Enhanced Pathogenicity of Diabetogenic T Cells Escaping a Non-MHC Gene-Controlled Near Death Experience¹

Type 1 diabetes (T1D)³ in both humans and NOD mice results from the autoimmune destruction of insulin-producing pancreatic β cells by T cells (1, 2). Similar to the case in humans, linkage analyses have demonstrated that the development and subsequent activation of diabetogenic T cells in NOD mice are controlled by multiple susceptibility (Idd) genes (2, 3). However, in both humans and NOD mice, the primary genetic component of T1D susceptibility is provided by certain MHC haplotypes (2, 3). In humans, specific combinations of HLA-DQ and HLA-DR class II alleles provide a large component of T1D susceptibility by mediating β cell autoreactive CD4 T cell responses (4). A particularly strong class II contributor to T1D in humans is the unusual DQ8 variant characterized by histidine and serine, rather than the more common proline and aspartic acid residues at positions 56 and 57 of the β-chain (4, 5). Similarly, T1D development in NOD mice is also dependent upon the unusual class II variants characterizing their H2^g7 MHC haplotype. Specifically, this entails a need for NOD APCs to homozygously express the unusual H2-A^g7 class II gene product (homologue of human DQ8) in the absence of H2-E (homologue of human DR) class II molecules.

Given the critical contributions of certain MHC class II molecules, it is not surprising that autoreactive CD4 T cells are essential for the development of T1D in both humans and NOD mice. Nonetheless, it is now clear that while they represent quite common variants shared by many strains lacking autoimmune proclivity, the K^d and/or D^b class I molecules encoded by the H2^g7 haplotype mediate autoreactive CD8 T cell responses that are also essential to T1D development (2, 6). There is also epidemiological evidence that the risk of T1D development in humans is increased when certain MHC class I alleles are expressed in conjunction with particular MHC class II susceptibility alleles, and perhaps other genes (7). It was subsequently shown that when transgenically expressed in NOD mice, the common human HLA-A2.1 class I variant stimulates β cell autoreactive CD8 T cell responses, resulting in an accelerated onset of T1D (8). This provided the first functional evidence that in the proper genetic context, some common human MHC class I variants are diabetogenic contributors.

An open question is what allows common MHC class I variants to stimulate diabetogenic CD8 T cell responses in some individuals, but not others. In this study, we tested whether the common class I variants of the H2^g7 haplotype aberrantly acquire autoimmune functions in NOD mice through complex interactions with non-MHC Idd susceptibility genes. In particular, we wished to determine whether non-MHC genes of NOD origin impair the tolerogenic mechanisms that would normally cause potentially diabetogenic H2^g7 MHC class I-restricted T cells to be deleted intrathymically or to be functionally anergized in the periphery.

NOD/LtDvs mice are maintained at The Jackson Laboratory (Bar Harbor, ME) by brother-sister mating. NOD mice separately carrying transgenes encoding the TCR α (Vα8)- or β (Vβ2)-chain of the β cell autoreactive CD8 T clone AI4 have been previously described (9). The C57BL/6.H2^g7 congenic stock (B6.H2^g7) is maintained at the N8 backcross generation (10). The AI4 α- and β-chain transgenes were congenically transferred to the B6.H2^g7 background. At the N7 backcross generation, the B6.H2^g7.AI4α and B6.H2^g7.AI4β stocks were both shown to be fixed to homozygosity for markers delineating all known B6-derived Idd loci (11). An intercross strategy was then used to produce either NOD or B6.H2^g7 mice carrying both the AI4 α- and β-chain transgenes (designated NOD.AI4 and B6.H2^g7.AI4). Lymphocyte-deficient NOD-scid mice (12) are maintained at the N11 backcross generation. MHC class I-deficient NOD.β₂m^null (β₂-microglobulin^null) congenic mice (13) are also maintained at the N11 backcross generation. Inbred mice with a mixed C57BL/6 and DBA/2 genetic background, including an H2^b MHC haplotype, and expressing a transgenic (Tg) TCR specific for a D^b class I-restricted lymphocytic choriomeningitis virus (LCMV) peptide (designated in this work as B6,D2.LCMV TCR Tg mice), have been previously described (14). Development of an N9 backcross stock of NOD.LCMV TCR Tg mice and an N10 NOD.scid.LCMV TCR Tg stock has also been previously described (15). All mice were housed under specific pathogen-free conditions and allowed free access to food (National Institutes of Health diet 31A; Purina, St. Louis, MO) and acidified drinking water.

T1D development was defined by glycosuric values of >3, as assessed with Ames Diastix (Miles Diagnostics, Elkhart, IN).

