Mature Mainstream TCRαβ⁺CD4⁺ Thymocytes Expressing L-Selectin Mediate “Active Tolerance” in the Nonobese Diabetic Mouse¹

The nonobese diabetic (NOD)³ mouse develops a spontaneous form of autoimmune insulin-dependent diabetes mellitus (IDDM) that closely resembles the human disease (1, 2). Autoreactive T lymphocytes play a major pathogenic role. Both CD4⁺, essentially belonging to the IFN-γ-producing Th1 subset, and CD8⁺ T cells have been implicated in the pathogenesis of IDDM (3, 4, 5, 6, 7, 8). Diabetogenic T cells, which transfer acute IDDM into immunoincompetent syngeneic recipients, are present in high frequency in the spleens of diabetic NOD mice (3, 9). In parallel to these effector cells, there is substantial evidence showing the presence, especially in young prediabetic NOD mice, of a subset of T lymphocytes mediating “active tolerance,” that is, exerting active control of or a down-regulatory effect on diabetogenic lymphocytes or their precursors. Thus, cotransfer experiments have shown that CD4⁺ T splenocytes from prediabetic animals fully prevent the transfer of disease by diabetogenic cells (10, 11). Moreover, diabetogenic T cells can transfer IDDM into syngeneic adult mice only if they are sublethally irradiated (2, 9, 10, 12). Effective depletion of CD4⁺ lymphocytes, as obtained after adult thymectomy and treatment with specific Abs, can effectively substitute for the irradiation (13). This further corroborates the CD4 nature of the protective cell. Lastly, cyclophosphamide, an alkylating agent that in different models, including delayed-type hypersensitivity (14) and experimental allergic encephalomyelitis (15, 16, 17, 18), has been shown to selectively affect T cell-dependent regulation, is able to trigger acute diabetes within 2 wk when injected into young prediabetic NOD mice (2, 12, 19). A mere toxic effect on β cells has been excluded by the work of Charlton et al., which showed that cyclophosphamide-induced IDDM is prevented by the transfusion of mononuclear cells from nondiabetic NOD mice (19).

These regulatory T cells have been evidenced not only in peripheral lymphoid organs but also in the thymus. Boitard et al. reported that, when cotransferred with diabetogenic T cells, total thymocytes from young prediabetic NOD mice successfully protected from disease as efficiently as did their spleen CD4⁺ cells (10). Moreover, Dardenne et al. showed that thymectomy performed in NOD females at weaning (3 wk of age) significantly increased the incidence of spontaneous IDDM, thus confirming the thymic origin of the regulatory population (20).

The aim of the present work has been to further characterize the phenotypic and functional properties of the thymic population capable of down-regulating diabetogenic effectors. In this vein we have used a series of T cell markers that dissect functionally distinct subsets of thymocytes (21). This issue is of particular interest given the recent data from our group (21) and from that of Baxter et al. (22) showing that NOD mice exhibit a marked deficiency in the recently described NK T cell subset (23, 24, 25, 26, 27). Based on the evidence that this subset produces massive amounts of IL-4 upon TCR cross-linking, one major question was to determine whether NK T cells could directly modulate in conventional cotransfer experiments the pathogenic ability of diabetogenic T cells.

NOD (K^d, I-A^g7, D^b) and NOD-scid mice were bred in our animal facilities under specific pathogen-free conditions; in females, spontaneous IDDM appears by 14 wk of age (80% incidence at 30 wk of age) and is preceded by insulitis at 4 to 6 wk. Colorimetric strips were used to monitor glycosuria (Glukotest, Boehringer Mannheim, Mannheim, Germany) and glycemia (Haemoglukotest and Reflolux F, Boehringer Mannheim). Mice were screened for glycosuria twice a week. When glycosuric the animals were also screened for hyperglycemia. Mice were considered diabetic when glycemia ≥3 g/liter was scored at two consecutive measurements.

