β Cells cannot directly prime diabetogenic CD8 T cells in nonobese diabetic mice

Abstract

Type 1 diabetes (T1D) is caused by the destruction of insulin-producing islet β cells. CD8 T cells are prevalent in the islets of T1D patients and are the major effectors of β cell destruction in nonobese diabetic (NOD) mice. In addition to their critical involvement in the late stages of diabetes, CD8 T cells are implicated in the initiation of disease. NOD mice, in which the β₂-microglobulin gene has been inactivated by gene targeting (NOD.β₂M^−/−), have a deficiency in CD8 T cells and do not develop insulitis, which suggests that CD8 T cells are required for the initiation of T1D. However, neither in humans nor in NOD mice have the immunological requirements for diabetogenic CD8 T cells been precisely defined. In particular, it is not known in which cell type MHC class I expression is required for recruitment and activation of CD8 T cells. Here we have generated transgenic NOD mice, which lack MHC class I on mature professional antigen-presenting cells (pAPCs). These “class I APC-bald” mice developed periislet insulitis but not invasive intraislet insulitis, and they never became diabetic. Recruitment to the islet milieu does not therefore require cognate interaction between CD8 T cells and MHC class I on mature pAPCs. Conversely, such an interaction is critically essential to allow the crucial shift from periislet insulitis to invasive insulitis. Importantly, our findings demonstrate unequivocally that CD8 T cells cannot be primed to become diabetogenic by islet β cells alone.

Many cell types are found in the insulitis lesions that precede type 1 diabetes (T1D), including professional antigen-presenting cells (pAPCs) such as dendritic cells (DCs), macrophages, B lymphocytes, and both CD4⁺ and CD8⁺ T lymphocytes. Each of these cell types is required for the development of T1D. However, their contributions to disease pathogenesis and, in particular, their role in the events that lead to the recruitment and priming of autoreactive, islet-specific T cells, have not been resolved. For the rational design of preventative strategies for T1D, we require a clear understanding of the earliest cellular interactions that lead to the recruitment, retention, and initial priming of diabetogenic CD8 T cells.

CD8 T Cells Are Required for Initiation of Insulitis.

CD8 T cells are essential for the development of T1D in both humans and in nonobese diabetic (NOD) mouse models. Not only are they important final effectors of β cell death (1, 2), they are also important in the early stages of disease as indicated by the finding that insulitis was completely absent in CD8-deficient [β2-microglobulin-deficient (β₂M^−/−)] NOD mice (3). Clearly, CD8 T cells are important early, but it remains somewhat controversial as to which cells, whether pAPCs or the β cells themselves, are sufficient and/or required to interact with the CD8 T cells for their retention in the islet milieu and their activation to become islet-invasive.

CD8 T Cells Require Cognate Interaction with MHC Class I to Become Islet-Invasive.

It has been shown that splenocytes from young prediabetic donors can transfer diabetes into MHC class I-sufficient but not MHC class I-deficient NOD.scid recipients, and this transfer depends on both CD4 and CD8 T cells (4, 5). To distinguish between the role of β cells and other cells in the islet tissue in priming CD8 T cells, Kay et al. (6) expressed a β₂M transgene under the control of the rat insulin promoter (RIP) in β₂M-deficient NOD mice, which restored MHC class I expression to the islet β cells but not to the APCs, and, despite the paucity of CD8 T cells, the mice developed invasive insulitis. The interpretation of this result was that MHC class I expression on islet β cells was sufficient to activate diabetogenic CD8 T cells responsible for initiating invasive insulitis.

Although islet β cell expression of MHC class I may be sufficient to initiate intraislet insulitis, it is certainly not requisite. NOD mice lacking MHC class I on islet β cells still developed intraislet infiltration (7). Currently, although we know that CD8 T cells are crucial effectors of β cell death and are important early in the development of insulitis, we do not have a clear understanding of the requirements for activation of these cells. We know only that activation can occur in the absence of β cell MHC class I expression (7).

