Minimal Impact of a De Novo–Expressed β-Cell Autoantigen on Spontaneous Diabetes Development in NOD Mice Free

Transgenic RIP-LCMV-GP H-2^b (B6-GP), RIP-LCMV-nucleoprotein H-2^b (B6-nucleoprotein) (backcross F_10–12 to C57BL/6J), and RIP-LCMV-nucleoprotein H-2^d (BALB/c-nucleoprotein) mice expressing LCMV-glycoprotein (GP) or -nucleoprotein in their pancreatic β-cells, respectively, and for the B6-nucleoprotein and BALB/c-nucleoprotein line also in the thymus, were previously generated and characterized in our laboratory (19). Screening was performed by Southern blot and PCR as described previously (19). C57BL/6J (B6), BALB/c, and NOD Lt/J (NOD) mice were obtained from The Jackson Laboratories (Bar Harbor, ME).

The complete cDNA gene for the LCMV-nucleoprotein coding regions was assembled from overlapping cDNA clones derived from the S RNA segment of LCMV (ARM clone 53b) (20). The LCMV-nucleoprotein cDNA was cloned into the Ins-plasmid as BamHI fragment. The Ins-plasmid contains the human insulin promoter (Ins) and the Hepatitis B virus polyadenylation signal (HBS) as a terminator of transcription. Transgenic Ins-LCMV-nucleoprotein NOD mice (NOD-nucleoprotein) were generated at the Scripps Transgenic and Embryonic Steam Cell core facility. Briefly, the SacI/HindIII linearized Ins-LCMV-nucleoprotein-HBS gene construct was microinjected into one of the pronuclei of fertilized NOD mouse eggs, which were subsequently implanted into the oviduct of pseudo-pregnant recipient NOD mice. Offspring were screened for presence of the transgene by PCR and slot blot hybridization. The primers nucleoprotein-sense (5′-CAGTTATAGGTGCTCTTCGC) and nucleoprotein-antisense (5′-AGATCTGGGAGCCTTGCTTTG) were used for the amplification of transgenic LCMV-nucleoprotein from mouse-tail DNA yielding a 289-bp PCR product (19). Slot blot hybridization was performed as described elsewhere (20) using a radiolabeled LCMV-nucleoprotein–specific probe.

Organs were snap-frozen and homogenized in Trizol reagent (Invitrogen), and total RNA was extracted. To prevent its degradation, pancreatic RNA was re-extracted a second time to ensure complete removal of RNases. Residual genomic DNA was eliminated by DNase digestion for 20 min with the TURBO-DNAfree kit (Ambion). Reverse transcription was carried out with 5 μg total RNA using the first-strand cDNA synthesis kit (Amersham Biosciences) with the oligo dT-primer provided in the kit according to the manufacturer's instructions. PCR primers were as follows: β-actin, 5′ GACGGCCAGGTCATCACTAT 3′ and 5′ ACATCTGCTGGAAGGTGGAC 3′; and nucleoprotein, 5′ CAGTTATAGGTGCTCTTCGC 3′ and 5′ AGATCTGGGAGCCTTGCTTTG 3′.

Thymus, pancreas, and brain were obtained from 6- to 8-week-old B6-nucleoprotein (n = 4), NOD-nucleoprotein (n = 4), and NOD (n = 4) mice. Pancreatic islets were isolated using a Ficoll gradient as previously described (21). Proteins from each organ were extracted from the tissues using the T-PER tissue protein extraction reagent (Pierce) as recommended by the manufacturer. Extracted proteins were used in an enzyme-linked immunosorbent assay (ELISA) as follows. Briefly, the wells were coated overnight at 4°C with 10 or 100 μg/ml proteins diluted in 1× PBS and extracted from the pancreatic islets or the thymus, respectively. Brain protein extracts were coated at both 10 and 100 μg/ml in 1× PBS. The wells were washed with 0.05% Tween 20 in PBS (PBS-T) and blocked with 2% BSA in PBS-T (saturation buffer) for 1 h at 37°C. After three washings, rabbit polyclonal anti-LCMV antibody (22) or guinea pig polyclonal anti-insulin antibody (Dako) both diluted 1:2,000 in saturation buffer were incubated for 2 h at 37°C. After three washings, biotinylated anti-rabbit or anti–guinea pig secondary antibodies (dilution 1:2,000) were added to the wells in saturation buffer for 1 h at 37°C. After washings, horseradish peroxidase–conjugated avidin (dilution 1:2,000; Vector Laboratories, Burlingame, CA) was used for 1 h at 37°C. The reactivity was revealed by using the 2,2′-azino-bis-[3-ethylbenzthiazoline-6-sulfonic acid] substrate (Sigma, St. Louis, MO). The absorbance was measured at 405 nm after 30 min.

