T-Cell Tolerance by Dendritic Cells and Macrophages as a Mechanism for the Major Histocompatibility Complex-Linked Resistance to Autoimmune Diabetes

We used 4.1-NOD mice and recombination activating gene-2 (RAG-2)-deficient 4.1-NOD mice expressing a transgenic, I-A^g7-restricted, β-cell-reactive TCR derived from a pancreatic islet-associated CD4⁺ T-cell clone (NY4.1), as previously described (18). The fine antigenic specificity of this TCR is unknown, but 4.1-CD4⁺ T-cells do not recognize recombinant GAD65, GAD67, preinsulin, ICAp69, phogrin, or IA-2 (S.T., P.S., A.A., J.V., P.S., unpublished observations). NOD.I-Eα^k mice were produced by backcrossing I-Eα^k-transgenic C57BL/6/129 mice (12) (Jackson Laboratory) with NOD mice for 10 generations. N8 mice were typed for microsatellites linked to known Idd loci, as described (19), and mice homozygous for NOD-derived Idd regions were used for further breeding. I-Aαβ^k-transgenic NOD mice (NOD.I-A^k) (13) were from G. Morahan (Walter and Eliza Hall Institute, Melbourne, Australia). I-Aβ^d-transgenic NOD mice (NOD.I-A^d) were from H. McDevitt (Stanford University, Palo Alto, CA) (16). NOD.I-A^g7PD mice were produced by backcrossing the I-Aβ^g7PD transgene of BALB/c.I-A^g7PD mice (20) (from O. Kanagawa, Washington University, St. Louis, MO) onto the NOD background for at least seven generations and typed for homozygosity at known Idd regions. H-2Ma-deficient B6/129 mice (21) were from L. van Kaer (Vanderbilt University, Nashville, TN). I-Aβ^b-deficient C57BL/6 mice were from Taconic Farms (Germantown, NY). H-2^g7-congenic C57BL/10 mice (B10.H2^g7) were from L. Wicker (Merck Research Labs, Rahway, NJ). 4.1-NOD mice (I-Aα^d/I-Aβ^g7) were crossed with NOD.I-Eα^k, NOD.I-A^d, NOD.I-A^g7PD, and C58/j (H-2^k) mice to produce 4.1-NOD.I-Eα^k, 4.1-NOD.I-A^d, 4.1-NOD.I-A^g7PD, and 4.1-(NODxC58/j) F1 mice, respectively. All transgenic mice studied here were hemizygous for both the MHC and TCR transgenes and were housed in a specific pathogen-free facility.

Diabetes was monitored by measuring urine glucose levels with Diastix and a glucometer. Animals were considered diabetic after at least two consecutive readings of 3+.

Chimeras were generated following standard protocols (18). In brief, recipient mice were treated with two doses of 500 rads 3 h apart from a ¹³⁷Cs source (Atomic Energy of Canada, Ottawa, ON) and then transfused with marrow cell suspensions (8–15 × 10⁶ cells/mouse). Chimeric mice were killed 6–8 weeks after transplantation.

The monoclonal antibodies (mAbs) anti-CD8α-PE (53-6.7), anti-CD4-fluorescein isothiocyanate (FITC) (RM4-5 or IM7), anti-Vβ11-FITC (RR3-15), anti-I-Aβ^b-biotin (AF6-120.1), anti-I-E^k-biotin (17-3-3), anti-I-Aβ^k/g7-biotin (10-3.6), anti-CD45R/B220-biotin or -PerCP (RA36B2), anti-CD11c-biotin (HL3), and anti-CD11b-biotin (M1/70) were from PharMingen (San Diego, CA). Anti-L3T4 (CD4)-biotin (YTS191.1) was from Cedarlane (Hornby, ON). Streptavidin-PerCP was from Becton Dickinson (San Jose, CA). Thymocytes and splenocytes were analyzed with a FACScan.

B-cells were purified by positive selection with B220-coated microbeads using the MiniMACS system (Miltenyi Biotec, Auburn, CA). Purity was >95% CD19⁺ or B220⁺ cells. To purify dendritic cells, spleens or thymi (n = 10) were dissected in 10 ml of 100 units/ml collagenase type IV in Hank’s buffered salt solution on ice and then incubated with 400 u/ml collagenase at 37°C for 30 min. Dendritic cells (CD11c⁺) were purified by two rounds of positive selection using anti-CD11c-coated microbeads (Miltenyi-Biotech). Purity was >95% CD11c⁺ cells. Peritoneal macrophages were collected by peritoneal lavage of mice injected intraperitoneally 3 days earlier with 1 ml of 3% thyoglicollate in PBS. The cells were >80% pure as determined by staining with anti-F4/80 and anti-CD11b mAbs.

