The Role of Nitric Oxide and the Unfolded Protein Response in Cytokine-Induced β-Cell Death

Sprague-Dawley rats were purchased from Harlan (Indianapolis, IN). RINm5F (rat insulinoma) cells were obtained from Washington University Tissue Culture Support Center (St. Louis, MO). Human islets were obtained from the Islet Cell Resource at Washington University (St. Louis, MO) and University of Miami (Miami, FL). RPMI-1640, CMRL-1066 tissue culture medium, l-glutamine, penicillin, and streptomycin were from GIBCO-BRL (Grand Island, NY). PERK^−/−, A/A, wild-type, and IRE1α^−/− mouse embryonic fibroblasts (MEFs) have previously been described (29,30). Empty pETFVA vector and pETFVA-eIF2α (S51A) were gifts from Dr. Randal Kaufman (University of Michigan), and pcDNA3-PERK (K621M) has been previously described (31). FCS was obtained from Hyclone Laboratories (Logan, UT). Human recombinant IL-1β and human and mouse recombinant IFN-γ were obtained from PeproTech (Rocky Hill, NJ). Prostaglandin J₂ ((PGJ₂) was from Cayman Chemical (Ann Arbor, MI). Enhanced chemiluminescence was purchased from Amersham Pharmacia Biotech (Piscataway, NJ). Neutral red, camptothecin, and tunicamycin were obtained from Sigma (St. Louis, MO). Caspase-3 fluorometric assay was from R&D Systems (Minneapolis, MN). N^G-monomethyl-l-arginine (NMMA) and (Z)-1(N,N-diethylamino) diazen-1-ium-1,2-diolate (DEA-NO) were purchased from Axxora (San Diego, CA). Rabbit anti–phospho-eIF2α and rabbit anti-eIF2α were from Stressgen (San Diego, CA). Rabbit anti–caspase-3 and rabbit anti-active caspase-3 were obtained from Cell Signaling Technology (Beverly, MA). Mouse anti-CHOP was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Rabbit anti–high-mobility group box 1 protein (HMGB1) was from BD Pharmingen (San Diego, CA). Horseradish peroxidase–conjugated donkey anti-rabbit and donkey anti-mouse antisera were obtained from Jackson Immunoresearch Laboratories (West Grove, PA). All other reagents were obtained from commercially available sources.

Islets were isolated from male Sprague-Dawley rats (250–300 g) by collagenase digestion as previously described (32). Islets were cultured overnight in complete CMRL-1066 (containing 2 mmol/l l-glutamine, 10% heat-inactivated FCS, 100 units/ml penicillin, and 100 μg/ml streptomycin) at 37°C under an atmosphere of 95% air and 5% CO₂ before experimentation. RINm5F cells and MEFs were removed from growth flasks by treatment with 0.05% trypsin and 0.02% EDTA for 5 min at 37°C, washed two times with RPMI-1640 or Dulbecco's modified Eagle's medium, respectively, and plated at the indicated densities.

RINm5F cells were transiently transfected using the Amaxa Nucleofector electroporator (Amaxa Biosystems, Gaithersburg, MD) with 2 μg pEGFP (enhanced green fluorescent protein) and pETFVA-eIF2α (S51A) or pcDNA3-PERK (K621M) as previously described (33). Using this method, we routinely obtain a transfection efficiency of 50–60%, as determined by EGFP expression.

Western blot analysis of RINm5F cells and islet lysates were performed using semidry transfer conditions as previously described (34). All primary antibodies were diluted 1:1,000. Horseradish peroxidase–conjugated donkey anti-rabbit and donkey anti-mouse secondary antibodies were used at 1:7,000 and 1:5,000, respectively. Antigens were detected by enhanced chemiluminescence according to the manufacturer's recommendations (Amersham Pharmacia Biotech, Piscataway, NJ).

Nitrite production was determined by adding 50 μl of the Greiss reagent to 50 μl culture supernatant (35). The absorbance was measured at 540 nm, and nitrite concentrations were calculated from a sodium nitrite standard curve.