Blood glucose values were determined using OneTouch Ultra (LifeScan, Milpitas, CA). Unless otherwise indicated, mice with blood glucose values >300 mg/dl were considered diabetic.

LCMV Armstrong clone 53b (LCMV) prepared by a single passage on BHK-21 cells was used throughout experiments. Unless otherwise indicated, mice were infected with a single dose of 1 × 10⁵ plaque forming units (pfu) i.p.

Cytotoxic T-lymphocyte (CTL) activity was measured in a standard 5- to 6-h ex vivo ⁵¹Cr release assay as described (23). In brief, MC57 (H-2D^b–positive fibroblast cell line), syngeneic NIT (β-cell insulinoma cell line, H-2^g7), or allogeneic BALB/c (H-2^d) target cells were infected with LCMV (multiplicity of infection = 0.1) or pulsed with major histocompatibility complex (MHC) class I–restricted LCMV peptides (10⁻⁵ mol/l). Assays using effector splenic lymphocytes at days 7 or 8 after infection used effector to target ratios of 50:1, 25:1, and 12.5:1, and those using CTL clones used ratios of 5:1 and 2.5:1.

Pancreata were dissected, and PDLNs were carefully removed. Pancreata were mechanically disrupted using glass slides. Material was pooled from two to five mice, and the lymphocytes were separated using a lympholyte gradient (CedarLane).

Flow cytometry antibodies were purchased from BD Pharmingen (San Diego, CA), eBioscience (San Diego, CA), Caltag Laboratories (Burlingame, CA), and R&D Systems (Minneapolis, MN). Single-cell suspensions were stained with MHC class I tetramers and fluorescein isothiocyanate–conjugated anti-CD8 antibody for 30 min at 4°C. For multicolor fluorescence-activated cell sorter analysis, single-cell suspensions were first stained with MHC class I tetramers for 15 min at room temperature and then with the antibody mix for 20 min at 4°C. Afterward, cells were washed and fixed in 2% paraformaldehyde. For intracellular stains, single-cell suspensions were restimulated for 5–6 h with NP396 or NRP-V7 peptide in the presence of brefeldin A (Sigma). Cells were stained for surface expression of CD8, CD3, CD44, and CD62L and were then fixed, permeabilized, and stained for intracellular IFN-γ and TNF. Samples were acquired either on a FACSCalibur or LSRII flow cytometer and analyzed using CellQuest (BD Biosciences, San Jose, CA) or FlowJo (version 8.1.1; Tree Star, Ashland, OR) software. The following MHC class I tetramers were used: H-2D^b refolded with the LCMV peptide NP396 (FQPQNGQFI) (24), or H-2K^d refolded with the NRP-V7 peptide (KYNKANVFL) or the negative control peptide TUM (KYQAVTTTL) (17). NRP-V7 tetramers were a kind gift of Rusung Tan (Vancouver, Canada).

Tissues were immersed in Tissue-Tek OCT compound (Sakura Finetek USA, Torrance, CA) and quick-frozen on dry ice. Six- to 10-μm cryomicrotome tissue sections were affixed to sialin-coated Superfrost Plus slides (Fisher Scientific, Pittsburgh, PA). Sections were fixed with 95% EtOH at −20°C and washed in PBS, and then endogenous avidin-biotin was blocked using the avidin/biotin blocking kit from Vector Laboratories. Primary anti-mouse CD8 (Pharmingen) and biotinylated secondary antibodies (Vector Laboratories) were incubated with the sections, and the color reaction was obtained by sequential incubation with avidin-peroxidase conjugate (Vector Laboratories) and diaminobenzidine–hydrogen peroxide.

