Self-Specific MHC Class II-Restricted CD4⁻CD8⁻ T Cells That Escape Deletion and Lack Regulatory Activity¹

The development of conventional αβ T lymphocytes relies on the interaction of their Ag receptor (TCR) with self peptide:self MHC complexes expressed on thymic epithelial cells. Failure to establish such interactions leads to the death of immature thymocytes by apoptosis (neglect). This interaction either rescues immature thymocytes from apoptosis and signals them to complete their maturation (intrathymic positive selection), or causes their physical elimination from the thymic microenvironment by precipitating the execution of the apoptotic program (clonal deletion). These opposed outcomes allow the generation of a highly diverse mature TCR repertoire that is depleted of T cells highly reactive to most intrathymically expressed self determinants (1, 2, 3, 4, 5). Besides clonal deletion, additional mechanisms can operate to neutralize autoreactive thymocytes. Those include induction of a state of functional inactivation termed anergy (6, 7, 8), reduction of TCR surface expression (9, 10), as well as TCR α-chain substitution (11, 12).

Despite these multiple mechanisms of intrathymic tolerance induction, functional autoreactive αβ T lymphocytes can be detected in the periphery of both human and unmanipulated laboratory animals (13). The possibility that many autoreactive T cells may not be tolerized at the immature stage simply because the self Ags they are able to react against are not expressed in the thymic microenvironment may have been overestimated because multiple tissue-specific Ags that are targets in major inflammatory autoimmune diseases are indeed significantly expressed in the thymic microenvironment and particularly in medullary epithelial cells (14).

Because many inflammatory autoimmune disorders involve MHC class II-restricted self Ag-specific αβ T cells (15), it appears fundamental to better understand the principles underlying the induction of central tolerance to self peptide:self MHC class II complexes, as well as exceptions to these principles.

In this study, we identify and characterize MHC class II-restricted, self Ag-specific αβ T cells that, due to lack of CD4 coreceptor expression, escape intrathymic deletion without being antagonized or acquiring suppressor/regulatory activity, populate and seem to persist in peripheral lymphoid organs and are capable of reacting to their cognate Ag in an IL-2-dependent fashion.

C57BL/6, B10.BR-H2^k2 H2-T18^a/SgSnJ (B10.BR), and B10.A-H2ⁱ⁵ H2-Tl8^a (5R) SgSnJ (B10.A(5R)) mice were obtained from The Jackson Laboratory (Bar Harbor, ME). The B10.A(5R) RAG-2-deficient mice were obtained from K. Bottomly (Yale University). The 1H3.1 TCR transgenic (Tg)⁴ mice (Vα1/Vβ6) were described elsewhere (16). The I-A^bβ/invariant chain (Ii) double-deficient mice were a gift of P. Marrack and J. Kappler (Howard Hughes Medical Institute (HHMI), Denver, CO). The Igκ-Eα and 107.1 I-Eα^d Tg mice were a gift of R. A. Flavell (HHMI, Yale University) (17, 18, 19). Mice were genotyped by PCR using tail genomic DNA and specific oligonucleotides.

Fluorescent labeled mAbs were used for multicolor staining. Briefly, 0.2 × 10⁶ cells were incubated in microtiter U-bottom plates with saturating concentration of labeled mAb in 20 μl, for 30 min on ice. Cells were washed twice and analyzed immediately. For two-step staining, cells were incubated first with purified mAbs in PBS, 2% FCS/0.1% NaN₃, followed by a F(ab′)₂ of goat anti-mouse Ig-FITC conjugate (Sigma-Aldrich, St. Louis, MO). The mAbs used were anti-Vβ6 FITC (clone RR4-7), anti-Vα2,3.2,8.3,11.1/.2 FITC (B20.1, RR3-16, B21.14, RR8-1), anti-CD45R/B220-PE (RA3-6B2), anti-CD86/B7-2 biotin (GL1), anti-CD5 FITC (53-7.3), anti-NK1.1 FITC (PK136) from BD PharMingen (San Diego, CA), anti-CD8α PE/FITC/CyChrome (53-6.7) from Life Technologies (Grand Island, NY), and anti-CD4 Quantum Red (H129.19) from Sigma-Aldrich. The Y3JP (mouse IgG2a, anti-I-A^b), 25-9-17 (mouse IgG2a, anti-I-A^b), Y-Ae (mouse IgG2b, anti-A^b + Eα52–68), Y17 (mouse IgG2b, anti-I-E), 53-6.72 and 2.43 (both rat IgG2b, anti-CD8), and 14.8 (rat IgG2b, anti-CD45RA (B220)) mAbs were affinity purified from hybridoma supernatants. A FACScan flow cytometer and the CellQuest software, both from BD Biosciences (Mountain View, CA), were used to collect and analyze the data. Nonviable cells were excluded using forward and side scatter electronic gating. For cell sorting, freshly isolated lymph node cells were triple stained for CD4, CD8, and Vβ6. CD4⁺ and CD4⁻ populations were sorted after gating on Vβ6⁺ cells using a FACStar^Plus station (BD Biosciences).

