Disruption of the γ-Interferon Signaling Pathway at the Level of Signal Transducer and Activator of Transcription-1 Prevents Immune Destruction of β-cells

The STAT-1^−/− mutation was produced by deleting 5.7 kb of the STAT-1 gene, resulting in a total inability to produce STAT-1 in homozygous mutant animals (26). These mice have been backcrossed for 10 generations in the C57BL/6 background. STAT-1^−/− mice were kept under specific pathogen free conditions. C57BL/6 mice (Harlan, Zeist, the Netherlands) were used as wild-type controls. All experimental designs were conducted with the approval of the Animal Ethics Committees of the Katholieke Universiteit Leuven (in vivo studies) and Université Libre de Bruxelles (in vitro studies).

Islets were obtained from 10-week-old male and female wild-type control and STAT-1^−/− mice as previously described (9,27). To obtain purified β-cells, islets were dissociated and sorted by FACStar flow cytometer (Becton Dickinson, Sunnyvale, CA), as previously described for rat β-cells (28). The preparations used in the present experiments contained 84 ± 3% and 84 ± 4% β-cells for wild-type and STAT-1^−/− mice, respectively. Cell culture was performed in Ham’s F-10 medium supplemented with 10 mmol/l glucose, 50 μmol/l isobutylmethylxanthine, 0.5% BSA (Boehringer Mannheim, Mannheim, Germany), and 2 mmol/l l-glutamine (Life Technologies, Paisley, Scotland). These culture conditions, originally developed for the maintenance of rat β-cells, have been validated for mouse islet or β-cell culture (10).

Whole islets were cultured in suspension, whereas fluorescence-activated cell sorter (FACS)–purified single β-cells were cultured attached to polylysine-coated dishes. The effect of cytokines was examined after 2 and 5 days of culture in the presence of recombinant murine IFN-γ (1,000 units/ml, 107 units/mg; PeproTech) and recombinant human IL-1β (50 units/ml, 38 units/ng; kind gift of Dr. C.W. Reynolds from the National Cancer Institute, Bethesda, MD). These concentrations of cytokines were selected based on our previous studies with mouse pancreatic islets and β-cells (9,10,29).

The viability of islet cell preparations was assessed after 2 and 5 days exposure to cytokines. Whole islets or isolated β-cells were incubated for 15 min with propidium iodide (10 mg/ml) and Hoechst 342 (20 mg/ml). The approximate percentage of dead islet cells was estimated by two individual observers, with one of them unaware of the sample identity (9,10). Evaluation of cell death in whole islets is complicated by the superposition between cells, and these observations should thus be considered as semiquantitative. The percentages of apoptotic cells in the single β-cell preparations were assessed by propidium iodide and Hoechst after 2 or 5 days exposure to cytokines, as previously described for mouse β-cells (9,10). This fluorescence assay for single β-cells is quantitative and has been validated by systematic comparisons with electron microscopy observations (30). The apoptosis and necrosis indexes were calculated as [(percent apoptotic or necrotic cells in experimental condition − percent apoptotic or necrotic cells in control)/(100 − percent dead cells in control)] × 100. The percent viability in control single β-cells (not exposed to cytokines) was in the range of 55–70%. These values were similar after 2 or 5 days of culture and were not significantly different between wild-type and STAT-1^−/− β-cells (data not shown).

Culture media were collected after 24 and 48 h, stored at −20°C and assayed for immune-reactive insulin. Insulin release in the culture media was determined by enzyme-linked immunosorbent assay (ELISA).

Intraperitoneal glucose tolerance tests (IPGTTs) were carried out in 8-week-old male wild-type and STAT-1^−/− mice after an overnight fast. Briefly, mice were injected intraperitoneally with 2 g/kg body wt d-glucose dissolved in sterile PBS. Blood samples were drawn at different time points (0, 15, 30, 60, 90, and 120 min) from the tail vein. Plasma glucose concentrations were assayed by a glucose oxidase method using an automatic glucometer (Glucocard, Menarini).

