γ-Interferon Signaling in Pancreatic β-Cells Is Persistent but Can Be Terminated by Overexpression of Suppressor of Cytokine Signaling-1

NODscid mice were bred in the animal facilities of the Walter and Eliza Hall Institute of Medical Research. IFN-γ^−/− and SOCS-1^−/− IFN-γ^−/− mice, on mixed C57BL/6-129Sv backgrounds, were obtained from Dr. W. Alexander of the Walter and Eliza Hall Institute of Medical Research.

NIT-1 insulinoma cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Life Technologies, Gaithersburg, MD) supplemented with nonessential amino acids (NEAA), 100 units/ml penicillin, 100 μg/ml streptomycin, and 10% fetal calf serum (FCS). NIT-1 cells were transfected with the SOCS-1 expression plasmid REP9SOCS-1 (see below) by electroporation as previously described (11), and stable transfectants were selected with 500 μg/ml Geneticin (Life Technologies).

Unless otherwise stated, recombinant murine IFN-γ (Genentech, South San Francisco, CA) was used at 100 units/ml; recombinant human IL-1β (Genzyme, Cambridge MA) was used at 10 units/ml; recombinant murine TNF-α (Genentech) was used at 250 units/ml; and recombinant murine IFNα (provided by Dr. S. Gerondakis, the Walter and Eliza Hall Institute of Medical Research) was used at 100 units/ml.

pGLB928LUC (provided by Dr. D. Paulnock, University of Wisconsin, Madison, WI) contains 928 bp of the murine guanylate-binding protein-1 promoter (containing GAS elements) (32) ligated to the luciferase gene. The SOCS-1 expression plasmid, REP9SOCS-1, was constructed by subcloning the SOCS-1 coding region from the genomic murine SOCS-1 clone 95-11-10 (obtained from Dr. T. Willson, the Walter and Eliza Hall Institute of Medical Research) into the mammalian expression vector REP9 (RSV promoter).

NIT-1 cells were transiently transfected with the STAT1-responsive reporter pGLB928LUC (±REP9SOCS-1) by electroporation. The total number of cells required for each experiment was electroporated in the same tube. They were then plated out into six-well plates at 8 × 10⁵ cells per well. After 1 day, transfected cells were stimulated with IFN-γ, and then a luciferase assay was performed. Luciferase assays were performed by harvesting the cells in 300 μl lysis buffer (1% Triton X-100 in 25 mmol/l glycylglycine, 15 mmol/l MgSO₄, 4 mmol/l EGTA, and 1 mmol/l dithiothreitol [DTT], pH 7.8). The samples were centrifuged, and 100 μl supernatant was added to 365 μl luciferase assay buffer (25 mmol/l glycylglycine, 15 mmol/l MgSO₄, 4 mmol/l EGTA, 16 mmol/l potassium phosphate, 1 mmol/l DTT, and 2 mmol/l ATP, pH 7.8). The reactions were performed at room temperature using a Lumat LB950 luminometer (Berthold, Wildbald, Germany), which injects 100 μl of 25 mmol/l luciferin in glycylglycine buffer (25 mmol/l glycylglycine, 15 mmol/l MgSO4, 4 mmol/l EGTA, and 10 mmol/l DTT, pH 7.8) into the extract and assay buffer. Luciferase activity was measured by calculating the light emitted during the initial 10 s of the reaction, and the values are expressed in arbitrary light units. In some experiments, SV40 early promoter–β-gal was cotransfected to determine whether transfection efficiency could confound the results. However, correction for β-gal expression made little difference to the results (data not shown).

Nuclear extracts were prepared according to the method of Schreiber et al. (33), and gel shifts were performed as described elsewhere (13). STAT1 DNA binding activity was measured using the GAS from the FcγRI promoter: 5′-GTACGAGATGTATTTCCCAGAAAA-3′.