For detection of IL-4 and IL-10, T cells from 1H3.1 TCR Tg B10.A(5R) Rag-2^−/− mice were stimulated with purified B10.A(5R) APCs with or without exogenous IL-2 (20 U/ml) for 20 h. Supernatants were collected, and cytokine secretion was determined by sandwich ELISA using standard techniques. Purified 11B11 (rat IgG1) and biotinylated BVD6-24G2 (BD PharMingen) mAbs were used to detect mouse IL-4. JES5-2A5 (BD PharMingen) and biotinylated SXC-1 (rat IgM) mAbs were used to detect mouse IL-10.

For proliferation assay, T cells were cultured in U-bottom 96-well plates for 3 days at 37°C in Click’s Eagle Hank’s amino acid medium (Irvine Scientific, Santa Ana, CA) supplemented with 5% FCS, 5 × 10⁻⁵ M 2-ME, 2 mM l-glutamine, and 50 μg/ml gentamicin. In some cases, CD4⁺ T cells were enriched to 90/95% by depletion of CD8⁺ and MHC class II⁺ cells using anti-CD8, anti-B220, and anti-I-A^b mAbs. T cells (30–50 × 10³/well) were stimulated using irradiated C57BL/6 splenocytes (2.5 × 10⁵/well), plus serial dilutions of Eα52–68 peptide (ASFEAQGALANIAVDKA) in 150 μl.

In most experiments, rIL-2 was used at 10–20 U/ml. LPS blasts were obtained by treating splenocytes with LPS (Sigma-Aldrich) for 2 days in culture. The anti-mouse CD28 mAb used was purified R2/60.1.21 (rat IgM). The stimulated cells were incubated in duplicate wells, and 1 μCi of [³H]thymidine/well was added to the culture during the last 12 h. The plates were then harvested, and cpm were determined. For inhibition experiments, mAbs were sterile filtered and added to cultures (3–5 μg/ml).

To test the suppression potential of CD4⁻CD8⁻ 1H3.1 T cells, T cells from 1H3.1 TCR Tg B10.A(5R) Rag-2^−/− mice were cultured in the presence of purified B10.A(5R) APCs and purified naive CD4⁺ 1H3.1 TCR Tg T cells labeled with CFSE (Molecular Probes, Eugene, OR) in the presence or absence of exogenous IL-2. The naive CD4⁺ to CD4⁻CD8⁻ T cell ratio was 1:1. Unlabeled naive CD4⁺ 1H3.1 TCR Tg T cells were used as control for TCR ligand accessibility constraint. The cultures were analyzed by flow cytometry after 3–4 days.

The 1H3.1 αβ TCR Tg⁺ (C57BL/6 × B10.BR)F₁ thymic lobes were excised after 15 days of gestations and cultured in 10% FCS RPMI 1640 medium using Transwell polycarbonate membrane (Corning Costar, Cambridge, MA). Medium was replaced every 2 days. After 8 days of culture, thymocyte suspensions were prepared and analyzed by immunostaining and flow cytometry.