Abs to CD4 (clone GK1.5), to CD8 (clone 53.6.7 and clone 3-155), to αβTCR (clone H57-597), to Vβ8TCR (clone F23.1), and to HSA (anti-CD24, clone J11d) were purified and fluoresceinated and/or biotinylated in our laboratory. PE-anti-CD4 (clone RM4.5), PE-anti-CD24 (clone M1/69), FITC- and biotin-anti-CD44 (clone 1 M7.8), biotin-anti-CD62L (MEL-14), PE-anti-HSA (clone M.1/69), and FITC-anti-CD8 (clone 53.6.7) were obtained from PharMingen (San Diego, CA). Biotin-anti-3G11 mAb was a gift from Dr. A. Bendelac (Department of Molecular Biology, Princeton University, Princeton, NJ). Surface markers were assessed by flow cytometry. Briefly, cells were stained in HBSS containing 5% heat-inactivated FCS (Techgen, Les Ulis, France) and 0.01% sodium azide and were incubated for 30 min with the appropriate concentration of biotin-labeled mAbs in 96-well round-bottom microplates on ice. After three washes, the cells were incubated for another 30 min with FITC-labeled mAbs and PE-labeled streptavidin or PE-labeled mAbs. For triple staining, during the second step, streptavidin-Tricolor (Caltag) was added to the appropriate FITC- and PE-labeled mAb combination. Control stainings were performed using isotype-matched, biotinylated FITC- or PE-labeled irrelevant Abs. Flow cytometry was performed on a FACScan flow cytometer (Becton Dickinson, Mountain View, CA). A minimum of 1 × 10⁴ events were acquired on a gate including viable lymphocytes. For acquisition and analysis, the software used was LYSYS II or, in more recent experiments, CellQuest (Becton Dickinson).

Thymi and spleens were carefully recovered from exsanguinated mice. For cell purification experiments, pools of 5 to 15 thymi or spleens were used. Mature double-negative and CD4⁺ thymocytes were enriched by means of complement depletion. Whole thymocyte suspensions were incubated at 37°C for 40 min with the IgM Abs 3-155 (rat anti-mouse CD8) and J11d (rat anti-mouse heat stable Ag, anti-HSA) plus complement (Low-Tox Rabbit Complement, Cedarlane, Ontario, Canada). Mature CD8⁺, immature CD4⁺CD8⁺, and the vast majority of immature HSA⁺CD4⁻CD8⁻ thymocytes were thus eliminated. Viable cell suspensions were recovered after density gradient centrifugation (J. Prep., Techgen, Les Ulis, France; 1400 rpm for 20 min). The cells recovered at the interface were washed twice in culture medium. To assess the depletion, HSA⁻CD8⁻ thymocytes were stained with a combination of HSA and CD8 Abs different from those used for the depletion (clones M.1/69 and clone 53.6.7, respectively). The depleted population obtained contained >95% HSA⁻CD8⁻ cells as assessed by flow cytometry.

When needed, HSA⁻CD8⁻ thymocytes were further subdivided on the basis of CD62L and 3G11 expression using magnetic activated cell sorting (MACS, Miltenyi Biotech, Bergisch-Gladbach, Germany) as previously described (21). Briefly, HSA⁻CD8⁻ thymocytes were incubated with the appropriate concentration of anti-CD62L and anti-3G11 mAbs. The cells were washed and incubated with the streptavidin-coated paramagnetic beads (Miltenyi Biotech) and passed through the column within the magnetic activated cell sorting device according to the manufacturer’s instructions. The purity of the CD62L 3G11-negative and -positive fractions recovered was analyzed by flow cytometry after staining with PE-labeled streptavidin. The purity of the sorted cells was in all cases >90% (see Fig. 1 A), and recovery ranged from 50 to 70%.

For some experiments thymocytes were recovered from 3-wk-old mice that received at birth a single injection (200 μg/mouse i.p.) of the hamster CD3 mAb 145 2C11 as described by Hayward and Schreiber (28). Cell staining showed that in these mice the thymocyte suspensions recovered contained 99% of HSA⁺ cells and no mature TCR⁺CD3⁺. With this method one can obtain highly purified HSA⁺ thymocytes while avoiding the use of CD4 and CD8 Abs (in conventional positive cell selection techniques) that may modify the functional capacity of the recovered population.

