wild-type mice or RIP-OVA Tg into MHC-II–sufficient OT-2 Tg. We observed clonal deletion of OT-2 T cells when the grafted thymi were from RIP-OVA Tg, but not from wild-type mice, as expected (Fig. 2A, center of the left panel, 2B). Knowing this, we then grafted thymi from wild-type mice or RIP-OVA Tg onto MHC-II–deficient OT-2 Tg. We observed no clonal deletion of OT-2 T cells in the thymi from RIP-OVA Tg (Fig. 2A, center of the right panel, 2B), suggesting that mTECs from RIP-OVA Tg alone were not sufficient for induction of clonal deletion of OT-2 T cells. Instead, BM-derived APCs [which could not induce clonal deletion using OVA produced themselves (Fig. 1F)] were required for the clonal deletion.

Using this system, we then examined the role of Aire in the clonal deletion. When BM-derived APCs were able to participate in tolerance induction (i.e., when thymi were grafted into MHC-II–sufficient OT-2 Tg), lack of Aire resulted in impairment of clonal deletion (Fig. 2A, right of the left panel, 2B), recapitulating the results described above (Fig. 1A, 1B). When MHC-II–deficient OT-2 Tg were used as recipients (i.e., when BM-derived APCs were unable to contribute to the tolerogenic events), clonal deletion of OT-2 T cells did not occur, irrespective of Aire expression (Fig. 2A, right of the right panel, 2B). Thus, BM-derived APCs were required for the Aire-dependent clonal deletion, most likely through Ag transfer from mTECs.

To gain insights into how Aire in mTECs affects clonal deletion of Ag-specific T cells through Ag transfer, we visualized the process of cognate interaction between mTECs and OT-2 T cells. In combination with Aire/GFP-KI mice in which Aire⁺ mTECs were marked with GFP (Aire^+/gfp), we have previously demonstrated that Aire⁺ mTECs from RIP-OVA Tg had direct interactions with OT-2 T cells (28). We investigated whether these interactions might be affected by the absence of Aire in mTECs by preparing RIP-OVA Tg in which the endogenous Aire locus was disrupted by homozygous insertion of the GFP gene (RIP-OVA Tg/Aire^gfp/gfp). FACS-sorted CD4⁺ OT-2 T cells were mounted on thymic slices from RIP-OVA Tg/Aire^+/gfp or RIP-OVA Tg/Aire^gfp/gfp, and the interaction between mTECs and OT-2 T cells was assessed using a two-photon laser microscope (Supplemental Fig. 2A) (28). In terms of the number of clusters (Supplemental Fig. 2B) and the duration of contact (Supplemental Fig. 2C), we found that OT-2 T cells were able to attach to Aire-marked mTECs, irrespective of Aire protein expression. Although there was a tendency for cluster numbers to be decreased in RIP-OVA Tg/Aire^gfp/gfp, precise comparison of cluster numbers between RIP-OVA Tg/Aire^gfp/gfp and RIP-OVA Tg/Aire^+/gfp was hampered by the morphological changes in mTECs due to their lack of Aire (19) and/or differences in the intensity of the GFP signals depending on the heterozygosity or homozygosity of the Aire/GFP-KI allele. Thus, we consider that Aire does not have a major impact on the mechanical interaction between mTECs and OT-2 T cells.

Recently, Batf3-expressing CD103⁺ DCs have been implicated in Ag transfer from mTECs to BM-derived APCs, with this action being Aire dependent (38). Therefore, we investigated the composition of BM-derived APCs in the thymi from Aire-deficient mice. Although thymic B cells were reduced in Aire-deficient mice (Supplemental Fig. 3A), the composition of DCs including Sirpα⁺ and Sirpα⁻ DCs remained unchanged without Aire (Supplemental Fig. 3B). When Aire-sufficient and Aire-deficient OT-2 Tg were compared, the overall composition of DCs was also unchanged, except for a reduction in CD103⁺CD11b⁻ DCs (Supplemental Fig. 3C). Thus, alteration in the composition of thymic DCs does not seem to be a major reason for the defect in Ag transfer in Aire-deficient mice.