Splenocytes were depleted of CD8⁺ T-cells using anti-CD8 mAb-coated beads or used to prepare pure CD4⁺ T-cells using the MiniMACs system (Miltenyi Biotec). Pancreatic islets were isolated as described (18). CD4⁺ T-cells (2 × 10⁴ cells) were incubated in duplicate with γ-irradiated (3,000 rad) NOD islet cells (10⁵/well) in 96-well plates for 3 days at 37°C in 5% CO₂, with or without rIL-2 (Takeda, Osaka, Japan). Alloreactivity of mature 4.1-T-cells was assessed by culturing 2 × 10⁴ CD4⁺ T-cells from RAG-2^−/− 4.1-NOD mice with 10⁵ allogeneic dendritic cells, peritoneal macrophages, or B-cells. All cultures were pulsed with 1 μCi of [³H]-thymidine during the last 18 h of culture. Specific proliferation was calculated by subtracting background (counts per minute of cultures only containing islet cells or APCs and counts per minute of cultures of T-cells alone) from islet cell- or APC-induced proliferation (counts per minute of cultures containing T-cells and either islet cells or APCs).

Dulling assays were done by co-culturing 10⁵ APCs with 10⁵ thymocytes from RAG-2-deficient 4.1-NOD mice in triplicate in ∪-bottomed 96-well plates at 37°C, 5% CO₂, for 20 h. At the end of the incubation period, cells were stained with anti-CD4-PE, anti-CD8-FITC, and anti-CD11c-biotin/streptavidin-PerCP, anti-CD11b-FITC, or anti-B220-PerCP and analyzed by flow cytometry.

Freshly isolated islet and islet-associated T-cells (2 × 10⁴) were activated with phorbol-myristate acetate (PMA) (10 ng/ml) and ionomycin (250 ng/ml) for 12 h at 37°C, 5% CO₂. The supernatants were assayed for interleukin (IL)-2, IL-4, and γ-interferon (IFN-γ) content by enzyme-linked immunosorbent assay (R&D Systems, Minneapolis, MN). Intracytoplasmic staining was done using the Cytofix/Cytoperm Plus (with GolgiStop) Kit (PharMingen) and the following mAbs: anti-IL2-PE (JES6-5H4), anti-IL4-PE (11B11), or anti-IFN-γ-PE (XMG1.2).

Mice were given two injections of 200 μl i.v. of a 4-mg/ml solution of 5-bromo-2′-deoxyuridine (BrdU; Calbiochem, La Jolla, CA) 4 h apart. The pancreata and spleens were collected 12 h later. Quantitation of the percentage of islet-infiltrating cells and splenocytes that incorporated BrdU was done on frozen tissue using a BrdU staining kit (Calbiochem). The degree of insulitis was determined on hematoxylin and eosin (H/E)-stained sections, using previously described criteria (18).

Statistical analyses were performed using the Mann-Whitney U and χ² tests.

Thymocytes expressing the 4.1-TCR undergo deletion in H-2^g7/b mice by engaging I-A^b molecules on bone marrow-derived APCs (18,19). To investigate whether this phenotype could also be triggered by other antidiabetogenic MHC class II molecules, we followed the fate of the 4.1-TCR in 4.1-TCR-transgenic NOD mice expressing an I-Eα^k transgene (4.1-NOD.I-Eα^k mice). Nontransgenic NOD mice transcribe and translate I-Eβ^g7 molecules but cannot express I-Eαβ^g7 heterodimers because they carry a nonproductive I-Eα gene (22). Because the introduction of I-Eα transgenes in NOD mice restores I-E expression and renders the mice completely or partially resistant to diabetes (17,23–26), we wondered whether I-E molecules could also trigger the deletion of 4.1-thymocytes.

One-third of 4.1-NOD.I-Eα^k mice (7 of 21) had thymocyte and splenic cytofluorometric profiles that were compatible with complete negative selection of 4.1-thymocytes, as compared with other TCR-transgenic models (27–30). These mice (herein referred to as “deleting”) displayed a significant reduction in the percentage of CD4⁺CD8⁻ thymocytes, a reduction in the percentage of CD4⁺CD8⁻ thymocytes expressing the transgene-encoded Vβ11 element, and an increase in the percentage of CD4⁻CD8⁻ thymocytes when compared with 4.1-NOD mice (Table 1 and Fig. 1A and B, upper panels). In the spleen, deleting 4.1-NOD.I-Eα^k mice had significantly fewer CD4⁺ T-cells and CD4⁺Vβ11⁺ T-cells than 4.1-NOD mice (Table 1 and Fig. 1A and B, lower panels). The few Vβ11⁺CD4⁺ T-cells that matured in deleting mice expressed significantly lower levels of the transgenic TCRβ chain on the cell surface than the Vβ11⁺CD4⁺ T-cells that matured in 4.1-NOD mice (Table 1, Fig. 1A and B, and data not shown). This suggested that in deleting 4.1-NOD.I-Eα^k mice, most CD4⁺ T-cells were selected on endogenous TCR chains that had bypassed allelic exclusion. The remaining two-thirds of 4.1-NOD.I-Eα^k mice (14 of 21) displayed thymocyte and splenic profiles that were similar to those seen in 4.1-NOD mice (Table 1 and Fig. 1A and B), where 4.1-thymocytes undergo positive selection. Individual mice were classified as deleting if the values corresponding to the parameters described above were at least 2 SDs above or below the corresponding values in 4.1-NOD mice. Therefore, I-Eα^k/I-Eβ^g7 heterodimers can trigger massive deletion of 4.1-thymocytes, albeit only in a fraction of the mice.