For MLDS induction, 8-week-old male STAT-1^−/− and wild-type mice were injected intraperitoneally with freshly dissolved streptozotocin (50 mg/kg) (Sigma, St. Louis, MO) for 5 consecutive days. Development of glucosuria was monitored with Clinistix (Bayer Diagnostics) and glucose concentrations of venous blood were determined using a glucometer (Glucocard, Menarini) as described above. Clinical diabetes was defined by positive glucosuria and hyperglycemia (blood glucose levels >250 mg/dl) on 2 consecutive days in nonfasted animals. Mice were followed until 40 days after the last injection of streptozotocin. At that time point, animals were killed and pancreata were removed for histology and insulin content measurement. In a separate set of wild-type and STAT-1^−/− animals, the pancreas was removed 3 days (experimental day 8) after the start of streptozotocin for histology.

Pancreata were removed from killed mice and dissected free of adipose tissue and lymph nodes in PBS at 4°C under a stereomicroscope. Half of each pancreas was snap-frozen in liquid nitrogen-chilled isopentane and stored at −20°C until processing. For histology, serial cryostat sections (5-μm thickness) were cut and stained with hematoxylin and eosin to assess pancreatic islet integrity and immune cell infiltration. The remaining half of the pancreas was used for insulin content determination. Briefly, protein was extracted overnight at 4°C in acid–ethanol. Insulin concentrations in the extracts were measured using an ultrasensitive rat insulin ELISA kit (Mercodia, Uppsala, Sweden) with rat insulin as standard as previously described (31).

Cultured islet cells (24 h of cytokine exposure) were snap-frozen in liquid nitrogen and stored at −80°C until RNA extraction (TRIzol reagent, Invitrogen). cDNA was created using Superscript II RT (Invitrogen) and quantitative PCR analysis was performed with the ABI prism 7700 Sequence detector (Applied Biosystems, Foster City, CA) as previously described (32). PCR conditions were 2 min at 50°C and 10 min at 94°C, followed by a total of 40–45 two-temperature cycles (15 s at 94°C and 1 min at 60°C). Primer and probe sequences for the determination of mouse cDNAs for housekeeping gene β-actin (33), IL-1β (33), IL-15 (33), IFN-γ (33), iNOS (33), MCP-1 (34), IP-10 (34), and MIP-3α (34) were as described in the indicated references. The target cDNAs present in each sample were corrected for the respective β-actin values. The primer and probe sequences used for determination of mouse cDNA for STAT-1 were: FW, 5′-CACCAGAACCGATGGAGCTT-3′; RV, 5′-TCCGGGACATCTCATCAAACT-3′; and TP, 5′-CACCCTTCTAGACTTCAGACCACAGACAACCT-3′.

The statistical software NCSS 2000 (Kaysville, UT) was used for statistical analysis. Data are expressed as means ± SE. Areas under the curve were determined for analysis of IPGTT, whereas the χ² test was performed for the incidence of MLDS-induced diabetes. ANOVA and Student’s t test were used for multiple comparisons whenever appropriate. Significance was defined as P ≤ 0.05. Unless indicated otherwise, the graphs show triplicate determinations from at least three independent experiments.

In an initial series of experiments, we confirmed the STAT-1–deficient state of the pancreatic islets from knockout mice. Thus, there was no STAT-1 mRNA expression under basal conditions of cell culture or after exposure to IL-1β plus IFN-γ in islets from STAT-1^−/− mice (Fig. 1). In contrast, STAT-1 expression was clearly upregulated in islets from wild-type mice after stimulation with IL-1β + IFN-γ (CYTK) (Fig. 1).

