Idd9 congenic mice develop a high degree of insulitis yet very few go on to develop diabetes 12. To test for differences in the cellular composition of Idd9 islet infiltrates we quantified the major leukocyte populations present in the infiltrate of age-matched 12–14 week old Idd9 congenic and NOD mice by flow cytometry. We observed a striking reduction in the % CD8 T cells in the Idd9 congenic infiltrates (p = 0.02), as shown in Table 1 and Figure 1(A–B). The reduction in CD8 T cell infiltration was also evident in situ, by staining pancreatic sections (Figure 1C). Expression of the activation/memory marker CD44 was equivalent in both CD4 and CD8 T cell populations within the islet infiltrate (Figure 1D) suggesting that the activation state of T cells recruited to the islets is unaffected in Idd9 congenic mice. No significant difference in the % CD8 T cells in secondary lymphoid organs or peripheral blood was observed in Idd9 congenic mice either at 6 weeks of age (Figure 1E), or in older mice (data not shown). Therefore, the decreased % CD8 T cells is specific to the islet environment.

Having found that CD8 T cell infiltration is reduced in Idd9 congenic islets we wanted to test whether the activation of islet-specific CD8 T cells in lymph nodes is inhibited in Idd9 congenic mice. We transferred CFSE-labeled splenocytes from a TCR transgenic strain (8.3NODScid) 13 expressing islet-specific CD8 T cells into Idd9 congenic and NOD recipients. The CFSE+CD8+ (Figure 2A) populations recovered from the panLN of Idd9 congenic and NOD recipients were numerically equivalent (Figure 2B) and underwent similar cell division, primarily in the panLN where the cognate antigen is present (Figure 2C). We also examined the expression of activation markers on adoptively transferred 8.3 T cells. CFSE+CD8+ cells expressed high levels of both CD44 (Figure 2D) and CD11a (Figure 2E) specifically in the panLN. However, there was no difference in the expression of CD44 and CD11a on transferred 8.3 T cells in Idd9 congenic and NOD mice. The activation of islet-specific CD8 T cells on the NOD background is therefore not inhibited in the Idd9 panLN.

To test whether the diabetogenic potential of the anti-islet immune response is reduced in Idd9 congenic mice we transferred splenocytes from 13 week old Idd9 congenic and NOD donors into NODScid recipients, or co-transferred 10 million cells from each strain, and monitored the incidence of diabetes. The transfer of both Idd9 congenic and NOD cells individually induced diabetes in 100 % of recipients (Figure 3A), demonstrating that Idd9 splenocytes have the capacity to induce disease. However, the onset of disease was delayed in the recipients of Idd9 cells (p = 0.002). Idd9 splenocytes are therefore fully capable of inducing diabetes, but do so with delayed kinetics compared to NOD splenocytes. The kinetics of diabetes onset in co-transferred recipients was intermediate between that seen in recipients of either type of splenocytes alone, demonstrating that although Idd9 congenic splenocytes are less pathogenic, there is no evidence that they are able to regulate diabetes induction. This conclusion is further supported by the fact that the diabetes incidence induced by 10 million NOD splenocytes was less at 13 weeks post-transfer (67 %, n = 6, data not shown) than the 100 % diabetes induced by co-transfer of 10 million NOD plus 10 million Idd9 splenocytes.

It had previously been suggested that the reason Idd9 congenic mice are protected against diabetes, despite having an extensive islet infiltrate, could be due to the priming of a non-pathogenic anti-islet T cell response that releases the protective cytokine IL4 12. To test whether IL4 is required for the protection mediated by Idd9 genes we bred Idd9 congenic mice deficient in IL4 (Idd9.IL4KO mice). As shown in Figure 3B, a high incidence of diabetes was seen in both NOD (74 % diabetes, n = 23) and NOD.IL4KO mice (74 %, n = 27), as previously described 14. However, Idd9.IL4KO mice developed the same low frequency of diabetes (3 %, n = 29) as Idd9 congenic mice (3 %, n = 33), demonstrating that IL4 expression is not required for diabetes protection in the Idd9 congenic strain.