Total RNA was isolated from rat islets or RINm5F cells using the RNeasy kit (Qiagen, Valencia, CA). First-strand cDNA synthesis was performed using oligo(dT) and the reverse transcriptase Superscript Preamplification System (Invitrogen, Carlsbad, CA), according to the manufacturer's specifications. A fragment of Xbp1 was then amplified using the following primers: 5′-AAACAGAGTAGCAGCGCAGACTGC-3′ (forward) and 5′-GGATCTCTAAAACTAGAGGCTTGGTG-3′ (reverse) (36). PCR products were digested with PstI for 1h at 37°C, and samples were analyzed on a 2% agarose gel containing ethidium bromide. IRE1α catalyzes the splicing of a 26-nucleotide fragment from the Xbp1 mRNA transcript that contains a PstI digestion site. Therefore, when IRE1α is active, Xbp1 is spliced, resulting in a fragment of ∼600 bp, while unspliced Xbp1 appears as two fragments of ∼300 bp due to the presence of the PstI restriction site. GAPDH was amplified using previously described primers (37) to serve as a control for cDNA amounts.

Cell viability was determined using the neutral red assay as previously described (19,38). Following treatment, RINm5F cells (1.0 × 10⁵ cells/200 μl RPMI-1640) were washed and then incubated in 200 μl fresh media containing neutral red (50 μg/ml) for 2 additional hours at 37°C. Following this incubation, the excess neutral red was discarded by washing and the cells were fixed with 1% formaldehyde-1% CaCl₂. The neutral red was extracted from the cells using 100 μl 50% ethanol/1% acetic acid lysing solution. The amount of neutral red in the lysate was measured at a wavelength of 540 nm. The percentage of dead cells was determined by comparing neutral red uptake of treated samples with uptake by the unstimulated control, where neutral red uptake by the unstimulated control cells is set at 100% viable (or the absence of cell death). For a limited number of experiments, the MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay was used as an index of cell viability (19).

Following treatment, RINm5F cells (2.0 × 10⁵ cells/400 μl RPMI-1640) or isolated rat islets (150 islets/400 μl CMRL-1066) were isolated, washed, and then lysed in 55 μl lysis buffer. Caspase-3 activity was determined using 50 μl lysate according to the manufacturer's specifications (Caspase-3 Fluorometric Assay Kit; R&D Systems). The relative fluorescence units were normalized to total protein content of sample determined by the bicinchoninic acid assay (Pierce, Rockford, IL). Treated samples were compared with unstimulated controls, and results are reported as fold increase over control.

The effects of NO on gene expression were evaluated by Affymetrix gene chip analysis. In brief, human islets (1,000/condtion) were treated for 3 h with or without NO donor DEA-NO (0.5 mmol/l). The islets were harvested by centrifugation and total RNA extracted using Trizol (as outlined by supplier, Gibco/BRL). The total RNA was further purified using the RNeasy Kit (Qiagen) and then used for expression profiling with Affymetrix U133 gene chips at the Washington University Gene Sequencing Facility (St. Louis, MO). Changes in mRNA levels, as determined by gene chip analysis, are the average of four experiments from four independent human islet preparations.

Statistical comparisons were made between groups using one-way ANOVA. Significant differences between groups, treated versus unstimulated controls (P < 0.05, unless otherwise indicated), were determined by Newman-Keuls post hoc analysis.