Cells were cultured in RPMI 1640 culture medium (Life Technologies, Gaithersburg, MD) supplemented with Glutamax, 10% FCS (Techgen), 0.05 mM β-ME, penicillin (100 IU/ml), and streptomycin (100 μg/ml). When needed, thymocyte suspensions were cultured for 60 h in the presence of human rIL-7 according to the method we previously described (29, 30). Briefly, total thymocyte suspensions (20 × 10⁶/well) were cultured in 24-well plastic plates (Costar, Cambridge, MA) in complete medium supplemented with rIL-7 (1000 U/ml; Sanofi, Labege, France). After 60 h of culture at 37°C in a humidified atmosphere containing 10% CO₂, viable cells were recovered by centrifugation (1400 rpm for 20 min) on a density gradient (J. Prep., Techgen). The cells at the interface were recovered, washed twice in complete medium, and used for in vitro (phenotype analysis and cytokine production) and in vivo assays. When needed, they were also used in magnetic sorting experiments (described above).

To assess cytokine production, lymphocyte suspensions were plated in triplicate (1–2 × 10⁵/well; 200 μl final volume) in 96-well round-bottom microplates (Nunc, Roskilde, Denmark) coated with 10 μg/ml of anti-αβTCR (H57-597). Supernatants were recovered at 48 h of culture at 37°C in a humidified atmosphere containing 5% CO₂ and stored at −80°C until tested. In control cultures coating with the αβTCR mAb was omitted, and no cytokines were detected in the supernatants.

IL-4 and IFN-γ in the supernatants were measured by means of two-site sandwich ELISAs, as previously described (21). The 11B11 (anti-IL-4) and AN18 (anti-IFN-γ) mAbs were used for coating the plates (capture mAbs), and biotinylated-BVD6 (anti-IL-4) and R46A2 (anti-IFN-γ) mAbs were used as second Abs. The 11B11 hybridoma was provided by W. Paul (National Institutes of Health, Bethesda, MD), and the BVD6, AN18 hybridomas were provided by A. O’Garra (DNAX, Palo Alto, CA). Briefly, the plates were coated overnight with the capture mAbs diluted in carbonate-bicarbonate buffer (pH 9.8; 0.1 M). After washing (PBS/0.1% Tween) and blocking (PBS/1% BSA), the test samples and the standards were distributed, and the plates were incubated for 2 h at room temperature. Following incubation with the biotinylated mAbs and subsequently with horseradish peroxidase-labeled streptavidin (Vector, Burlingame, CA), the plates were washed, and ortho-phenylene-diamine was used as a substrate (Sigma, St. Louis, MO). The cytokine concentrations were expressed as nanograms per milliliter, based on a calibration curve established for each assay using serial dilutions of recombinant standards. Mouse rIL-4 and IFN-γ were obtained from R&D (Minneapolis, MN). The sensitivities of the IL-4 and IFN-γ assays were 0.2 and 0.1 ng/ml, respectively.

Depending on the experiment, two sorts of recipients were used, adult irradiated NOD males or, in more recent experiments, NOD-scid. As already reported by several groups, including ours, adult 6- to 8-wk-old NOD males have to be sublethally irradiated (750 rad) to be used as recipients of cell transfers (9, 10). In such recipients the various transferred cells were injected 24 h following the irradiation according to our conventional protocol (10). Depending on the experiment, the animals received either a single cell population or, in the case of cotransfer experiments, a mixture of two distinct cell populations. The precise cell numbers used varied depending on the experiments and are detailed in Results.

In all the experiments, diabetogenic cells were recovered by gentle disruption from the spleen of overtly diabetic NOD females. Pools of 7 to 10 diabetic mouse spleens were used. It has been well established by several laboratories, including ours, that diabetes transfer is exclusively mediated by T cells (3, 9, 31, 32). Thus, the numbers of diabetogenic spleen cells to inject were deduced according to the proportions of CD3⁺ cells, as assessed by FACS analysis, scored in each pooled spleen cell preparation. This is referred to in the text as T cell equivalents.

Other transferred populations included whole thymocyte suspensions and purified thymocyte subsets recovered from either fresh thymocyte suspensions or thymocytes recovered after a 60-h in vitro culture in the presence of IL-7.

Sections of paraffin-embedded or frozen pancreata were used. For conventional histopathology, sections were stained with hematoxylin and eosin to score mononuclear cell infiltration as follows: grade 0 = normal islets, grade 1 = focal or peripheral insulitis (lymphocytes surrounding the islet, but no destruction of endocrine cells as assessed by labeling with anti-insulin Abs), and grade 2 = invasive destructive insulitis.

When appropriate, results were analyzed using the χ² test.

