Cytokine-induced β-cell death is an important event in the pathogenesis of type 1 diabetes. The transcription factor nuclear factor-κB (NF-κB) is activated by interleukin-1β (IL-1β), and its activity promotes the expression of several β-cell genes, including pro- and anti-apoptotic genes. To elucidate the role of cytokine (IL-1β + γ-interferon [IFN-γ])-induced expression of NF-κB in β-cell apoptosis, rat β-cells were infected with the recombinant adenovirus AdIκB(SA)2, which contained a nondegradable mutant form of inhibitory κB (IκB(SA)2, with S32A and S36A) that locks NF-κB in a cytosolic protein complex, preventing its nuclear action. Expression of IκB(SA)2 inhibited cytokine-stimulated nuclear translocation and DNA-binding of NF-κB. Cytokine-induced gene expression of several NF-κB targets, namely inducible nitric oxide synthase, Fas, and manganese superoxide dismutase, was prevented by AdIκB(SA)2, as established by reverse transcriptase–polymerase chain reaction, protein blot, and measurement of nitrite in the medium. Finally, β-cell survival after IL-1β + IFN-γ treatment was significantly improved by IκB(SA)2 expression, mostly through inhibition of the apoptotic pathway. Based on these findings, we conclude that NF-κB activation, under in vitro conditions, has primarily a pro-apoptotic function in β-cells.
Prolonged exposure of rodent and human β-cells to the cytokines interleukin-1β (IL-1β) and γ-interferon (IFN-γ) leads to functional suppression and eventual cell death, mostly by apoptosis (rev. in 1,2,3). The β-cell outcome after immune-mediated injury, i.e., apoptosis or survival with functional recovery, probably depends on the complex pattern of dozens of genes up- or downregulated in parallel and/or sequentially (4,5).
Cytokine-induced modifications in gene expression are mediated by the activation of different transcription factors. The transcription factor nuclear factor-κB (NF-κB) seems to regulate several genes affected by cytokines in β-cells, as suggested by a recent oligonucleotide array analysis of cytokine-treated β-cells performed by our group (5) and by gel shift analysis and transient transfections with promoter-luciferase reporter constructs for inducible nitric oxide synthase (iNOS) (6,7,8), Fas (8a), and manganese superoxide dismutase (MnSOD) (9). Whereas iNOS and Fas contribute to β-cell death, MnSOD is probably involved in β-cell defense against the toxic oxygen free radicals generated during exposure to cytokines (4). Preponderance of the anti- or pro-apoptotic effects of NF-κB depends on both the cell type and the type of inducer tested (10). It remains to be determined which is the main role for NF-κB in the context of cytokine-induced β-cell apoptosis.
NF-κB is present in an inactive form in the cytoplasm, where it is bound to inhibitory κB (IκB) proteins (rev. in 11). After stimulation by different agents, such as cytokines, bacterial products, and viruses and their proteins (rev. in 12), IκB is phosphorylated at two NH2-terminal Ser residues (S32 and S36) by an IκB kinase complex, leading to its degradation at the proteasome level. The released NF-κB complex can then enter the nucleus and stimulate or inhibit the transcription of genes containing NF-κB binding sites in their promoter regions (11). Mutation of the amino acids 32 and 36 of IκB prevents its degradation and consequent NF-κB activation (13). Indeed, overexpression of an IκB mutant form, AdIκB(SA)2, with serine 32 and 36 substituted by alanine residues, specifically blocked NF-κB activation, without affecting other transcription factors (14). Moreover, infection of intestinal epithelial cells (14) and rat hepatic stellate cells (15) with a recombinant adenovirus expressing AdIκB(SA)2 prevented the expression of several cytokine-induced genes whose transcription depends on NF-κB.