Peripheral blood leukocytes, thymocytes, or splenic leukocytes from 4- to 6-wk-old mice were enumerated and characterized by multicolor FACS analyses (FACSCalibur; BD Biosciences, San Jose, CA) using the CellQuest 3.0 data reduction system. Expression of the Tg AI4 TCR α- and β-chain was detected by respective use of red fluorescent PE- or green fluorescent FITC-conjugated Abs specific for Vα8 (B21.14) or Vβ2 (B20.6) elements. Three-color flow cytometric analyses were performed to determine the proportions of CD4/CD8 double-positive (DP) and CD8 single-positive (SP) thymocytes from the indicated mice that were TCRαβ and/or TCR Vα8 positive. CD8 expression was detected with the mAb 53-6.72 conjugated to a blue fluorescent allophycocyanin (AP) tag. CD4 expression was detected with the mAb GK1.5 conjugated to either PE or FITC. The presence of TCRαβ complexes on CD4/CD8 DP thymocytes was detected with the FITC-conjugated mAb H57-597. This Ab was also used to assess TCRαβ expression levels on peripheral splenic T cells. Splenic T cells expressing the Tg AI4 TCR were detected by staining with the PE-labeled TCR Vα8-specific Ab mixed with the AP-labeled CD8 Ab. DP thymocytes expressing the LCMV TCR were quantified with CD4-, CD8-, and Vβ8 (F23.2)-specific mAbs, respectively, conjugated to FITC, AP, and PE.

CD8 T cells were purified from the spleens of 5- to 6-wk-old NOD.AI4 and B6.H2^g7.AI4 mice by the previously described magnetic bead-based negative selection approach (15). Flow cytometric analyses indicated CD8 T cell purity was >85%. Triplicate aliquots of 5 × 10⁴ CD8 T cells were seeded into flat-bottom 96-well plates in 200 μl of the previously described medium (16) also containing 2 × 10⁵ irradiated (2000 R) NOD or B6.H2^g7 splenic leukocytes as APC and the previously described AI4 mimotope peptide (YFIENYLEL) (17) at concentrations ranging from 0 to 100 nM. Following a 48-h incubation at 37°C in a 95% air/5% CO2 humidified atmosphere, the cultures were pulsed with 1 μCi/well [³H]thymidine for an additional 24 h. The cultures were then harvested, and [³H]thymidine incorporation was determined using a LKB Betaplate 1205 system (LKB Instruments, Gaithersburg, MD). Data are presented as mean cpm ± SEM of triplicate cultures.

Thymocyte suspensions were prepared from B6,D2.LCMV TCR Tg and NOD.LCMV TCR Tg mice by collagenase digestion to retain the thymic epithelium cells and APCs. Cells were incubated for 24 h with D^b-binding LCMV agonist (KAVYNFATM) or irrelevant control (ASNENMETM) peptides at concentrations ranging from 0 to 10 μM. The proportion of remaining viable DP thymocytes was then determined by FACS analysis by staining with propidium iodide and CD4, CD8, and Vβ8 Abs. The intracellular Nur77 staining was performed, as previously described (18), using anti-Nur77 (12.14) and FITC anti-IgG1 (A85-1) (BD Biosciences).

Bone marrow from B6.H2^g7.AI4 and NOD.AI4 mice was depleted of mature CD4 and CD8 T cells, as previously described (8). The indicated bone marrow cells (2 × 10⁷) were injected i.v. into 4- to 6-wk-old NOD and B6.H2^g7 recipients that had been sublethally irradiated (600 R from a ¹³⁷Cs source) and subsequently injected i.p. with 0.5 mg of the CD40L-specific mAb MR1 on days 0, 3, and 14 postreconstitution. Chimeras were assessed for AI4 T cell development at 6 wk postreconstitution.

Thymocytes were pooled from three to five 4- to 5-wk-old NOD.AI4 and B6.H2^g7.AI4 mice and an aliquot stained with a CD4/CD8/Vα8 Ab mixture to assess numbers of AI4 CD8 SP T cells by flow cytometry. A thymocyte mixture (200 μl) containing 1 × 10⁶ AI4 CD8 SP T cells of each donor type was injected i.v. into either NOD.scid or sublethally irradiated (650 R) (NOD × B6.H2^g7)F₁ recipients. At 6 wk posttransfer, recipient spleens were assessed by flow cytometry for levels of each AI4 CD8 T cell population. AI4 CD8 T cells of NOD or B6.H2^g7 origin were respectively identified by expression of the CD45.1 and CD45.2 markers. For the competition assays in (NOD × B6.H2^g7)F₁ recipients, the data were corrected for endogenous CD45/Vα8/CD8-staining levels.