The UPR is a survival pathway that is activated in response to cellular stress. When cells are unable to sufficiently respond to and overcome the stress, a programmed death pathway leads to cellular demise. NO, the primary mediator of cytokine-induced β-cell damage and death (18,19,39,40), has been shown to activate the UPR in a number of cell types, including β-cells (20,21,41,42). Consistent with these previous studies, treatment of RINm5F cells (Fig. 1A) or rat islets (data not shown, reference 43) with IL-1 results in the time-dependent production of NO that is first apparent after a 12-h incubation, with an ∼10-fold increase in nitrite formation after a 24-h incubation. Consistent with the time-dependent production of NO, IL-1 stimulates CHOP expression (Fig. 1B) in RINm5F cells and eIF2α phosphorylation in rat islets (Fig. 1C) that is first apparent after a 12-h incubation. Importantly, the NOS inhibitor NMMA, which prevents IL-1–induced NO production, also attenuates CHOP expression and eIF2α phosphorylation in response to IL-1. When active, IRE1α catalyzes the processing of a 26-nucleotide fragment (containing a PstI restriction enzyme site) from Xbp-1 mRNA. Treatment of rat islets for 24 h with IL-1 results in IRE1α activation, as evidenced by the processing of Xbp1 mRNA (accumulation of the spliced form; Fig. 1D). NO mediates IRE1α activation, as Xbp-1 processing is attenuated by NMMA. The N-linked glycosylation inhibitor tunicamycin, which stimulates CHOP expression, eIF2α phosphorylation, and Xbp1 splicing (Fig. 1B–D), was used as a positive control for UPR activation in these experiments.

Similar to rodent islets and RINm5F cells, NO also activates the UPR in human islets. Treatment of human islets for 3 h with the NO donor compound DEA-NO results in a three- to sixfold increase in the accumulation of the UPR-responsive mRNAs encoding CHOP, ATF-4, and gadd34 (Fig. 1E). In addition, DEA-NO also stimulates an ∼sixfold increase in the accumulation of hsp 70 mRNA in human islets. Overall, these findings, consistent with previous reports by our laboratory and others (20,21,42), show that cytokine treatment stimulates UPR activation in a NO-dependent manner in rodent and human islets.

Using a biochemical assay of membrane integrity, the effects of 6-, 12-, and 24-h exposures to IL-1 on RINm5F cell viability were examined. As shown in Fig. 2A, a 24-h incubation with IL-1 results in a 23% reduction in RINm5F cell viability. NO is the mediator of cytokine-induced β-cell death, as IL-1–induced nitrite production correlates temporally with cell death (Fig. 2B) and NMMA prevents IL-1–induced β-cell death. UPR activation following treatment with tunicamycin is associated with a loss of RINm5F cell viability (Fig. 2A), and this action is independent of β-cell production of NO. Taken together, these findings correlate cytokine-induced iNOS expression and NO production with activation of an ER stress response and the induction of a death cascade in β-cells.

Because NO activates the UPR and prolonged UPR activation is associated with cell death, the role of UPR activation as a potential mediator of cytokine-induced β-cell death was further investigated by examining the effects of cytokines and UPR activators on caspase-3 activity in RINm5F cells and rat islets. This executioner caspase is activated by both the intrinsic and extrinsic apoptotic pathways (44,45). Following a 24-h incubation, tunicamycin enhances caspase-3 activity in RINm5F cells (2.4-fold) and rat islets (2.5-fold; Fig. 3A and B). The activation of caspase-3 following tunicamycin treatment correlates with the ability of this UPR activator to induce RINm5F cell death (Fig. 2). Using the MTT assay as a second methodology to examine cell death, a 24-h treatment with tunicamycin induces the death of 30.0 ± 7.3% of RINm5F cells. The caspase-3 inhibitor IV (100 μmol/l; CalBiochem, San Diego, CA), which attenuates caspase-3 activity in islets and RINm5F cells (19), reduces tunicamycin-induced RINm5F cell death to 11.8 ± 5.1%. These findings suggest that caspase-3 is activated and contributes to the loss of β-cell viability in response to ER stress activators such as tunicamycin.