Islets were isolated from mice by intraductal digestion of the pancreas with collagenase P (Roche Diagnostics, Basel) as previously described (13) and separated on a BSA density gradient. Islets were handpicked to purity and cultured in CMRL-1066 (Life Technologies) supplemented with 100 units/ml penicillin, 100 μg/ml streptomycin, and 10% FCS.

Total RNA was purified from primary islets and NIT-1 cells using RNAzol B (Tel-Test, Friendswood, TX). mRNA analysis was performed by reverse transcriptase (RT)–polymerase chain reaction (PCR). First-strand cDNA synthesis from 1 μg total RNA was performed in a 20 μl reaction using MLV reverse transcriptase (Promega, Madison, WI). cDNAs of interest were amplified from 1 μl RT product per PCR using REDTaq DNA polymerase (Sigma, St. Louis, MO). The primers 5′-GTGGGCCGCCCTAGGCACCA-3′ and 5′-CTCTTTGATGTCACGCACGATTTC-3′ were used to amplify a 530-bp fragment of the murine β-actin sequence; 5′-ATGGTCCTCTGCGTACAG-3′ and 5′-TGACAAGCAGTTAGAGTC-3′ amplified a 440-bp fragment of the murine CIS sequence; 5′-ATGGTAGCACGCAACCAGGTG-3′ and 5′-CTCCAGCAGCTCGAAAAGGCA-3′ amplified a 460-bp fragment of the murine SOCS-1 sequence; and 5′-CCATGACCCTGCGGTGCCTGG-3′ and 5′-GGAATTTATATTCTTCCAAGT-3′ amplified a 593-bp fragment of the murine SOCS-2 sequence. PCRs were optimized such that the number of cycles that were used allowed for the analysis of products within the linear part of the amplification reaction. Amplification of β-actin (22 cycles), CIS (28 cycles), SOCS-1 (28 cycles), and SOCS-2 (30 cycles) was performed for 20 s at 96°C, 30 s at 55°C, and 40 s at 72°C. PCR products were resolved on an agarose gel, transferred onto a nylon membrane, and Southern hybridized with [α³²P]dATP-labeled cDNA probes (Megaprime DNA Labeling; Amersham Pharmacia Biotech, Uppsala, Sweden) for CIS, SOCS-1, SOCS-2, or β-actin (SOCS cDNA probes were provided by Dr. W. Alexander). Radiation counts for SOCS RT-PCR products were quantitated by Phosphorimager (Molecular Dynamics, Sunnyvale, CA) analysis using ImageQuant version 5.0 and were corrected against β-actin.

NIT-1 cells (2 × 10⁷) were lysed on ice for 30 min in 1 ml KALB lysis buffer (34) supplemented with 1 mmol/l phenylmethylsulfonyl fluoride, 1 μg/ml pepstatin A, and 2 μg/ml aprotinin, and then they were centrifuged for 10 min at 4°C. The clarified supernatant was collected and incubated with 3 μg anti–SOCS-1 antibody (2E1; anti–SOCS-1 antibodies were provided by Dr. J.-G. Zhang, the Walter and Eliza Hall Institute of Medical Research) on ice for 1 h. Forty μl of 50% protein G-Sepharose gel slurry was added, with a further incubation on a rotating wheel for 2 h at 4°C. The protein G beads were washed three times with KALB lysis buffer, and then the immunoprecipitate was eluted with 2× protein sample buffer (8% SDS, 40% 2-mecaptoethanol, and 0.5 mol/l Tris-HCl, pH 6.6, bromophenol blue). The immunoprecipitates were resolved on a 14% polyacrylamide gel, transferred onto a nitrocellulose membrane, and probed with a biotinylated anti–SOCS-1 antibody (4H1), followed by a secondary horse radish peroxidase–conjugated streptavidin (Amersham Pharmacia Biotech). SOCS-1 protein was visualized with Lumi-Light (Roche Diagnostics).