To test whether cytokine-induced activation of NF-κB in pancreatic β-cells is mostly pro- or anti-apoptotic, we infected β-cells with AdIκB(SA)2 overexpressing the IκB “super-repressor.” This approach prevented cytokine-induced nuclear translocation of NF-κB and expression of the NF-κB–dependent genes iNOS, Fas, and MnSOD. Furthermore, it prevented cytokine-induced apoptosis in β-cells. Our data indicate that NF-κB is mostly pro-apoptotic in β-cells.
RESEARCH DESIGN AND METHODS
Isolation and culture of β-cells.
Islets were isolated from male Wistar rats. The rats were housed according to the Guidelines of the Belgian Regulations for Animal Care, and the experimental protocol was approved by the Ethical Committee for Animal Experiments of the Vrije Universiteit Brussel. Islets were isolated, dissociated, and purified to single β-cells that were ∼95% viable, as previously described (16). β-Cells were cultured in Ham’s F10 medium supplemented with 10 mmol/l glucose, 50 mmol/l IBMX (3-isobutyl-l-methylxantine), 1% bovine serum albumin (Boehringer Mannheim, Mannheim, Germany), 0.1 mg/ml streptomycin (Continental Pharma, Puteaux, Belgium), 0.075 mg/ml penicillin (Laboratoires Diamant, Brussels, Belgium), and 2 mmol/l l-glutamine (Gibco, Paisley, U.K.) (17). For mRNA and nuclear or total cellular protein extraction, single β-cells were reaggregated for 3 h in a rotatory shaking incubator (18). For viability experiments (see below), β-cells (1 × 104 cells per well) were cultured in Falcon 96-well microtiter plates (Becton Dickinson, Franklin Lakes, NJ) containing 200 μl medium. For nuclear translocation experiments, 5 × 104 β-cells were plated in 8-well poly-l-lysine–coated chamber slides. During prolonged culture, the culture medium was changed every 3 days.
At 24 h postinfection, cytokines were added. The effects of cytokines after transfection with the different adenoviral constructs were examined after 1 h (nuclear translocation and DNA binding), 24 h (mRNA and protein abundance), 3 days (nitrite concentration), and 6 days (viability) of culture in the presence of recombinant murine IFN-γ (1,000 units/ml, 10 units/ng; Holland Biotechnology, Leiden, the Netherlands) and recombinant human IL-1β (50 units/ml, 38 units/ng; a gift from Dr. C.W. Reynolds from the National Cancer Institute, Bethesda, MD). The concentration of cytokines and the time points of sampling were selected based on our previous studies with rodent pancreatic islets and β-cells (2,8,19,20). Every 3 days, fresh cytokines were added in parallel to medium change.
Transfection with recombinant adenoviruses.
After overnight incubation, part of the β-cells were either infected with control virus (AdLuc or AdGFP) or with AdIκB(SA)2 expressing the NF-κB super-repressor. The recombinant replicative-deficient adenovirus containing a mutated nondegradable IκB (AdIκB(SA)2) was prepared as previously described (14). The recombinant adenoviruses AdGFP and AdLuc were used as controls for transfection with AdIκB(SA)2. β-Cells were infected at a multiplicity of infection (MOI) of 7.5 for 2 h at 37°C.
Nuclear extracts and electrophoretic mobility shift assay.
DNA binding by nuclear protein from β-cells (4 × 105 β-cells) or rat insulinoma (RIN) cells (4 μg protein) and supershift analysis with anti-p65 (see below) were performed as previously described (9). The sequence of the NF-κB consensus oligonucleotide was 5′-AGCTTCAGAGGGGACTTTCCGAGA (upper strand shown).
Immunofluorescence and immunoblotting.
Immunoblotting and immunofluorescence were done as previously described (21). Anti-p65 (sc-372; Santa Cruz Biotechnology, Santa Cruz, CA) was used at a 1:100 dilution, anti-IκB (sc-371; Santa Cruz Biotechnology) at 1:500, anti-iNOS (BD Transduction Labs, Los Angeles) at 1:3,000, and anti-hemaglutinin (Roche Diagnostics, Mannheim, Germany) at 1:100.
mRNA isolation and reverse transcriptase–polymerase chain reaction.