The generation and maintenance of LCMV-specific T-cell clones in long-term cultures have been previously described (25,26). In brief, the NP396-specific NOD T-cell clone 9-A-8 and the GP33-specific NOD T-cell clone 13-G-8 were isolated from LCMV-infected NOD recipients and stimulated frequently with irradiated NP396 or GP33 peptide-loaded syngeneic macrophages, respectively, and recombinant human IL-2 (20 units/ml).

NOD-nucleoprotein transgenic mice expressing the LCMV-nucleoprotein in pancreatic β-cells were generated directly on the NOD background (Fig. 1A). Four NOD-nucleoprotein transgenic lines (lines 6152, 6156, 6159, and 6171) were established (Fig. 1B; data not shown). Mice derived from lines 6152 and 6171 expressed LCMV-nucleoprotein in the pancreas and in the thymus (Fig. 1C and D; data not shown); lines 6156 and 6159, however, did not express LCMV-nucleoprotein in the pancreas (data not shown). The expression of LCMV-nucleoprotein in the pancreas correlated with the speed of diabetes induction after infection with LCMV. In fact, only mice expressing nucleoprotein in the pancreas (derived from lines 6152 and 6171) were susceptible to LCMV-induced diabetes (Fig. 2A). Pancreata from these mice analyzed by immunohistochemistry 1 week after infection with LCMV also revealed the early infiltration of islets by CD8 T-cells (Fig. 2B).

Diabetes in the new NOD-nucleoprotein transgenic mice occurred rapidly within 2 weeks of infection with LCMV (Fig. 2A), compared with the slow onset (3–6 months) in B6-nucleoprotein mice (19). The rapid onset of diabetes in NOD-nucleoprotein transgenic mice correlated well with the comparatively higher frequency of NP396-specific T-cells, which made up 2% of CD8 T-cells in NOD-nucleoprotein versus 0.8% in B6-nucleoprotein lines (Table 1). Despite the rapid onset of diabetes, NOD-nucleoprotein transgenic mice did exhibit some degree of tolerance to the nucleoprotein antigen, as evidenced by the lower NP396-specific CD8 T-cell frequency and cytotoxic activity compared with NOD nontransgenic mice (Table 1). Because nucleoprotein is expressed in the thymus of NOD-nucleoprotein transgenic mice (Fig. 1C), nucleoprotein-specific tolerance could be explained by thymic deletion of high avidity nucleoprotein-specific CD8 T-cells. Similar observations have already been made for B6-nucleoprotein transgenic mice (19). In addition, other tolerance mechanisms, such as peripheral tolerance, may also be involved in deletion of nucleoprotein-specific CD8 T-cells.

The precise time course of spontaneous diabetes development in NOD-nucleoprotein transgenic females with (derived from line 6171) or without pancreatic nucleoprotein expression (derived from lines 6156 and 6159), and in transgene-negative littermates (NOD-nucleoprotein⁻) was followed to understand the impact of the de novo–expressed nucleoprotein islet antigen on the spontaneous disease. Interestingly, overall incidence of diabetes was not significantly increased (log-rank test, P < 0.1392), and the kinetics of development were only slightly accelerated (Fig. 2C). Thus, expression of the foreign nucleoprotein antigen in pancreatic β-cells only had a minimal impact on the course of spontaneous diabetes development in NOD-nucleoprotein transgenic mice. This was surprising because NOD-nucleoprotein mice did not exhibit complete tolerance to nucleoprotein (Table 1) and rapid induction of diabetes had occurred after LCMV infection (Fig. 2A), demonstrating that there were no other defects or epigenetic phenomena that would prevent destruction of β-cells in these mice.

We next determined whether introduction of nucleoprotein into the spontaneous autoimmune process in NOD mice would result in the accumulation of nucleoprotein-specific CD8 lymphocytes. NP396-specific CD8 lymphocytes were therefore isolated at different time points from the PDLNs and spleens during spontaneous diabetes development and enumerated by NP396/D^b tetramer staining (Fig. 3A and B). As expected, NP396-specific CD8 T-cells were not detected in the PDLNs or spleens of transgene-negative littermates (NOD-nucleoprotein⁻) (Fig. 3A and B, white circles). However, in NOD-nucleoprotein transgenic mice expressing nucleoprotein in the pancreas (NOD-nucleoprotein⁺, derived from line 6171) NP396-specific CD8 T-cells were detected in the PDLNs but not the spleens (Fig. 3A and B, black circles). NP396-specific T-cells were found in the PDLNs of 9 of 13 mice between 13 and 18 weeks of age and were less frequently detected (2 of 9 mice) in 9-week-old NOD-nucleoprotein⁺ mice (Fig. 3A). The frequency of NP396-specific CD8 T-cells in the PDLN was quite low (0.3–0.7% NP396/D^b tetramer⁺ CD8 T-cells) compared with the frequencies arising after systemic LCMV infection (2–3% of CD8 T-cells; data not shown). In addition, not all mice examined had detectable NP396-specific CD8 T-cells (Fig. 3A).