When dispersed islet cells are used as targets in cytotoxicity assays, they undergo a high rate of spontaneous death that can complicate experimental interpretations. Thus, an alternative assay was developed in which intact islets were used as targets because this greatly enhances cell survival. Intact NOD pancreatic islets (10/well) were allowed to adhere in 96-well plates by a 7- to 10-day incubation at 37°C in low-glucose DMEM medium. These islets were then labeled with 5 μCi/well ⁵¹Cr for 3 h at 37°C. Islets were washed and overlaid with 100 μl of medium containing various numbers of either NOD.AI4 or B6.H2^g7.AI4 T cells that had been preactivated for 72 h with the mimotope peptide at a concentration of 10 nM. For establishing E:T ratios, each islet was assumed to contain ∼800 cells. A minimum of three wells was established for each T cell type and E:T ratio. Controls consisted of at least six wells of labeled NOD islets cultured in the absence of T cells. Following an overnight incubation at 37°C, the radioactivity in two fractions from each well was measured. The first fraction was the culture supernatant, and the second was obtained by solubilizing the remaining islets in that well with 100 μl of 2% SDS. The percentage of ⁵¹Cr release for each well was calculated by the formula: (supernatant cpm)/(supernatant + SDS lysate cpm). This allowed us to normalize for the fact that due to variation in the sizes of individual islets, the total levels of ⁵¹Cr incorporation in each well could differ greatly. In turn, percentage of specific lysis was calculated by subtracting the mean percentage of ⁵¹Cr release from islets cultured in medium alone from the percentage of ⁵¹Cr release of islets cultured with a given type and number of T cells. Data are presented as the mean percentage of specific lysis ± SEM. In other experiments, NOD splenocytes were Con A activated for 48 h, labeled with ⁵¹Cr, pulsed with the AI4 mimotope peptide at a concentration of 10 nM, and used as targets for NOD.AI4 or B6.H2^g7.AI4 cytotoxic T cells in the previously described assay system (19).

The technique used is described elsewhere (20, 21, 22). Briefly, 3- to 4-wk-old NOD.AI4 or B6.H2^g7.AI4 mice were anesthetized, and thoraxes were opened to expose the thymic lobes. Each thymic lobe was injected with 10 μl of 350 μg/ml filtered FITC (Sigma-Aldrich, St. Louis, MO) in sterile PBS, randomly labeling thymocytes in a range of 30–60%. The incision was closed with surgical staples, and the mice were warmed until fully recovered from anesthesia. Mice were killed 24 h later, and their thymocytes and splenocytes were counted and analyzed by FACS with anti-CD4 PE/anti-CD8 AP and anti-CD8 AP/anti-Vα8 PE Ab mixtures. Recent thymic emigrants in spleens were identified and quantitated as live-gated FITC⁺ cells expressing either CD4 or CD8. Relative thymic emigration rates in NOD.AI4 or B6.H2^g7.AI4 mice were calculated by dividing the numbers of live splenic FITC⁺ CD4 and CD8 cells by total numbers of FITC⁺ thymocytes.

The technique used is a modified version of an earlier described protocol (16). Total RNA was purified from 5–10 × 10⁶ MACS-purified CD8 T cells using the Ambion purification kit (Austin, TX). One microgram of each RNA sample was reverse transcribed (RT) into single-strand cDNA using Moloney murine leukemia virus reverse transcriptase and random hexamer primers, followed by digestion with DNase I (Ambion) for 30 min at 37°C. To determine the TCR Vα gene families used by each CD8 T cell population, cDNAs were amplified by PCR by using the previously decribed Cα primer paired with 1 of 20 Vα primers that each amplify all members of a particular Vα gene family (23). The quality of the DNase and RT reactions was assessed by amplifying β-actin cDNA from the RT⁺ and RT⁻ samples with the following primers: β-actin F, TGGAATCCTGTGGCATCCATGAAA, and β-actin R, TAAAACGCAGCTCAGTAACAGTCC. PCR products (0.5–0.8 kbp) were resolved in 1.5% agarose gels and detected by ethidium bromide staining.

Two previously developed genetic resources allowed us to test whether allelic variants of non-MHC genes interactively support or inhibit the development of diabetogenic CD8 T cells that use common MHC class I variants as their restriction elements. One of these was a T1D-resistant stock of B6 mice congenic for the H2^g7 MHC haplotype (designated B6.H2^g7) (10). The other was a NOD stock transgenically expressing the TCR from the K^d MHC class I-restricted T cell clone AI4 that contributes to the earliest phases of autoimmune β cell destruction leading to T1D (designated NOD.ΑI4 mice) (9). The AI4 TCR transgenes were congenically transferred to the B6.H2^g7 stock.