We also examined the contribution of BM-derived APCs to the production of Tregs using a thymus graft experiment. In the presence of OVA expression from Aire-sufficient RIP-OVA Tg, production of Tregs was markedly reduced when the thymi were grafted onto MHC-II–deficient OT-2 Tg compared with the recipients of MHC-II–sufficient OT-2 Tg (Fig. 2C, centers of the left and right panels, 2D), suggesting an important role of BM-derived APCs in the induction of Ag-specific Tregs. However, it is also important to mention that small but significant numbers of clonotypic Tregs were produced in the thymi from Aire-sufficient RIP-OVA Tg when grafted into MHC-II–deficient OT-2 Tg (Fig. 2C, compare left and center of the right panel, 2D). Thus, in contrast to the clonal deletion, mTECs (including Aire⁺ mTECs) were able to directly induce Ag-specific Tregs without any contribution from BM-derived APCs, although the efficiency was rather low. These results suggested a differential contribution of mTECs to the clonal deletion and production of Tregs in RIP-OVA Tg. It was noteworthy that lack of Aire resulted in impaired production of Tregs in both the presence and absence of cooperation with BM-derived APCs (Fig. 2C, 2D). Thus, Aire in mTECs controlled the production of Tregs, and the effect of Aire on this process was more evident when BM-derived APCs were able to participate.

Given that BM-derived APCs play an essential role in clonal deletion through Ag transfer, we investigated whether B cells are an essential component in this process. For this purpose, double Tg (RIP-OVA Tg crossed with OT-2 Tg) were further crossed onto B cell–deficient mice [Ighm-deficient mice (21)]. Even in the absence of B cells, we observed clonal deletion of OT-2 T cells (Fig. 3A, top and middle). Production of Ag-specific Tregs was also not affected (Fig. 3A, bottom). Similarly, when thymic DCs were depleted by utilizing the CD11c-DTA Tg system (24) in double Tg, we observed no impairment of clonal deletion and Treg production (Fig. 3B), although the efficiency of DC depletion with this system may not be maximal. These results suggest a redundant role of thymic B cells and thymic DCs in receiving OVA from mTECs and for presenting it for clonal deletion and Treg production. However, proof of this hypothesis would require the generation of RIP-OVA Tg lacking both B cells and DCs, although the significance of other thymic APCs such as macrophages still remains unaddressed in this experimental setting.

MHC-II⁺ eTACs derived from BM functionally inactivate CD4⁺ T cells through a mechanism that does not require Tregs (9, 10). Beyond eTACs, Aire is also expressed in other BM-derived APCs such as B cells (8) and DCs, although Aire expression in the latter remains controversial (39, 40). We investigated whether Aire in eTACs and/or other BM-derived APCs including thymic B cells and thymic DCs play a role in the clonal deletion and/or production of Tregs in RIP-OVA Tg by transferring Aire-deficient OT-2 BM cells into RIP-OVA Tg. Both clonal deletion (Fig. 3C, top and middle) and production of Tregs (Fig. 3C, bottom) were not affected by the absence of Aire in BM-derived cells.

The results obtained so far using RIP-OVA Tg in combination with OT-2 Tg suggested that BM-derived APCs, irrespective of their Aire expression, are essential for clonal deletion, whereas BM-derived APCs are not an absolute requirement for production of Tregs. Remarkably, Aire in mTECs plays a critical role in both clonal deletion and production of Tregs. We then investigated whether these findings would also be applicable to a different type of Tg model in which cell types expressing self-antigen and/or those responsible for Ag presentation differ from those of RIP-OVA Tg. For this purpose, we established a Tg model in which OVA expression was confined to Aire-expressing cells. This was achieved by inserting a membrane-bound form of the OVA gene, as for RIP-OVA Tg (18), into the Aire locus with concomitant disruption of Aire expression (Aire/OVA-KI; Supplemental Fig. 4A, 4B). In contrast to RIP-OVA Tg, the expression pattern of OVA faithfully reflected the expression of Aire itself (Fig. 4A, left): mTEC^high expressed a much higher level of OVA than did mTEC^low. Direct comparison of the expression levels of OVA in mTECs from Aire/OVA-KI together with RIP-OVA Tg highlighted the differential pattern of OVA expression by mTECs between the two strains (Fig. 4A, right). Consistent with a recent report, thymic B cells (8) together with cDCs also expressed a rather tiny amount of OVA in comparison with mTEC^high.