We previously studied intrathymic negative selection imposed by an endogenously assembled and expressed self peptide:MHC class II complex in vivo (20). We used Tg C57BL/6 (I-A^b+/I-Eα⁻) mice expressing the Eα52–68:I-A^b complex-specific 1H3.1 αβ TCR (21, 22) bred to B10.A(5R) (I-A^b+/I-Eα⁺) mice as well as to several C57BL/6 I-Eα^d Tg lines expressing the Eα52–68:I-A^b complex in distinct cell types (17, 18, 19). As a result of tolerance induction, 1H3.1 TCR/I-Eα double-Tg mice displayed severely reduced thymic cellularity, and their lymph node cells reacted to anti-CD3 stimulation, but remained unresponsive to B10.A(5R) APCs that constitutively express the Eα52–68:I-A^b complex (20). However, a proliferative response to B10.A(5R) splenocytes could be observed in the presence of exogenous IL-2 (Fig. 1,A). This response was in part Ag specific because the Y-Ae mAb, which also specifically reacts to the Eα52–68:I-A^b complex (23), reduced the IL-2-dependent proliferation when added to the culture (Fig. 1 B). In contrast with Y-Ae, the isotype-matched (IgG2b) Y17 anti-I-E mAb did not reduce the reactivity. Additionally, the 25.9.17 mAb that reacts to multiple peptide:I-A^b complexes, but not to the Eα52–68:I-A^b complex (24), had no effect. This confirmed that the response to B10.A(5R) APCs observed in the presence of IL-2 relied in part on the specific recognition of the Eα52–68:I-A^b complex, and therefore suggested that a T cell population expressing the 1H3.1 αβ TCR escaped clonal deletion in the 1H3.1 TCR/I-Eα double-Tg animals. Interestingly, IL-2-dependent, Eα52–68:I-A^b complex-responsive T cells were detected in 1H3.1 TCR Tg B10.A(5R) mice, in 1H3.1 TCR/Igκ-Eα double-Tg mice that express I-Eα on B cells and dendritic cells at a lower level than in B10.A(5R) mice (19), and also in 1H3.1 TCR/107.1 double-Tg mice that express I-Eα on all MHC class II-positive cells, but at a higher level than in B10.A(5R) mice (17, 18). Thus, peripheral lymphoid organs from 1H3.1 αβ TCR Tg I-Eα⁺ mice contain T cells able to specifically react to APCs presenting the Eα52–68:I-A^b complex in the presence of exogenous IL-2.

Vβ6⁺ T cells were clearly detectable in secondary lymphoid organs as well as in the blood of 1H3.1 TCR/I-Eα double-Tg mice. These cells, which display a virtually normal TCR expression level on their surface, were essentially CD4/CD8 double negative, with few CD4⁺ T cells (20). Both specific proliferative responses and cytokine production by CD4⁺ 1H3.1 αβ TCR Tg naive T cells are inhibitable by anti-CD4 mAbs (unpublished observation). Therefore, the fact that no significant inhibition was observed in the presence of the anti-CD4 mAb GK1.5 (Fig. 1,B) suggested that the IL-2-dependent response of lymph node cells from 1H3.1 TCR/I-Eα double-Tg mice to the Y-Ae epitope was not dependent on CD4⁺ 1H3.1 TCR Tg T cells. CD4⁻CD8⁻ Vβ6⁺ cells were indeed responsible for the IL-2-dependent, specific response to B10.A(5R) APCs because, after sorting of both CD4⁻CD8⁻ Vβ6⁺ and CD4⁺ Vβ6⁺ lymph node cells from 1H3.1 αβ TCR/I-Eα double-Tg mice, a dose-dependent response to the Eα52–68 peptide in the presence of IL-2 was observed only for the CD4⁻CD8⁻ Vβ6⁺ subset (Fig. 1,C). The Vβ6⁺ CD4⁺ subset displayed a background response that was not specific because it was not dependent on the peptide dose and most likely corresponds to the non-Y-Ae blockable fraction of the response detected in the inhibition experiments (see Fig. 1,B). The data in Fig. 1 C also demonstrate that the CD4⁻CD8⁻ Vβ6⁺ and Vβ6⁺ CD4⁺ subsets were not suppressing the response of each other in the absence of IL-2 because neither of them responded to B10.A(5R) APCs after separation.

Accordingly, the phenotypic analysis of 1H3.1 TCR/I-Eα double-Tg lymph node cells cultured in the presence of B10.A(5R) APCs plus IL-2 revealed that the proliferating cells were CD4⁻CD8⁻ Vβ6⁺ (not shown). Finally, the definitive evidence that CD4⁻CD8⁻ Vβ6⁺ cells seen in 1H3.1 αβ TCR/I-Eα double-Tg mice can express the complete 1H3.1 αβ TCR is that such cells were readily detectable in spleen, lymph nodes, and peripheral blood of recombination-activating gene-deficient (Rag-2^−/−) 1H3.1 αβ TCR Tg B10.A(5R) mice (Fig. 2,A). In addition, splenocytes and lymph node cells from these mice readily proliferated in response to the Y-Ae epitope in the presence of exogenous IL-2 and maintained a CD4⁻CD8⁻ Vβ6⁺ cell surface phenotype (Fig. 2 B). Thus, CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells populate peripheral lymphoid organs of 1H3.1 TCR Tg/I-Eα⁺ mice and are capable of responding to the Eα52–68:I-A^b complex in an IL-2-dependent fashion.