The frequency of NP396-specific CD8 T-cells was then compared with those specific for another autoantigen, IGRP, as detected by NRP-V7/K^d tetramer staining (Fig. 3C and D). The average frequencies of NRP-V7–and NP396-specific CD8 T-cells in the PDLNs of 13- to 18-week-old mice were very similar, representing 0.52 and 0.54% of CD8 T-cells, respectively. However, NRP-V7–specific CD8 T-cells were found in a higher proportion of NOD (NOD-nucleoprotein⁺ and NOD-nucleoprotein⁻) mice, especially at early time points during diabetes development (nine of nine mice with NRP-V7/K^d tetramer⁺ staining at 9 weeks of age) (Fig. 3C). In addition, NRP-V7–but not NP396-specific CD8 T-cells were found systemically in the spleen of NOD-nucleoprotein⁺ and NOD-nucleoprotein⁻ mice (Fig. 3D), as reported previously (15).

Although the frequency of NP396–or NRP-V7–specific CD8 T-cells in the PDLNs was relatively low (Fig. 3), these cells were enriched in the pancreas itself (Fig. 4A). Lymphocytes were isolated from pooled pancreas samples from naive NOD-nucleoprotein transgenic mice expressing nucleoprotein in the pancreas (NOD-nucleoprotein⁺, derived from line 6171) or nontransgenic littermates (NOD-nucleoprotein⁻). The frequency of NP396–and NRP-V7–specific CD8 T-cells in the pancreas was much higher than in the PDLNs (Figs. 4A and 3A and C). However, NRP-V7–specific CD8 T-cells accumulated in the pancreas to fourfold higher levels than NP396-specific T-cells (Fig. 4A).

To analyze whether the increased accumulation of NRP-V7–specific T-cells in the pancreas was paralleled by a more activated phenotype of these cells and/or different expression of homing markers, we compared expression of activation markers (CD62L, CD44, and CD122) (Fig. 4B, left) and expression of CXCR3 (Fig. 4B, right) on NP396–and NRP-V7–specific CD8 T-cells. It has been shown that activated T_H1 but not T_H2 cells express CXCR3 (27,28) and that pancreatic β-cells in NOD mice produce CXCL10, the ligand for CXCR3 (29). Therefore, increased CXCR3 expression on activated T-cells might well correlate with increased migration and accumulation of these cells into the pancreas. NP396–and NRP-V7–specific CD8 T-cells showed an activated/memory phenotype (CD62L^loCD44^hiCD122⁺); however, NRP-V7–specific T-cells showed stronger up-regulation of CD44, which is indicative of a stronger activation/memory status (Fig. 4B). In addition, 40% of NRP-V7–specific but only 5% of NP396-specific CD8 T-cells expressed CXCR3 (Fig. 4B), and this correlated well with the higher frequencies of NRP-V7–specific T-cells found in the pancreas (Fig. 4A). Furthermore, the increased accumulation of NRP-V7–specific CD8 T-cells in the islets correlated well with the higher frequencies of these cells found systemically in the spleen compared with NP396-specific CD8 T-cells. Such NP396-specific CD8 T-cells were not found in the spleen of NOD-nucleoprotein transgenic mice expressing nucleoprotein in the pancreas (Fig. 3B), whereas NRP-V7–specific cells were detected in the spleen early, at 9 weeks of age, in NOD mice (Fig. 3D). Thus, differences in systemic frequencies of CD8 T-cells specific for either antigen in the spleen was reflected in differences in their frequencies in the islets, whereas identical levels were found in the pancreatic node.