DCs are known to be the primary cell type responsible for the capture, processing, and presentation of antigen to CD8 T cells (8). They are also involved in cross-presentation of autoantigens, a process that occurs after immature DCs internalize antigen derived from damaged, apoptotic, or dead cells. Once mature, DCs are able to induce activation of CD8 T cells with specificity for the cross-presented antigen. Antigen that enters the cross-presentation pathway is thought to be released from damaged or dying cells but not from healthy cells (9). For this reason, antigens expressed on normal, non-pAPCs, such as β cells, are probably ignored by CD8 T cells. From approximately weaning age, NOD mice develop an inflammatory lesion in the pancreatic tissue around the islet. This insulitic lesion progresses through a series of checkpoints to become islet-invasive and finally β cell-destructive. Although it has been shown that pAPCs within islet tissue are sufficient to capture and cross-present β cell antigens to CD8 T cells in the draining lymph node (10–12), this cross-presentation has not been demonstrated as an absolute requirement for the development of intraislet invasion in the NOD spontaneous diabetes model. It has previously been proposed that MHC class I expression on the islet β cells is important for the retention of CD8 T cells within the islet (13) and that MHC class I expression on β cells alone is sufficient for the priming of diabetogenic T cells and the development of intraislet invasion (6).

In this work we describe NOD mice, which have been genetically modified to lack MHC class I expression on their mature pAPCs (class I APC-bald NOD mice) but express MHC class I normally in all other tissues, including islet β cells (class I APC-normal NOD mice). The CD8 T cells in these mice are able to mount MHC class I-restricted immune responses. The mice have been compared with their MHC class I-sufficient wild-type littermates to determine whether CD8 T cells can be recruited and activated to invade the islet tissue in the absence of cognate interaction with MHC class I on pAPCs.

Results

Class I APC-Bald NOD Mice Have Functional CD8 T Lymphocytesxs.

Thymocytes of class I APC-bald NOD mice developed into mature CD8 single-positive cells despite the deficiency in MHC class I on the surface of thymic APCs. CD8 T cells were present in the periphery at ≈10% of the normal number, and they were functional in a cytotoxic lymphocyte assay (data not shown). Spleens of class I APC-bald NOD mice (n = 4) contained 1.1 ± 0.6 × 10⁶ CD8 T cells compared with 10.8 ± 1.1 × 10⁶ CD8 T cells per spleen of class I APC-normal NOD mice (n = 4). CD4 thymocyte development and peripheral CD4 T lymphocyte absolute cell numbers were similar in class I APC-bald mice (24.5 ± 10.0 × 10⁶ per spleen; n = 4) and class I APC-normal mice (26.0 ± 6.5 × 10⁶ per spleen; n = 4). Splenic CD8 T cells of class I APC-bald NOD mice were conventional TCRαβ⁺ and CD8αβ⁺), and they were not enriched for TCRγδ- or CD8αα-bearing T cells, indicating that the niche normally occupied by conventional CD8 T cells was not filled by CD8 T cells of an unconventional phenotype (data not shown). Natural killer T (NKT) cell development, which relies largely on the interaction between the invariant NKT cell T cell receptor (TCR) and the β₂M-dependent CD1d molecule on thymocytes (14, 15), was unaffected, as predicted, because the MHC H-2IE (IE) promoter is not active in double-positive mouse thymocytes on which these cells are selected (data not shown).

The Majority of pAPCs of Class I APC-Bald NOD Mice Are Deficient in Surface MHC Class I.

Flow cytometric analysis of class I APC-bald NOD mice and their wild-type littermates confirmed that β₂M deletion and the lack of expression of MHC class I K^d and D^b (D^b is not shown) were confined to pAPCs. Single-cell suspensions of thymi, spleen, and lymph nodes were stained for both MHC class I K^d and cell surface markers specific for DCs, B cells, macrophages, and T cells. Although MHC class I expression on T lymphocytes was similar in both class I APC-bald NOD and wild-type littermates, it was ablated in the majority of DCs and B lymphocytes of class I APC-bald NOD mice (Fig. 1). Only 10.9 ± 4.5% of conventional CD11c⁺ B220⁻ DCs and 14.9 ± 3.9% of total splenic B lymphocytes were MHC class I-positive (n = 6). The majority of activated, MHC class II-expressing macrophages were MHC class I-deficient.

Fig. 1.