When Aire/OVA-KI were crossed with OT-1 Tg (Fig. 4B) and OT-2 Tg (Fig. 4C), we observed clonal deletion of clonotypic CD8⁺ and CD4⁺ T cells, respectively. Because we disrupted the endogenous Aire locus by insertion of the OVA gene in Aire/OVA-KI, mice harboring one copy of the OVA gene together with one Aire-null allele (Aire^−/OVA) lacked expression of the Aire protein. When Aire^−/OVA were crossed with OT-1 Tg or OT-2 Tg, we observed clonal deletion of OT-1 or OT-2 cells, respectively (Fig. 4B, 4C, top and middle), to an extent similar to that observed for Aire^+/OVA, suggesting that Aire was dispensable for the clonal deletion in this model. These results implied that lack of Aire does not affect the processing and/or presentation of a prefixed self-antigen of OVA that is required for clonal deletion. In contrast, production of Ag-specific Tregs (Foxp3⁺ clonotypic OT-2 T cells) was reduced by the lack of Aire (Fig. 4C, bottom).

We observed more robust clonal deletion in Aire/OVA-KI than that in RIP-OVA Tg (compare Fig. 1A, bottom, and Fig. 4C, middle) after crossing with OT-2 Tg. Although the overall frequency of Tregs among clonotypic CD4⁺ T cells was similar in two Tg models (compare Fig. 1C and Fig. 4C, bottom), the expression levels of TCR-Vβ5 in Foxp3⁺ cells were lower in Aire/OVA-KI than in RIP-OVA Tg (Fig. 4D), suggesting that the fate of autoreactive T cells in Aire/OVA-KI is directed more toward clonal deletion than to Treg development.

In addition to the difference in the cell types expressing OVA as a source of self-antigen, different types of APCs (i.e., mTECs and BM-derived APCs) might account for the differential requirement for Aire in the clonal deletion of autoreactive T cells in two Tg models. Because thymic B cells expressed Aire and Aire⁺ B cells were able to directly present model self-antigens driven by the Aire promoter in the BAC Tg system (8), we first examined the contribution of BM-derived APCs to the clonal deletion in our Aire/OVA-KI. For this purpose, BM cells from double Tg (Aire/OVA-KI crossed with OT-2 Tg) were transferred into wild-type recipient mice to eliminate the contribution of OVA produced by mTECs as a source of self-antigen, as demonstrated in Fig. 1F. We observed clonal deletion of OT-2 T cells (Fig. 5A, center) and production of Ag-specific Tregs with small but significant numbers (Fig. 5B, center) in the thymi of recipient mice, indicating that Aire (OVA)-expressing BM-derived APCs were able to directly induce both clonal deletion and production of Tregs, in marked contrast to the RIP-OVA Tg model (Fig. 1F). Of note, lack of Aire in BM-derived APCs (i.e., BM cells from Aire^−/OVA crossed with OT-2 Tg) did not impair the clonal deletion (Fig. 5A, right) or production of Tregs (Fig. 5B, right) in the wild-type recipients.

Unexpectedly, a lack of B cells also did not affect the clonal deletion (Fig. 5C) and production of Tregs (Fig. 5D) in this setting: BM cells from double Tg (Aire/OVA-KI crossed with OT-2 Tg) that had been further crossed onto B cell–deficient mice [Ighm-deficient mice (21)] showed clonal deletion and production of Tregs to a level comparable to that seen in an Ighm-sufficient background (compare Fig. 5A, 5B and 5C, 5D). Although our real-time PCR showed only a tiny amount of OVA expression from thymic cDCs (Fig. 4A), this might be sufficient for the clonal deletion and production of Tregs in the absence of thymic B cells.

Interestingly, Aire-deficient recipients that had received BM cells from double Tg (Aire/OVA-KI crossed with OT-2 Tg) showed slightly weaker clonal deletion (Fig. 5E, middle) and production of fewer Tregs (Fig. 5E, bottom) compared with the Aire-sufficient recipient mice that had received the same BM cells. These results suggested that Aire in mTECs contributes to the creation of a thymic niche that is optimal for tolerance induction.