As discussed above, pancreatic NP396-specific CD8 T-cells in naïve NOD-nucleoprotein transgenic mice with pancreatic nucleoprotein expression (NOD-nucleoprotein⁺, derived from line 6171) showed an activated phenotype (Fig. 4B). Hence, it was surprising that the de novo activation of NP396-specific CD8 T-cells had such a minimal impact on spontaneous diabetes, despite the fact that infection with LCMV induces rapid diabetes. We therefore analyzed whether these activated cells were also functional. We infected NOD-nucleoprotein⁺ transgenic mice with LCMV or left them untreated and compared the ability of the different CD8 T-cell populations in the pancreas to secrete IFN-γ after a 5-h restimulation period with NP396 peptide (Fig. 5A). Both CD8 T-cell populations produced IFN-γ rapidly, a hallmark of antigen-experienced T-cells (Fig. 5A, inset). We then compared the functional avidity of pancreatic NP396-specific CD8 T-cells from naïve or day-7–infected NOD-nucleoprotein⁺ mice by stimulating these cells directly ex vivo with graded doses of NP396 peptide followed by quantification of intracellular IFN-γ production. NP396-specific CD8 T-cells from naïve NOD-nucleoprotein⁺ mice had similar functional avidity to those generated after infection with LCMV (Fig. 5A).

We then compared the overall frequency of NP396-specific CD8 T-cells in the spleen, PDLNs, and pancreas of naive and LCMV-infected NOD-nucleoprotein⁺ transgenic mice and included transgene-negative littermates as negative controls (Fig. 5B). The frequency of NP396-specific CD8 T-cells in all organs tested was much higher in day 7 LCMV-infected than in naïve NOD-nucleoprotein⁺ transgenic mice (Fig. 5B). This was particularly apparent in the pancreas, where LCMV infection resulted in an approximate sevenfold increase in the frequency of NP396-specific CD8 T-cells compared with those detected in naïve mice (Fig. 5B). Furthermore, the total number of NP396-specific CD8 T-cells generated in the spleen after LCMV infection was much higher even than those specific for NRP-V7 (Fig. 5C). Thus, accumulation of autoaggressive CD8 T-cells in the islets and the resulting impact on disease kinetics strongly correlates with the total number of antigen-specific T-cells present. The frequency of NP396-specific CD8 T-cells that we only found locally in the PDLNs of naïve NOD-nucleoprotein⁺ mice was possibly too low to systemically re-circulate and significantly impact the course of the disease.

We then analyzed whether systemic increase of antigen-experienced NP396-specific CD8 T-cell numbers in naïve NOD-nucleoprotein⁺ transgenic mice would result in diabetes induction because after LCMV infection, the number of systemic NP396-specific CD8 T-cells increased drastically and the mice developed diabetes rapidly. We therefore generated two LCMV-specific clones, NP396-specific CD8 T-cell clone 9-A-8 and GP33-specific CD8 T-cell clone 13-G-8, from LCMV-memory NOD mice as described in research design and methods. To determine specificity and function of the different CD8 T-cell clones, we restimulated the clones for 5 h with irradiated macrophages, which were left untreated or loaded with the different peptides, and we then measured intracellular IFN-γ secretion (Fig. 6A). NP396-specific CD8 T-cell clone 9-A-8 secreted IFN-γ after restimulation with NP396-loaded but not GP33-loaded macrophages, whereas GP33-specific CD8 T-cell clone 13-G-8 exclusively secreted IFN-γ after restimulation with GP33-loaded but not NP396-loaded macrophages (Fig. 6A). Specificity and functionality of NP396- or GP33-specific T-cell clones 9-A-8 and GP33 were additionally confirmed by NP396/D^b and GP33/D^b tetramer staining, respectively, and by lysis of peptide-loaded MC57 or NIT target cells (data not shown). To analyze the influence of increased systemic numbers of antigen-experienced autoaggressive T-cells on diabetes induction, T-cell clones were restimulated overnight with irradiated peptide-loaded macrophages, followed by intravenous injection of 3 × 10⁵ NP396-specific 9-A-8 or 3 × 10⁵ GP33-specific 13-G-8 T-cells into naïve NOD-nucleoprotein transgenic mice expressing nucleoprotein in the pancreas (NOD-nucleoprotein⁺, derived from line 6171) (Fig. 6B). Four weeks after transfer of NP396-specific T-cells, one-half of the mice had developed diabetes, whereas all of the mice having received GP33-specific T-cells remained diabetes free (Fig. 6B). Thus, increasing systemic numbers of activated autoantigen-specific CD8 T-cells increased spontaneous diabetes incidence in naïve NOD-nucleoprotein⁺ transgenic mice. In parallel, we analyzed the influence of higher systemic NP396-specific T-cell numbers on diabetes kinetics in an inflammatory environment, e.g., during LCMV infection. NOD-nucleoprotein⁺ transgenic mice were infected with LCMV, and 4 days later they were either left untreated or injected intravenously with 1 × 10⁵ NP396-specific 9-A-8 T-cells (Fig. 6B). NOD-nucleoprotein⁺ mice that received NP396-specific T-cells developed diabetes faster and with a higher incidence than LCMV-infected mice that had not received T-cells (Fig. 6B). In summary, increasing systemic numbers of antigen-experienced NP396-specific T-cell in a naïve or inflammatory environment resulted in accelerated diabetes development and increased diabetes incidence.

