Cytokines Tumor Necrosis Factor-α and Interferon-γ Induce Pancreatic β-Cell Apoptosis through STAT1-mediated Bim Protein Activation*

Type 1 diabetes is characterized by local inflammation (insulitis) in the pancreatic islets causing β-cell loss. The mitochondrial pathway of apoptosis is regulated by the balance and interaction between Bcl-2 members. Here we clarify the molecular mechanism of β-cell death triggered by the pro-inflammatory cytokines tumor necrosis factor (TNF)-α and interferon (IFN)-γ. The combination of TNF-α + IFN-γ induced DP5, p53 up-regulated modulator of apoptosis (PUMA), and Bim expression in human islets and rodent β-cells. DP5 and PUMA inactivation by RNA interference partially protected against TNF-α + IFN-γ-induced β-cell apoptosis. DP5 knock-out mice had increased β-cell area, and isolated islets from these mice were resistant to cytokine exposure. Bim expression was transcriptionally regulated by STAT1, and its activation triggered cleavage of caspases. Silencing of Bim protected rodent and human β-cells to a large extent against TNF-α + IFN-γ, indicating a major role of this BH3-only activator protein in the mechanism of apoptosis. Our data support a highly regulated and context-dependent modulation of specific Bcl-2 members controlling the mitochondrial pathway of β-cell apoptosis during insulitis.

Immune-mediated pancreatic ␤-cell loss, mainly due to apoptosis, is a hallmark of type 1 diabetes (T1D) 5 (1,2). In the early stages of the disease, infiltrating macrophages and T-cells release pro-inflammatory cytokines such as interleukin (IL)-1␤, tumor necrosis factor (TNF)-␣, and interferon (IFN)-␥, which, together with cell-to-cell death effectors (granzyme B, FasL, etc.), contribute to the induction of ␤-cell apoptosis and the buildup of insulitis (1)(2)(3)(4). In vitro studies demonstrated that the combination of different pro-inflammatory cytokines, but not each of them alone, activates ␤-cell apoptosis (1). It is conceivable that the cytokine combination and distribution in the vicinity of the ␤-cells vary during T1D development (3). The individual genetic background, immune assault timing, and degree of islet infiltration may also affect cytokine composition during insulitis. Therefore, a clear understanding of the apoptotic ␤-cell pathways activated downstream of different cytokine combinations is needed to individualize therapies aiming to prevent ␤-cell destruction in T1D.
Apoptosis was originally described as a physiological mechanism of cell death that allows cell turnover and tissue reorganization, but later evidence indicated that it is also an important mechanism of cell demise during viral infection and autoimmune diseases (5,6). Different protein modulators, effectors, and pathways regulate the decision to undergo apoptotic cell death (6). Apoptosis can be activated by two major mechanisms: the extrinsic and intrinsic pathways (6). The extrinsic pathway is characterized by engagement of death receptors and caspase-8 cleavage/activation. In the second mechanism (intrinsic), the mitochondria play a key role in the triggering of cell death. Transcriptional and post-transcriptional modulation and protein-protein interaction of Bcl-2 members determine the cell outcome in this pathway (7)(8)(9). After an apoptotic stimulus, the sensitizer Bcl-2 proteins (DP5, Bik, Bad, and/or Noxa) are transcriptionally or post-transcriptionally activated and interact through their Bcl-2 homology 3 (BH3) domain with the anti-apoptotic Bcl-2 members (Bcl-2, Bcl-XL, Bcl-W, and A1). This interaction releases BH3-only activator proteins (Bid, Bim, and/or PUMA) that directly bind and induce conformational changes in the multichannel pro-death proteins Bax and Bak (9 -11). Activated Bax translocates from the cytosol to the mitochondria and together with Bak forms pores in the mitochondrial membrane, releasing pro-apoptotic proteins such as cytochrome c to the cytoplasm (10,12) and/or apoptosis-inducing factor (AIF) to the nucleus (13). Once in the cytoplasm, cytochrome c interacts with apoptotic protease activat-ing factor to form the apoptosome, leading to pro-caspase cleavage and activation and subsequent cell death (14).
We have recently shown that the pro-inflammatory cytokines IL-1␤ ϩ IFN-␥ induce the BH3-only sensitizer DP5 and the BH3-only activator PUMA, leading to ␤-cell death (9,15,16). Less is known about the pathway of apoptosis and Bcl-2 proteins modulated by TNF-␣ ϩ IFN-␥. It was previously described that mouse islets lacking the transcription factor STAT1 are resistant to TNF-␣ ϩ IFN-␥-induced apoptosis (17), but the downstream molecular effector(s) remain(s) unknown. Against this background, we presently performed an extensive study using human islets, DP5 knock-out mice, rat fluorescence-activated cell sorting (FACS)-purified primary ␤-cells, and INS-1E cells to clarify the mechanisms underlying TNF-␣ ϩ IFN-␥-induced ␤-cell demise. The findings obtained indicate that TNF-␣ ϩ IFN-␥ utilize the BH3-only activator Bim as a key pro-apoptotic effector downstream of STAT1 induction, suggesting a characteristic modulation of Bcl-2 pathways by different inflammatory mediators.