We then wanted to test whether Idd9 protection maps to radiation-sensitive cells of the immune system, or to non-immune cells. We lethally irradiated (950 Rad) young Idd9 congenic and NOD mice and reconstituted them with bone marrow from 6 week old NOD mice. Approximately 70 % chimerism was achieved using this protocol [see Additional file 1], therefore while some host hematopoetic cells remain the majority are donor derived. The degree of chimerism was equivalent in the two recipient strains, therefore previous reports describing the competitive advantage of NOD hematopoetic stem cells in allogeneic recipients are not relevant in this case 15. Whereas 83 % (n = 12) of NOD mice developed diabetes by 20 weeks following bone marrow transfer, only 8 % (n = 12) of Idd9 congenic mice developed diabetes (p = 0.0004, Figure 3C). Idd9 congenic mice therefore retain their protection against diabetes in the presence of NOD-derived immune cells. Since no evidence of immune regulation was observed in Figure 3A, this suggests that Idd9 resistance may map to non-lymphoid cells.

To determine whether the chimeric mice exhibit the same characteristics of disease protection as Idd9 congenic mice, we used NOD mice expressing the Thy1.1 allele (NOD.NONThy1.1 mice) as bone marrow donors and tracked the infiltration of donor CD8 T cells in islets. Six weeks following irradiation and bone marrow reconstitution, we isolated islet-infiltrating cells from recipients. While there was no difference in the % Thy1.1+CD4+ cells in the islets of Idd9 congenic and NOD recipients, the %Thy1.1+CD8+ cells was significantly reduced (p = 0.01) in the islets of Idd9 congenic compared to NOD mice (Figure 3D). Both the reduced % CD8 T cells in the islets of Idd9 congenic mice and the protection against diabetes therefore appear to correlate with expression of Idd9 gene products in radiation-resistant cells.

The results of the bone marrow reconstitution experiments suggested the possibility that at least a component of the protective effect of Idd9 genes may be mediated by non-immune cells. To determine whether protection is localized to the islet tissue we tested whether islets from Idd9 congenic mice are resistant to autoimmune destruction. Since we observed a reduced % CD8 T cells in Idd9 congenic islets, we hypothesized that islet resistance to CD8-mediated autoimmunity may be specifically affected. We transplanted islets isolated from 5 week old Idd9 congenic and NOD mice under the kidney capsule of immunodeficient NODScid recipients. Islets from young mice with minimal insulitis damage were used as donors and the islets were cultured for several days before transplant to deplete tissue-resident leukocytes. The transplanted mice were injected with splenocytes from 8.3NODScid TCR transgenic mice and then five days later grafts were taken and the extent of destruction determined by histology (Figure 4A). We observed that Idd9 islet tissue remained intact while the 8.3 CD8 T cells destroyed NOD islets. To quantitate the extent of graft destruction we scored the number of healthy islets in each graft using glucagon staining to reveal the presence of islet remnants. Healthy islets show the typical distribution of glucagon-positive cells scattered around the islet periphery, whereas islets in which the core of target insulin-producing cells has been destroyed exhibit a collapsed glucagon-positive remnant. Control grafts, in recipients where no CD8 T cells were transferred, from both Idd9 congenic and NOD donors were healthy. In grafts from mice that were injected with islet-specific CD8 T cells, on average only 42 % of NOD islets, compared to 88 % Idd9 congenic islets (p = 0.004), were scored as healthy (Figure 4B). Islets from Idd9 congenic mice therefore possess an intrinsic capacity to resist CD8-mediated autoimmune destruction.

NOD islets can be induced to undergo cell death in vitro by treatment with TNF in combination with IFNγ 16. In order to determine whether cytokine-responsiveness is altered in Idd9 congenic islets we tested whether sensitivity to in vitro cytokine-induced cell death is reduced. Since islet cells can undergo cell death when dispersed into single cells we chose to stain live intact islets with a fluorescent live/dead viability stain and electronically quantitate the % dead cells from confocal images. As shown in Figure 4C–D, cytokine treatment induced a significantly greater amount of cell death in NOD islets compared to Idd9 islets (p = 0.01). Islets from Idd9 congenic mice therefore display altered cytokine-responsiveness compared to NOD and are resistant to the cytotoxic effects of TNF/IFNγ treatment in vitro.