Thymic negative selection of autoreactive T cells generally occurs at the CD4/CD8 DP stage of development when TCR expression first occurs (24). Hence, we compared the number of DP thymocytes in the NOD.AI4 and B6.H2^g7.AI4 stocks. The thymic profiles of the two Tg strains differed at several levels. Macroscopically, the thymi of B6.H2^g7.AI4 mice contained many fewer cells (4.25 ± 1.03 × 10⁶, n = 9) than those from the NOD.AI4 (8.37 ± 1.21 × 10⁷, n = 9) or B6.H2^g7 (1.48 ± 0.082 × 10⁸, n = 4) strains. Moreover, as shown in Fig. 1, the number of AI4 DP thymocytes was much less in B6.H2^g7.AI4 than NOD.AI4 mice (31-fold reduction). The number of thymic AI4 CD8 SP T cells was also less in B6.H2^g7.AI4 than NOD.AI4 mice, albeit to a lesser extent than observed in the DP compartment (4-fold difference). These observations indicate non-MHC genes regulate the development of MHC class I-restricted autoreactive T cells.

The significantly lower numbers of AI4 DP thymocytes in B6.H2^g7 than NOD genetic background mice could be explained by either decreased levels of positive selection or increased levels of negative selection in the former stock. Positive and negative selection are primarily mediated by nonhemopoietically derived thymic epithelial cells and bone marrow-derived APCs, respectively (25, 26). Thus, we devised a reciprocal bone marrow chimera protocol to test our hypotheses.

The approach used was to inject the recipients i.p. with a CD40L-specific mAb on days 0, 3, and 14, relative to their receipt of a sublethal radiation dose, and injection of T cell-depleted bone marrow cells. This protocol prevents graft-vs-host and host-vs-graft responses and also preserves a significant portion of the recipient’s bone marrow-derived APCs from radiation-induced deletion. Consequently, the thymus of the bone marrow chimeras harbors both recipient- and donor-type APCs. Overall levels of hemopoietic reconstitution were assessed at 6 wk posttransfer by determining the number of leukocytes staining positive for the CD45.1 (NOD type) vs CD45.2 (B6.H2^g7 type) marker. The level of reconstitution was comparable in the two experimental groups (NOD.AI4→B6.H2^g7, 74.6 ± 1.4% donor type and B6.H2^g7AI4→NOD, 77.5 ± 2.2% donor type) (Fig. 2,A). Regardless of recipient type, significantly higher proportions and total numbers of AI4 TCR-expressing DP thymocytes developed from hemopoietic precursors of NOD than B6.H2^g7 origin (Fig. 2 B). Because it occurs independently of the thymic epithelial environment, the higher level of AI4 T cell development in NOD than B6.H2^g7 mice does not result from an enhancement of positive selection controlled by non-MHC genes in the former strain. However, these results also indicate the numerical differences in AI4 DP thymocytes between the two Tg stocks are independent of the genetic background of the bone marrow-derived APCs responsible for inducing negative selection. Thus, it appears that a T cell-intrinsic defect(s) most likely contributes to impaired negative selection of diabetogenic AI4 T cells in NOD mice. It should also be noted that a somewhat greater number (2-fold) of NOD marrow-derived AI4 DP thymocytes was found in B6.H2^g7 compared with syngeneic recipients. Although the thymic environment in each group contains bone marrow-derived APCs from both strains, the thymic epithelium cells that are the most efficient mediators of positive selection are entirely recipient type. Therefore, the observed difference between the two chimeras suggests that positive selection may actually be more efficient in the thymic environment of B6.H2^g7 than NOD mice. However, a T cell-intrinsic negative selection defect(s) under non-MHC gene control appears to be the major reason that higher levels of diabetogenic AI4 T cell develop in NOD than B6.H2^g7 mice.

We ascertained whether the non-MHC gene-controlled defect impairing negative selection of AI4 T cells in NOD mice is an anomaly limited to this clonotype. This was done by determining whether antigenic stimulation in vitro differentially deleted DP thymocytes from NOD and B6,D2 stocks that transgenically express the same D^b-restricted nonautoimmune LCMV-reactive TCR (15). This same type of assay could not be used in the AI4 system because due to the fact they have already undergone efficient negative selection, very few AI4 TCR-expressing CD4/8 DP thymocytes remain present in the B6.H2^g7 background stock. As depicted in Fig. 3 A, antigenic stimulation results in the negative selection of LCMV-reactive DP thymocytes from the B6,D2, but not the NOD genetic background stock (19.1 vs 61.2% viable DP thymocytes remain present following culture with 10 μg/ml LCMV peptide for 24 h).