NIT-1 insulinoma cells were liberated from tissue culture plates with 0.2% trypsin and 10 mmol/l EDTA in phosphate-buffered saline (37°C for 4 min), washed twice with DMEM NEAA + 10% FCS, and then allowed to recover for 30 min at 37°C. A biotinylated anti–H-2D^b monoclonal antibody (28-14-8; Pharmingen, San Diego, CA) was used to stain for class I MHC, followed by phycoerythrin-conjugated streptavidin (Caltag, Buringame, CA). Analysis was performed on a FACScan (Becton Dickinson, Franklin Lakes, NJ). The method for dispersing islets into single cells for flow cytometric analysis is described elsewhere (4). Islet cells were stained for class I MHC expression as described for NIT-1 cells. The β-cell population was identified by their high flavin adenine dinucleotide autofluorescence (4).

NIT-1 cells were seeded into 12-well plates at a density of 10⁵ cells per well. After cytokine treatment, apoptosis was quantitated according to the method described by Nicoletti et al. (35), which measures the fragmentation of nuclei. Briefly, the cells were harvested and then incubated overnight at 4°C in 50 μg/ml propidium iodide in 0.1% sodium citrate with 0.1% Triton X-100, which stains nuclear DNA. The cells were then analyzed on a FACScan. Apoptotic bodies were identified by their reduced DNA content.

Statistical comparisons were made using two-way analyses of variance. Calculations were performed with the aid of Prism version 2.0a (GraphPad Software, San Diego, CA).

The kinetics of IFN-γ signaling in NIT-1 insulinoma cells were examined in a luciferase reporter assay for STAT1 activity (Fig. 1A). NIT-1 cells were transiently transfected with the IFN-γ–responsive luciferase reporter pGLB928LUC (containing GAS elements) and then treated with IFN-γ over a 24-h time period. IFN-γ treatment of NIT-1 cells resulted in maximal luciferase expression (∼20-fold over untreated) after 8 h, which remained unchanged after 24 h, suggesting the persistence of activated STAT1. This was confirmed when STAT1 activity was directly analyzed by gel shift (Fig. 1B). Correlating with luciferase expression, STAT1 homodimers were detected within minutes but then persisted as long as the cytokine was present, even after 24 h. Expression of IFN-γ–responsive genes, such as class I MHC, also remained maximal in NIT-1 cells for as long as the cytokine was present (not shown). The same kinetics of signaling were also evident in primary pancreatic islets after IFN-γ stimulation. The induction of both STAT1 activation, measured by gel shift, and class I MHC expression, measured by flow cytometry, persisted throughout the time that IFN-γ was present. These findings demonstrate that IFN-γ signaling is prolonged in pancreatic β-cells and is not terminated in the manner seen in Bud-8 human fibroblasts (17).

The discovery of the SOCS family of proteins has provided a better understanding of how cytokine signaling is negatively regulated. It has been shown in cell lines such as M1 monocytic and NIH3T3 fibroblast cells that SOCS-1 expression is induced by IFN-γ, whereas overexpression of SOCS-1 blocks IFN-γ signaling (21,22). Induction of SOCS-1 is thought to function in an intracellular negative-feedback loop to switch off IFN-γ-receptor signaling. The transient nature of IFN-γ signaling in some cell types may be a result of negative regulation by SOCS-1. The persistence of IFN-γ signaling in pancreatic β-cells could therefore be due to lack of SOCS-1 expression. Thus, we analyzed β-cells for the expression of SOCS genes.

The expression of SOCS genes in primary murine islets was analyzed by RT-PCR after treatment with IFN-γ, IL-1β, or TNF-α (Fig. 2A). Two members of the family, CIS and SOCS-2, were transiently expressed in primary islets after treatment with all three cytokines. SOCS-1 was also expressed after treatment with IFN-γ but not IL-1β or TNF-α. Islets from NOD mice were also analyzed for SOCS transcripts because the presence of proinflammatory cytokines during the progression of diabetes may also induce SOCS gene expression. Both CIS and SOCS-2 message were present in the islets of mice at an early stage of disease (50 days), whereas SOCS-1 was not detected until 70 days (data not shown).