Proliferation assays using splenic CD4⁺ T-cells from deleting 4.1-NOD.I-Eα^k mice as responders and irradiated NOD or NOD.I-Eα^k islet cells as a source of antigen and APCs confirmed the absence of β-cell-reactive CD4⁺ T-cells in these mice (Fig. 2A). Surprisingly, the splenic CD4⁺ T-cells of nondeleting 4.1-NOD.I-Eα^k mice also displayed proliferative unresponsiveness to islet stimulation in vitro (Fig. 2B, left panel). This unresponsiveness was caused by anergy (rather than deletion) because the addition of rIL-2 to the cultures restored the CD4⁺ cells’ proliferative responsiveness to antigen (Fig. 2B, right panel). The splenic CD4⁺ T-cells from both 4.1-NOD and 4.1-NOD.I-Eα^k mice did not express activation markers, including CD25, CD69, or CD44 (not shown), suggesting that they are naïve and quiescent.

Studies of 4.1-TCR-transgenic (NODxC58/j) F1 mice, which are H-2^g7/k heterozygous and can thus express I-Eα^k/I-Eβ^k and I-Eα^k/I-Eβ^g7 heterodimers, indicated that these two phenotypes (deletion and anergy) were not an artifact of transgenesis. As shown in Table 1 and Fig. 1C, 4.1-(NODxC58/j) F1 mice (also referred to as 4.1-H-2^g7/k mice) displayed the same two phenotypes. Altogether, 8 of 14 mice had a phenotype compatible with complete deletion of 4.1-thymocytes, whereas the remaining 6 mice had a phenotype compatible with positive selection (Table 1 and Fig. 1C; mice classified into “deleting” or “nondeleting” groups based on the criteria described above). The prevalence of deleters in the 4.1-(NODxC58/j) F1 strain was slightly greater than in the 4.1-NOD.I-Eα^k strain, but the differences did not reach statistical significance. The thymocyte deletion in 4.1-(NODxC58/j) F1 mice was triggered by endogenous I-E^k molecules and/or I-Eα^k/I-Eβ^g7 heterodimers, rather than by I-A^k molecules, because deletion was not seen in I-Aα^k/I-Aβ^k-transgenic 4.1-NOD mice (Table 1 and Fig. 1D). Deletion of 4.1-thymocytes in 4.1-(NODxC58/j) F1 and 4.1-NOD.I-Eα^k mice was not mediated by Vβ11-binding superantigens, because non-TCR-transgenic (NODxC58/j) F1 and NOD.I-Eα^k mice had as many thymic and splenic Vβ11⁺CD4⁺ T-cells as NOD mice (data not shown). This is not surprising because neither (NODxC58/j) F1 nor NOD.I-Eα^k mice express mouse mammary tumor virus superantigens that are tolerogenic for Vβ11⁺ T-cells (31). Furthermore, deletion required expression of both chains of the 4.1-TCR, because 4.1-TCRβ-transgenic (NODxC58/j) F1 mice had as many Vβ11⁺CD4⁺ thymocytes as 4.1-TCRβ-transgenic NOD mice (73 ± 8 vs. 75 ± 3%, respectively). Together, these data indicate that I-Eα^k/I-Eβ^g7 molecules, regardless of whether they are transgene-encoded, can efficiently tolerize thymocytes expressing the 4.1-TCR.

To investigate whether the I-E^k-induced deletion of 4.1-thymocytes was triggered by hematopoietic cells, we followed the fate of 4.1-CD4⁺ thymocytes in lethally irradiated NOD mice transfused with marrow from deleting or nondeleting 4.1-NOD.I-Eα^k mice. The phenotype of the bone marrow donors was a good, but imprecise, predictor of the phenotype seen in the chimeras. Altogether, 7 of 8 chimeras reconstituted with marrow from deleting donors deleted 4.1-thymocytes (P < 0.01 vs. 4.1-NOD → NOD chimeras), whereas 16 of 20 chimeras reconstituted with marrow from nondeleting donors displayed nondeleting flow cytometric profiles. Therefore, the phenotype of the chimeras is largely imprinted in the donor marrow, but marrow from single nondeleting or deleting donor mice can give rise to both nondeleting and deleting chimeras.

To determine the effects of thymocyte tolerance on 4.1-CD4⁺ T-cell-induced diabetes, we monitored 4.1-NOD.I-Eα^k mice for the development of diabetes (Table 2). As shown in Table 2, 4.1-NOD.I-Eα^k mice were nearly as diabetes-resistant as non-TCR-transgenic NOD.I-Eα^k mice, regardless of whether they were deleting or nondeleting.