Nur77, a member of the orphan steroid receptor family, has been shown to play a key role in negative selection (27, 28). We evaluated by intracellular staining the level of Nur77 expression in DP thymocytes from the NOD and B6,D2 LCMV TCR Tg-expressing stocks following antigenic stimulation for 24 h. As shown in Fig. 3 B, Nur77 was more readily induced and expressed by a larger proportion of DP thymocytes from B6,D2 than NOD genetic background mice. Therefore, in two different models, NOD-derived non-MHC genes increased resistance to Ag-driven negative selection.

Given they negatively select AI4 T cells with far greater efficiency than the NOD background stock, we were surprised to find that all of an initial cohort of B6.H2^g7.AI4 female mice developed T1D by 4 wk of age (Fig. 4,A). Analyses of larger numbers of B6.H2^g7.AI4 females indicated that none remained disease free for greater than 6 wk of age (data not shown). This is significantly faster than the already accelerated rate of T1D characterizing the NOD.AI4 stock (Fig. 4,A). Although there were 31- and 4-fold reductions in numbers of DP and CD8 SP AI4 T cells in the thymi of B6.H2^g7 compared with NOD genetic background mice, only a 1.7-fold difference was found in the spleen (Fig. 4 B). Thus, we evaluated several possible explanations for why, relative to thymic levels, AI4 T cells are enriched in the periphery of B6.H2^g7 background mice and efficiently mediate T1D.

The narrowing of AI4 T cell numbers in the periphery of NOD.AI4 and B6.H2^g7.AI4 mice might reflect differential peripheral expansion in these two strains. If it were an Ag-driven expansion, the peripheral AI4 T cells would express activation markers such as CD69. Therefore, we assessed the activation status of AI4 T cells in the spleen and pancreatic lymph nodes of NOD and B6.H2^g7 mice. The proportion of splenic AI4 T cells that expressed CD69 was low (0.2–2.3%) in both strains. Compared with the splenic populations, the proportion of CD69-positive AI4 T cells was marginally higher in pancreatic lymph nodes (0.6–6.0%), but once again no strain differences were observed. These results exclude the possibility that AI4 T cell numbers become more similar in the periphery than thymi of NOD and B6.H2^g7 mice through differential Ag-driven expansion.

In certain circumstances, naive T cells can undergo Ag-independent homeostatic expansion (29). An ability to undergo greater levels of homeostatic expansion could account for why the numbers of B6.H2^g7 AI4 T cells more closely approach those of NOD origin in the periphery, but not the thymus. To test this hypothesis, equal numbers (1 × 10⁶) of SP CD8 thymocytes from B6.H2^g7.AI4 and NOD.AI4 donors were coinjected into either NOD.scid or sublethally irradiated (NOD × B6.H2^g7)F₁ recipients. Six weeks posttransfer, the proportion of AI4 CD8 T cells of each donor type was evaluated by staining splenocytes with an anti-CD8/Vα8/CD45 mixture. These analyses indicated that CD8 AI4 T cells of both donor types repopulated the peripheral lymphoid system to the same extent (Table I). Thus, there is no homeostatic expansion advantage of B6.H2^g7 over NOD AI4 CD8 T cells.

We next tested whether more efficient thymic export accounts for greater peripheral enrichment of AI4 T cells in B6.H2^g7 than NOD genetic background mice. Recent thymic emigrants (RTE) have been shown to express the CD103 (integrin α_E) protein (30). Thus, we used the CD103 marker to identify RTEs in spleens of B6.H2^g7.AI4 and NOD.AI4 mice. When normalized to numbers of AI4 CD8 SP thymocytes, there was a smaller proportion of CD103⁺CD8⁺ AI4 T cells in the spleens of NOD (15%) than B6.H2^g7 (37%) genetic background mice (Fig. 4,C). However, because CD103 has been described to identify CD8 RTE in humans and not yet in mice, we further tested our hypothesis by the well-established FITC thymic injection protocol (20, 21, 22). The results confirmed the B6.H2^g7.AI4 stock has a greater daily thymic export rate than NOD.AI4 mice (Fig. 4 D). Thus, the narrowing number of peripheral AI4 T cells in NOD and B6.H2^g7 mice could be explained by a greater rate of thymic emigration in the latter stock.