It has been previously shown that total thymic cells collected from young prediabetic NOD mice could protect from the transfer of diabetes in adult irradiated recipients (10). The present experiments confirmed these results and extended them to NOD-scid recipients. As shown in Table I, in both adult irradiated recipients and NOD-scid mice 50 × 10⁶ total thymocytes from NOD females fully inhibited the diabetogenic capacity of 5 × 10⁶ spleen T cell equivalents (see Materials and Methods) from diabetic NOD mice. This protective effect was dose dependent, since 5 × 10⁶ thymocytes failed to impede diabetes development. Protector cells were collected from both young (3 wk) and adult (8-wk-old) prediabetic animals. In all cases the thymocyte population transferred included 80 to 85% of double-positive CD4⁺CD8⁺ cells and 15 to 20% of mature single-positive and double-negative cells.

Subsequently, experiments were designed to determine whether the thymocyte population endowed with the protective ability belonged to the immature or mature thymic compartment. Complement depletion experiments were thus performed using an Ab to the heat-stable Ag, which is expressed on immature double-positive thymocytes. As detailed in Table II, the mature HSA⁻ population reproducibly retained the protective activity; as few as 7.5 × 10⁶ HSA⁻ thymocytes were sufficient to abolish the pathogenic capacity of 5 × 10⁶ diabetogenic T cell equivalents. CD8 single-positive thymocytes were not essential for this effect, since an additional complement-mediated depletion with an Ab to CD8 did not modify the protective capacity of HSA⁻ thymocytes (Tables II and III). These data also argue against the fact that the protection afforded by 50 × 10⁶, but not 5 × 10⁶, thymocytes could be due to a dilution effect of the diabetogenic population.

Cotransfer experiments were also performed using suspensions of HSA⁺ thymocytes collected from 3-wk-old NOD mice. As shown in Table II, HSA⁺ thymocytes cells were totally incapable at transferring protection.

As for total thymocytes, the protective ability was evidenced with purified mature thymocytes collected not only at 3 wk but also at 9 to 10 wk of age (Table III).

There is extensive evidence in the literature that mature HSA⁻CD8⁻ thymocytes constitute a heterogeneous population. One may distinguish within this population mainstream TCRαβ⁺CD4⁺ cells from the discrete subset of recently characterized thymocytes sharing NK and TCRs, namely NK T cells (23, 24, 25, 26, 27). A major characteristic that distinguishes these two subsets is their differential expression of membrane receptors such as CD44, 3G11, and CD62L (21, 24, 25, 26, 27, 33, 34, 35, 36). As we previously reported (21), 60 to 70% of HSA⁻CD8⁻ cells are CD62L⁺ and correspond to the mainstream population that exclusively includes the CD4⁺ single-positive thymocytes (Fig. 1 A). In contrast, the CD62L⁻ population that, depending on the age of the thymus donors, comprises 15 to 20% of the cells, includes both double-negative and CD4⁺ single-positive thymocytes showing an intermediate αβTCR expression typical of NK T cells. Thymic NK T cells, as opposed to CD4⁺ mainstream thymocytes, also lack the 3G11 marker (21, 34) (data not shown).

Thus, in further immunomagnetic cell sorting experiments we used the L-selectin receptor and the 3G11 marker that are expressed at high density on mainstream single-positive CD4⁺ but not on NK T thymocytes. One representative experiment is detailed in Figure 1 to show the phenotypic distribution of CD62L and 3G11 before and after sorting. As we previously reported (21), this purification method allows high enrichment of the CD62L⁻3G11⁻ NK-like T subset that concentrates the capacity to rapidly produce high amounts of IL-4 upon TCR cross-linking. This latter functional capacity is exclusively attributed to NK T cells in the thymus (21, 24, 26, 27, 29).

The present results reproducibly showed that it is the CD4⁺CD62L⁺3G11⁺ and not the CD62L⁻3G11⁻ population (the subset that includes NK-like T cells (Fig. 1,A) that retains the protective ability of HSA⁻CD8⁻ cells (Table III). To maintain, as much as possible, homogeneous experimental conditions, the absolute numbers of sorted cells used for cotransfer experiments were deduced from the proportion of the subset under analysis in the original HSA⁻CD8⁻ population.