Histological studies confirmed that deleting 4.1-NOD.I-Eα^k mice did not develop insulitis, as expected (Table 2 and Fig. 3D). Surprisingly, however, nondeleting 4.1-NOD.I-Eα^k mice developed significant insulitis (Table 2 and Fig. 3E), despite the fact that their peripheral CD4⁺ T-cells were unresponsive to antigenic stimulation in vitro (see above). The insulitis lesions of nondeleting 4.1-NOD.I-Eα^k mice were caused by 4.1-CD4⁺ T-cells rather than by endogenous (nontransgenic) T-cells, because non-TCR-transgenic NOD.I-Eα^k mice are completely resistant to insulitis (Table 2 and Fig. 3B). To ascertain whether these lesions were caused by an accumulation of anergic, nonproliferating 4.1-CD4⁺ T-cells in islets, we investigated whether the islet-infiltrating cells of nondeleting 4.1-NOD.I-Eα^k mice could incorporate exogenous BrdU in situ. Surprisingly, islets from BrdU-injected, nondeleting 4.1-NOD.I-Eα^k mice contained as many BrdU⁺ cells as islets from BrdU-injected 4.1-NOD mice (10 ± 1.4 vs. 8.6 ± 0.8%) (Fig. 3F and G). Thus, in nondeleting 4.1-NOD.I-Eα^k mice, 4.1-CD4⁺ T-cells are responsive to antigenic stimulation in vivo, yet they are unable to cause disease.

To determine whether the inability of the islet-infiltrating T-cells of these mice to trigger β-cell damage was caused by the recruitment of immune-deviated 4.1-CD4⁺ cells, we compared the ability of their islet-associated T-cells to secrete IL-2, IFN-γ, and IL-4 upon stimulation with PMA and ionomycin. The islet-associated T-cells of 4.1-NOD and nondeleting 4.1-NOD.I-Eα^k mice produced IL-2 and IFN-γ (232 ± 130 vs. 66 ± 15 pg/ml and 1,387 ± 735 vs. 205 ± 43 pg/ml, respectively) but virtually undetectable levels of IL-4 (9 ± 5 vs. 6 ± 1 pg/ml). These results were confirmed by intracytoplasmic staining of islet-associated CD4⁺ T-cells with anti-IL-2, anti-IFN-γ, and IL-4 mAbs (data not shown). Thus, despite their ability to proliferate against islet antigen in vivo and to secrete pro-inflammatory cytokines, the “anergic” 4.1-CD4⁺ T-cells of nondeleting 4.1-NOD.I-Eα^k mice cannot cause clinically significant β-cell damage in vivo. This is in contrast to the exquisite diabetogenic activity of the islet-associated 4.1-CD4⁺ T-cells of 4.1-NOD mice, which are highly cytotoxic to β-cells, even in the absence of B-cells and endogenous T-cells (32).

We next asked whether the ability of I-A^b (18) and I-E to induce 4.1-thymocyte tolerance was shared by other antidiabetogenic class II molecules (i.e., I-A^d). We crossed 4.1-NOD mice (expressing I-Aα^d and I-Aβ^g7) with I-Aβ^d-transgenic NOD mice (NOD.I-A^d) to produce 4.1-NOD.I-A^d mice (expressing I-Aα^d, I-Aβ^g7, and I-Aβ^d molecules). As shown in Table 1 and Fig. 4, these mice displayed cytofluorometric profiles compatible with negative selection of 4.1-thymocytes. When compared with 4.1-NOD mice, 4.1-NOD.I-A^d mice had significant reductions in the percentages of CD4⁺CD8⁻ and Vβ11⁺CD4⁺CD8⁻ thymocytes and a significant increase in the percentage of CD4⁻CD8⁻ thymocytes (Table 1 and Fig. 4A, left panels). Deletion was incomplete but was seen in all of the mice that were studied. As was the case for 4.1-NOD.I-Eα^k mice, deletion of 4.1-thymocytes in 4.1-NOD.I-A^d mice was not triggered by endogenous superantigens because non-TCR-transgenic NOD.I-A^d and NOD mice had comparable numbers of Vβ11⁺CD4⁺ thymocytes and splenocytes (not shown). Studies of bone marrow chimeras indicated that I-A^d-induced deletion of 4.1-thymocytes was also triggered by bone marrow-derived APCs (eight of eight 4.1-NOD.I-Ad → NOD chimeras had a deleting phenotype versus none of seven chimeras generated with marrow from 4.1-NOD mice, P < 0.01).

The splenic CD4⁺ T-cells of 4.1-NOD.I-A^d mice did not proliferate in response to islet antigen, even in the presence of rIL-2 (Fig. 4B). This suggested that the peripheral frequency of mature 4.1-CD4⁺ T-cells in these mice was too low to be detected in short-term proliferation assays. To determine whether 4.1-thymocyte deletion in 4.1-NOD.I-A^d mice affected the natural history of diabetes, we followed 19 female 4.1-NOD.I-A^d mice for the development of diabetes. As shown in Table 2, very few mice developed diabetes, and those that did, did so significantly later than 4.1-NOD mice (Table 2). Surprisingly, the mice that did not become diabetic developed insulitis (Table 2). The insulitic T-cells of these mice were actively proliferating in situ because they efficiently incorporated BrdU (data not shown). However, they were not immune-deviated because their cytokine profiles were similar to those displayed by the insulitic T-cells of 4.1-NOD mice (described above and data not shown). Therefore, 4.1-NOD.I-A^d mice export a very small number of 4.1-CD4⁺ T-cells to the periphery, and although these cells can trigger insulitis, they cannot efficiently cause diabetes.