Evidence presented in this report shows that introduction of a novel autoantigen into the spontaneous disease process in the NOD mouse had no significant impact on incidence of type 1 diabetes and only led to a minimal acceleration of the disease course. This was quite unexpected, because antigenic spreading is responsible for relapses in the EAE model for multiple sclerosis and was therefore thought of as a hallmark event in the pathogenesis of T-cell-mediated autoimmune disorders. One could hypothesize, based on these findings, that the autoimmune process in the NOD model is already driven by a variety of antigens at the time when nucleoprotein-specific T-cells are being activated in the PDLNs. Therefore, the impact of one additional antigen is rather limited. Conversely, attempting therapeutic elimination of antigenic specificities that behave like the response to the nucleoprotein neo-antigen will probably have only a small impact in reducing disease incidence and severity, if this elimination occurs in the latter part of the pre-diabetic phase. Although the early elimination of certain autoreactive T-cells (anti–insulin-1 or anti-IGRP) by using knockout mice (13,30) or peptide treatment (17) has been successful in preventing diabetes, this might not have a high impact later during the pre-diabetic phase or in recent-onset diabetic patients. In addition, other studies have demonstrated that elimination of the T-cell response to another autoantigen, GAD, does not impact disease course (12,31).

An important consideration is that different autoantigens appear to take on differential roles in the autoimmune process. Some are suited as “drivers,” whereas others act as followers, with a lesser impact on disease. The precise hierarchy might vary inter-individually and might also depend on the properties of the antigen. The intracellular compartmentalization of islet antigens, for example, might play an important role in establishing such a hierarchy. It is thus possible that only certain autoantigens are capable of inducing “driver” populations that significantly impact disease; possible examples of such autoantigens are IGRP (17) and proinsulin (32). This consideration will be important for devising immune-based preventions in type 1 diabetes, and we propose that induction of regulatory lymphocytes that can act as bystander suppressors for multiple antigenic autoaggressive specificities rather than elimination of targeted specificities might be a more effective strategy as long as the initiating autoantigens and driver epitopes, if existent, remain elusive in human type 1 diabetes pathogenesis (8,9,33,34).

We would argue that LCMV-nucleoprotein has antigenic properties that make it suitable as a driver antigen, as evidenced by the rapid diabetes onset after LCMV infection. The lack of detection of systemic NP396-specific CD8 T-cells during the spontaneous disease course but their high frequency after systemic LCMV infection in the NOD-nucleoprotein⁺ mice is therefore surprising. The fact that the frequency of NP396-specific T-cells found in the PDLNs of responder mice was similar to the frequency of NRP-V7–specific T-cells argues that NP396 can be cross-presented efficiently. Why, however, NP396-specific CD8 T-cells were only present in ∼70% of mice remains elusive and will be the subject of future studies.