Similar to Vβ6⁺CD4⁺ T cells, which populate the periphery of normal 1H3.1 αβ TCR Tg C57BL/6 mice, freshly isolated CD4⁻CD8⁻ Vβ6⁺ cells from secondary lymphoid organs of 1H3.1 TCR Tg B10.A(5R) Rag-2^−/− mice were essentially CD44^lowCD62L^highCD25⁻ (Fig. 3,A), and therefore displayed a phenotype most consistent with a naive T cell status (25). When the deleting mice were not deficient in RAG molecules, CD4⁻CD8⁻ Vβ6⁺ cells were also CD44^lowCD62L^high and lacked surface expression of the activation markers CD25 and CD69 (Fig. 3,B). In sharp contrast, many CD4⁺ Vβ6⁺ cells displayed an activated/memory phenotype: they were CD44^highCD62L^low. Besides, a significant fraction of them were clearly positive for CD25 and CD69 surface expression (Fig. 3,B). These features indicated that among residual CD4⁺ Vβ6⁺ cells, there were cells that had encountered their Ag in vivo, most likely in the thymus, and were functionally altered because sorted CD4⁺ Vβ6⁺ were not specifically responsive to their cognate Ag (Fig. 1 C). Some Vβ6⁺ CD4⁺ cells certainly expressed endogenously rearranged TCR α-chains because this population displayed an increased frequency of Vα (2, 3.2, 8.3, 11.1/.2)⁺ cells relative to mature Vβ6⁺ CD4⁺ cells from nondeleting 1H3.1 TCR Tg mice (25–30% vs 3–6%). In contrast, the frequency of Vα (2, 3.2, 8.3, 11.1/.2)⁺ cells among CD4⁻CD8⁻ 1H3.1 T cells was rather lower (from 1 to 2.5%). Collectively, the data indicate that peripheral CD4⁻CD8⁻ 1H3.1 TCR Tg T cells that are not subject to deletion in vivo are characterized by a naive T cell phenotype. This observation is consistent with the fact that the 1H3.1 αβ TCR is entirely dependent on CD4 coengagement, most likely through lck kinase recruitment, to signal T cells for activation upon confrontation of the Eα52–68:I-A^b complex (our unpublished observation).

Because appropriate costimulation is required for efficient activation of naive αβ T cells to occur (15), we asked whether the response of CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells to the Eα52–68:I-A^b complex could be restored by supplying additional costimuli. T cells from 1H3.1 TCR Tg B10.A(5R) Rag-2^−/− mice incubated in the presence of B10.A(5R) APCs plus the R2/60.1.21 anti-CD28 mAb mounted a specific proliferative response that was 25–30% of that seen in the presence of exogenous IL-2 (Fig. 4,A). A similar result was obtained when CD4⁻CD8⁻ 1H3.1 TCR Tg Rag-2^−/− T cells were stimulated using irradiated B cell blasts derived from B10.A(5R) splenocytes using LPS treatment; the response was 20–25% of that seen in the presence of IL-2 (Fig. 4,A). These experiments indicated that additional costimulation, represented in this study by further engagement of CD28 or by exposure to B7-1/2^high B10.A(5R) APCs, could partially restore the response to the Eα52–68:I-A^b complex. In contrast, increased strength of the signal delivered through the 1H3.1 αβ TCR also induced detectable activation of CD4⁻CD8⁻ 1H3.1 TCR Tg Rag-2^−/− T cells. This was evidenced by the fact that truncation variants of the Eα52–68 peptide (such as Eα52–67 and Eα54–66) that trigger an increased response of CD4⁺1H3.1 αβ TCR Tg T cells (26) could induce a detectable dose-dependent proliferative response (Fig. 4 B). We conclude that, in the absence of IL-2, additional signal 2 or strengthened signal 1 could partially substitute for the absence of CD4 coreceptor engagement by naive 1H3.1 αβ TCR Tg T cells. Thus, CD4⁻CD8⁻ 1H3.1 TCR Tg T cells display a certain level of responsiveness and differ from both CD4⁺ and CD8⁺ αβ T cells that are inactivated in vivo upon recognition of their cognate ligand and remain totally unresponsive to strong stimuli, such as clonotypic Ab cross-linking even in the presence of exogenous IL-2 (8, 27).