Comparison of the amino acid sequences of the class II molecules that can tolerize 4.1-thymocytes suggested an association between tolerogenic activity and the presence of certain residues between positions 56 and 67 of the different β-chains. No other regions in I-Aα or I-Aβ chains had sequence motifs that were either shared by most deleting class II molecules or absent from nondeleting class II molecules. This association was intriguing because this motif contains proline and aspartic acid at positions 56 and 57, which are epidemiologically associated with the ability of protective MHC molecules to afford diabetes resistance in both humans and mice (3,4).

To investigate the independent contribution of these two residues to the 4.1-tolerogenic activity of antidiabetogenic MHC class II molecules, we followed the fate of the 4.1-TCR in I-A^g7PD-transgenic 4.1-NOD mice (4.1-NOD.I-A^g7PD). Unlike I-A^g7 molecules, I-A^g7PD molecules carry proline and aspartic acid at I-Aβ chain positions 56 and 57 and have a strong antidiabetogenic activity in NOD mice (3,4). Flow cytometry studies of 4.1-NOD.I-A^g7PD mice revealed that, unlike I-A^b, I-E, and I-A^d, I-A^g7PD did not cause massive deletion of 4.1-thymocytes (Table 1 and Fig. 4A, right panels). However, these mice had reduced CD4⁺-to-CD4⁻CD8⁻ thymocyte ratios when compared with 4.1-NOD mice (Table 1 and Figs. 1 and 4), and their peripheral CD4⁺ T-cells were less responsive to antigenic stimulation than the peripheral CD4⁺ T-cells of 4.1-NOD mice (in the absence but not the presence of rIL-2) (Fig. 4B). Interestingly, 4.1-NOD.I-A^g7PD mice developed a significantly reduced incidence and a significantly delayed onset of diabetes when compared with 4.1-NOD mice (Table 2). The mice that did not develop diabetes developed insulitis. As was the case for 4.1-NOD.I-A^d and nondeleting 4.1-NOD.I-Εα^k mice, the insulitic T-cells of these mice were not immune-deviated (not shown). Therefore, the presence of proline and aspartic acid at I-Aβ positions 56 and 57 has a protective effect against 4.1-CD4⁺ T-cell-induced diabetes through a mechanism that does not involve massive deletion of 4.1-thymocytes or their immune deviation.

The above data show that different MHC class II molecules can tolerize a single, highly diabetogenic I-A^g7-restricted TCR. We have previously shown that one of these class II molecules, I-A^b, can trigger the deletion of 4.1-thymocytes in the thymic medulla but cannot restrict their positive selection in the thymic cortex (19). To determine whether the same was true for I-A^d, I-E^d, and I-A^g7PD, we investigated whether these molecules could mediate the positive selection of 4.1-thymocytes on thymic epithelial cells. We followed the fate of the 4.1-TCR in lethally irradiated NOD (I-A^g7), BALB/c (I-A^d and I-E^d), or BALB/c.I-A^g7PD mice (I-A^d/g7PD and I-E^d) that had been reconstituted with 4.1-NOD marrow. These chimeras expressed I-A^g7, I-A^g7PD, and/or I-A^d and I-E^d molecules on radioresistant thymic epithelial cells but only expressed I-A^g7 on the radiosensitive, bone marrow-derived APCs of the thymic medulla. As shown in Fig. 5A (upper right panel), thymi of 4.1-NOD → BALB/c chimeras had profiles that were similar to those seen in MHC class II-deficient 4.1-NOD mice (4.1-I-A^0/0 in Fig. 5A, lower right panel), where 4.1-thymocytes undergo massive developmental arrest. In contrast, thymi from 4.1-NOD → BALB/c.I-A^g7PD mice, which expressed I-A^g7PD, I-A^d, and I-E^d molecules on thymic epithelial cells and I-A^g7 on marrow-derived APCs, had profiles compatible with positive selection of the 4.1-TCR; 4.1-NOD → BALB/c.I-A^g7PD chimeras had more thymocytes (22 ± 4 vs. 5 ± 0.7 × 10⁶; P < 0.003) and greater CD4⁺CD8⁺-to-CD4⁻CD8⁻ thymocyte ratios than 4.1-NOD → BALB/c chimeras (3.2 ± 0.5 vs. 1 ± 0.3; P < 0.007) (Fig. 5A). As expected, splenic CD4⁺ T-cells from 4.1-NOD → BALB/c.I-A^g7PD chimeras proliferated almost as well as splenic CD4⁺ T-cells from 4.1-NOD mice in response to islet antigen, both in the presence and absence of rIL-2 (Fig. 5B). Therefore, 4.1-thymocytes can recognize I-A^g7PD (but not I-A^d or I-E^d) molecules on thymic epithelial cells. This demonstrates that, as was the case for I-A^b (19), deleting class II molecules can restrict the negative selection of 4.1-thymocytes in the medulla but not their positive selection in the thymic cortex.