Next, we confirmed that treatment with IL-1β in combination with IFN-γ significantly decreased cell viability of wild-type islets after 48 h of cytokine exposure (Fig. 2A). After 5 days of cytokine exposure, only 22 ± 13% wild-type viable cells were left compared with 93 ± 1% viable cells in control conditions (Fig. 2B). Islets from STAT-1^−/− mice showed resistance to cytokine-induced islet destruction, as reflected by fully preserved islet viability after 48 h (Fig. 2A) and 5 days of cell culture (Fig. 2B). In parallel, insulin release into the medium by wild-type islets decreased significantly upon exposure to the cytokine combination IL-1β plus IFN-γ, whereas insulin secretion was fully preserved in STAT-1^−/− islets exposed to cytokines (Fig. 2C and D).

To discriminate between apoptosis and necrosis and to investigate the importance of the absence of STAT-1 in the β-cells themselves in the observed protection against cytokine-induced islet cell death, FACS-purified β-cells from wild-type and STAT-1^−/− mice were studied (Table 1). In accord with the results obtained in whole islets, exposure of wild-type β-cells to cytokines induced a clear increase in the apoptotic index at both 2 and 5 days (Table 1), but there was no increase in the necrosis index (data not shown), confirming that apoptosis is the main form of cytokine-induced β-cell death in this model (8). On the other hand, STAT-1^−/− purified β-cells were totally protected against cytokine-induced apoptosis, with apoptotic indexes around zero after 2 and 5 days (Table 1).

Under basal conditions, wild-type and STAT-1^−/− islets expressed IL-1β, MCP-1, MIP-3α, IP-10, and iNOS mRNAs at comparable levels (Fig. 3), but there was no expression for IL-15 in STAT-1^−/− islets in basal culture conditions.

Exposure of wild-type islets to inflammatory cytokines induced a marked upregulation of several genes encoding for chemokines and iNOS. A similar upregulation was observed in STAT-1^−/− islets for MCP-1 and MIP-3α, but induction of IL-15, IP-10, and iNOS was completely prevented by the STAT-1 disruption (Fig. 3).

We next examined the effects of STAT-1 disruption in vivo in a model of immune-mediated β-cell destruction, namely MLDS. To assure that the basal conditions were similar between wild-type and STAT-1^−/− mice, we first demonstrated that deletion of STAT-1 does not affect normal glucose tolerance. IPGTTs, performed in wild-type and STAT-1^−/− mice 8 weeks of age, showed similar responses as reflected by the area under the glucose curve (Fig. 4). In addition, no differences were noted between control and knockout pancreatic insulin content (164 ± 15 vs. 148 ± 46 pmol/mg pancreas in wild-type and STAT-1^−/− mice, respectively).

Second, we excluded differences in the sensitivity to a single high dose of streptozotocin (150 mg/kg). Blood glucose levels and diabetes incidence were similar in wild-type and STAT-1^−/− mice 48 h after high-dose streptozotocin (394 ± 52 mg/dl [100%, 6/6] vs. 386 ± 24 mg/dl [100%, 6/6], respectively; NS). Further follow-up of the animals up to 7 days confirmed stable high blood glucose level in wild-type and^−/− mice (data not shown).

A clear difference in sensitivity to MLDS was, however, observed between wild-type and STAT-1^−/− mice. At the start of the five daily injections of streptozotocin, plasma glucose levels were 98 ± 17 mg/dl in wild-type mice and 95 ± 6 mg/dl in STAT-1^−/− mice (NS). On experimental day 10, glucose levels were significantly higher in wild types than in STAT-1^−/− mice and increased rapidly on day 17, continuing to rise until reaching a plateau level after 30 days. On the other hand, glucose levels remained virtually constant in streptozotocin-treated STAT-1^−/− mice until day 45 (Fig. 5A).