The TNFR2 gene lies within the Idd9 congenic interval and the NOD allele generates a protein with 5 amino acid variants compared to the diabetes resistant B6 strain 17. TNFR2 is therefore a potential candidate gene for the altered responsiveness to TNF/IFNγ we observe in Idd9 islets. TNFR2 mRNA expression has been shown to be induced in islet cells during diabetes progression 18. To test whether TNFR2 protein is expressed in islet β cells of diabetes susceptible and resistant strains, we analysed dispersed islets from 5 week old NOD and Idd9 congenic mice by flow cytometry. Using the characteristic autofluoresence in the FL1 channel as a β cell marker 19, we found that approximately 3 % β cells are TNFR2 positive (Figure 5A). Immunohistochemical staining of pancreatic sections from 12 week old Idd9 congenic and NOD mice confirmed that a sub-population of islet β cells in both strains express TNFR2 protein, whereas no staining occurred using the secondary antibody alone, or in islets from B6.TNFR2KO mice (Figure 5B).

One of the key changes that occurs in β cells in response to cytokines is the upregulation of the death receptor Fas 20, and Fas upregulation may play a role in the process of autoimmune islet cell death in vivo. We therefore tested whether TNFR2 expression in islets is required for Fas upregulation following cytokine treatment. Intact islets from B6 mice were treated for 48 hours with medium, TNF, IFNγ, or with a mixture of TNF and IFNγ, then dispersed and stained with antibodies to Fas or isotype control for analysis by flow cytometry. Fas induction occurred only in the presence of IFNγ and TNF, not with either cytokine alone (Figure 6A). To test the role of TNFR2 in Fas induction, we then treated islets from B6 and B6.TNFR2KO with TNF/IFNγ, or with medium alone. As shown in Figure 6B–C, Fas upregulation following IFNγ/TNF treatment was signficantly reduced in B6.TNFR2KO islets compared to B6 islets (p = 0.0003). Fas upregulation was not completely abolished in B6.TNFR2KO islets following TNF/IFNγ-treatment, therefore it is likely that TNFR1 also contributes to Fas upregulation. However, TNFR2 clearly mediates Fas upregulation in β cells in response to cytokines.

We therefore tested whether Fas induction is also reduced in Idd9 congenic islets. Intact islets from 5 week old donors were again used to minimize insulitis damage and were treated with IFNγ, or a mixture of IFNγ and TNF. As shown in Figure 6D–E, Fas expression was significantly greater in NOD compared to Idd9 islets (p = 0.03) following treatment with IFNγ/TNF, further supporting the idea that NOD islets exhibit an enhanced responsiveness to cytokines and that this trait is corrected by expression of protective Idd9 gene variants. To test whether TNFR2 signals are involved in mediating this this enhanced responsiveness in NOD islets, we also pretreated islets with a blocking anti-TNFR2 antibody, or with an isotype control, to test whether this would ablate the increased cytokine responsiveness in NOD islets. While Fas upregulation was significantly inhibited in Idd9 islets by addition of the blocking TNFR2 antibody (36 % decrease, p = 0.02), Fas induction was only slightly decreased in NOD islets by the presence of the blocking antibody (13 % decrease). This suggests that TNFR2-mediated Fas upregulation in NOD islets is refractive to inhibition compared to Idd9 islets.

TNFR2 signaling in other cell types is known to be mediated via adaptor proteins, particularly RIP and TRAF2, that link the cytoplasmic domain of the receptor to downstream intracellular signaling molecules 21. We therefore determined whether these adaptors are expressed in islets from Idd9 congenic and NOD mice, and tested whether cytokine treatment differentially affects their expression. Immunohistochemical staining of pancreatic sections from 12 week old mice with antibodies to TRAF2 revealed a distinctive staining pattern concentrated in a peri-nuclear compartment (Figure 7A). We then showed by immunoblotting that both TRAF2 (Figure 7B) and RIP (Figure 7C) are expressed in isolated islets. Treatment of intact islets with TNF/IFNγ for 2 days increased the amount of RIP present by approximately 2-fold, while TRAF2 expression was not altered. However, significant differences in the expression of adaptor molecules between Idd9 congenic and NOD islet lysates were not observed. Therefore, while differential signaling downstream of TNFR2 may occur in NOD and Idd9 congenic islets, it is not likely to be primarily exerted at the absolute level of RIP or TRAF2 protein.