Because the IL-2-dependent, Eα52–68:I-A^b complex-responsive T cells were initially detected in secondary lymphoid organs of the TCR/I-Eα double-Tg mice, the possibility existed that such cells developed extrathymically. In such case, T cells may not confront the deleting ligand in the most efficient context known to mediate clonal deletion, that is, in contact with interdigitating (dendritic) cells that are MHC class II^high and are present throughout the thymus with the highest concentration at the corticomedullary junction (28). For instance, the analysis of thymectomized male mice reconstituted with bone marrow cells from HY TCR Tg female mice has revealed that self-reactive T cells can develop extrathymically; they are found in the spleen and the liver and are not subjected to deletion (29). To examine the issue of extrathymic differentiation in our system, we looked for CD4⁻CD8⁻ Vβ6^high cells in the thymus of TCR Tg/Y-Ae⁺ mice. We repeatedly detected CD4⁻CD8⁻ cells among Vβ6^high thymocytes in 1H3.1 αβ TCR Tg B10.A(5R) Rag-2^−/− mice (Fig. 5,A). Such cells were less frequent, yet present (1–5%), within the thymus of normal (I-Eα⁻) 1H3.1 αβ TCR Tg Rag-1^−/− mice. When total thymocytes were exposed to Eα52–68 in culture, a dominant Vβ6⁺CD4⁺ population expanded in the case of normal 1H3.1 αβ TCR Tg mice, whereas no expansion was detected among thymocytes cultured from 1H3.1 αβ TCR Tg B10.A(5R) Rag-2^−/− mice (not shown). However, adding IL-2 to the culture induced expansion of CD4⁻CD8⁻Vβ6⁺ cells from the thymus: CD4⁻CD8⁻Vβ6⁺ cells represented 90–95% of all cells after 1 wk of culture when 1H3.1 αβ TCR Tg B10.A(5R) Rag-2^−/− thymocytes were stimulated using C57BL/6 APCs, Eα52–68, and IL-2 (Fig. 5,B). This indicated that CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells escaping negative selection are present intrathymically. Finally, CD4⁻CD8⁻ Vβ6^high cells were also detected in fetal thymic organ culture derived from 1H3.1 αβ TCR Tg⁺ (C57BL/6 × B10.BR) F₁ (i.e., I-A^b+/I-Eα⁺) fetuses (Fig. 5 C), demonstrating that, on a deleting background, thymic CD4⁻CD8⁻ Vβ6⁺ cells can truly result from intrathymic development as opposed to in situ accumulation of cells that differentiate extrathymically and recirculate.

Because CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells are detected in 1H3.1 TCR Tg B10.A(5R) Rag-2^−/− mice, their intrathymic development definitely does not involve, either in a direct or indirect manner, engagement of alternate TCR α-chains resulting from endogenous rearrangements. That is, escape from intrathymic deletion by CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells results neither from the rescue of 1H3.1 αβ TCR Tg thymocytes already expressing endogenously rearranged, additional TCR α-chain, nor from induction of endogenous rearrangement upon recognition of the deleting ligand. The detection of CD4⁻CD8⁻ Vβ6⁺ cells in the thymus of 1H3.1 TCR Tg B10.A(5R) Rag-2^−/− mice further indicates that the development of these cells was independent of the emergence of precursors of other T cell lineages because in such mice, all the endogenous TCR loci remain unrearranged.

Reaggregation culture have shown that cortical epithelial cells presenting the cognate Ag can induce immature thymocytes to differentiate into CD4⁻CD8⁻ T cells able to secrete IL-10 and suppress the proliferative response of naive CD4⁺ T cells in the presence of IL-2 (30). We were not able to detect suppressive activity among CD4⁻CD8⁻ 1H3.1 TCR Tg Rag-2^−/− T cells in the presence or absence of IL-2. We also failed to detect secretion of IL-4 or IL-10 by these cells by ELISA (not shown). Commitment of 1H3.1 thymic precursors to nonconventional αβ T cell lineages was also a possibility because there are several precedents in the literature. For instance, immature thymocytes have been shown to be driven along the CD4⁻CD8⁻ NK T cells pathway (31, 32, 33, 34), or to acquire features of intraepithelial lymphocytes (35). It appears not to be the case in this study because we did not observe expression of either NK1.1 or CD8αα molecules on CD4⁻CD8⁻ 1H3.1 TCR Tg T cells (not shown). Differentiation into the CD25⁺ suppressor/regulatory subset (36) was also unlikely because, as mentioned above, CD4⁻CD8⁻ 1H3.1 TCR Tg Rag-2^−/− T cells from thymus or spleen did not display suppressive activity and were constantly CD4⁻, CD25⁻ (Fig. 3 A), and CTLA-4⁻ (not shown).