To determine the nature of the tolerogenic APC population, we compared the ability of purified thymic dendritic cells, splenic dendritic cells, splenic B-cells, and peritoneal exudate macrophages from several mouse strains to downregulate CD4 and CD8 co-receptors from monospecific 4.1-CD4^hiCD8^hi thymocytes (an ex vivo assay of thymocyte tolerance referred to as “dulling” [33–35]). We focused these studies on H-2^b+ mouse strains because I-A^b triggers complete deletion of the 4.1-TCR in 100% of the mice. As shown in Fig. 6A and B, dendritic cells from C57BL/6 (I-A^b) mice, but not from NOD (I-A^g7), I-Aβ^b-deficient C57BL/6 (I-A^−/−) or C57BL/10.H2^g7 (I-A^g7) mice, triggered dulling of double-positive thymocytes from RAG-2-deficient 4.1-NOD mice, which can only express the 4.1-TCR. To ascertain whether the dulling activity of I-A^b molecules was peptide-specific, we also tested thymic and splenic dendritic cells from H-2Ma-deficient C57BL/6 mice. Professional APCs from H-2Ma-deficient C57BL/6 mice express normal levels of I-A^b (Fig. 6D), but these molecules present a very limited repertoire of peptides (mostly invariant chain-derived [CLIP]). As shown in Fig. 6A, H-2Ma-deficient dendritic cells could not downregulate CD4 and CD8 from 4.1-CD4⁺CD8⁺ thymocytes. Thus, the ability of 4.1-thymocytes to recognize I-A^b molecules on dendritic cells is peptide-specific.

Subsequent experiments with macrophages and splenic B-cells indicated that the ability of dendritic cells from C57BL/6 mice to downregulate CD4 and CD8 from 4.1-CD4⁺CD8⁺ thymocytes was not a property of all professional APCs. Dulling was efficiently induced by peritoneal exudate macrophages but not by splenic B-cells (Fig. 6C), despite the fact that the levels of I-A^b on these two APC types were virtually identical (Fig. 6C). Finally, proliferation and cytokine secretion assays using splenic 4.1-CD4⁺ T-cells from 4.1-NOD mice as responders and the different APC types discussed above revealed that the peptide/I-A^b complex that triggers dulling of immature 4.1-thymocytes cannot trigger functional responses from mature 4.1-CD4⁺ T-cells (data not shown). Together, these data suggest that 4.1-thymocytes undergo deletion by recognizing a tolerogenic peptide/I-A^b complex on dendritic cells and macrophages but not B-cells.

Bone marrow-derived APCs expressing the antidiabetogenic class II molecule I-A^b trigger the complete deletion of thymocytes bearing the highly diabetogenic, I-A^g7-restricted 4.1-TCR (18,19). Because the diabetes resistance afforded by protective MHC class II molecules (including I-A^b) in non-TCR-transgenic mice is also triggered by bone marrow-derived cells (5–11,17), the above observation suggested a general mechanism for the MHC class II-linked resistance to autoimmunity. Here, we have tested the ability of three different antidiabetogenic class II molecules (I-Eα^k/I-Eβ^g7, I-A^d, and I-A^g7PD) to tolerize 4.1-CD4⁺ T-cells and/or to prevent 4.1-CD4⁺ T-cell-induced diabetes. We show that 4.1-CD4⁺ T-cells undergo different forms of tolerance in mice expressing I-Eα^k, I-A^d, or I-A^g7PD molecules, owing to the ability of immature 4.1-CD4⁺CD8⁺ thymocytes to recognize protective peptide/MHC class II complexes on dendritic cells and macrophages but not on B-cells or thymic epithelial cells. These data provide but one explanation as to how structurally diverse MHC class II molecules might provide dominant resistance against type 1 diabetes.

Expression of transgenic (and wild-type) I-Eα^k molecules in 4.1-NOD mice restored I-E expression and triggered the deletion of 4.1-CD4⁺ thymocytes in approximately one-third of the mice. The remaining two-thirds of mice positively selected 4.1-CD4⁺ T-cells, but these cells were unresponsive to antigenic stimulation in vitro. Experiments with bone marrow chimeras indicated that the deletion of 4.1-thymocytes in 4.1-NOD.I-Eα^k mice was a partially stochastic process that was mediated by bone marrow-derived cells. The mechanisms underlying the incomplete penetrance of this phenotype (deletion) are unclear. However, because the incomplete penetrance of the deleting phenotype seen in 4.1-NOD.I-Eα^k mice was also seen in 4.1-(NODxC58/j) F1 mice, it is unlikely that it can be accounted for by introduction of a genetic contaminant derived from the I-Eα^k donating strain. Furthermore, we have indirect evidence suggesting that it may be the result of differences in the number of APCs expressing I-E molecules (data not shown). It is also worth noting that whereas overexpression of an I-Eα^d transgene completely prevented diabetes in non-TCR-transgenic NOD mice, expression of this transgene at physiological levels did not completely abrogate diabetes susceptibility in NOD mice (17).