It is intriguing that CD8 T-cells specific for NP396 were only found in the PDLNs at levels comparable with those specific for NRP-V7. As mentioned earlier, no NP396-specific CD8 T-cells were found in the spleen of naïve NOD-nucleoprotein⁺ transgenic mice, whereas T-cells specific for NRP-V7 were readily detectable. Thus, there appears to be a stark difference in systemic load of CD8 T-cells generated to the nucleoprotein versus the IGRP autoantigen. This systemic difference is ultimately reflected in fourfold lower numbers of NP396-specific T-cells in the islets. One could conclude from these findings that numbers of autoaggressive CD8 T-cells generated in the PDLNs in response to autoantigens released from destroyed β-cells and transported there via lymphoid drainage of the pancreas do not necessarily correlate with disease severity. Rather, the overall systemic load of autoaggressive CD8 T-cells is indicative of their impact on the ongoing autoimmune process. This notion is supported by our finding that NRP-V7–specific CD8 T-cells in naïve NOD and NOD-nucleoprotein⁺ transgenic mice, as well as NP396-specific CD8 T-cells in NOD-nucleoprotein⁺ transgenic mice after LCMV infection exhibited high systemic levels and both appeared to strongly impact the disease course (17,19) (Fig. 5). Why NRP-V7–specific T-cells are present systemically remains elusive at this point. A likely conclusion is that priming to IGRP is not only occurring in the PDLNs but also in other sites, for example, the islets themselves. This would explain their high numbers among infiltrates. However, this is not a sufficient explanation for their comparatively high frequency in the spleen. It is tempting to speculate that this is due to differences in trafficking mediated by differential chemokine receptor expression. For example, more NRP-V7–specific CD8 T-cells were found to express CXCR3 than NP396-specific T-cells. A recent report by Shigihara et al. (35) associated elevated serum CXCL10 (ligand for CXCR3) levels with type 1 diabetes progression in NOD mice. Thus, CXCR3 expression might enable NRP-V7–specific T-cells to leave the pancreas and enter the systemic circulation. In contrast, NP396-specific CD8 T-cells showed very low levels of CXCR3 expression and thus could not enter the systemic circulation. In summary, the reduced levels of CXCR3 expression and the overall lower frequencies of NP396–versus NRP-V7–specific CD8 T-cells in naïve NOD-nucleoprotein⁺ transgenic mice might therefore well explain the lack of accelerated spontaneous diabetes. Previous reports have shown that LCMV infection induced T_C1/T_H1 responses accompanied by strong upregulation of CXCR3 on activated T_C1/T_H1 cells and early upregulation of pancreatic CXCL10 levels (36–38). These findings also explain the accelerated and high incidence of diabetes seen after LCMV infection of NOD-nucleoprotein⁺ transgenic mice (Fig. 2A). LCMV infection induced strong activation and expansion of NP396-specific T-cells and upregulation of CXCR3 (Fig. 5B; data not shown), enabling NP396-specific T-cells to leave the pancreatic lymph node, enter the systemic circulation, and migrate toward the pancreas. This process was followed by rapid type 1 diabetes development. In the end, peripheral priming of a large number of autoaggressive CD8 T-cells, accompanied with increased CXCR3 expression, might be necessary for diabetes development.

Taken together, therapeutic elimination of only one or a few antigenic specificities will not necessarily have a major impact on an established autoimmune process, unless a driver antigen is chosen, to which high numbers of autoaggressive T-cells are generated systemically. Whether spontaneous autoimmunity in the NOD is driven by a multitude of differential specificities or a few driver clones (2) that are present in the PDLNs and islets in high numbers remains unclear. Induction and acceleration of type 1 diabetes could be the result of a simple numbers game and differential CXCR3 expression of activated autoantigen-specific T-cells, as evidenced by rapid induction of diabetes in NOD-nucleoprotein⁺ transgenic mice after LCMV infection or after adoptive transfer of antigen-experienced NP396-specific T-cell clones.

From a therapeutic point of view, as long as autoaggressive driver specificities remain elusive, induction of bystander regulatory T-cells that can act locally in the PDLNs and affect multiple autoaggressive specificities (9) and even trigger spreading of regulatory functions (39) might be a more promising strategy than therapeutic elimination of autoaggressive T-cells.

M.G.V.H. was supported by National Institutes of Health Grants AI-44451, DK-51091, and U-19 AI-51973 and by a Juvenile Diabetes Research Foundation (JDRF) grant. M.M.M. is a recipient of a Swiss National Science foundation postdoctoral fellowship. A.E.J is a recipient of a JDRF postdoctoral fellowship. D.B. is a recipient of a European Marie-Curie Outgoing fellowship.

We thank Dr. Rusung Tan (University of British Columbia, Vancouver, BC, Canada) for his kind gift of NRP-V7 and TUM tetramers and Dr. Michael Bruns for providing us with the rabbit polyclonal anti-LCMV antibody.