We next investigated whether mTECs from Aire/OVA-KI could induce clonal deletion directly without involvement of BM-derived APCs by employing the thymus graft experiment used for RIP-OVA Tg (Fig. 2). When thymi from Aire/OVA-KI (+/OVA) were grafted into MHC-II–deficient OT-2 Tg, we observed clonal deletion of OT-2 T cells to a degree similar to when they were grafted into MHC-II–sufficient OT-2 Tg (Fig. 6A, top and middle, 6B), suggesting that Aire/OVA-KI is a model in which direct presentation of self-antigen by mTECs can occur independently from BM-derived APCs. Consistent with the dispensable role of Aire for clonal deletion in Aire/OVA-KI crossed with OT-2 Tg (Fig. 4C), grafting of fetal thymi from Aire^−/OVA into MHC-II–deficient OT-2 Tg also resulted in clonal deletion of OT-2 T cells (Fig. 6A, top and middle, 6B); for some unknown reason, Aire-deficient thymi [Aire/OVA-KI (−/OVA)] grafted into MHC-II–sufficient OT-2 Tg became severely atrophic, and we were unable to evaluate the composition of the thymocytes (Y. Mouri and Mitsuru Matsumoto, unpublished observations). Thus, neither Aire in mTECs nor that in BM-derived APCs (Fig. 5A) is required for clonal deletion in the Aire/OVA-KI model.

Similar to RIP-OVA Tg (Fig. 2C, 2D), production of Tregs was significantly reduced when MHC-II–deficient OT-2 Tg were used as recipients, in comparison with MHC-II–sufficient OT-2 Tg recipients (Fig. 6A, 6B, bottom). It was noteworthy that lack of Aire in mTECs resulted in a small but significant reduction of Tregs in the Aire/OVA-KI model in the absence of any contribution from BM-derived APCs (i.e., MHC-II–deficient OT-2 Tg were used as recipients) (Fig. 6A, bottom of the right panel, 6B). In contrast, lack of Aire in mTECs resulted in rather augmented clonal deletion under these conditions (Fig. 6A, middle of the right panel, 6B), again suggesting a differential requirement of Aire for clonal deletion and production of Tregs. Overall, the results from two Tg models suggest that the requirement for Aire in mTECs for clonal deletion was conditional: Aire was indispensable for the clonal deletion in RIP-OVA Tg but not in Aire/OVA-KI. In contrast, Aire regulates the production of Tregs in both models. Accordingly, we suggest a differential requirement for Aire in the clonal deletion and production of Tregs, depending on the cell types expressing self-antigen and/or the types of thymic APCs involved in the tolerance induction, as summarized in Supplemental Fig. 4C.

The results obtained from the Tg models suggested that Aire contributes to the creation of a thymic microenvironment suitable for clonal deletion and/or the production of Tregs. Because the final steps of SP T cell maturation take place within the medulla (41), and subsequent migration of mature SP T cells to the periphery might be controlled by Aire-dependent production of chemokines (13, 42), we examined the maturation status of SP T cells from Aire-deficient mice in a polyclonal setting. As SP T cells differentiate from the semimature to mature stage within the medulla, they begin to express CD62L and Qa-2 with concomitant loss of CD24 expression (43). When these maturation-associated cell markers were examined on SP T cells, Aire-deficient mice contained lower percentages of CD62L^highCD24^low SP T cells among both CD4SP (Fig. 7A) and CD8SP T cells (Fig. 7B) compared with those from control mice. Percentages of Qa-2^high SP T cells among both CD4SP and CD8SP T cells were also reduced in Aire-deficient mice (Fig. 7A, 7B). One noteworthy finding was that the altered maturation profiles of SP T cells were not merely secondary to autoimmunity in Aire-deficient mice, because reduced percentages of mature CD62L^highCD24^low/Qa-2^high CD4SP T cells in the thymus were also observed in the Aire-deficient OT-2 Tg we used for the current study (Fig. 7C).