The NOD variant of TNFR2 contains a mutation (C436Y) 17 adjacent to the TRAF2 consensus binding motif (426-SXEE-429) 22 and flanking sequences are thought to modulate TRAF2 binding 23. We therefore tested whether altered recruitment of TRAF2 by TNFR2 variants could potentially explain the differential responsiveness to TNF/IFNγ that we observed in the islets of Idd9 congenic mice. We examined the kinetics of colocalization of TNFR2 and TRAF2 in dispersed islet cells following cytokine stimulation in vitro. Dispersed islet cells isolated from young donors were treated with TNF and IFNγ for 0, 30 and 60 minutes and stained sequentially with antibodies to TNFR2, TRAF2 and insulin for analysis by confocal microscopy. DAPI was used as a nuclear dye, and an ethidium bromide derivative used to exclude dead cells from the analysis. Additional cells were stained in parallel minus either the TNFR2 or TRAF2 primary antibody and non-specific staining was not observed. These controls were also used to set the staining threshold for analysis. The colocalization coefficient (M1) of TNFR2 with TRAF2 was then determined for insulin positive TNFR2 expressing β cells at each timepoint. The majority of cells were insulin positive and very few dead cells were observed. TNFR2 positive cells were defined as more than 10 pixels above threshold per cell.

At the zero and 30 minute timepoints the colocalization coefficient in both Idd9 congenic and NOD cells was equivalent (approximately 0.6), as shown in Figure 7D–E. At these timepoints, TRAF2 was seen to be primarily concentrated in a peri-nuclear compartment in both strains. TRAF2 remained confined to the peri-nuclear compartment in NOD islet cells at 60 minutes. However, after 60 minutes of cytokine treatment few Idd9 cells exhibited this staining pattern and a weak diffuse cytoplasmic staining was more commonly observed instead. Our quantitative analysis indicated that at the 60 minute timepoint the colocalization coefficient was significantly greater in NOD cells compared to Idd9 cells (p = 0.02). Colocalization appeared strongest in this peri-nuclear compartment, and therefore the redistribution of TRAF2 in Idd9 cells at 60 minutes after cytokine treatment is likely to account for the decreased M1 value. TNFR2 activation is associated with TRAF2 translocation to the ER 24. The data therefore suggests that TNFR2 signaling is aberrantly sustained in NOD islet β cells following TNF exposure.

NOD.B10Idd9R28 mice (line 1104) containing B10 genome across all three Idd9.1/9.2/9.3 intervals (48 cM), referred to here as Idd9 congenic mice, were obtained from Taconic Farms (Idd9R28, line 1104) and maintained in the rodent breeding colony at The Scripps Research Institute (TSRI). NOD and NODScid mice were obtained from the rodent breeding colony at TSRI. 8.3NODScid mice 13 were kindly provided by Dr. Pere Santamaria (University of Calgary), and a colony is maintained at TSRI. NOD.IL4KO mice were obtained from Jackson (stock 004222) and crossed with Idd9 congenic mice. NOD, NODIL4KO, Idd9 and Idd9IL4KO mice were generated by intercross of appropriate progeny and used as internal controls. Fluorescent genotyping using the polymorphic markers D4Mit63, D4Mit72, D4Mit28 and D4Mit180 was used to ensure retention of the entire Idd9 congenic interval during intercross. B6.TNFR2KO mice were obtained from Jackson (stock 002620). Female mice were used for all experiments. All live animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) and the Animal Research Committee (ARC) and were conducted in accordance with institutional guidelines for animal care and use.

Diabetes incidence was determined by monitoring blood glucose levels weekly using Glucometer Elite strips. Mice with two successive blood glucose levels greater than 300 mg/dl were considered diabetic.

Pancreas tissue was taken from 12 week old Idd9 congenic and NOD mice and, after removal of all panLN, immediately cut into small pieces and digested in 1 mg/ml collagenase P (Roche) in complete DMEM for 20 minutes at 37° with agitation. After washing, the digested pancreas was layered over Histopaque 1077 (Sigma) and the islets recovered from the interface after centrifugation. Following further washes, the islets were treated with 0.05 % trypsin/EDTA (Gibco) to create a single cell suspension. Leukocytes released from the trypsinized islets were incubated in DMEM at 37° for between 60 and 90 minutes to allow recycling of cleaved surface molecules. Cells were then counted by trypan blue exclusion and stained for analysis by flow cytometry. An antibody to CD45 was included in the staining in order to compare leukocyte subpopulations as a percentage of infiltrating cells, independently of the extent of infiltration. All antibodies used for flow cytometry were obtained from BD Pharmingen. Staining was performed according to standard procedures. A FACSCaliber dual laser cytometer was used in conjunction with CellQuest software for data acquisition and analysis.