In contrast to cell death induced by the UPR activator tunicamycin, cell death induced by IL-1 does not correlate with caspase-3 activation. IL-1 fails to activate caspase-3 in RINm5F cells (Fig. 3A) or rat islets (Fig. 3B) following exposures as long as 24 h. These observations are consistent with our previous studies showing that extended cultures up to 48 h with IL-1 also fail to stimulate caspase-3 activity in RINm5F cells (19). There is a slight 30–40% increase in caspase-3 activity in rat islets treated for 24 h with IL-1; however, this increase was variable and did not achieve statistical significance. While IL-1 failed to stimulate caspase-3 activation, it did stimulate NO production by RINm5F cells (Fig. 2B) and rat islets (Fig. 3C). These findings dissociate cytokine-induced NO production and UPR activation from the mechanisms by which cytokines mediate β-cell death. Importantly, known inducers of the UPR stimulate islet and RINm5F cell death by a process that is dependent on caspase-3 activation, while IL-1 fails to stimulate caspase-3 activation even though NO production in response to IL-1 is an effective activator of the UPR in islets. Since NO has been reported to inhibit caspase-3 activity (46–48), we also examined the effects of IL-1 and the NOS inhibitor NMMA on caspase activation. In the presence of NMMA, IL-1 still failed to activate caspase-3 (Fig. 3A and B). Although IL-1 is capable of activating the UPR in a NO-dependent manner, these findings suggest that the mechanisms responsible for β-cell death under conditions of prolonged UPR activation (tunicamycin) differ from cytokine-induced β-cell death, as IL-1 fails to stimulate caspase-3 activation or caspase-dependent β-cell death.

To further explore the relationship between cytokine-induced UPR activation and cytokine-induced β-cell death, we examined the effects of IL-1 on the processing of caspase-3 from the inactive procaspase to the active mature protease. As shown in Fig. 4A, a 24-h incubation of RINm5F cells with IL-1 fails to stimulate processing of caspase-3 to its active mature form. In contrast to the effects of IL-1, the peroxisome proliferator–activated receptor γ ligand PGJ₂ stimulates the processing of caspase-3 to the active mature protease. Importantly, we have recently shown that PGJ₂ activates the UPR and induces β-cell apoptosis in a caspase-3–dependent manner (33).

Cells undergoing necrosis are known to release HMGB1 (49–52). Indeed, we have recently shown that cytokines stimulate HMGB1 release from rat and human islets and RINm5F cells in a NO-dependent manner (19). In contrast, inducers of apoptosis fail to stimulate HMGB1 release from rat and human islets and RINm5F cells (19,51). Consistent with a role for caspase-3 in β-cell death in response to UPR activators, PGJ₂ fails to stimulate HMGB1 release from rat islets, while IL-1 stimulates HMGB1 release from rat islets (Fig. 4B). These findings, using a second UPR activator, provide additional evidence dissociating the activation of the UPR from cytokine-mediated β-cell death.

While cytokines stimulate UPR activation in islets and RINm5F cells in an NO-dependent fashion, experimental evidence presented in Figs. 2–4 indicates that the mechanisms of cytokine-induced β-cell death differ from the pathway responsible for UPR-mediated β-cell death. One potential explanation for these differences is in the ability of NO to inhibit caspase-3 activity. As shown in Fig. 5, tunicamycin stimulates a threefold increase in RINm5F cell caspase activity following a 12-h incubation. Importantly, tunicamycin-induced caspase-3 activation is prevented by the NO donor sodium DEA-NO, while DEA-NO alone fails to activate caspase-3 activity in RINm5F cells.