With respect to the requirement for MHC class II molecules, we observed that some CD4⁻CD8⁻ Vβ6⁺ T cells could be detected in the periphery of 1H3.1 αβ TCR Tg I-Aβ^b−/− mice (typically 2–3% of Vβ6⁺ cells) and could expand in vitro in the presence of Ag plus IL-2 (Table I). Thus, some CD4⁻CD8⁻ Vβ6⁺ T cells may develop in the absence of I-Aαβ^b heterodimers. The 1H3.1 αβ TCR Tg Ii^−/− combination was convenient to examine the role of I-Aαβ^b molecules (although at a diminished surface expression level) because in such mice, the development of CD4⁺1H3.1 αβ TCR Tg T cells is deficient (37) and therefore cannot outcompete that of CD4⁻CD8⁻αβ TCR Tg T cells. We observed that expression of I-Aαβ^b molecules could promote the emergence of 1H3.1 TCR Tg CD4⁻CD8⁻ T cells. The fact that CD4⁻CD8⁻ Vβ6⁺ T cells from 1H3.1 αβ TCR Tg Ii^−/− mice displayed a high proliferative response relative to 1H3.1 αβ TCR Tg and 1H3.1 αβ TCR Tg I-Aβ^b−/− mice (Table I) suggests that the emergence of CD4⁻CD8⁻ T cells expressing the complete 1H3.1 αβ TCR is more efficient in 1H3.1 αβ TCR Tg Ii^−/− mice. The data support the notion that I-Aαβ^b MHC class II molecules can efficiently support the emergence of CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells. In regular 1H3.1 αβ TCR Tg mice, the low fraction of CD4⁻CD8⁻ Vβ6⁺ T cells may possibly reflect outcompetition by CD4⁺1H3.1 αβ TCR Tg T cells. Alternatively, it is possible that in Ii-deficient mice, I-A^b molecules present self peptide(s) that is not, or poorly presented in the presence of Ii and that acts as appropriate ligand(s) for the development of CD4⁻CD8⁻ 1H3.1 T cells. To this respect, the surface amount of CD5 was higher on CD4⁻CD8⁻ Vβ6⁺ T cells from 1H3.1 TCR Tg Ii^−/− mice than on CD4⁻CD8⁻ Vβ6⁺ T cells from normal 1H3.1 TCR Tg mice, yet lower than that observed in 1H3.1 TCR Tg B10.A(5R) (i.e., Y-Ae⁺) mice (Fig. 6). Because the expression level of CD5 is determined by the signaling strength involved in selection (38), the data suggest that exposure to self peptide:I-A^b complexes causes a high intensity signaling in CD4⁻CD8⁻ Vβ6⁺ T cells from 1H3.1 TCR Tg Ii^−/− mice relative to those of normal 1H3.1 TCR Tg mice.

Finally, naive CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells appeared to persist over time in the spleen and lymph nodes of 1H3.1 TCR/I-Eα double-Tg or 1H3.1 TCR Tg B10.A(5R) Rag-2^−/− mice because they were still abundant at ∼6 mo of age. This may suggest that peripheral self peptide:self MHC class II complexes were accessible to provide naive CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells with the appropriate signals needed for their persistence. The data make the additional point that cellular interactions with B lymphocytes are presumably not required for such peripheral persistence to occur.

Apart from a severe reduction in thymic size, a major feature of 1H3.1 TCR Tg/I-Eα⁺ mice was the presence of a significant number of CD4⁻CD8⁻ Vβ6⁺ cells both in the thymus and in the periphery (20). The evidence that these cells contain true 1H3.1 αβ TCR Tg T cells comes from the observation that CD4⁻CD8⁻ Vβ6⁺ cells are abundant in the thymus, spleen, and lymph nodes of 1H3.1 αβ TCR Tg B10.A(5R) mice with a deficient recombinase activity (Rag-2^−/−). Thus, on an I-Eα⁺ genetic background, lack of coreceptor expression allows MHC class II-restricted autoreactive αβ T cells with a normal expression level of TCR to escape intrathymic deletion and populate peripheral lymphoid organs.