Freshly isolated pancreas tissue from 12–14 week old Idd9 congenic and NOD mice was snap frozen in OCT medium (CD8 staining) or fixed in 10 % NBF and paraffin-embedded (TNFR2 and TRAF2 staining). For CD8 staining, sections (10 μm) were cut from three different levels from each pancreas, each level separated by 300 μm, and air dried at room temperature overnight. Absolute ethanol was used for fixation, and 10 % normal goat serum, or BSA, was used as blocking agent. Anti-insulin (Dako, Carpinteria, 1/400) and anti-CD8 (BD Pharmingen, San Diego, 1/100) primary antibodies were incubated overnight at 4°. Texas Red conjugated anti-guinea pig (1/100) and Fluoroscein conjugated anti-rat (1/200) secondary antibodies (both from Vector) were incubated for 1 hour at room temperature in the dark. Topro-3 was used to stain nuclei and was added to anti-fade component A (Molecular Probes, Eugene OR) when mounting the slides. Slides were stored at -80° until analyzed using Bio-Rad MRC1024 laser scanning confocal microscope and 40 × oil objective lens, using Bio-Rad LaserSharp (v3.2) software to collect images. For colocalization studies, polyclonal goat anti-TNFR2 (R&D systems, 1/10 dilution), donkey anti-goat-fluoroscein (Molecular Probes, Eugene OR, 1/50), rabbit anti-TRAF2 (Leinco, 1/10-1/50 dilution), donkey anti-rabbit-Cy 5 (Jackson, 1/50), anti-insulin and donkey anti-guinea pig-Texas Red (Jackson, 1/50) were added sequentially for co-localization studies. DAPI was included in the mounting medium (Vector). Sections from B6.TNFR2KO mice stained with anti-TNFR2 antibody in parallel did not show specific staining. The TRAF2 antibody used for staining sections gave a single band when used for western blotting, demonstrating its specificity for the TRAF2 protein. Slides were stored at 4° until analyzed using a Bio-Rad (Zeiss) Radiance 2100 Rainbow laser scanning confocal microscope and 60× objective with 2.5× magnification. Zeiss LSM Examiner software used to determine M1 colocalization coefficient. Non-fluorescent TNFR2 and TRAF2 staining was performed on 12–14 week old pancreas sections as above except using DAB (Vector) to develop the signal.

Whole splenocytes from 4–6 week old 8.3NODScid mice were used in all experiments. Where CFSE labeling was required whole splenocytes were incubated in PBS at 5 × 10'7 cells/ml with 5 mM CFSE (Molecular Probes, Eugene OR) for 10 min at 37°. The cells were then washed twice in cold PBS and 20 million labeled cells were injected iv into recipient mice. The recipients of CFSE-labeled 8.3NODScid splenocytes were aged between 6 and 9 weeks of age, and were age-matched in each experiment. For the transplant experiments, Lympholyte M (Cedarlane Laboratories, Ontario, Canada) was used in some cases, according to the manufacturer's instructions, to remove dead cells and red blood cells, and 30 million unlabeled 8.3NODScid splenocytes were injected iv per mouse. A sample of the injected cells was labeled with antibodies to CD8 and Vβ 8.1/8.2 to confirm that all CD8 T cells injected were positive for the 8.3 TCR (CD8+ gated cells were 97–99 % Vβ 8.1/8.2+). To determine the activation phenotype the cells were also labeled with antibodies to CD44. The CD8+Vβ 8.1/8.2+ cells were typically approximately 40 % CD44hi. For CFSE transfer experiments panLN and ingLN from recipient mice were taken 4 days post-injection and single cell suspensions prepared in sterile filtered Hanks medium containing 5 % FBS. The cells were then stained with antibodies to CD8, CD44 and CD11a for analysis by flow cytometry.

Donor splenocytes from 13 week old Idd9 congenic and NOD mice were lysed to remove red cells and injected iv into NODScid recipient mice. Either 20 million Idd9 congenic or NOD splenocytes were transferred into recipients, or 10 million of each population was co-transferred. Blood glucose levels were monitored for 13 weeks following transfer when all recipient mice were diabetic.

Idd9 congenic and NOD mice 6 weeks of age were lethally irradiated (950–1,000 Rad) and injected with bone marrow cells prepared from 5 week old NOD donors. Bone marrow was isolated from the leg bones of donor mice and resuspended at 50 million cells/ml. Following irradiation, recipients were injected iv with 10 million bone marrow cells in PBS, and maintained on antibiotics in the drinking water.