Although more efficient thymic emigration allows the few AI4 T cells that avoid negative selection in B6.H2^g7 mice to reach levels in the periphery approaching that seen in NOD mice, the numbers of such diabetogenic effectors are still significantly less in the former stock. Yet, T1D is more rapidly accelerated in B6.H2^g7.AI4 than NOD.AI4 mice. Consequently, we hypothesized the AI4 CD8 T cells that do seed the periphery of B6.H2^g7 genetic background mice might be more functionally aggressive than those of NOD origin. To initially test this hypothesis, we compared the ability of AI4 T cells from both stocks to proliferate when stimulated by a recently identified mimotope peptide (17). There is evidence that some diabetogenic CD8 T cells in NOD mice are initially activated when encountering pancreatic β cell-derived peptides that are cross-presented by APCs (6). However, APCs in NOD mice are characterized by several functional defects (31, 32, 33). For this reason, we compared the ability of purified NOD- and B6.H2^g7-derived CD8 AI4 T cells to proliferate in response to the mimotope peptide presented by APCs from both strains. B6.H2^g7-derived AI4 T cells had a greater capacity than those of NOD origin to proliferate in response to mimotope stimulation regardless of the APC population used (Fig. 5 A).

Given their greater Ag-driven proliferative capacity, we reasoned that AI4 T cells from B6.H2^g7 mice might also exhibit greater cytotoxic activity than those of NOD origin. Hence, we performed a cytotoxicity assay with AI4 T cells from both stocks against NOD pancreatic islets. As expected, the control pancreatic islets from MHC class I-deficient NOD.β₂m^−/− mice were not killed by AI4 T cells of either NOD or B6.H2^g7 origin (data not shown). There was no nonspecific killing of standard MHC class I-positive NOD islets by irrelevant CD8 T cells from NOD.scid.LCMV TCR Tg mice (Fig. 5,B). However, class I-expressing NOD islets were killed to a significantly greater extent by AI4 T cells of B6.H2^g7 than NOD origin at E:T ratios of 10:1, 2:1, and 0.4:1 (Fig. 5,B). It was only at a 50:1 E:T ratio that there was no statistical difference between the ability of NOD- or B6.H2^g7-derived AI4 T cells to kill NOD islets. Furthermore, Con A-activated NOD splenocytes pulsed with the AI4 mimotope peptide were killed to a significantly greater extent by AI4 T cells of B6.H2^g7 than NOD origin at all E:T ratios tested (Fig. 5 C). These collective results indicate that despite undergoing much higher levels of negative selection in the thymus, compared with those generated in NOD mice, AI4 CD8 T cells of B6.H2^g7 origin show a greater rate of thymic emigration, a greater proliferative response to Ag presentation, and finally, more potent killing of pancreatic islet cells. Such functions explain the initially paradoxical finding that while demonstrating a very potent central tolerance process, B6.H2^g7.AI4 mice develop dramatically accelerated T1D.

Our studies indicated that differences in both the selection and functional activation of AI4 effectors in NOD or B6.H2^g7 mice are T cell intrinsic. The surface level of TCR expression is known to be a key to Ag responsiveness and to greatly differ at various stages of T cell development (for review, see Ref.34). TCR expression is lower on immature thymocytes than peripheral, mature naive T cells. As a result, some antigenic peptides can activate mature T cells, but at the same concentration cannot induce the level of signaling required to trigger intrathymic deletion of the same clonotype (35). Thus, variable levels of TCR expression on DP thymocytes and/or mature T cells could account for the differential ability of diabetogenic AI4 effectors to be selected and functionally activated in NOD and B6.H2^g7 background mice. Splenic AI4 T cells from the two Tg stocks exhibited equivalent levels of Vβ2 expression. However, many NOD-derived peripheral AI4 T cells were found to express the Vα8 chain at much lower levels than those of B6.H2^g7 origin (Fig. 6,A). One possible explanation was an overall reduction in TCR expression on a subset of NOD-derived AI4 T cells. This was ruled out because Vα8^high and Vα8^low AI4 T cells from NOD mice stained equivalently with a generic TCRαβ-specific Ab (Fig. 6,B). We therefore tested the alternative possibility that the Vα8^low subpopulation of AI4 T cells in NOD mice results from a dilution effect elicited by coexpression of endogenous TCR Vα chains. TCR Vα mRNA transcripts expressed by purified CD8 T cells from both NOD.AI4 and B6.H2^g7.AI4 mice were assessed by RT-PCR. Transcripts using most of the 20 Vα chain family members were detected in each population (Fig. 6 C). However, certain Vα chains (such as Vα1, 2, 3, 9, 13, 15, 16, and 17) were expressed at higher levels by AI4 CD8 T cells of NOD than B6.H2^g7 origin. To exclude the possibility of detecting nonproductive Vα rearrangements, the RT-PCR results were correlated with protein expression analyses. This was done by staining AI4 T cells from NOD and B6.H2^g7 mice with the only two commercially available Abs that can recognize endogenous TCR Vα elements (Vα2, Vα3) derived from each strain. NOD AI4 T cells were found to express higher levels of Vα2 and Vα3 TCR chains on their surface than those of B6.H2^g7 origin (data not shown). These collective results indicate the clonotypic AI4 TCR is expressed at a lower density on mature CD8 T cells from NOD.AI4 than B6.H2^g7.AI4 mice due to enhanced expression of endogenous Vα chains. This reduced expression is likely to be the reason that NOD AI4 T cells exhibited lower Ag-driven proliferative and cytotoxic responses than those of B6.H2^g7 origin.