Loss of MHC class I is limited to the APC compartment in class I APC-bald NOD mice. Class I APC-bald NOD mice (n = 6), class I APC-normal NOD mice (n = 5), and NODβ₂M^−/− (n = 6) were analyzed for H-2K^d expression by FACS. Total conventional DCs (CD11c⁺), B lymphocytes (B220⁺ or CD19⁺), and macrophages (I-A^g7+ F4/80⁺ CD11b⁺) are shown, and the mean percentages (±SD) of H-2K^d-sufficient cells are indicated. All APC cell types from class I APC-bald NOD mice were largely deficient in MHC class I. In contrast, as expected, splenic T lymphocytes (CD4⁺, TCR⁺; Far Right) remain MHC class I-sufficient.

APCs of Class I APC-Bald NOD Mice Are Unable to Activate CD8 T cells in Vivo.

Because the majority but not all of the pAPCs of class I APC-bald NOD mice were deficient in MHC class I, we endeavored to determine whether those remaining MHC class I-sufficient APCs were capable of priming naïve CD8 T cells. We transferred carboxy-fluorescein diacetate succinimidyl ester (CFSE)-labeled, highly diabetogenic, monoclonal CD8 T lymphocytes, bearing the 8.3 TCR specific for the β cell antigen IGRP (16), into both class I APC-bald NOD mice and class I APC-normal NOD mice. Six days after transfer, 8.3 TCR-positive T cells were detected in the pancreatic lymph nodes of both hosts, but they were unable to divide in the class I APC-bald mice (Fig. 2). The percentage of 8.3 TCR-positive CD8 T cells that had gone through at least one cellular division was 5.4 ± 2.1 (n = 6) in class I APC-bald NOD mice and 53.1 ± 21.6 (n = 5) in class I APC-normal NOD mice. Hence, APCs in the pancreatic lymph nodes of class I APC-bald NOD mice were unable to activate a diabetogenic CD8 T cell clone.

Fig. 2.

APCs of class I APC-bald NOD mice are unable to activate CD8 T cells in vivo. CFSE-labeled 8.3 TCR transgenic CD8 T cells were injected into either class I APC-normal mice (n = 5) or class I APC-bald NOD mice (n = 6). Six days after transfer, the percentage of proliferating 8.3 TCR⁺ CD8 T cells (R1) in the pancreatic lymph nodes of class I APC-normal mice (Upper) and class I APC-bald mice (Lower) was determined by flow cytometry. This result was representative of two independent experiments.

Destructive Insulitis and Diabetes Are Inhibited in Class I APC-Bald NOD Mice.

Insulitis in class I APC-bald NOD female mice was less severe than that observed in nondiabetic class I APC-normal NOD female mice. In striking contrast to the invasive infiltrate observed in class I APC-normal NOD mice (Fig. 3A), the insulitis lesions observed in class I APC-bald NOD mice were unique in that the infiltrate was consistently peripheral to the islet (Fig. 3F). In the pancreata from 6 of 7 class I APC-bald NOD mice, the insulitis lesion was limited to the periislet milieu, whereas in all (6 of 6) pancreata from class I APC-normal NOD female mice the lesion included invasive insulitis (P = 0.05; Fig. 4A). None of the islets of >300-day-old NOD.β₂M^−/− mice had insulitis. Not surprisingly, class I APC-bald NOD mice did not develop diabetes (0 of 12) compared with their class I APC-normal littermates (8 of 27). The content of the insulitis lesions found in class I APC-bald NOD mice was examined in more detail by immunofluorescence staining (Fig. 3 B–E and G–J), which confirmed that inflammatory cells were confined to the periphery of the islet and included abundant CD4 T cells, macrophages, and B lymphocytes but very few CD8 T cells.

Fig. 3.

The islets of class I APC-bald NOD female mice have noninvasive insulitis lesions. (A and F) Hematoxylin and eosin stain of pancreata from >300-day-old, nondiabetic, class I APC-bald NOD mice showed infiltrates limited to the periislet region. Pancreatic sections were stained by immunofluorescence with anti-insulin antibody (red) and anti-CD8 antibody (green). Sections were additionally costained with anti-CD4 (B and G), anti-B220 (C and H), anti-CD11c (D and I), or anti-CD11b (E and J). The noninvasive infiltrate of pancreata from class I APC-bald NOD mice contained CD4 T cells, B cells, DCs, and macrophages, but it lacked CD8 T cells.

Fig. 4.