Although 73% of wild-type mice developed diabetes (defined as positive glucosuria and blood glucose levels >250 mg/dl on 2 consecutive days) by 40 days after the last streptozotocin injection, only 39% of STAT-1^−/− mice reached the diabetic level at the end of the follow-up period (P ≤ 0.05 vs. controls) (Fig. 5B), suggesting that absence of STAT-1 provides a partial protection against development of diabetes. That this protection was only partial is confirmed by the pancreatic insulin content determination at the end of the experiment, where a clearly higher insulin level is still present in the STAT-1^−/− pancreata versus controls (21.8 ± 3.9 vs. 5.8 ± 1.7 pmol/mg pancreas, P ≤ 0.01), although a 85% decrease was observed compared with the levels present before streptozotocin administration (see above). This was also reflected by the histological image at the end of the study where mostly small islets were present in both groups (data not shown).

Pancreatic β-cells are particularly susceptible to the cytotoxic effects of inflammatory cytokines, secreted by infiltrating immune cells during the course of insulitis (8). Cytokine-mediated β-cell death results from changes in the global gene expression profile of β-cells (1,2). Therefore, autoimmune diabetes can be seen as a failure of intracellular signaling pathways in the β-cells, pushing these cells toward death instead of survival and making these molecular pathways increasingly relevant as therapeutic targets. The cytokine IFN-γ, in synergy with other stimuli, triggers in the β-cells the expression and secretion of a large number of proinflammatory cytokines, chemokines, costimulatory molecules, and enzymes, which are known to control cellular immune reactions that contribute to β-cell destruction (1,11). For many of these genes, mRNA expression is regulated at the transcriptional level by binding of IFN-γ-activated STAT-1 (itself upregulated in β-cells after costimulation with inflammatory cytokines) to respective binding sites (γ-activated sequences) within the target promoters (1,11).

In the present study, we used mice with a knockout of the transcription factor STAT-1 to determine whether JAK-STAT activation is essential for cytokine-stimulated gene expression and β-cell death in vitro and to investigate whether these findings are translated into protection against in vivo β-cell destruction. We found that absence of STAT-1 in whole islets protects them against IFN-γ (studied in combination with IL-1β)–induced cell death. Moreover, the observation of complete protection of FACS-purified β-cells from STAT-1^−/− mice against cytokine-induced apoptosis confirms that the protection is due to the absence of STAT-1 in the β-cells themselves and not to its absence in putative passenger leukocytes. In line with our observations, other studies have recently shown that β-cell-specific overexpression of suppressor of cytokine signaling (SOCS) proteins 1 and 3, which block IFN-γ activation of the JAK-STAT signaling cascade, prevents diabetes in different mouse models (35–37). Overexpression of SOCS-1, however, impairs both STAT-1 tyrosine phosphorylation and toll-like receptor signaling (38), whereas upregulation of SOCS-3 blocks NF-κB activation (39) and modifies signaling via the insulin receptor (40) in β-cells. Significantly, our data demonstrate that interfering exclusively with IFN-γ-activated JAK-STAT signaling protects insulin-secreting β-cells from cytokine-elicited inflammatory responses and death in vitro. We have previously shown that IL-1β contributes to β-cell death via activation of NF-κB (3,5). Taken together with the present data, we suggest that STAT-1 and NF-κB are key transcription factors regulating the gene networks that trigger β-cell apoptosis.

We have previously demonstrated that expression of IL-15, IP-10, and iNOS is regulated by IFN-γ, whereas the expression of MCP-1 and MIP-3α is IL-1β–dependent (7). In the case of IFN-γ, many responsive genes are induced through interaction of phosphorylated STAT-1 homodimers and an inducible enhancer termed γ-activated sequences. Other potential downstream effectors responsible for gene induction by IFN-γ are IRF-1 (41), the extracellular signal-regulated kinases 1 and 2 (42), NF-κB (p65/p50) (43), and CCAAT/enhancer-binding protein-β. An important observation in the present study is that STAT-1–deficient islets, in contrast to wild-type islets, are unable to upregulate expression of mRNAs for IL-15, IP-10, and the enzyme iNOS after cytokine exposure, suggesting that JAK/STAT activation is crucial for such responses. On the other hand, no inhibitory effects of STAT-1 deficiency were observed on cytokine-elicited MCP-1 and MIP-3α expression in β-cells.