EXPERIMENTAL PROCEDURES
Cell Culture and Treatment-Human islets were isolated in Pisa (Italy) from non-diabetic organ donors, with the approval of the local ethical committee. The islets were isolated by enzymatic digestion and density gradient purification (18) and cultured in M199 medium containing 5.5 mM glucose. Donor age was 57 Ϯ 14 years, and the preparations contained 50 Ϯ 16% ␤-cells (n ϭ 10), as evaluated by immune staining for insulin (methods as in Ref. 19). Primary islets were also isolated from adult Wistar rats (Charles River Laboratories Belgium, Brussels, Belgium), DP5 knock-out mice previously generated on 129SV/C57BL/6 mixed background (20), or wild type littermate mice. Animals were bred and used in accordance with the guidelines of the Belgian Regulations for Animal Welfare or with the Canadian Council of Animal Care guidelines. All experimental protocols used were approved by the Ethical Committee for Animal Experiments of the ULB or by the Animal Care Committee of Mount Sinai Hospital. Primary rat ␤-cell isolation and culture were carried out as published previously (21,22). Islet isolation was achieved by collagenase digestion, and primary rat ␤-cells were obtained after hand picking of islets using a stereomicroscope, islet dispersion, and FACS sorting (FACSAria, BD Biosciences). Sorted cells were plated on polylysine-coated dishes and precultured in Ham's F-10 (Invitrogen, Paisley, UK) medium for 48 h before any further experimental procedures. Mouse islet isolation was performed as described previously (23). The insulin-producing INS-1E cell line (a kind gift from Professor C. Wollheim, Centre Médical Universitaire, Geneva, Switzerland) was cultured in RPMI 1640 medium (Invitrogen) supplemented with 5% fetal calf serum (FCS), 0.1 mM sodium pyruvate, 1 mM Hepes, and 0.5 M 2-mercaptoethanol (Invitrogen) (24). Recombinant rat (100 units/ml for INS-1E cells and 500 units/ml for primary rat ␤-cell), mouse (500 units/ml), or human (500 units/ml) IFN-␥ (R&D Systems, Abingdon, UK), recombinant murine TNF-␣ (1000 units/ml, Innogenetics, Gent, Belgium), human recombinant IL-1␤ (50 units/ml for mouse islets and 10 units/ml for INS-1E cells, a kind gift from Dr. C. W. Reynolds, NCI, National Institutes of Health, Bethesda, MD), and IL-1 receptor antagonist (IL-1ra; 1000 units/ml, R&D Systems) were used. Cytokine concentrations were selected based on our previous time course and dose-response studies (1,24).
Western Blot-After cell culture and treatment, cells were lysed using the Laemmli buffer, and total protein extracts were extracted and resolved by SDS-PAGE, transferred unto a nitrocellulose membrane, and immunoblotted with the antibodies indicated in supplemental Table S1. The proteins were revealed with a secondary anti-rabbit or anti-mouse horseradish perox-idase-labeled anti-IgG (1/5000, Lucron Bioproducts, De Pinte, Belgium). The protein bands were visualized using an enhanced chemiluminescence (ECL) kit (SuperSignal West Femto, Thermo Scientific, Pierce) and quantified using the Aida1D analysis software (Fujifilm, London, UK). The intensity values for the proteins were corrected by the values of the housekeeping protein ␣-tubulin and are shown as -fold induction versus the control sample (considered as 1).
Cell Viability Assay-The percentage of ␤-cell death was determined after a 15-min incubation with the DNA binding dyes propidium iodide (PI, 5 g/ml; Sigma) and Hoechst 33342 (HO, 10 g/ml; Sigma) (19,24,27,28). A minimum of 500 cells was counted in each experimental condition. Viability was eval- uated by two independent observers (one being blind to sample identity). The agreement between findings obtained by the two observers was 90%. Results are expressed as the percentage of dead cells, calculated as the number of apoptotic plus necrotic cells/total number of cells counted ϫ 100. Apoptosis was also measured by an enzyme-linked immunosorbent assay (ELISA), using a cytoplasmic histone-associated DNA fragment detection kit (Roche Diagnostics, Vilvoorde, Belgium). Results are expressed as arbitrary units of optical densities. Apoptosis was confirmed by additional methods including mitochondrial Bax translocation, cytochrome c release, and caspase-3 and -9 cleavage as suggested (29).
Real Time RT-PCR-After exposure to cytokines, INS-1E cells, rat primary ␤-cells, and human islets were lysed and harvested; poly(A) mRNA was isolated using the Dynabeads mRNA DIRECT kit (Invitrogen) and reverse-transcribed as described previously (22,24). Quantitative RT-PCR was done by using SYBR Green fluorescence on a LightCycler machine (Roche Diagnostics, Manheim, Germany), and val-ues were compared with a standard curve. Expression values of the genes of interest were corrected by the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and were normalized by the control value in the experimental series considered as 1. GAPDH expression is not modified by the present cytokine combinations in INS-1E cells (27) or when compared with the stable housekeeping genes OAZ1 and SRP14 (30) in primary ␤-cells or human islets (supplemental Fig. S1, A and B). Primer sequences for rat and human PUMA, Bim, DP5, OAZ1, SRP14, and GAPDH and rat Bcl-XL, Bcl-2, STAT1, and PTPN2 are described in supplemental Table S3.
Immunofluorescence-␤-Cells were plated on polylysinecoated coverslips and treated with cytokines for at least 48 h. The cells were then fixed with 4% paraformaldehyde and permeabilized with 0.3% Triton X-100. The fixed cells were incubated overnight at 4°C with the following primary antibodies:   DP5 2), and cells were allowed to recover for 36 h; cytokines were then added, and cell death was measured by HO/PI after 16 h of cytokine addition. Data shown are means Ϯ S.E. of 5 independent experiments. ANOVA followed by paired t test with the Bonferroni correction: *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. E, immunostaining results for insulin in pancreas sections from wild type and DP5 knock-out mice were summed using an optimized positive pixel count algorithm and normalized per total pancreas area (square millimeters) analyzed per mouse (n ϭ 4). Two-tailed paired t test: *, p Ͻ 0.05. F, islets from DP5 knock-out or wild type mice were exposed to the cytokine combination TNF-␣ ϩ IFN-␥ for 48 h. The viability of islet cell preparations was assessed by HO/PI. Cell death was counted in at least 24 islets derived from four mice per condition. Bar, 60 m. Two-tailed paired t test: **, p Ͻ 0.01. Data shown are representative of 3-5 independent experiments. Chromatin Immunoprecipitation (ChIP) Assay-The ChIP assay was performed as described (15). Briefly, extracts were precleared by a 1-h incubation with protein A/herring sperm DNA at 4°C. Samples were incubated overnight at 4°C with anti-STAT1 antibody (Santa Cruz Biotechnology, 1:200) or preimmune goat serum as negative control. Immunocomplexes were conjugated with protein A/herring sperm DNA and washed. Samples were then eluted, proteinase K-treated, and incubated overnight at 65°C in high salt solution to reverse the cross-link reaction. DNA fragments were analyzed by standard PCR, and input DNA was analyzed simultaneously for normalization. PCR was performed with the following primer pair for the STAT1 binding site Ϫ686 to Ϫ385 of the Bim promoter: 5Ј-TCCCACCACCAGCTCTGCAC-3Ј and 5Ј-GCTCCCT-TGGTTTGCGGAGC-3Ј.