Insulitis severity. (A) A total of 208 islets of age-matched class I APC-normal NOD (n = 6), 159 islets of class I APC-bald NOD (n = 7), and 76 islets of NOD.β₂M^−/− (n = 3) were scored for the degree of insulitis as described in Materials and Methods. Bars: white, grade 0; horizontal lines, grade 1; gray, grade 2; vertical lines, grade 3; black, grade 4. (B) Class I APC-bald NOD mice did not develop invasive insulitis even when reconstituted with CD8 T cells (n = 8). Negative control NOD.scid mice (n = 2) and CD4⁺NOD.scid mice (n = 2) did not develop invasive insulitis, whereas positive control CD4⁺NOD.scid mice reconstituted with CD8 T cells (n = 4) did develop invasive insulitis. Splenocytes from class I APC-bald mice were capable of mediating invasive insulitis in NOD.scid recipient mice with class I-sufficient APCs. NOD.scid recipient mice developed invasive insulitis when reconstituted with either class I APC-bald (n = 3) or class I APC-normal (n = 5) splenocytes. (C) Immunofluorescent staining of invasive islet infiltrates in NOD.scid recipients of class I APC-normal splenocytes (i), class I APC-bald splenocytes (ii), and NOD CD4 and CD8 T cells (iii) demonstrates the presence of CD4 and CD8 T cells. Islet infiltrates of NOD.scid recipients of CD4 T cells alone were noninvasive (iv). All sections were stained with an anti-insulin antibody (red), an anti-CD8 antibody (green), and an anti-CD4 (blue).

Reconstitution of APC-Bald NOD Mice with CD8 T Cells Does Not Restore Invasive Insulitis.

To determine whether the failure of APC-bald NOD mice to develop invasive insulitis was the result of the inability of the class I-deficient APCs to prime diabetogenic CD8 T cells, APC-bald NOD mice, or control CD4⁺NOD.scid mice were reconstituted with CD8 T cells from young NOD donors. None of the APC-bald recipients developed invasive insulitis (0 of 8; Fig. 4B), indicating that the absence of invasive insulitis in APC-bald NOD mice was not the result of a deficient CD8 T cell repertoire. In contrast, all of the positive control NOD.scid recipients of CD4 and CD8 T cells (4 of 4) and none of the negative control NOD.scid that received CD4 T cells alone developed invasive insulitis (n = 2).

T Cells from APC-Bald NOD Mice Are Diabetogenic in the Presence of Class I-Sufficient APCs.

Further evidence demonstrating that the failure of APC-bald NOD mice to develop invasive insulitis was not caused by a deficiency in the CD8 T cell repertoire was shown in adoptive transfer experiments. Splenocytes from APC-bald NOD mice were transferred into NOD.scid recipients that had class I-sufficient APCs capable of priming the CD8 T cells. Both invasive insulitis and diabetes developed in NOD.scid recipients of APC-bald NOD splenocytes (3 of 6) with kinetics and severity similar to those in recipients of APC-normal donor splenocytes (2 of 3) (Fig. 4B). Furthermore, both CD4 and CD8 T cells were observed in the intraislet invasions of NOD.scid recipients that received either APC-bald or APC-normal donor splenocytes (Fig. 4C). Taken together, these results strongly support the argument that the absence of diabetes in APC-bald NOD mice is the result of a deficiency in class I-restricted antigen presentation and not a defect in the CD8 T cell repertoire.