As expected, expression of the AI4 TCR Vα8 chain was lower on CD4/8 DP thymocytes than mature CD8 T cells in both NOD and B6.H2^g7 mice (Fig. 6 D). However, the lowered efficiency of allelic exclusion originally observed in mature T cells also resulted in significantly lower expression of the AI4 Vα8 chain on CD4/8 DP thymocytes of NOD than B6.H2^g7 origin. The same observation was made in the LCMV TCR model: NOD.LCMR TCR Tg DP thymocytes or mature CD8 T cells express the Vα2 chain at a lower level than the corresponding cells in B6,D2.LCMV TCR Tg mice (DP Vα2 mean fluorescence intensity (MFI): 97.4 ± 5.1 vs 202.8 ± 18.3, and splenic CD8 Vα2 MFI: 270.5 ± 1.2 vs 351.1 ± 7.3). This indicates that the lowered level of TCR Vα expression in NOD mice compared with a control strain is not unique to our AI4 TCR model. As described above, a lower level of clonotypic TCR expression could result in less successful achievement of the signaling threshold required to induce negative selection of AI4 DP thymocytes in NOD than B6.H2^g7 AI4 DP mice. The fact that mature NOD AI4 T cells express the TCR at higher levels than thymocytes most likely accounts for the ability of the former to be activated and induce T1D when encountering their cognate β cell autoantigen. Although the higher levels of TCR expression at all stages of T cell development may allow more efficient intrathymic negative selection of the AI4 clonotype in B6.H2^g7 mice, our data suggest the same feature endows the mature effectors escaping deletion with a greater cytotoxic capacity than those of NOD origin.

The present study shows that dysfunctions controlled by non-MHC genes contribute to less efficient negative selection of diabetogenic CD8 T cells in NOD mice than other strains expressing the same common class I variants characterizing the H2^g7 haplotype. Standard non-Tg B6.H2^g7 mice are T1D resistant (10). Because of this, and the fact that they negatively select the AI4 clonotype far more efficiently than NOD mice, we anticipated the B6.H2^g7.AI4 stock would also be T1D resistant. However, instead, the B6.H2^g7.AI4 stock was found to develop T1D even more rapidly than the already accelerated rate characterizing NOD.AI4 mice. This apparently results from the fact that the few AI4 T cells escaping negative selection in the B6.H2^g7 stock are exported from the thymus more efficiently and are more functionally aggressive than those of NOD origin. In outcross studies with NOD, resistant mouse strains have been found to harbor some genetic variants that contribute to T1D susceptibility to a greater extent than the corresponding NOD allele (2). Hence, our data suggest that more efficient thymic export and enhanced functional aggressiveness of β cell autoreactive T cells represent disease subphenotypes controlled by T1D susceptibility genes derived from overall resistant strains. Our findings with the AI4 clonotype are not the first example of an NOD-derived β cell autoreactive TCR being even more pathogenic when expressed in a normally T1D-resistant strain. The H2-A^g7 class II-restricted BDC2.5 TCR displays an even more drastic difference, as, unlike the AI4 TCR, it induces only a marginal incidence of T1D in NOD mice, but a very high rate of disease when expressed on the B6.H2^g7 background (36).

Our results also provide insight to the long debated issue of the stage of T cell development in which tolerance is broken in NOD mice. There have been reports that the breakdown of immunological tolerance in NOD mice is a peripheral rather than an intrathymic event. Another group addressed the issue of peripheral tolerance induction in NOD mice through use of a multitransgenic stock that produces CD8 T cells expressing a TCR specific for a model viral Ag expressed in pancreatic β cells (37, 38). This group found that unlike what happens in BALB/c and B10.D2 mice expressing the same transgenes, the NOD stock does not peripherally delete the high avidity CD8 T cells specific for the pancreatic β cell-expressed viral Ag. Verdaguer et al. (39) demonstrated a non-MHC genetic factor(s) contributes to impaired peripheral deletion in NOD mice of T1D relevant NY8.3 CD8 T cells. Pathogenic activation in NOD mice of the β cell autoreactive CD4 T cell clonotype BDC2.5 has been reported to result from peripheral tolerance induction defects (40). Together, these studies strongly argue for defective induction of peripheral T cell tolerance in NOD mice. Although a peripheral tolerance defect is indubitably an important flaw contributing to T1D in NOD mice, it is unlikely to be the sole factor. Indeed, our current data indicate an impaired ability to induce central tolerance is also an important diabetogenic factor in NOD mice.