Poly(A)+ RNA was isolated from β-cells (0.5 × 105 cells) using oligo(dT)25-coated polystyrene Dynabeads (Dynal, Oslo). The reverse transcription reaction and subsequent polymerase chain reaction (PCR) were performed as previously described (19,20). The primers used for iNOS, MnSOD, and glyceraldehyde-phosphate dehydrogenase (GAPDH) have been described previously (9,19). The primers for Fas were 5′-GAATGCAAGGGACTGATAGC (forward primer) and 5′-TGGTTCGTGTGCAAGGCTC (reverse primer). The number of cycles was selected to allow linear amplification of the cDNA under study. The ethidium bromide–stained agarose gels were photographed under ultraviolet-transillumination using a Kodak Digital Science DC40 camera (Rochester, NY). The PCR band intensities on the image were quantified by Biomax 1D image analysis software (Kodak) and expressed as pixel intensities (optical density). The target cDNAs present in each sample were corrected for their respective GAPDH values. Expression of the “housekeeping” gene GAPDH is not affected by exposure to cytokines (22).
Determination of nitrite concentration.
Culture media were collected after 72 h for colorimetric nitrite determination (23). Nitrite was not determined at subsequent time points, because after 6 days, β-cells exposed to cytokines have a significant decrease in viability (see results).
Assessment of β-cell viability and function.
The percentage of viable, apoptotic, and necrotic β-cells was determined after 6 days of exposure to cytokines, the required time to detect significant increases in β-cell death (24,25). For this purpose, β-cells were incubated for 15 min with propidium iodide (PI; 10 mg/ml) and Hoechst 342 (20 mg/ml) (26). The cells were examined by inverted microscopy with ultraviolet excitation at 340–380 nm. Viable cells were identified by their intact nuclei with blue fluorescence (Hoechst 342), necrotic cells by their intact nuclei with yellow-red fluorescence (Hoechst 342 + PI), and apoptotic cells by their fragmented nuclei, exhibiting either a blue (Hoechst 342; early apoptosis) or yellow-red fluorescence (Hoechst 342 + PI; late apoptosis) (26). This fluorescence assay is quantitative for single β-cells and has been validated by systematic comparisons with electron microscopy observations (26). This method has been used to evaluate apoptosis/necrosis in rat (25,26), mouse (20,27), and human (24) β-cells. The use of purified β-cells in these experiments provides a homogeneous cell population (>95% β-cells), decreasing the problem of apoptosis/necrosis detection in non–β-cells, which is inherent to studies performed in whole islets. Determination of β-cell insulin content and release was performed 48 h after infection, as previously described (18).
The results are presented as the means ± SE. Statistical differences between the groups were determined by Student’s t test or, when indicated, by analysis of variance (ANOVA) followed by multiple t tests with the Bonferroni correction.
The effect of adenovirus-mediated IκB(SA)2 overexpression on β-cell function and cytokine-induced NF-κB activity.
In an initial series of experiments, green fluorescent protein (GFP) was used as reporter to determine the number of successfully infected cells. The day after incubation of the β-cells with the adenoviruses, at a MOI of 7.5, we observed that >75% of the β-cells expressed the target protein (n = 5). However, GFP was found to influence β-cell survival (see below), forcing us to add an additional control adenovirus expressing luciferase, which did not affect β-cell survival (see below). MOIs >20 caused β-cell death (data not shown).