Discussion

To determine whether diabetogenic CD8 T cells require cognate interaction with pAPCs for the initiation and progression to T1D, we generated class I APC-bald NOD mice. These mice expressed MHC class I in all cell compartments except in mature pAPCs. These cells were “bald” for MHC class I because their β₂M gene had been permanently deleted. In terms of the ability to assess the role of CD8 T cells in the initiation of T1D, the class I APC-bald NOD mice have several distinct advantages over both β₂M-deficient NOD mice and adoptive transfer models with NOD mice. Because the β₂M-deficiency in NOD.β₂M^−/− mice is systemic, it causes not only widespread deficiency in MHC class I and CD8 T cells but also the effective deletion of other molecules that rely on pairing with β₂M such as CD1, the iron transporter HFE, and the neonatal FcR. It is important to realize that the other traits of β₂M^−/− mice such as complete NKT cell deficiency, iron accumulation in tissues, and a block in placental transfer of maternal immunoglobulins to the fetus, may also contribute to the inhibition of diabetes in NOD.β₂M^−/− mice (3, 17, 18). However, because the deletion of β₂M in class I APC-bald NOD mice is limited to the mature pAPCs, we were able to avoid the problems inherent in the β₂M^−/− mice. Unlike NOD.β₂M^−/− mice, class I APC-bald NOD mice had normal development of NKT cells and also of CD4 T and B lymphocyte subsets in terms of peripheral numbers and proportions. The most crucial advantage of the class I APC-bald NOD mice over the NOD.β₂M^−/− model is that they did develop CD8 T cells, albeit with reduced numbers. These cells persisted throughout the life span of the mice, indicating that the level and distribution of MHC class I expression on peripheral non-pAPCs and immature pAPCs were sufficient to maintain their long-term survival. The class I APC-bald NOD mouse model therefore allowed us to investigate the spontaneous early MHC class I-dependent cellular interactions that are required for the recruitment and activation of diabetogenic CD8 T cells. In particular, they allowed us to distinguish between the roles of pAPCs and islet β cells in providing these signals.

Class I APC-bald NOD mice not only failed to develop diabetes but their islets were completely free of invasive infiltrate, even in very old mice. However, unlike NOD.β₂M^−/− mice, which remained completely insulitis-free, a noninvasive insulitis was consistently observed in a small percentage of islets in >300-day-old class I APC bald NOD mice. This noninvasive insulitis lesion contained very few CD8 T cells, but it was abundant in all of the other major immune cell types found in the NOD infiltrates, including CD4 T lymphocytes, macrophages, and DCs.

Although it has previously been shown that a similarly reduced number of CD8 T cells is sufficient to cause diabetes in HLA-A2 transgenic NOD.β₂M^−/− mice (19), the repertoire in these mice may have been more diabetogenic than that in the APC-bald NOD mice. It remained possible that the repertoire in APC-bald NOD mice may have been deficient in crucial diabetogenic clones because of reduced selection through the APC-bald thymus. To address this question, we reconstituted APC-bald NOD mice with normal CD8 T cells from young NOD donors. Despite reconstitution of the CD8 compartment in APC-bald NOD mice, invasive insulitis was never observed. This result indicates that the absence of invasive insulitis and diabetes in APC-bald NOD mice was not the result of a deficiency in the CD8 repertoire but rather a defect in antigen presentation. Further evidence was provided by experiments in which splenocytes from the class I APC-bald NOD mice were transferred into a NOD.scid environment. NOD.scid recipients of APC-bald splenocytes developed both intraislet-invasive insulitis and diabetes. This result indicated that despite the reduced numbers of CD8 T cells observed in APC-bald NOD mice, the repertoire was fully capable of causing diabetes in an environment with class I-sufficient antigen presentation.

There are two important conclusions that can be drawn from these data. First, the activation of CD8 T cells by mature pAPCs is not required for the initiation of insulitis because periislet insulitis occurred in APC-bald NOD mice. Furthermore, the complete absence of CD8 T cells within the islet and the paucity of CD8 T cells observed within the periislet infiltrate strongly suggest that any contribution to the development of periislet insulitis by CD8 T cells did not occur within the lesion itself. It is possible that CD8 T cells within the pancreatic lymph node (PLN) contributed to the development of periislet insulitis through interaction with MHC class I expression either on immature pAPCs or on non-APCs. Certainly, naïve 8.3 T cells were found to home to the PLN of class I APC-bald NOD mice. Nevertheless, it is difficult to envisage how naïve CD8 T cells could cause the recruitment of inflammatory cells to the islets of these mice given that the highly diabetogenic 8.3 CD8 T cells did not undergo activation and division in the PLN of class I APC-bald NOD mice.