Consistent with a protective role of the UPR in prolonging survival in response to cellular stress, recent studies have shown that inhibition of UPR transducer activity increases the susceptibility of cells to ER stress (23). To further probe cell viability in response to cytokines and UPR activators, wild-type, Perk^−/−, A/A (expressing S51A eIF2α mutant), and IRE1α^−/− MEFs were treated for 24 h with IL-1 and IFN-γ or with tunicamycin. The combination of cytokines was required because IL-1 alone fails to stimulate NO production or the death of MEFs (data not shown). In combination, IL-1 and IFN-γ stimulate an ∼10- to 15-fold increase in NO production (Fig. 6A) and the death of 17.3% of MEFs (Fig. 6B). Importantly, there is no statistically significant difference in the levels of cell death or the levels or NO production in response to IL-1 and IFN-γ when comparing wild-type, Perk^−/−, A/A, or IRE1α^−/− MEFs (Fig. 6). As expected, tunicamycin, which failed to stimulate NO production, reduced the viability of wild-type MEFs by 41.7%. Inhibition of the PERK signaling cascade, either by expression of a mutant eIF2α that cannot be phosphorylated (A/A) or via the absence of PERK, further enhanced the level of tunicamycin-induced cell death to 75–80% of the MEFs. We also examined the effects cytokines and tunicamycin on the viability of RINm5F cells or RINm5F cells expressing dnPERK mutant (K621M) or the S51A eIF2α mutant (Fig. 6C, data presented as percent increase in IL-1–and tunicamycin-induced death). Similar to MEFs, 19% of RINm5F cells are killed following a 24-h incubation with IL-1, and this level of death is not modified in cells expressing either the S51A eIF2α mutant or cells expressing dnPERK (K621M). In contrast to IL-1, tunicamycin induces the death of 26% of RINm5F cells following a 24-h incubation, and the level of death is enhanced by 77% in cells expressing the S51A eIF2α mutant. In RINm5F cells expressing dnPERK, the level of tunicamycin-induced death is enhanced approximately twofold (Fig. 6C). While the increase in tunicamycin-induced death of RINm5F cells expressing UPR transducer mutants may appear to be somewhat attenuated when compared with the increase in death in the corresponding MEFs, it is the expected increase in the level of RINm5F cell death when the 60% transfection efficiency of the UPR transducer mutants is taken into account in these assays. These findings, which support previous studies indicating that UPR-mediated cell death is enhanced in the absence of a functional PERK response (23,33), provide direct evidence that cytokine-induced death can occur independent of UPR activation, as deletion of either PERK or IRE1α or expression of the S51A eIF2α mutant does not alter the levels of death in response to cytokine treatment.

The purpose of this study was to evaluate the role of ER stress and the subsequent activation of the UPR as a mechanism by which cytokine treatment causes β-cell death. The UPR is a conserved cellular response designed to alleviate multiple forms of cellular stress due to oxidative or nitrosative damage, aberrant accumulation of proteins in the ER, and stress associated with virus infection (12,22). If cellular stress is not eliminated or cells fails to recover from the stressful insult, apoptosis ensues (12,22,23). β-Cells are highly sensitive to perturbations in the response to ER stress. UPR activators, such as thapsigargin, have been shown to induce β-cell apoptosis (24). In addition, mice deficient in PERK develop diabetes that is due to the apoptotic loss of β-cells (26,53) and β-cell mass is reduced in mice expressing a S51A eIF2α mutant (23), which is not capable of being phosphorylated by PERK or other eIF2α kinases (23). NO, a known activator of the UPR (20,21,42), has been implicated as an inducer of ER stress in β-cells, and UPR activation in response to this ER stress has been reported to mediate β-cell apoptosis (20,21). In support of this hypothesis, cytokines stimulate eIF2α phosphorylation, IRE1α activation, and UPR gene expression in a NO-dependent fashion that correlates with cytokine-induced β-cell death (20,21).