With the use of two Tg models expressing a common and prefixed OVA self-antigen in different types of cells, and involving different Aire-expressing APCs in the induction of tolerance, we have demonstrated distinct roles of mTECs and BM-derived cells in the clonal deletion and/or production of Tregs together with the requirement for Aire in each process. Aire in mTECs was required for clonal deletion when BM-derived APCs present self-antigen through the Ag-transfer system (RIP-OVA Tg model), whereas it was dispensable for this purpose when mTECs were able to directly present self-antigen to autoreactive thymocytes (Aire/OVA-KI model). The latter finding implies that Aire in mTECs does not have a major impact on posttranscriptional control of the expression of self-antigens required for clonal deletion. In contrast to clonal deletion, production of Tregs was dependent on the function of Aire in mTECs, but not in BM-derived cells. Thus, the combination of these cell types that express self-antigens and APCs responsible for the presentation of self-antigens determines not only the fate of autoreactive thymocytes (i.e., clonal deletion versus development into Tregs) but also the requirement for Aire in these tolerogenic events.

Although Tg models expressing a surrogate self-antigen of OVA (12, 13) or hen egg lysozyme (11) under the control of the rat insulin promoter have been commonly used for the study of Aire-dependent tolerance induction in the thymus, the specific cell types that express these model self-antigens have not been fully characterized. For example, because RIP-OVA Tg crossed with OT-2 Tg showed Aire-dependent negative selection (12, 13), it was somewhat unexpected that OVA expression in mTECs was lower in mTEC^high than in mTEC^low, which does not follow the pattern of Aire expression. Furthermore, OVA expression in BM-derived cells had not been taken into account when the role of Aire in Ag transfer from mTECs to BM-derived APCs was proposed in the previous study (12). Another unexpected feature in this model was the abundant expression of anti-sense OVA transcript in mTECs. Given that no obvious changes in the levels of the OVA sense transcript in both mTECs and BM-derived cells were observed in the absence of Aire, it was necessary to investigate additional mechanisms of tolerance induction involving Aire.

We have definitively demonstrated that Aire plays a role in self-antigen transfer from mTECs to BM-derived APCs in the RIP-OVA Tg model using thymus grafting. The importance of self-antigen transfer for establishment of self-tolerance has also been suggested by extensive sequencing of the TCRα-chain in a fixed TCRβ Tg model (38). In that study, the roles of mTECs and BM-derived APCs were studied using BM chimeras in which MHC-II expression in mTECs was knocked down by short hairpin RNA for CIITA driven by the Aire promoter (44) and BM chimeras into which MHC-II–deficient BM cells had been transferred, respectively. That study found that approximately half of the Aire-dependent deletion involved BM-derived APCs, as was the case in the RIP-OVA Tg model. However, the specific cell types acting as the source of the corresponding self-antigens (i.e., mTECs and/or BM-derived cells) remained unknown because no particular model self-antigens with a defined expression profile were introduced in that experiment. In this context, our analysis using the RIP-OVA Tg model clearly demonstrated that cells expressing self-antigens and thymic APCs responsible for elimination of T cells specific to the corresponding self-antigen are distinct, further emphasizing the importance of cross-talk between mTECs and BM-derived APCs through the transfer of self-antigens (29, 37).

Although lack of Aire in mTECs impaired the induction of tolerance involving Ag transfer, we found using two-photon laser microscopic analysis that OT-2 T cells were able to attach to Aire-marked mTECs (i.e., GFP⁺ mTECs from Aire/GFP-KI) and form clusters with a comparable size and stability in both Aire-sufficient and Aire-deficient RIP-OVA Tg. These results suggested that Aire does not spatially control the process of cluster formation by self-antigen (OVA)–expressing mTECs (i.e., RIP⁺ mTECs), OT-2 T cells, and BM-derived APCs; although BM-derived APCs were not visualized in the present imaging study, we have previously demonstrated that Aire⁺ mTECs and DCs have close contact in the thymic medulla (26). Instead, Aire may function for efficient killing of OT-2 T cells that have already encountered self-antigens and completed cluster formation. Currently, the process of OT-2 T cell death cannot be monitored long enough by two-photon laser microscopic analysis (28). Therefore, the precise action of Aire in the tolerogenic events occurring through the Ag transfer system awaits further investigation.