Although β-cells constitute the majority of cells within an islet, other cell types are also present. To confirm that β-cells express SOCS genes, the NIT-1 insulinoma cell line was used as a pure source of β-cells. SOCS gene expression was analyzed in NIT-1 cells after treatment with IFN-γ, IFNα, IL-1β, or TNF-α (Figs. 2B and C). As in primary islets, CIS and SOCS-2 were transiently induced by IFN-γ, IL-1β, and TNF-α but not IFNα. SOCS-1 was transiently expressed after IFNα treatment, peaking at 2 h and then decaying. SOCS-1 was also induced by IFN-γ and reached maximal expression by 4 h but was not downregulated within the time course. This prolonged SOCS-1 expression after IFN-γ treatment was also observed in primary islets. Although IL-1β or TNF-α alone did not induce SOCS-1 expression in NIT-1 cells, both were able to do so in combination with a low concentration of IFN-γ, suggesting synergistic interactions between these cytokines (Fig. 2D).

IFN-γ signaling was found to be persistent in β-cells, despite the induction of SOCS-1 mRNA expression. The presence of message, however, does not necessarily indicate translation into protein. Therefore, we analyzed NIT-1 cells for SOCS-1 protein expression by immunoprecipitation/Western blotting (Fig. 2E). SOCS-1 protein was first detected 1 h after IFN-γ treatment, which increased after 4 and 24 h, thus correlating with SOCS-1 mRNA expression.

SOCS-1 is expressed in β-cells in response to IFN-γ, but IFN-γ signaling remains persistent. To examine whether SOCS-1 is functional in β-cells, the ability of overexpressed SOCS-1 to suppress IFN-γ signaling was studied in NIT-1 insulinoma cells. The effect of transfected SOCS-1 was analyzed in a luciferase reporter assay for STAT1 activation. NIT-1 cells were transiently transfected with the IFN-γ–responsive luciferase reporter pGLB928LUC, with or without the cotransfection of a SOCS-1 expression vector. Cotransfection of SOCS-1 into NIT-1 cells effectively blocked the induction of luciferase by IFN-γ, even when high concentrations of IFN-γ (500 units/ml) were used (Fig. 3A). Furthermore, NIT-1 cells that were stably transfected with SOCS-1 did not upregulate class I MHC expression in response to IFN-γ (RS1.1, -1.6, and -1.7) (Fig. 3B).

When different SOCS-1–transfected NIT-1 clones were analyzed for IFN-γ responsiveness, we found that IFN-γ signaling was unaffected in some clones (RS1.2 and -1.4), as measured by the upregulation of class I MHC expression (Fig. 4). These clones expressed transfected SOCS-1 at levels similar to endogenous SOCS-1 in untransfected NIT-1 cells after IFN-γ treatment (Fig. 3C). In contrast, the level of SOCS-1 overexpressed in the IFN-γ-unresponsive clones RS1.1, -1.6, and -1.7 was much greater than endogenous SOCS-1. The ability of SOCS-1 to suppress IFN-γ signaling in β-cells thus depends on constitutive overexpression of high levels of the protein. These results suggest that in β-cells, endogenous SOCS-1 is expressed at levels (even after IFN-γ treatment) insufficient to terminate IFN-γ signaling.

IFN-γ, in combination with other proinflammatory cytokines such as TNF-α, is cytotoxic to β-cells (11,31,36). Therefore, we investigated whether SOCS-1 overexpression could inhibit IFN-γ–induced cell death in NIT-1 insulinoma cells. The SOCS-1–overexpressing NIT-1 clones RS1.1 and -1.7 and parental NIT-1 cells were treated with either IFN-γ or TNF-α or both. Cell death was then measured in a flow cytometic analysis for nuclear fragmentation (Fig. 4). Consistent with our previous findings (11), IFN-γ alone did not induce cell death in parental NIT-1 cells but significantly enhanced the cytotoxicity of TNF-α for these cells. However, this enhancement was not observed in NIT-1 cells that constitutively overexpress SOCS-1 (RS1.1 and -1.7). SOCS-1 overexpression therefore protects NIT-1 cells from IFN-γ–enhanced death from TNF-α.