Although there are previous reports that NOD thymocytes are less prone to negative selection than those from autoimmune resistant strains (41, 42, 43), the stimuli used to induce deletion may not have been as physiologically relevant as the AI4 TCR Tg system. For example, Kishimoto and Sprent (42) used the procedure of incubating thymocytes or injecting mice with nonphysiologic doses of Abs that stimulate TCR and CD28 signaling. They found a deletional defect in NOD mice that was most prominent in a specific population of semimature thymocytes. This result was contested by another group that used the same experimental approach (44), but could not reproduce the results of Kishimoto and Sprent (42). Thus, while the use of TCR and CD28 stimulatory Abs can give useful information, this approach cannot provide final and definitive proof of the physiologic process of self-reactive T cell fate. Lesage et al. (43) found that non-MHC genes contributed to an impaired ability of NOD mice to intrathymically delete CD4 T cells expressing a TCR specific for a hen egg lysozyme peptide expressed as a model Ag in pancreatic β cells. However, it remains questionable whether the levels and sites of expression of the Tg hen egg lysozyme molecules in this system are equivalent to the AI4 Ag naturally targeted in T1D, and hence, whether the development of T cells with a potential to recognize epitopes derived from these two types of proteins is influenced in similar ways. For these reasons, our current study provides the first evidence that a naturally T1D relevant class I-restricted autoreactive T cell clonotype evades intrathymic negative selection when developing in NOD mice, but not in an MHC-matched control strain. This defect is T cell intrinsic because NOD.AI4 thymocytes are not negatively selected when developing in the presence of B6.H2^g7-derived APC. However, it cannot be completely ruled out that NOD APCs also contribute to defective negative selection of AI4 T cells. This is because the proportion of host-derived APCs in the reciprocal chimeras we analyzed was ∼25%, which may be insufficient to modulate the development of AI4 T cells. Several recent studies have indicated that the transcription factor activity induced by the orphan steroid receptor Nur77 is an important component in the triggering of thymocyte negative selection (18, 45). Using the nonautoimmune LCMV TCR model, we show the greater resistance of NOD than B6 thymocytes to Ag-driven negative selection may be due to less efficient induction of Nur77.

Variations in expression levels of the clonotypic AI4 TCR may provide one explanation for the differences between the NOD and B6.H2^g7 genetic background stocks. Due to less efficient α-chain allelic exclusion, AI4 T cells developing in NOD mice are more prone to expressing additional TCR specificities than those of B6.H2^g7 origin. As a result, at all stages of development, the surface density of the clonotypic AI4 TCR is lower on NOD than B6.H2^g7 T cells. It has been previously reported that some peptides that can activate mature T cells at the same dose fail to trigger the signaling levels required to induce the negative selection of clonotype-matched CD4/8 DP thymocytes due to their lower levels of TCR expression (35). Hence, their lower level of clonotypic TCR expression could result in less successful achievement of the signaling threshold required to induce the negative selection of AI4 DP thymocytes in NOD than B6.H2^g7 mice. In both strains, AI4 TCR density is greater on mature T cells than DP thymocytes, but with higher levels continuing to be maintained in the B6.H2^g7 background stock. These levels of TCR expression are clearly sufficient to allow mature AI4 T cells from both strains to activate their effector functions following Ag engagement, but also result in a more vigorous response by those of B6.H2^g7 origin. If this overall scenario is correct, an implication is that the same phenotype, in this case relatively high TCR expression levels, could have double-edged sword effects that contribute to T1D resistance at one level of T cell development, but at another actually promote β cell destruction. This may be an important consideration in understanding the pathogenic basis of T1D, and perhaps other autoimmune diseases, because due to steric interference events and other factors, potential autoreactive T cells may first encounter an agonist ligand at different developmental time points.

There is epidemiological evidence that particular common class I variants can also increase the risk of T1D development when expressed in some humans (7), but the circumstances under which they do so have not been fully elucidated. Our current data indicate a contributing factor may be the presence of non-MHC genetic variants that impair the intrathymic negative selection of T cells with a potential to recognize pancreatic β cell Ags presented by the class I molecule in question. If this premise is correct, the identification of such non-MHC genetic factors would be of importance in more accurately predicting progression to T1D in humans known to possess high risk MHC haplotypes. Thus, we are now mapping non-MHC loci that cosegregate with levels of AI4 T cells in F₂ progeny derived from an intercross of the NOD.AI4 and B6.H2^g7.AI4 TCR Tg stocks. This may ultimately lead to the identification of human non-MHC homologues that exert similar functions.