As expected, deletion of 4.1-thymocytes in deleting 4.1-NOD.I-Eα^k mice afforded the mice resistance to both insulitis and diabetes. Although nondeleting 4.1-NOD.I-Eα^k mice (bearing anergic 4.1-CD4⁺ T-cells) were also diabetes resistant, they developed insulitis. Surprisingly, the insulitis lesions of these mice were caused by actively proliferating 4.1-CD4⁺ T-cells. This indicated that 4.1-CD4⁺ T-cells maturing in nondeleting 4.1-NOD.I-Eα^k mice were only anergic to in vitro but not in vivo stimuli. The diabetes resistance of nondeleting 4.1-NOD.I-Eα^k mice was not caused by MHC-induced differentiation of 4.1-CD4⁺ T-cells into Th2 cells or by recruitment of regulatory Th2 cells to islets, because their islet-associated T-cells produced IL-2 and IFN-γ but not IL-4. Furthermore, it was not caused by the inability of 4.1-CD4⁺ T-cells to recognize islet autoantigen, because 4.1-CD4⁺ T-cells from 4.1-NOD recognized NOD.I-Eα^k (I-A^g7+/I-E⁺) and NOD (I-A^g7+/I-E⁻) islets equally well (S.T., P.Se., A.A., J.V., P.Sa., unpublished observations). These data suggested that in nondeleting 4.1-NOD.I-Eα^k mice, the anergyzing stimulus was sufficiently strong to prevent the differentiation of 4.1-CD4⁺ T-cells into β-cell-cytotoxic T-cells but not strong enough to prevent their proliferation in response to strong in vivo stimuli (i.e., autoantigen-loaded dendritic cells). The fact that nondeleting 4.1-NOD.I-Eα^k mice develop insulitis (unlike non-TCR-transgenic NOD.I-Eα^k mice) does not necessarily argue against a role for thymocyte tolerance as a mechanism for the MHC-linked resistance to diabetes in non-TCR-transgenic mice. The high frequency of 4.1-thymocytes in 4.1-TCR-transgenic mice may, in some mice, overwhelm the tolerogenic machinery that in non-TCR-transgenics would target a significantly smaller population of 4.1-like T-cells.

Although the overall three-dimensional structure of I-A^b and I-E^k molecules is similar, the amino acid sequences of their α- and β-chains are very different (36). This prompted us to investigate whether the ability of these two MHC molecules to tolerize 4.1-CD4⁺ thymocytes might be shared by other antidiabetogenic MHC class II molecules, such as I-A^d and I-A^g7PD. The 4.1-NOD.I-A^d mice displayed a phenotype that was similar to that seen in deleting 4.1-NOD.I-Eα^k mice, with two major differences—the deleting phenotype of 4.1-NOD.I-A^d mice was incomplete, but its penetrance through the mouse population was complete. Unlike deleting 4.1-NOD.I-Eα^k mice, 4.1-NOD.I-A^d mice developed insulitis, and a small percentage of mice even developed diabetes. The mechanisms underlying these phenotypic differences are unknown but may be related to differences in the levels or timing of expression of I-E and I-A on thymic APCs. Subsequent studies of 4.1-NOD.I-A^g7PD mice suggested that this ability of protective class II molecules to prevent 4.1-CD4⁺ T-cell-induced diabetes was not unique to I-A^b, I-E^k, and I-A^d. The 4.1-NOD.I-A^g7PD mice displayed reduced CD4⁺-to-CD4⁻CD8⁻ thymocyte ratios when compared with 4.1-NOD mice, and 4.1-CD4⁺ T-cells maturing in 4.1-NOD.I-A^g7PD mice consistently showed a slight reduction in proliferative responsiveness to antigen, which could be corrected by the addition of IL-2. The 4.1-NOD.I-A^g7PD mice developed moderate insulitis but a significantly reduced incidence (and delayed onset) of diabetes. It could be argued that the antidiabetogenic effect of the transgenic MHC class II molecules tested here was an artifact of transgenesis (i.e., by causing a reduction in the percentage of B-cells) (37). However, this is highly unlikely because 4.1-thymocyte tolerance is caused by both transgenic and wild-type MHC class II molecules. We cannot exclude the possibility that the different degrees of tolerance observed in this study reflect the effect of different levels of expression of the diverse transgenes in the relevant cell types. However, the fact that two different wild-type MHC class II haplotypes (H-2^b and H-2^k) have different degrees of tolerogenic activity on 4.1-thymocytes argues, in part, against this possibility.