To confirm that physiological levels of SOCS-1 do not terminate IFN-γ signaling in β-cells, the effect of SOCS-1 deficiency on IFN-γ signaling was examined. SOCS-1^−/− mice die by 3 weeks of age (37,38), but they can be rescued by removing the IFN-γ gene (39,40). Islets from SOCS-1^−/− IFN-γ^−/−, IFN-γ^−/−, and wild-type mice were stimulated with IFN-γ over a 96-h time course. The β-cell population was analyzed by flow cytometry for class I MHC expression as a marker of IFN-γ signaling (Fig. 5A). Maximal class I MHC expression on the β-cells of all three genotypes was observed 72 h after the addition of IFN-γ, which remained at the same level after 96 h. Similarly, we found that IFN-γ-induced STAT1 activation was also not altered by SOCS-1 deficiency (Fig. 5B). The levels of activated STAT1 at 0.5, 4, and 24 h after IFN-γ treatment were the same in SOCS-1^−/− islets as in controls. Therefore, SOCS-1 does not modulate the kinetics or intensity of IFN-γ signaling in β-cells.

Although SOCS-1 does not modulate the kinetics of IFN-γ signaling in β-cells, it may regulate the sensitivity of β-cells to IFN-γ. The sensitivity of SOCS-1^−/− β-cells to IFN-γ was assessed by measuring the upregulation of class I MHC as a readout for responsiveness to the cytokine. Pancreatic islets were treated with 0–1,000 units/ml IFN-γ, and then the β-cell population was analyzed for class I MHC expression by flow cytometry (Fig. 5C). Class I MHC expression on both SOCS-1^−/− and wild-type β-cells was upregulated by IFN-γ in a dose-dependent manner. However, SOCS-1 deficiency did not augment the sensitivity of β-cells to IFN-γ. At all concentrations (1–1,000 units/ml), IFN-γ induced the same level of class I MHC on SOCS-1^−/− β-cells as on wild-type cells. These results confirm that physiological levels of SOCS-1 do not negatively regulate IFN-γ signaling in β-cells.

We have also begun to analyze death of SOCS-1^−/− β-cells induced by cytokine combinations. These data also indicate that the contribution of IFN-γ to cytokine-induced β-cell death and iNOS induction is not increased with SOCS-1 deficiency (M.M.W. Chong and T.W.H. Kay, unpublished data).

The decay of phosphorylated STAT1 is also required for the termination of IFN-γ signaling. This has been shown to be a result of a nuclear protein tyrosine phosphatase (17). We investigated whether this mechanism was intact in β-cells. The expression of SOCS-1 is directly regulated by STAT1 (41) and could be used as a readout for STAT1 activity. Consistent with persistent STAT1 activation in the presence of continuous exposure to IFN-γ (Fig. 1), maximal SOCS-1 expression was sustained in NIT-1 insulinoma cells for as long as IFN-γ was present (Fig. 5). SOCS-1 expression was not downregulated even after 1 week (not shown). Removal of the cytokine, however, resulted in rapid decay of SOCS-1 transcription back to basal levels within 2–4 h (Fig. 6). This suggests that phosphorylated STAT1 is rapidly inactivated on removal of IFN-γ, at least to levels insufficient to stimulate gene transcription. A mechanism to inactivate phosphorylated STAT1 therefore exists in β-cells.

IFN-γ signaling in some cell types is transient (16,17), suggesting the existence of mechanisms that negatively regulate signaling in these cells. In pancreatic β-cells, however, we have found that IFN-γ signaling is persistent. Signaling is only terminated on removal of the cytokine. Persistence of signaling may be important in the cytotoxicity of IFN-γ for β-cells because the induction of many mediators of β-cell damage is dependent on prolonged treatment with the cytokine. For example, treatment with IFN-γ (plus IL-1β) for 24–48 h is required to induce Fas and iNOS expression in β-cells (3,4,11,31,36,42).