In the current study, we have used a biochemical approach to investigate the role of UPR activation in cytokine-mediated β-cell death. Consistent with previous studies, we show that IL-1 stimulates the production of NO by islets and insulinoma cells in a time-dependent manner that correlates with the activation of the UPR. UPR activation is mediated by β-cell production of NO as IL-1–induced eIF2α phosphorylation, CHOP expression, and Xbp1 processing (IRE1α activation) are attenuated by the NOS inhibitor NMMA. While cytokine-induced NO production correlates with cytokine-induced UPR activation, the UPR does not appear to participate in cytokine-induced β-cell death. This conclusion is supported by three findings. 1) β-Cell death in response to UPR activators is associated with a 2.5- to 3-fold increase in caspase-3 activity and is attenuated by the caspase-3 inhibitor IV. In contrast, IL-1 fails to activate caspase-3; however, this cytokine does induce the death of RINm5F cells and rat islets in an NO-dependent manner (Figs. 2 and 3 and reference 19). 2) The peroxisome proliferator–activated receptor γ ligand PGJ₂ induces β-cell apoptosis by a pathway that is associated with UPR activation and that is dependent on the activation of caspase-3 (33). Consistent with these previous findings, PGJ₂ stimulates the proteolytic cleavage of caspase-3 to the active protease, whereas IL-1 fails to stimulate caspase-3 cleavage. We observed similar differences in the response of β-cells to apoptosis inducers, such as camptothecin and staurosporine (induce caspase-3–dependent β-cell apoptosis), as compared with cytokines that stimulate caspase-3–independent β-cell necrosis (19). 3) NO is a known inhibitor of caspase-3 activity (via S-nitrosothiol formation at the active-site cysteine) (46–48), and we show that NO inhibits tunicamycin-induced caspase-3 activity in RINm5F cells (Fig. 5). Taken together, these findings dissociate the mechanisms associated with cytokine-induced β-cell death from cytokine-induced UPR activation. Cytokines appear to kill β-cells by NO-dependent necrosis that is caspase-independent and associated with the release of the necrosis marker HMGB1 (Fig. 4 and reference 19). In contrast, UPR activators stimulate β-cell apoptosis that is mediated in part by caspase-3.

If our conclusion that the UPR does not mediate cytokine-induced β-cell death is correct, then inhibition of UPR transducer activity should not modulate the levels of cell death in response to cytokines. Specifically, this would support the hypothesis that NO is an activator of the UPR, and while UPR activation correlates temporally with cell death, it is not the mediator of cytokine-induced cell death. In contrast, if the UPR participates in cytokine-mediated damage, then inactivation of UPR transducer activity would be expected to enhance cell death. We have previously shown that inhibition of PERK enhances caspase-3 activation in response to PGJ₂ (33), and Scheuner et al. (23) demonstrated a 57% increase in tunicamycin-induced apoptosis of MEFs expressing a S51A mutation. Consistent with this hypothesis, there is no difference in the levels of cell death in wild-type, PERK^−/−, S51A (A/A), and IRE1α^−/− MEFs following a 24-h incubation with IL-1 and IFN-γ. In contrast, there is a 31% increase in the levels of tunicamycin-induced death in PERK^−/− and a 35% increase in A/A MEFs compared with wild-type controls. Similar to MEFs, tunicamycin-induced death of RINm5F cells expressing dnPERK or eIF2α S51A mutant is enhanced compared with cells expressing the vector control, while the levels of NO-mediated IL-1–induced death are not modified. Importantly, these findings show that cell death mediated by NO is not affected by the absence of UPR transducer activity. In contrast, removal of the protective actions of UPR, specifically the PERK pathway (including A/A MEFs), renders cells more susceptible to ER stress–mediated cell death.

Our findings provide biochemical evidence that UPR activation correlates in a temporal fashion with the stimulatory actions of cytokines on NO production and that NO is an activator of the UPR in β-cells. Importantly, this is only a correlation, as UPR activation does not appear to contribute to cytokine-induced β-cell death. In view of these findings, new questions emerge regarding UPR activation and the role of this protective response in cytokine-mediated β-cell damage. In a series of studies, we have shown that β-cells have a unique ability to recover from the damaging actions of cytokines (54–56). The addition of a NOS inhibitor to islets treated for 24 h with IL-1 results in the time-dependent recovery of insulin secretion and mitochondrial oxidative metabolism that is complete in 8 h (56). This recovery response is activated by NO, is associated with enhanced expression of heat shock proteins and chaperones that assist in protein folding (data not shown), and requires de novo gene expression (54,55). We hypothesize that NO-induced UPR activation does not function as a mediator of β-cell death but may function to assist in the recovery of β-cells from NO-mediated damage. This is one potential mechanism we are currently exploring in our laboratory.

This work was supported by National Institutes of Health Grants DK52194 and DK068839 (to J.A.C.) and an American Heart Association predoctoral fellowship award (to K.T.C.).

We thank Colleen Kelly Bratcher for expert technical assistance.