An alternative explanation for the occurrence of periislet insulitis lesions in class I APC-bald NOD mice but not in NOD.β₂M^−/− mice is one that does not involve CD8 T cells. The failure of NOD.β₂M^−/− mice to develop insulitis may conceivably result not from the absence of CD8 T cells but rather from some other effect of deleting β₂M. One possible explanation is that CD4 T cells in class I APC-bald NOD mice responded to peptides derived from β₂M presented in the context of MHC class II or to β₂M/MHC class I chaperoned peptides shed from β cells. This response could not of course have occurred in NOD.β₂M^−/− mice. This explanation would certainly fit with the findings of Kay et al. (6), who showed that insulitis occurred when β₂M was restored to the islets of NOD.β₂M^−/− mice. In these mice, insulitis was more severe than that seen in the class I APC-bald NOD mice, despite the almost complete absence of CD8 T cells. This finding may be explained by the absence of β₂M expression in the thymus and the consequent lack of CD4 tolerance to β₂M in RIP.β₂M NOD mice. The absence of CD8 thymic selection and the inability to detect thymic β₂M in RIP.β₂M NOD mice do suggest that the CD4 thymocytes with high affinity for β₂M-derived peptides may have escaped thymic censorship. Such CD4 T cells, with high affinity for MHC class II-presented peptides derived from β₂M, or β₂M/MHC class I-chaperoned peptides shed from β cells, could have become primed in RIP.β₂M NOD mice and initiated a more severe invasive insulitis than that observed in class I APC-bald NOD mice. If this interpretation of the data is correct, it brings into question the current dogma that CD8 T cells are essential for the initiation of insulitis.

The second and more important conclusion from these results is that MHC class I expression on islet β cells is not sufficient to activate CD8 T cells to become islet-invasive. No CD8 T cells could be found within the islets of class I APC-bald NOD mice. Despite ample MHC class I expression on all of the islet cells, very few CD8 T cells could be identified even around the periphery of the islet in the periislet insulitis. If MHC class I expression on islet β cells is important at all for the recruitment and retention of CD8 T cells within the islet, as has been proposed (6), then it is an event that must occur downstream of the cognate interaction between CD8 T cells and mature pAPCs. Our observations in NOD mice suggest that the crucial initiating pathogenic events of recruitment and retention of diabetogenic CD8 T cells within the islet are absolutely dependent on previous cognate interaction between CD8 T cells and MHC class I on mature pAPCs but not on β cells.

In summary, we have addressed the cellular interaction requirements for diabetogenic CD8 T cells. A unique mouse model was established in which the expression of MHC class I was ablated exclusively in mature pAPCs in vivo. The ablation was most effective in mature DCs, and it resulted in complete protection from invasive insulitis and therefore diabetes. Interestingly, we found that mild insulitis around the islets persisted, but it could not become islet-invasive in the absence of MHC class I on the majority of pAPCs. These data clearly demonstrate that islet β cells cannot provide the required activation signals to naïve CD8 T cells either to recruit and retain them within the islet tissue or to activate them to become diabetogenic. This finding demonstrates unequivocally that CD8 T cell activation by mature pAPCs is an absolute requirement for the development of intraislet invasive insulitis that ultimately leads to β cell destruction and T1D.

Materials and Methods

Mice.

Mice were housed in microisolator cages in the clean conventional facility at the Precinct Animal Centre, The Alfred Medical Research and Education Precinct, The Alfred Hospital, Melbourne, Australia. Mice were fed a standard diet of irradiated mouse cubes (Specialty Feeds, Perth, Western Australia). All mouse experiments were scrutinized and approved by the Animal Ethics Committee of The Alfred Medical Research and Education Precinct. NOD.β₂M^−/− mice were a kind gift from Thomas Kay (6). NOD.8.3 TCR transgenic mice were used with the kind permission of Pere Santamaria (16).

Generation of Class I APC-Bald NOD Mice.

NOD.β₂M^−/− with the fβ₂M^a transgene mice have been described in ref. 20. Briefly, a β₂M^a transgene incorporating two loxP sites (fβ₂M^a) was injected into NOD zygotes to generate fβ₂M^a transgenic founder mice. One founder was crossed with NOD.β₂M^−/− mice as described in ref. 20. Wild-type β₂M, transgenic β₂M, and the β₂M-null locus were identified in three primer PCRs using the following primers: 5′-TCT GGA CGA AGA GCA TCA GGG-3′ (neoF), 5′-AGG GGT AAT TGC TCA GCT CTC-3′ (3β₂M^aintronA⁺), and 5′-CAG TAG ACG GTC TTG GGC TC-3′ (β₂Mexon2). The incidence of diabetes in NOD.β₂M^−/−.fβ₂M^a mice, which for the purposes of this work are referred to as class I APC-normal NOD, was not significantly different from NOD.β₂M^+/− mice (20).