Members of the SOCS family of proteins have been shown to act as negative regulators of the Jak-STAT pathway by acting on the cytokine receptor or associated Jak kinase (19). In particular, SOCS-1 has been implicated in the negative regulation of IFN-γ signaling, both in vitro and in vivo. Overexpression of SOCS-1 has been shown to inhibit IFN-γ responsiveness in various cell lines (20,21,22). Analysis of SOCS-1–deficient mice has also suggested a role for SOCS-1 in regulating the IFN-γ-signaling pathway. These mice succumb to complex inflammatory pathology but can be rescued by IFN-γ deficiency (39,40), implicating IFN-γ overproduction, IFN-γ hypersensitivity, or both in the SOCS-1^−/− phenotype. In SOCS-1^−/− IFN-γ^−/− mice injected with IFN-γ, it was recently reported that STAT1 activation was prolonged in liver (43). STAT1 activation was also prolonged in SOCS-1^−/− hepatocytes treated with IFN-γ in vitro. Furthermore, in some cells of the hematopoietic lineage, SOCS-1 deficiency has been shown to result in hypersensitivity to IFN-γ (39,44). Consistent with studies in other cell types, we found that overexpression of SOCS-1 suppressed responses to IFN-γ in NIT-1 insulinoma cells, such as IFN-γ–induced STAT1 activation, class I MHC upregulation, and cell death. By analyzing the levels of transfected SOCS-1 required to suppress IFN-γ signaling, we found that physiological levels of SOCS-1, even after induction by IFN-γ, are insufficient to terminate IFN-γ signaling in β-cells. Although IFN-γ signaling was found to be persistent, we have indirectly shown that STAT1 activity is rapidly downregulated in β-cells on removal of IFN-γ, potentially by the nuclear protein tyrosine phosphatase, as previously described (17).

As IFN-γ has been shown to be an important mediator of β-cell death both in vitro and in vivo, the blockade of IFN-γ signaling remains important in designing therapies for the prevention of autoimmune β-cell destruction. We have demonstrated that overexpression of SOCS-1 (at high levels) terminates the otherwise persistent IFN-γ signaling in β-cells and may therefore be a useful strategy for protecting β-cells from the cytotoxic effects of the cytokine. Cottet et al. (45) recently demonstrated that IFN-γ activation of the Jak-STAT pathway is also involved in the suppression of insulin expression and glucose-stimulated insulin secretion by CDM3D insulinoma cells. This effect of IFN-γ on insulin-producing cells may also be dependent on persistent signaling because maximal suppression of insulin expression only occurred after prolonged exposure (48 h) to IFN-γ. However, it was found that IFN-γ–induced inhibition of both insulin expression and secretion could be prevented by overexpression of SOCS-1. SOCS-1 overexpression therefore protects β-cells from many effects of IFN-γ. Another potential benefit of manipulating SOCS-1 expression is that responses to multiple cytokines may be suppressed. Overexpression of SOCS-1 in other cell types has been shown to block signaling by a diverse range of cytokines, including members of the IL-6 cytokine family (20), type 1 interferons (22), IL-4 (46), and TNF-α (47). Thus, SOCS-1 overexpression could protect β-cells from multiple cytokines thought to be important in the progression of autoimmune diabetes.

This work was supported by the National Health and Medical Research Council, Australia, (Regkey 973002) and the Juvenile Diabetes Research Foundation’s Center grant for research into “Immune mechanisms of β-cell life and death.” We are grateful to Dr. Warren Alexander for mice and other reagents, Dr. Donna Paulnock for the pGLB928LUC plasmid, Dr. Tracy Willson for the genomic SOCS-1 clone, and Dr. Jian-guo Zhang for assistance with the detection of SOCS-1 protein. We also thank Rima Darwiche for her excellent technical assistance.