Putative nonspecific viral effects on basal and glucose-stimulated β-cell functions were determined by incubating the virus-infected β-cells at different glucose concentrations (n = 3–4 independent experiments, evaluated in duplicate). At 48 h after infection (MOI 7.5), the insulin content (in nanograms per 103 cells) was similar in all conditions under study: noninfected β-cells, 31.7 ± 5.0; AdLuc, 26.3 ± 4.9; and AdIκB(SA)2 26.2 ± 5.9 (P > 0.05 vs. control for all comparisons). The functional activity of the β-cells was further evaluated by assessment of insulin release in response to glucose. The fraction of the total cellular insulin secreted during a 2-h incubation in the presence of 2.5 or 20 mmol/l glucose, respectively, was: 0.8 ± 0.2 and 8.9 ± 2.0% in noninfected β-cells; 0.9 ± 0.3 and 9.2 ± 1.3% in AdLuc-infected β-cells; and 0.7 ± 0.2 and 8.5 ± 1.6% in AdIκB(SA)2-infected β-cells (P > 0.05 vs. control for all comparisons). Insulin mRNA abundance was also unaffected after infection (n = 2; data not shown). These findings suggest that at an MOI of 7.5, neither AdLuc nor AdIκB(SA)2 compromise the synthesis and secretion of insulin by β-cells.
The AdIκB(SA)2-specific effects were determined 2 days after infection with AdIκB(SA)2. The expression level of IκB(SA)2 exceeded the endogenous IκB expression by more than fivefold, and the IκB(SA)2 mutant protein was well preserved after the addition of a combination of cytokines (IL-1β and IFN-γ), whereas endogenous IκB was degraded (Fig. 1). Cytokine signaling downstream of IκB was severely affected by expressing IκB(SA)2. Although the cytokines caused nuclear translocation of the p65 subunit of NF-κB, in noninfected β-cells or β-cells infected with control viruses, the subcellular localization of NF-κB remained cytosolic in AdIκB(SA)2-infected cells (Fig. 2A). Moreover, gel shift analysis of control β-cells exposed to cytokines showed specific DNA binding of NF-κB, which was absent in AdIκB(SA)2-infected β-cells (Fig. 2B). Specificity was established by causing supershift of the binding complex when incubated with antiserum directed against the NF-κB subunit p65 as well as by competition between labeled and cold target sequence, in 50-fold molar excess (data not shown). The observations described above indicate that AdIκB(SA)2 successfully infects primary rat β-cells and leads to overexpression of IκB(SA)2, which is effective in preventing NF-κB activation and nuclear binding, without affecting β-cell–specific functions.
The effect of NF-κB activity on cytokine-induced gene expression and β-cell survival.
To investigate the downstream effects of inhibiting NF-κB signaling, we evaluated cytokine-induced gene expression in β-cells. Gene expression of MnSOD, Fas, and iNOS was switched on after the addition of cytokines to uninfected β-cells or to β-cells infected with the control viruses AdGFP or AdLuc (Fig. 3). On the other hand, the abundance of these mRNAs remained low in cytokine-treated β-cells infected with AdIκB(SA)2 (Fig. 3). After cytokine exposure, the relative abundance of iNOS, Fas, and MnSOD mRNAs (optical density corrected per GAPDH abundance) was, respectively, 4-, 8-, and 10-fold lower in β-cells infected with AdIκB(SA)2 compared with the control conditions (n = 3). Expression of the housekeeping gene GAPDH was not affected by exposure to cytokines or by infection with the different adenoviruses used in this study (Fig. 3). Inhibition of cytokine-induced iNOS mRNA expression by AdIκB(SA)2 was confirmed at the protein (Fig. 1) and reaction product level (see below). Whereas iNOS protein was absent in non–cytokine-treated β-cells, it was abundant after cytokine treatment of uninfected, AdGFP-infected (Fig. 1), or AdLuc-infected cells (data not shown). On the other hand, there was hardly any detectable iNOS protein expression in AdIκB(SA)2-infected cytokine-treated β-cells (Fig. 1). The iNOS enzymatic activity was estimated by measurements of medium nitrite (a stable product of nitric oxide [NO] oxidation) accumulation during a 72-h exposure to cytokines (50 and 1,000 units/ml for IL-1β and IFN-γ, respectively). Basal nitrite production, in the absence of cytokines, was low (means ± SE) and similar in all experimental groups under study (65 ± 20 pmol/103 cells × 72 h for noninfected control β-cells). In the presence of cytokines, nitrite production was similar in noninfected β-cells (222 ± 26 pmol/103 cells × 72 h, n = 4; P < 0.01 vs. respective control that was not exposed to cytokines, ANOVA) and in β-cells infected with AdLuc or AdGFP (241 ± 23 and 211 ± 24 pmol/103 cells × 72 h, respectively) but was reduced by more than twofold in β-cells overexpressing IκB(SA)2 (92 ± 29 pmol/103 cells × 72 h, n = 4; P < 0.01 vs. uninfected cells and AdLuc infected cells; P < 0.05 vs. AdGFP infected cells, ANOVA).