The extensive MHC promiscuity of the 4.1-TCR during thymocyte development appears to result from its ability to recognize tolerogenic peptide/MHC class II complexes on dendritic cells and macrophages. The diabetogenic 4.1-TCR is not a classic alloreactive TCR, however, because it is only promiscuous for MHC molecules expressed on dendritic cells and macrophages, not on B-cells or thymic epithelial cells. It should also be pointed out that 4.1-thymocytes undergo neither positive nor negative selection in mice expressing deleting class II molecules (I-A^b [19], I-A^d, and I-E) exclusively on thymic epithelial cells. The only exception to this is the ability of the 4.1-TCR to undergo weak positive selection in chimeras whose thymic epithelial cells express I-A^g7PD (but not I-A^g7), as has been reported for another β-cell-reactive TCR (BDC-2.5) (20). It is also clear that the MHC promiscuity of 4.1-thymocytes is peptide-specific, because H-2Ma-deficient thymic and splenic dendritic cells, which can almost exclusively express I-A^b molecules bound to CLIP, cannot trigger dulling of 4.1-CD4⁺CD8⁺ thymocytes. When taken together, these data suggest that the MHC-induced deletion of 4.1-thymocytes is mediated by one or more peptides that are selectively expressed in thymic bone marrow-derived APCs, as we had previously suggested for I-A^b (19).

Comparison of the amino acid sequences of the α- and β-chains of the nondeleting and deleting MHC class II molecules that we have tested so far did not reveal any obvious sequence homologies with the exception of I-Aβ or I-Eβ chain positions 55–67. Deleting MHC class II molecules (I-A^q, I-A^d, I-A^b, and I-E) share a homologous 56–67 motif that is absent in weakly tolerogenic (I-A^k and I-A^g7PD) or nontolerogenic (I-A^g7 and I-A^s) class II molecules. This sequence contains proline and aspartic acid at positions 56 and 57 (as opposed to histidine and serine in I-A^g7) and two insertions at positions 65 and 67. Because I-A^b, I-A^d, and I-A^q have an identical 56–67 sequence but variable tolerogenic activity, the tolerogenicity of this motif may be modulated by other factors, such as other β-chain residues, the nature of the α-chains, or the timing and levels of expression of the different MHC molecules. This hypothetical structure/function association is tantalizing, as the MHC-linked resistance to type 1 diabetes in both humans and mice is associated with polymorphisms at and around position 57 (3,4). The hypothetical contribution of residues at β-chain positions 56–67 to 4.1-thymocyte deletion is compatible with the recently described influence of amino acid substitutions at positions 56, 57, and 61 (different in I-A^g7 vs. I-A^d, I-A^k, and I-E^k) on structure (38,39). Our data are therefore compatible with a model in which structural differences conferred by polymorphisms between positions 56 and 67 would allow antidiabetogenic class II molecules to present tolerogenic peptides that cannot be presented at all by I-A^g7 and only inefficiently by I-A^k and I-A^g7PD (to certain highly diabetogenic T-cells).

In sum, our results lend support to the hypothesis that the MHC-associated resistance of non-TCR-transgenic mice (and perhaps humans, via DQB1*0602) to autoimmune diabetes might be mediated by thymocyte tolerance. Because it is highly unlikely that all I-A^g7-restricted diabetogenic effector T-cells in NOD mice will also recognize the “4.1-thymocyte-deleting” peptide(s) in the context of antidiabetogenic MHC class II molecules, we propose that 4.1-CD4⁺ T-cells are representative of a group of highly diabetogenic and MHC-promiscuous T-cells that, for as yet unknown reasons, would play a critical role in diabetogenesis. This would explain the presence of mildly insulitogenic, but not diabetogenic, T-cells in congenic NOD.H-2^g7/b, NOD.H-2^g7/q, or NOD.H-2^g7/nb1 mice (14,40,41); I-A^d-transgenic NOD mice (16); NOD mice reconstituted with bone marrow from I-Eα-transgenic NOD mice (11); and I-A^k-transgenic NOD mice (13,42). This hypothesis does not rule out the involvement of other mechanisms of MHC-associated protection from diabetes (i.e., immunoregulation), especially because the phenomena described here was observed with a single TCR. However, it is compelling because the geography and timing of negative selection of diabetogenic thymocytes in 4.1-NOD mice are in agreement with the widely accepted notion that the factors underlying the MHC-linked resistance to diabetes reside in the bone marrow and are linked to polymorphisms at around I-A/I-Eβ position 57 (3,4). Furthermore, it is attractive because it invokes a single mechanism by which multiple, structurally diverse MHC molecules can afford dominant resistance to a single autoimmune disease.

This work was supported by the Canadian Institutes of Health Research. P.Sa. is a Senior Scholar of the Alberta Heritage Foundation for Medical Research (AHFMR). A.A. is supported by the AHFMR and the Juvenile Diabetes Foundation International; P.Se. is supported by the AHFMR; and J.V. was funded by a fellowship from the Canadian Diabetes Association.

We thank H. McDevitt, O. Kanagawa, L. van Kaer, and L. Wicker for providing mice; T. Utsugi for providing Takeda rIL-2; S. Bou, M. Deuma, and S. Culp for technical assistance; Y. Yang and J. Elliott for helpful comments on the manuscript; and L. Bryant for flow cytometry.