STAT-1 actions are cell- and context-dependent and may lead to opposite outcomes, such as increasing survival or apoptosis (44). Studies using cells and mice deficient in STAT-1 have shown that the transcription factor mediates most immune/inflammatory effects of IFN-γ, including the induction of immune effector and inflammatory genes (26,45,46) and both pro- and antiapoptotic genes (14). Of note, recent studies have identified several IFN-γ–induced but STAT-1–independent genes in diverse cell types (47–49), leading to a currently ongoing reevaluation of the role of STAT-1 for IFN-γ signaling. We found that a targeted deletion of STAT-1 protein in mice did not affect the β-cell mass homeostasis, as demonstrated by normal IPGTT and comparable pancreatic insulin content to wild-type controls, but causes a partial protection against β-cell destruction by multiple low doses of streptozotocin. On the other hand, no direct protection was observed against the toxic effects of a high dose of streptozotocin on β-cells, suggesting that absence of STAT-1 prevents β-cell death in MLDS-induced diabetes by interfering with immune-mediated β-cell death, not by interfering with the alkylating actions of streptozotocin. Because these mice were deficient for STAT-1 at both the β-cell and immune-system levels, the present data do not allow us to conclude whether the observed protection was due to beneficial effects of STAT-1 absence on the immune system or β-cells or, more probably, in both. Of note, the degree of immune-cell infiltration in the islets after MLDS treatment was similar between wild-type and STAT-1^−/− mice, suggesting that under similar immune attack β-cells from STAT-1^−/− mice are more resistant to cell death.

These data, together with our previous observations, suggest a dual role for IFN-γ signaling in β-cell destruction. Thus, we have previously shown that disruption of the transcription factor IRF-1 not only failed to prevent cytokine-induced β-cell death in vitro (9), but it also aggravated MLDS-induced diabetes in vivo (50). This is in marked contrast with the present findings and suggests that although STAT-1–regulated genes have mainly a proapoptotic role in β-cells, genes regulated by the downstream transcription factor IRF-1 have a protective role against β-cell destruction. New experiments, including microarray analysis, are now required to identify the nature of these STAT-1–and/or IRF-1–regulated genes.

In conclusion, we have demonstrated a central role of STAT-1 in inflammation-induced β-cell death. Disruption of the STAT-1–mediated signaling cascade prevents upregulation of several chemokines and iNOS and completely prevents IFN-γ (studied in combination with IL-1β)-mediated β-cell death in vitro. Moreover, blocking STAT-1 partially protects β-cells against destruction in vivo. These data suggest that STAT-1 may become a therapeutic target in our efforts to increase β-cell resistance against immune destruction.

This work was supported by the JDRF Center for Prevention of Beta Cell Destruction in Europe Grant 4-2002-457; the Flemish Research Foundation Fonds Voor Wetenschappelijk Onderzoek (FWO) Grants G.0084.02 and G.0233.04, a doctoral scholarship for H.C., a clinical fellowship for C.M, and a postdoctoral fellowship for C.G. and P.M.; the Belgian Research Foundation Fonds National de la Recherche Scientifique; the Belgian Program on Interuniversity Poles of Attraction initiated by the Belgian State Grant IUAP P5/17; the Katholieke Universiteit Leuven Grant GOA 2004/10; and an Action de Recherche Concertée de la Communauté française de Belgique.

The authors thank the personnel of the laboratory of Experimental Medicine and Endocrinology at KUL (in particular J. Depovere, D. Valckx, and J. Laureys) and of the Laboratory of Experimental Medicine at ULB (in particular M. Neef, J. Schoonheydt, and M. Urbain) for their excellent technical support and constructive discussions.