Finally, we evaluated whether cytokine-induced NF-κB expression is also required for the induction of β-cell death. β-Cell survival was not affected when cells were infected with AdLuc or AdIκB(SA)2 (Fig. 4). Expression of GFP, however, significantly decreased β-cell viability, mainly through the induction of apoptosis (Figs. 4A and B). GFP has recently been shown to modify survival and/or function of other cell types (28), raising doubts about the utility of this protein as a “nonfunctional control” for viral transfections or as a tool to sort viable β-cells. As previously described (25), IL-1β + IFN-γ caused the death of >50% of the exposed rat β-cells (Fig. 4A). β-Cell death was caused by apoptosis (Fig. 4B) as well as by necrosis (Fig. 4C). In regard to overexpression of IκB(SA)2-rescued β-cells from the cytokine-induced death pathway, the number of viable β-cells in the presence of cytokines was two- to fourfold higher in AdIκB(SA)2-infected β-cells compared with cells from the different control groups (Fig. 4A). These beneficial effects of the IκB superrepressor were due to inhibition of both cytokine-induced apoptosis (Fig. 4B) and necrosis (Fig. 4C).
To investigate the role of NF-κB in cytokine-induced β-cell gene expression and apoptosis, we infected purified rat β-cells with recombinant adenoviruses expressing the nondegradable AdIκB(SA)2, a super-repressor of NF-κB (13,14). Infection with AdIκB(SA)2 induced an at least fivefold higher expression of IκB(SA)2 compared with the endogenous IκB, and the protein was not degraded after the addition of cytokines. This confirms the resistance of NF-κB super-repressor to cytokine-induced phosphorylation and consequent degradation (13,14).
The adenovirus-mediated expression of the IκB super-repressor prevented cytokine-induced NF-κB translocation to the β-cell nucleus, as judged by cellular immunostaining for p65 and gel shift analysis. This led to inhibition of the expression of iNOS, Fas, and MnSOD mRNAs. AdIκB(SA)2 also prevented, to a large extent, cytokine-induced iNOS protein expression and NO release. We have previously suggested, based on studies with transient transfections with promoter-luciferase reporter constructs, that NF-κB activation is crucial for cytokine-induced iNOS (8), MnSOD (9), and Fas (8a) expression. For MnSOD, the reporter construct data were not confirmed by the use of nonspecific NF-κB chemical inhibitors (i.e., pyrrolidine dithiocarbamate [PDTC]). Thus, PDTC prevented IL-1β–induced iNOS, but not MnSOD, expression (7). The present observations, based on a specific NF-κB blocker (13,14), validate the promoter construct studies and confirm that NF-κB activation and nuclear translocation are necessary steps for cytokine-induced iNOS, MnSOD, and Fas mRNA expression in primary β-cells.
The inhibition of Fas gene expression is probably not of primary importance for the β-cell fate in our model system, because purified β-cells do not express a functional ligand protein for Fas (FasL) (29; D.L. and D.L.E., unpublished data). However, during insulitis in animal models of autoimmune diabetes, FasL-expressing mononuclear cells invade the islets and may induce apoptosis in β-cells expressing Fas (Fas is probably upregulated in these cells by the local production of cytokines) (30,31,32). It is noteworthy that Fas expression (32,33) and apoptosis (33) have also been reported in human β-cells in the early stages of type 1 diabetes. In this context, the blocking of NF-κB could have a beneficial effect by preventing upregulation of Fas expression in β-cells.