The cytokine-producing ability (i.e., IL-4 and IFN-γ) upon TCR cross-linking of the various cell populations cotransferred is detailed in Figure 1 B. The CD4⁺CD62L⁺3G11⁺ thymocyte subset that concentrated the protective ability was, as expected, a poor IL-4 producer compared with the CD62L⁻3G11⁻ NK-like T cell-enriched subset.

We have previously reported that, upon in vitro culture in the presence of IL-7, mature thymocytes significantly expand (29, 30). The population recovered after 60 h of culture in the presence of IL-7 exclusively included mature single-positive or double-negative T cells, with a significant proportion of NK-like T cells within the CD62L⁻3G11⁻ subset (29, 30). We explored whether the in vitro IL-7 treatment could modulate the protective capacity of mature thymocytes. In a first series of experiments total thymocytes were incubated with IL-7, using already described experimental conditions (29, 30), namely, 60 h of culture at high cell density (20 × 10⁶/well) in the presence of 1000 U/ml of human rIL-7. The distribution of CD62L⁺3G11⁺ among αβTCR⁺ and Vβ8⁺ thymocytes before and after IL-7 culture is shown in Figure 2.

As few as 1 to 2 × 10⁶ of these IL-7 cultured thymocytes were sufficient to reproducibly protect from the transfer of disease afforded by diabetogenic splenocytes in either adult irradiated NOD recipients (Table IV) or NOD-scid mice (data not shown). By comparison, 50 × 10⁶ total thymocytes (including about 10–15% mature cells) or 5 × 10⁶ purified HSA⁻CD8⁻ thymocytes were needed. The same results were obtained when purified mature HSA⁻CD8⁻ thymocytes, instead of total thymocytes, were used for the IL-7 cultures (Table IV). Here again, as few as 2 × 10⁶ cultured cells very effectively protected from disease transfer.

To further characterize the thymocyte subset responsible for the protection observed, cultured thymocytes were sorted on the basis of CD62L and 3G11 expression. As described above for fresh uncultured thymocytes, the protective activity was retained in the CD62L⁺3G11⁺ population. Interestingly, as few as 0.5 × 10⁶ of IL-7 cultured CD62L⁺3G11⁺ thymocytes were sufficient to provide full protection from diabetes transfer by 5 × 10⁶ diabetogenic T cells (Fig. 3,A). At variance, CD62L⁻3G11⁻ cells that include CD4⁻ and CD4⁻CD8⁻ NK-like T thymocytes were totally ineffective in conferring protection (Fig. 3,A). As detailed in Figure 3 B, a majority of the islets from the pancreas of protected animals showed a peripheral noninvasive/nondestructive insulitis.

The cytokine-producing ability (i.e., IL-4 and IFN-γ) upon TCR cross-linking of the various IL-7 cultured cell populations used in cotransfers is detailed in Table V. The TCRαβ⁺CD62L⁺3G11⁺ thymocyte subset that concentrated the protective ability remained, a poor IL-4 producer compared with the CD62L⁻3G11⁻ NK-like T cell-enriched subset. It is interesting to emphasize that when cultured in the presence of IL-7, CD62L⁺3G11⁺ thymocytes acquire the capacity to produce massive amounts of IFN-γ. Indeed, the values were as high as those measured in the supernatants from stimulated cultures of TCRαβ⁺CD62L⁻3G11⁻ cells that are highly enriched in NK-like T thymocytes.

The present results confirm and further extend already published data obtained in our laboratory (10), showing that thymocytes from young prediabetic NOD mice are capable of protecting from diabetes transfer when coinjected with diabetogenic effectors into adult irradiated NOD recipients or NOD-scid. This inhibitory capacity is exclusively expressed by mature CD8⁻ thymocytes, as demonstrated by depletion experiments using Abs specific to HSA and CD8. Moreover, experiments performed using cell suspensions including 99% of immature HSA⁺ thymocytes clearly showed their total incapacity to inhibit disease transfer by diabetogenic cells. Additional evidence for the major role of mature thymocytes in this system came from experiments using thymocytes cultured in the presence of IL-7. IL-7 is a major T cell growth and maturation factor that, when added to high cell density thymocyte cultures, promotes the proliferation of most mature TCR⁺CD3⁺ thymocytes (29, 37). The protective ability of thymocytes from young NOD mice recovered at the end of an IL-7 culture was maintained and even potentiated, since as few as 0.5 to 2 × 10⁶ cells (as opposed to 50 × 10⁶ total thymocytes or 5 × 10⁶ purified HSA⁻CD8⁻ thymocytes) were sufficient to fully prevent the pathogenic effect of 5 × 10⁶ diabetogenic effectors.