A transgene (pIE-Cre) containing the murine MHC class II IEα promoter (21) and Cre recombinase was made by replacing exons 1–4 of the IEα gene with the 1.2-kb MluI fragment of plasmid pMC-Cre [gift from Richard Murray, DNAX, Palo Alto, CA (ref. 22)]. This construct was injected into NOD zygotes, and three transgenic founders were generated. Two of three founder lines expressed Cre in a tissue-specific manner, and one of these lines was chosen for further study. The incidence of diabetes in NOD.β₂M^+/− pIE-Cre transgenic mice was not significantly different from the diabetes incidence in NOD.β₂M^+/− mice in the colony (data not shown). The NOD.β₂M^−/−.pIECre line was crossed with the NOD.β₂M^−/−.fβ₂M^a line. The resulting NOD.β₂M^−/− offspring carrying both the pIE-Cre transgene and the fβ₂M^a transgene are referred to in this work as class I APC-bald NOD mice. The pIE-Cre transgene was identified by PCR with the following primer pair: 5′-TCG CTA CCC ATC TTC CAG AG-3′ (specific for the IE promoter) and 5′-CGA ACC TCA TCA CTC GTT GCA TCG A-3′ (specific for Cre).

T Cell Transfer Studies.

CFSE labeling of diabetogenic T cells.

CD8 T lymphocytes from NOD.8.3 TCR transgenic mice, bearing the TCR specificity for the β cell antigen IGRP (16), were enriched from pooled spleen and lymph nodes by depletion of non-T cells with magnetic beads (Pan T cell isolation kit; Miltenyi Biotech, Auburn, CA).

Transgenic 8.3 TCR T cells were labeled with 5 μM CFSE (Molecular Probes, Eugene, OR) at a cell density of 50 × 10⁶ cells/ml in Hanks' balanced salt solution (HBSS; Invitrogen, Carlsbad, CA) at room temperature for 5 min. Cells were washed three times in HBSS and adjusted to ≈10 × 10⁶ CD8 T cells per 0.2 ml. Donor 8.3 TCR CD8 T cells were derived from either male or female NOD8.3 mice that were 21–24 days old. Recipient class I APC-bald NOD mice (n = 6) were 88 ± 17 days old, and recipient class I APC-normal NOD mice (n = 5) were 81 ± 18 days old. All recipient mice were female. Recipient mice were injected i.p., and 6 days after transfer, pancreatic lymph node cells were isolated and labeled with an anti-CD8 monoclonal antibody and an MHC K^d tetramer reagent, which binds specifically to the 8.3 TCR. The tetramer was a phycoerythrin conjugate made by the National Institutes of Health, Core Tetramer Facility (Emory University, Atlanta, GA) by using the peptide NH₂-VYLKTNVFL-COOH (synthesized by Mimotopes, Victoria, Australia). Cellular division of CD8⁺ tetramer⁺ donor T cells, as indicated by dilution of CFSE, was assessed on a FACScalibur flow cytometer (Becton Dickinson, San Jose, CA) and analyzed with CellQuest Pro software (Becton Dickinson). Five thousand CD8⁺ tetramer⁺ events were analyzed per lymph node.

CD8 T cell reconstitution.

Female class I APC-bald mice and NOD.scid mice, which had previously been reconstituted with 5 × 10⁶ prediabetic CD4 T cells (referred to as CD4⁺NOD.scid), were reconstituted i.p. with a total of 4–5 × 10⁶ purified CD8 T cells. T cells were enriched from pooled spleens and lymph nodes (inguinal, brachial, axillary, mesenteric, and submandibular) by removing non-T cells with magnetic beads (Pan T cell isolation kit). CD8 T cells were purified to no less than 95% by FACS sorting (FACSAria; Becton Dickinson). Donor CD8 T cells were derived from female NOD mice that were 22–35 days old. Recipient class I APC-bald mice were 64–136 days old, and recipient CD4⁺NOD.scid mice were 42–103 days old. At 114–163 days after transfer, pancreata were collected and scored for histological analysis.