The Aire/OVA-KI model revealed Aire-independent clonal deletion of OT-2 T cells: OVA expressed by mTECs or BM-derived APCs was able to independently induce clonal deletion in the absence of Aire. In this model, expression of OVA by mTECs faithfully followed the endogenous expression of Aire, as expected. Given that almost all mTECs pass through a phase of Aire expression during their differentiation program (45), we assume that the subset of mTECs expressing OVA in Aire/OVA-KI is much broader than that in RIP-OVA Tg. In view of the fact that individual mTECs transcribe only a small subset of TRAs among the total TRA repertoire (46–48), it is possible that expression of OVA by a broad subset of mTECs at a given time masks the requirement for Aire in the induction of tolerance in the Aire/OVA-KI model.

OVA produced by BM-derived APCs (i.e., cDCs and B cells) from Aire/OVA-KI, but not from RIP-OVA Tg, was able to induce tolerance induction. Because the level of OVA expression in BM-derived APCs from RIP-OVA Tg was rather higher than that from Aire/OVA-KI, it remains unknown why OVA expressed by BM-derived APCs from RIP-OVA Tg was unable to induce the tolerance induction. One possible explanation is that an anti-sense OVA transcript might be produced in BM-derived APCs, as we observed for mTECs in RIP-OVA Tg. Anti-sense OVA transcripts in BM-derived APCs may regulate the activity of the sense OVA transcripts to which they bind because of their complementarity (34).

Although the role of mTECs together with the significance of Aire in mTECs for the production of Tregs have been well studied (14, 15, 49, 50), the role of BM-derived APCs in the production of Tregs has not been fully addressed. Using thymus grafting, we have clearly demonstrated that BM-derived APCs play an important role in the production of Tregs in two Tg models. The magnitude of Treg reduction resulting from lack of Aire in mTECs was more obvious when cooperation with BM-derived APCs was available in both Tg models (i.e., when MHC-II–sufficient OT-2 Tg were used as recipients in thymus grafting), suggesting that BM-derived APCs maximize the effect of Aire in mTECs in a cooperative manner for production of Tregs.

Because some types of BM-derived APCs (e.g., thymic B cells and eTACs) express Aire, we examined the role of Aire in BM-derived APCs by transferring the BM cells from Aire-deficient OT-2 Tg into an Aire-dependent tolerance model of RIP-OVA Tg (Fig. 3C). We also assessed the requirement for Aire in BM cells by transferring the BM cells from double Tg lacking Aire [Aire/OVA-KI (−/OVA) crossed with OT-2 Tg] into wild-type recipient mice (Fig. 5A). So far, we have not observed any effect of Aire deficiency in BM-derived APCs on the tolerogenic events. However, this does not imply that Aire in BM-derived APCs is unrelated to tolerance induction. In contrast to the two Tg models in which we expressed a prefixed OVA self-antigen model, thymic B cells and eTACs expressed a wide variety of TRAs whose spectrum was changed by the lack of Aire (8, 9). Because self-antigens expressed by BM-derived APCs were able to induce clonal deletion using self-antigens expressed independently, at least in the Aire/OVA-KI model, we assume that lack of Aire in BM-derived APCs might have resulted in the defect of tolerance induction observed in other experimental systems in which model self-antigens were not prefixed.

The fate of thymocytes undergoing either clonal deletion or development into Tregs is controlled by the strength of the interaction between self-antigens and the corresponding TCRs (31, 32, 51). When overall tolerogenic events were compared between the two Tg models we studied, there was an inverse correlation between clonal deletion and the production of Tregs: Aire/OVA-KI showed stronger clonal deletion by sacrifice of Treg production, whereas in RIP-OVA Tg production of Tregs was more prominent relative to clonal deletion. Interestingly, it has recently been suggested that Aire enforces autoreactive T cells into the Treg lineage (16), which may represent another mechanism for the creation of an Aire-controlled thymic microenvironment. Notably, both clonal deletion and production of Tregs were decreased when BM cells from double Tg (Aire/OVA-KI crossed with OT-2 Tg) were transferred into Aire-deficient mice, compared with transfer into Aire-sufficient mice (Fig. 5E). Thus, Aire in mTECs might create a thymic niche that is optimal for the qualitative nature of the interaction between self-antigens and the corresponding TCRs for tolerance induction, which might be associated with the Aire-dependent final maturation process of CD4/CD8 SP thymocytes (Fig. 7).

We thank F. Hirota for technical assistance, the laboratory members for helpful discussion, and Dr. L. Klein for critical reading of the manuscript.

The authors have no financial conflicts of interest.