The reason for absence of CD4 expression is unclear. It could theoretically be due to actively reduced CD4 surface expression upon confrontation of the deleting ligand. However, most studies describing altered coreceptor expression report a partial reduction rather than a complete loss of the expression (39, 40, 41, 42, 43), and in the case of complete extinction, re-expression could be observed after in vitro stimulation (9). In contrast, CD4⁻CD8⁻ 1H3.1 αβ TCR Tg Rag-2^−/− cells remained double negative in all situations, even after activation with Eα52–68 plus IL-2. In addition, we failed to detect CD4 expression by intracellular staining of CD4⁻CD8⁻ 1H3.1 αβ TCR Tg Rag-2^−/− cells (not shown). It is therefore possible that CD4⁻CD8⁻ 1H3.1 αβ TCR Tg cells emerge intrathymically as true double-negative cells. Indeed, in 1H3.1 αβ TCR Tg Rag-1^−/− mice on a nondeleting (I-Eα⁻) background, these cells are seen both intrathymically (3–5% of Vβ6^high thymocytes) and in the periphery (1–3% of Vβ6⁺ splenocytes) (not shown). Interestingly, elegant experiments have indicated that MHC class II-restricted αβ T cells can mature intrathymically at a normal rate without progressing through the CD4⁺CD8⁺ stage (44). In this study, the authors pointed out that the surface expression of TCR typically seen at the early stages of thymic development in TCR Tg mice may promote the maturation of thymocytes lacking coreceptors. It is possible that such a phenomenon operates in our system.

Unlike most αβ T cells that have been inactivated in vivo (27, 41, 45, 46), CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells retained a naive phenotype. For instance, when anergized in vivo, MHC class II-restricted, influenza virus hemagglutinin-specific CD4⁺ αβ T cells acquired an activated/memory cell phenotype and remained unresponsive to Ag or clonotypic Ab even in the presence of IL-2 (27). Similarly, intrathymically anergized CD8⁺ T cells were refractory to activation by clonotypic Ab with or without IL-2 (8). In our system, the addition of IL-2 most likely complements for the lack of CD4 engagement without which 1H3.1 αβ T cells are unable to react to Eα52–68 (unpublished observation). We believe the non-Y-Ae blockable proliferation seen when IL-2 is added to total 1H3.1 TCR Tg/I-Eα⁺ splenocytes is most likely due to CD4⁺Vβ6⁺ cells because this subset includes a sizable fraction of CD25⁺ and therefore IL-2-responsive cells.

The extensive analysis of T cells from male mice carrying the male Ag-specific, D^b-restricted, HY αβ TCR revealed that the accumulation of CD4⁻CD8⁻ HY αβ TCR Tg T cells corresponds to the expression of the transgene-encoded αβ TCR into cells of the γδ lineage (40, 47). This phenomenon may well take place in the case of 1H3.1 TCR/I-Eα⁺ double-Tg mice. However, the abundance of CD4⁻CD8⁻ Vβ6⁺ cells in the thymus and periphery of 1H3.1 αβ TCR Tg B10.A(5R) Rag-2^−/− mice indicates that the emergence of CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells is not necessarily dependent on the development of other T cell lineages, because in these mice, all endogenous TCR loci remain unrearranged, and therefore precursors committed to other lymphocyte lineages cannot emerge.

The fine interactions that drive positive selection of CD4⁻CD8⁻ 1H3.1 αβ TCR Tg cells remain unclear. We observed that the MHC class II molecules expressed in Ii^−/− mice could assist their development. In the absence of I-Aβ^b molecules, these cells were still detectable, suggesting that some of them may develop independently of the expression of the restricting MHC element, as it has been observed in the HY TCR Tg model (48, 49). Alternatively, one cannot formally exclude that the few CD4⁻CD8⁻ Vβ6⁺ cells seen on the 1H3.1 αβ TCR Tg I-Aβ^b−/− background were selected on rare chimeric MHC class II molecules generated upon assembly of I-Aα^b and I-Eβ^b chains. On the deleting background, it is possible that the Eα52–68:I-A^b complex itself supports the intrathymic positive selection of 1H3.1 αβ TCR Tg T cells lacking coreceptors. The elevated CD5 expression level displayed by these cells supports this idea. Indeed, epitopes structurally related to the Eα52–68:I-A^b complex are major contributor to the positive selection of CD4⁺ 1H3.1 αβ TCR Tg T cells on a C57BL/6 background (26), and recognition of a covalent configuration of the Eα52–68:I-A^b complex (50) was able to support intrathymic positive selection of CD4⁺1H3.1 αβ TCR Tg T cells in vivo (51). In the same line of argument, it was documented that the development of CD4⁻CD8^low HY TCR Tg T cells could be driven by recognition of their cognate Ag on bone marrow-derived cells (49). It was also observed that male, but not female, epitopes could support the extrathymic development of HY TCR Tg T cells after reconstitution of thymectomized recipients with bone marrow (29).