Splenocyte transfer studies.

NOD.scid mice (age 112–139) were reconstituted with pooled, unfractioned splenocytes and LN cells (inguinal, axillary, mesenteric, submandibular) from either class I APC-bald (2% CD8) or class I APC-normal (16% CD8) donors. Adult donor mice were aged-matched. Recipient mice received splenocytes equivalent to 5 × 10⁶ CD8 T cells, injected i.p. in 0.2 ml of PBS. At 100 days after transfer, pancreata were collected for histological analysis.

Diabetes and Pancreas Histology.

Mice were observed daily for signs of diabetes, and they were tested for glycosuria with Clinistix (Bayer, Elkhart, IN). Pancreas tissue was preserved for histology in 10% neutral-buffered formalin (Sigma, St. Louis, MO) before embedding in paraffin wax. Five-micrometer sections of pancreas were stained with hematoxylin and eosin. Sections were sampled from three separate levels of fixed tissue separated by 100 μm. Between 10 and 100 islets were scored from each pancreas. Insulitis severity was scored on a scale of 0–4: 0, no infiltrate; 1, periislet infiltrate; 2, circumferential accumulation of inflammatory cells; 3, intraislet infiltration; 4, severe structural derangement and complete loss of β cells.

Immunofluorescence of Pancreatic Islets.

APC-bald and APC-normal NOD mice.

The phenotype of insulitis lesions was examined by immunofluorescent staining. Tissues were embedded in Tissue-Tek OCT compound (Bayer), frozen, sectioned, and fixed in acetone. Sections from female class I APC-bald NOD mice and female nondiabetic class I APC-normal NOD mice were incubated with fluorescent primary antibodies specific for mouse CD4 (CT-CD4; Caltag, Buckingham, U.K.), CD8a (CTCD8a; Caltag), CD11c (HL3; Becton Dickinson), CD45R/B220 (RA3-6B2; Becton Dickinson), MHC class I K^d (SF1-1.1; Becton Dickinson), and CD11b (M1/70.15; Becton Dickinson). Insulin was stained with polyclonal guinea pig anti-swine insulin antiserum (DAKO, Carpinteria, CA), which was detected with a goat anti-guinea pig IgG antiserum conjugated to Texas red (AbCam, Cambridge, U.K.).

T cell transfer studies.

The phenotype of insulitis lesions in the T cell transfer studies was determined in OCT-embedded frozen tissue sections fixed in 4% paraformaldehyde in PBS, with a phycorerythrin-conjugated anti-mouse CD4 antibody (CT-CD4; Caltag) and a FITC-conjugated anti-mouse CD8 antibody (5H10; Caltag). Insulin was stained with polyclonal guinea pig anti-swine insulin antiserum (Dako), which was detected with an Alexa Fluor 647-conjugated goat anti-guinea pig antibody (Molecular Probes).

Flow Cytometry.

Single cell suspensions from thymus, spleen and lymph nodes were labeled with anti-mouse CD16/CD32 (FcIII/II receptor) and antibodies specific to mouse cell surface antigens: CD4 (CT-CD4; Caltag), CD8 (CT-CD8a; Caltag), CD11b (M1/70.15), CD11c (HL3), CD44 (IM7), TCRβ (H57-597), CD45R/B220 (RA3-6B2), CD19 (ID3), CD21 (7G6), CD23 (B3B4), IgM (II/41), F4/80 (CI:A3-1; Caltag), D^b (KH95), K^d (SF1-1.1), and I-A^k (10-3.6; cross-reactive with I-A^g7). All primary antibodies were obtained from Becton Dickinson unless specified otherwise. Biotinylated primary antibodies were detected with a streptavidin-PerCP (SAv-PerCP; Becton Dickinson) conjugate. Cell data were acquired on a FACScalibur flow cytometer and analyzed with CellQuest Pro software.

Abbreviations

APC

antigen-presenting cell

β₂M

β₂-microglobulin

CFSE

carboxy-fluorescein diacetate succinimidyl ester

DC

dendritic cell

IE

major histocompatibility complex H-2IE

NKT cell

natural killer T cell

NOD

nonobese diabetic

pAPC

professional antigen-presenting cell

RIP

rat insulin promoter

T1D

type 1 diabetes

TCR

T cell receptor.