Islets were hand picked from 5–7 week old Idd9 congenic and NOD donors. Typically 150–250 islets were isolated from each pancreas, and 9–12 donors of each strain were used for each experiment. Islets were isolated following pancreas inflation with collagenase through the common bile duct, as previously described 50. Islets were then cultured in hydrophobic 35 mm Petri dishes for 7 days in 10 % CO2 at 37°, and the media changed every 3 days. RPMI 1640 containing 10 % FCS and supplemented with glutamine and antibiotics was used throughout the procedure. Approximately 250 islets were transplanted under the left kidney capsule of 11 week old NODScid recipients, as previously described 50. The grafts were allowed to revascularize for seven days before transfer of splenocytes from 8.3NODScid mice. 5–7 days after splenocyte transfer the grafts were removed and fixed in 10 % NBF for paraffin embedding. Sections (4 μm) were cut and H&E stained for analysis of graft integrity and infiltration.

To score the extent of graft destruction adjacent sections were stained with anti-insulin and anti-glucagon (Dako, Carpinteria, 1/400) and scored as either healthy (normal distribution of predominantly insulin-positive cells with scattered glucagon-positive cells around the islet periphery) or collapsed/destroyed (few or no insulin-positive cells, and the predominance of glucagon-positive cells giving the islet a 'collapsed' appearance in glucagon-stained sections).

Islets were isolated from 5 week old mice and allowed to recover in culture medium for 4–6 days before being stimulated with 1,000 U/ml IFNγ and 1,000 U/ml TNF (BD Pharmingen, San Diego, CA), with each cytokine individually, or with medium alone, for 2 days. Cells were lysed with RIPA buffer (containing 20 mmol/liter Tris, pH 7.5, 1 mmol/liter EDTA, 140 mmol/liter NaCl, 1% NP-40, 1 mmol/liter orthovanadate, 1 mmol/liter PMSF, and 10 μg/ml aprotinin) and 20–150 ug protein loaded per lane on a 12 % gel for western blot. Mouse anti-RIP (BD Transduction Labs, San Diego, CA) and rabbit anti-TRAF2 (Leinco, St. Louis, Missouri) antibodies were used for immunodetection (Cell Signaling Technology). All membranes were stripped and reblotted with a mouse mAb to actin to confirm equal protein loading (ICN Biomedicals). Densitometry was performed using ImageJ software (Nih).

To induce islet cell death, isolated islets from 5 week old donors were cultured with 1,000 U/ml INFγ and 1,000–5,000 U/ml TNF for 6 days. 50–100 islets were used per assay. Islets were then incubated with the Molecular Probes (Eugene, OR) Live/dead stain in RPMI-1640 for 45 minutes at 37'. Intact, unfixed islets were then transferred to microscope slides, excess buffer removed using a stretched Pasteur pipette, and a cover slip sealed over the islets. Confocal images using 40× oil immersion lens of Z-stack sections 50–100 μm through each islet were taken. The number of live and dead cells in each field was determined using ImageJ software.

Hand-picked islets from 5 week old donor Idd9 congenic and NOD mice, or 4–6 month old B6 and B6.TNFR2KO mice, were cultured in complete RPMI-1640 media for 4–6 days to allow recovery and depletion of tissue-resident leukocytes. Islets were then treated with medium, IFNγ, TNF, or IFNγ +TNF for 48 hours. Cytokines were used at 1,000 U/ml unless otherwise stated. For blocking experiments islets were cultured with blocking anti-TNFR2 antibody (Pharmingen), or hamster IgG isotype control, for 1–2 hours prior to addition of cytokines. Islets were dispersed using 0.25% trypsin/EDTA for 4 min, washed and allowed to recover at 37' for 1 hour. Staining was carried out for 30 min at 4'C using PE-labeld anti-TNFR2 (clone HM102, Caltag), or with biotinylated anti-Fas (Jo2, Pharmingen) and SA-APC, or the relevant isotype control. 7AAD was added for 10 min at RT to exclude dead cells, and the samples analyzed immediately.

In all cases where a p value is shown, a two-tailed, unpaired Students' t-test was used to gauge significance, except for comparisons of diabetes incidences in which case the Kaplan-Meier survival test was used.