The immunopathologic significance of these protective cells is still uncertain. It is tempting, however, to relate them to the resistance to diabetes transfer that develops in NOD mice by 3 wk of age (12), the acceleration of diabetes onset after thymectomy at this same age (i.e., weaning) (20), and the induction of acute diabetes upon cyclophosphamide treatment (19, 38). The main question is then to characterize more precisely the nature of these regulatory cells that are not only found in the thymus, but also in the periphery, i.e., the spleen (F. Lepault, unpublished observations). Indirect data have suggested that they could express a preferential Th2 phenotype. Thus, NOD mice are kept disease free by systemic treatments with IL-4 or IL-10 (39, 40, 41), and a Th1/Th2 shift is evidenced in mice protected from disease by different means, including the administration of β cell autoantigens (2, 42, 43, 44), the injection of CFA or Calmette-Guérin bacillum (2, 45), or the islet-directed transgenic expression of a mutated I-A^g7 (46, 47). However, more direct evidence is lacking, since Th2 cell lines or clones cannot protect from disease transfer by diabetogenic cells (7, 8). Recent data have suggested a major role for TGF-β in this protection. Thus, TGF-β-producing clones are protective in cotransfer, and an Ab to TGF-β is able to break the tolerance induced upon oral insulin administration (48, 49).

Mature CD8⁻ thymocytes constitute a heterogeneous population including TCRαβ⁺ cells derived from both the mainstream and the nonmainstream differentiation pathways. Among nonmainstream thymocytes, a discrete subset was recently described that expresses receptors specific of both T cells (TCR) and NK cells (NK1.1, Ly 49), hence termed NK T cells. NK T cells include CD4⁺ single-positive as well as double-negative cells, which share an unusually restricted Vα14 Vβ8 TCR repertoire. Major additional distinctive features are, first, their restriction by nonpolymorphic MHC class I molecules, CD1 being one major ligand (23, 24, 25, 26, 27), and second, their expression of an activated cell phenotype, CD44⁺, CD62L⁻ (L-selectin, MEL-14⁻), 3G11⁻, CD122⁺ (the β-chain of the IL-2 receptor). Lastly, upon TCR cross-linking, NK T cells produce massive amounts of different cytokines, such as IL-4 and IFN-γ (23, 24, 25, 26, 50). Some authors have suggested that, through the production of IL-4, NK T cells promote Th2-type responses. This was supported by the data showing the absence of IgE production following anti-IgD polyclonal activation in β₂m KO mice that totally lack NK T cells (51). However, the uniqueness of this effect has been recently questioned, since in the same experimental conditions CD1 knockout mice, which also lack NK T cells, can mount IgE responses (52).

Although the NOD strain does not express the NK1.1 allele, NK T cells can be reliably detected using a combination of Abs to CD44, CD62L, 3G11, and Vβ8. Using this approach, we could verify in the thymus of normal C57BL/6 mice that the αβTCR⁺CD44⁺CD62L⁻3G11⁻ subset was NK1.1 positive and concentrated the ability to rapidly produce IL-4 upon TCR cross-linking (21). We showed a clear-cut deficit in the numbers of NK T thymocytes in 3- and 8-wk-old prediabetic NOD mice (21). This quantitative abnormality correlated with a major defect in the IL-4-producing capacity of NOD thymocytes (21, 39). Concordant data have been published by Baxter et al., who showed that CD4⁻CD8 cells are significantly decreased in both the thymus and the periphery of NOD mice (22).

It was thus relevant to directly assess the cotransfer behavior of purified mainstream- vs nonmainstream (NK T)-derived thymocytes. To address this question we took advantage of the absence of L-selectin on NK T cells contrasting with its presence in a large proportion of mature thymocytes (21, 25, 26, 27). Mature NOD CD4⁻CD8⁻ thymocytes were further selected by means of immunomagnetic cell sorting using MEL-14 Ab. The results clearly indicated that the TCRαβ⁺ thymocytes capable of protecting in cotransfer were CD62L⁺ or, alternatively, depended on an TCRαβ⁺CD62L⁺ subset to exert their protective ability. The frequency and the kinetics of diabetes transfer in recipients coinjected with TCRαβ⁺CD62L⁻ NK-like T thymocytes were identical with those seen when injecting diabetogenic cells alone.