Expression of alternate TCR α-chains can allow autoreactive αβ T cells to escape intrathymic deletion (52, 53, 54, 55, 56). It is assumed that, in such a situation, α-chains compete for pairing with the Tg β-chain. This reduces the surface expression level of the self-reactive TCR and lowers the sensitivity to the cognate ligand. It was also reported that TCR α-chain substitution could allow developing T cells to escape deletion (11, 12). This could occur by internalization of the Tg α-chain and its replacement by endogenously generated alternate α-chains. Re-expression of the original α-chain could then be observed in vitro (12). Based on the phenotype of the 1H3.1 αβ TCR Tg B10.A(5R) Rag-2^−/− mice mentioned above, we conclude that neither of these two phenomena were required for CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells to escape clonal deletion.

The escape from intrathymic clonal deletion in the absence of coreceptor expression has at least two implications with respect to αβ T cell development. First, the presence of CD4⁻CD8⁻ T cells in the lymphoid organs of the TCR/I-E double-Tg mice indicates that surface expression of CD4 coreceptor molecules constitutes neither a requirement nor a signal for mature thymocytes carrying a MHC class II-restricted αβ TCR to leave the thymic compartment and to home to and populate the periphery. Second, different experimental systems have shown that TCR-MHC ligand interactions involved in the intrathymic deletion of a given T cell are less stringent than those involved in the activation of its mature counterpart (57, 58, 59, 60, 61). For instance, in the lymphocytic choriomeningitis virus TCR Tg system, lymphocytic choriomeningitis virus epitope variants, which are poorly or not recognized by peripheral TCR Tg T cells, can respectively cause complete and partial intrathymic deletion (58, 62). Our observation indicates that although the intrathymic cell-cell interaction involved in deletion has low stringency, it is not low at the point that it cannot be influenced by the coreceptor. At least for the 1H3.1 specificity, it appears that the lack of CD4 coreceptor expression can change the outcome of the interaction, that is, convert negative into positive selection. The likely, but not necessarily correct explanation is that the recruitment of the lck kinase, which associates with the tail of CD4, is required to signal 1H3.1 TCR Tg thymocytes for deletion. Remarkably, the margin created by the absence of CD4 engagement was not so narrow that an increase in the density of the deleting complexes on bone marrow-derived cells (the most efficient stromal cell type in mediating intrathymic clonal deletion) would cause intrathymic deletion of CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells. This is illustrated by the detection of CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells both in 1H3.1 αβ TCR/Igκ-Eα double-Tg, 1H3.1 αβ TCR Tg B10.A(5R), and 1H3.1 αβ TCR/107.1 double-Tg mice, which respectively have a low, normal, and elevated expression level of the Eα52–68:I-A^b complex on dendritic cells.

There is ample evidence that naive αβ T cells require repeated interactions of their TCR with self peptide:self MHC complexes for their peripheral persistence (reviewed in Refs. 63 and 64). Memory αβ T cells appear less dependent on MHC contacts for their physical persistence, but such interaction appears required for their functionality (65). In the case of 1H3.1 TCR Tg/I-Eα⁺ mice, we have been able to detect a significant and functional population of naive CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells in spleen and various lymph nodes at least for an observation period of 5/6 mo. This contrasts with the fact that under conditions of impaired peripheral survival, the number of MHC class II-restricted T cells in spleen and lymph nodes is already declining at the adult age (66). Thus, CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells appear able to receive appropriate survival signals in vivo. If recognition of the Eα52–68:I-A^b complex itself plays a role in the intrathymic emergence of CD4⁻CD8⁻ 1H3.1 αβ TCR Tg T cells, as suggested by their high CD5 expression level, it is plausible that it also contributes to the peripheral persistence of these cells. Consistent with this notion is the fact that the covalent configuration of the Eα52–68:I-A^b complex could support both the intrathymic maturation and the peripheral maintenance of naive CD4⁺1H3.1 αβ TCR Tg T cells (51). Although CD4⁻CD8⁻ αβ T cells have been found to be active both in murine and human autoimmune pathologies (67, 68, 69) and may theoretically originate from such an escape from intrathymic clonal deletion, we did not notice signs of autoimmune response in our system during the time of the study.

In conclusion, we document in this work the existence of MHC class II-restricted αβ T cells specific for a self peptide:self MHC class II complex that lacks coreceptor expression, escapes clonal deletion in vivo, and remains able to specifically react to their cognate ligand in the presence of IL-2. Such cells can develop in the thymus independently of other T cell lineages, do not show signs of regulatory/suppressor activity, retain a fully naive phenotype, and persist over time in the periphery.

We thank Kim Bottomly (Yale University) for the B10.A(5R) Rag-2-deficient mice; Richard A. Flavell (HHMI, Yale University) for the Igκ-Eα and 107.1 I-Eα^d Tg mice; and Charles Annicelli and Joanne Daugherty for help with animal care. We also thank X. He (Yale University) for discussion.