Prolonged exposure of rodent pancreatic β-cells to combinations of cytokines leads to cell death by both apoptosis and necrosis (20,25,27). Whereas cytokine-induced necrosis is probably NO-mediated, apoptosis seems mostly NO-independent in mouse and human β-cells (20,24). We have recently suggested that the “non-NO” signal for β-cell apoptosis is mediated by an intricate pattern of up- and downregulation of the expression of dozens of genes (4,5). As suggested by microarray analysis and promoter studies, several of these genes are regulated by NF-κB (5). We observed that AdIκB(SA)2 prevents both the necrotic and apoptotic components of cytokine-induced β-cell death. As discussed above, the inhibition of necrosis is probably related to a AdIκB(SA)2-induced decrease in iNOS expression and NO production. On the other hand, the observation that the IκB super-repressor also prevents β-cell apoptosis indicates that at least part of the pro-apoptotic gene expression pattern in these cells is regulated by NF-κB. To further investigate this issue, we intend to perform additional microarray analysis of gene expression in β-cells exposed to cytokines with or without previous infection with AdIκB or AdLuc.
After the submission of this article, another study was published indicating that an IκB repressor protects human islets against IL-1β–induced NO production, inhibition of insulin release (apparently mediated by NO), and Fas-triggered apoptosis (34). These results are apparently similar to the present observations. There are, however, some methodological differences between these two studies. We discriminated among living, apoptotic, and necrotic cells on the basis of their nuclear features and provided an absolute number of apoptotic/necrotic cells, whereas Giannoukakis et al. (34) evaluated cell death by measuring whole-islet caspase 3 activity. In addition, our experiments were performed on purified rat β-cells; thus, we could identify the nature of the cells undergoing apoptosis. Because Giannoukakis et al. (34) analyzed cell death in whole human islet preparations, it is unclear whether the cells apparently undergoing apoptosis in their experiments are indeed β-cells.
The present data indicate that the transcription factor NF-κB has a mostly pro-apoptotic role in β-cells exposed to cytokines. This in vitro data provide a rationale for targeting NF-κB by genetic intervention in attempts to prevent β-cell death in the early stages of insulitis and after islet transplantation. However, it needs to be remarked that the anti- and pro-apoptotic roles of NF-κB in different tissues may vary depending on the type of assault inflicted on the target cells (10). Furthermore, there may be differences between in vitro (as in the present study) and in vivo conditions. Thus, additional in vivo experiments are required to evaluate whether blocking β-cell NF-κB activation will indeed protect these cells against the complex immune response leading to β-cell death in islets transplanted in animal models of type 1 diabetes.
This work was supported by grants from the Juvenile Diabetes Foundation International, Fund for Scientific Research Flanders (FWO), and by a Shared Cost Action in Medical and Health Research of the European Community. H.H. was supported by a postdoctoral research fellowship (Fund for Scientific Research, Flanders, Belgium) and a Career Development Award (Juvenile Diabetes Foundation International). The technical assistance from the Diabetes Research Center personnel involved in β-cell purification is gratefully acknowledged.
Address correspondence and reprint requests to H. Heimberg, Diabetes Research Center, Vrije Universiteit Brussel, Laarbeeklaan 103, B-1090 Brussels, Belgium. E-mail: firstname.lastname@example.org.
H.H. and Y.H. contributed equally to this work.
Received for publication 6 October 2000 and accepted in revised form 13 July 2001.
ANOVA, analysis of variance; FasL, ligand protein for Fas; GAPDH, glyceraldehyde-phosphate dehydrogenase; GFP, green fluorescent protein; IFN-γ, γ-interferon IκB, inhibitory κB; IL-1β, interleukin-1β; iNOS, inducible nitric oxide synthase; MnSOD, manganese superoxide dismutase; MOI, multiplicity of infection; NO, nitric oxide; PCR, polymerase chain reaction; PI, propidium iodide.