These results apparently conflict with those recently reported by Baxter et al. (22). However, these authors used a totally different strategy and concentrated on whole unseparated double-negative thymocytes, which include both mainstream (CD62L⁺) and nonmainstream NK-like T thymocytes (CD62L⁻). The injection of these double-negative thymocytes into 4-wk-old diabetes-prone unmanipulated NOD recipients fully prevented the development of the spontaneous disease (22). On the basis of our present results we would tend to ascribe this protective capacity to the CD62L⁺CD4⁻CD8⁻ subset that is known to include the precursors of the TCRαβ⁺CD62L⁺CD4⁺ subset.

This conclusion is also in keeping with the fact that in the periphery, i.e., the spleen of adult NOD mice, CD4⁺ protective cells also express high levels of L-selectin (F. Lepault, manuscript in preparation) while diabetogenic T splenocytes concentrate among CD62L⁻ (53). These data are also concordant with those showing no or very low levels of L-selectin among intraislet T lymphocytes despite the expression of vascular addressins on the postcapillary venules of inflamed NOD islets (54). Thus, CD62L, initially described as a homing receptor, that is, an addressin ligand (55), seems to provide a unique marker to separate, in both the thymus and the periphery, effector from regulatory cells mediating active tolerance.

Interestingly, this is also the case in PVG.RT1^u rats that, following adult thymectomy and repeated low-dose irradiation, develop autoimmune T cell-mediated IDDM. In these animals disease is effectively prevented by the transfer of TCRαβ⁺CD4⁺RT6⁺CD45RC^lowCD62L⁺ cells (56, 57, 58, 59). Although upon activation, these cells exhibited a preferential Th2 phenotype (producing IL-4 but not IFN-γ) a neutralizing Ab to rat IL-4 was unable to reproducibly abrogate the protection. In this context, it is interesting to note that in our NOD model and confirming our previously published results (21), upon activation by means of TCR cross-linking the protective TCRαβ⁺CD62L⁺ thymocyte subset produced minimal amounts of IL-4 compared with the nonprotective TCRαβ⁺CD62L⁻ NK T subset.

Although the importance of CD4⁺CD62L⁺ T cells in the active tolerance that protects young NOD mice from diabetes, their relationship with the NK T subset remains obscure. It is tempting to see a relationship between the early NK T cell defect consistently found in NOD mice and the failure of the regulatory/protector cells concomitant to diabetes onset. However, the present results argue against a direct role of NK T cells in protection, since they can be physically separated from the subset inhibiting diabetes transfer. It is also difficult to see them as precursors of the protective population, since, at 3 wk of age, the NOD thymus contains a large amount of protector cells and yet is very deficient in NK T cells. More likely would be the involvement of NK T cells in maintenance or long term homeostasis of the protector cells.

One crucial, still unsolved issue concerns the specificity as well as the pathways driving the intrathymic generation of the regulatory T cells described in the various models discussed above. In a series of very elegant experiments, the groups of Le Douarin and Coutinho have demonstrated, initially in birds (quail to chick grafts) and subsequently in mice, that the thymic epithelium could actively select for cells mediating active tolerance (60, 61). It is tempting to relate these observations to various reports showing the expression of different autoantigens within the thymus (59).

To conclude, the NOD mouse is protected from disease onset during the first 2 or 3 mo of life by CD4 regulatory T cells expressing the CD62L⁺ phenotype. Interestingly enough, the thymus-derived CD4⁺ subset identified has the potential to control peripheral, fully differentiated pathogenic autoimmune effectors. Although these CD4 T cells are physically distinct from NK-like T cells, a relationship between these two populations cannot at this point be formally excluded.

We thank I. Cisse for managing the specific pathogen-free NOD mouse colony, C. Gouarin and J. Primo for their excellent technical assistance, D. Broneer for revising the manuscript, and M. Netter and M. Lillié for the art work. We are especially indebted to Sanofi Compagny for providing human rIL-7.