Bim Modulates Apoptosis in Pancreatic ␤-Cells
Statistical Analysis-Data are expressed as means Ϯ S.E. of the indicated number of independent experiments. A significant difference between experimental conditions was assessed by a two-tailed paired t test or ANOVA followed by paired t test with the Bonferroni correction. p values Ͻ 0.05 were considered statistically significant.

The Cytokine Combination TNF-␣ ϩ IFN-␥ Triggers ␤-Cell
Apoptosis through the Mitochondrial Pathway of Cell Death-␤-Cell exposure to TNF-␣ ϩ IFN-␥ induced mitochondrial Bax translocation, cytochrome c release, and DNA fragmentation (Fig. 1, A and B; supplemental Fig. S2, A-D), but not nuclear AIF translocation (Fig. 1A; supplemental Fig. S2A). In primary rat ␤-cells and INS-1E cells, AIF is anchored to the inner mitochondrial membrane and needs to be actively processed to translocate to the nucleus (13); this process occurs independently of mitochondrial Bax translocation and cytochrome c release. Consistently with cytochrome c release from the mitochondria, TNF-␣ ϩ IFN-␥ triggered the cleavage and activation of caspases -9, -7, and -3 (Fig. 1C). Total pro-caspase-7 protein expression was up-regulated after cytokine combination (Fig.  1C), but the ratio of cleaved/pro-caspase 7 increased by 5-7fold after 16 -24 h (p Ͻ 0.05; data not shown). Taken together, these results suggest that the intrinsic pathway of apoptosis is activated in ␤-cells following exposure to TNF-␣ ϩ IFN-␥. Importantly, we confirmed that TNF-␣ combined with IFN-␥ induces cleavage of caspase-3 and triggers apoptosis in dispersed human islets (Fig. 2, A and B; supplemental Fig. S2E). This apoptotic effect of TNF-␣ ϩ IFN-␥ was not prevented by IL-1ra in INS-1E cells. Conversely, IL-1ra protected against IL-1␤ ϩ IFN-␥-induced cell death (Fig. 2C). Thus, TNF-␣ ϩ IFN-␥ directly induce apoptosis in our experimental model, without the need of IL-1␤ production by the target cells (31).
PUMA, a BH3-only Activator, Participates in TNF-␣ ϩ IFN-␥-induced ␤-Cell Death-PUMA plays a major role for pancreatic ␤-cell apoptosis in the context of IL-1␤ ϩ IFN-␥ exposure (16). PUMA transcript expression was induced by ϳ6-fold in human islet preparations, primary rat ␤-cells, and INS-1E cells following treatment with TNF-␣ ϩ IFN-␥ (Fig. 4, A-C). Upregulation of PUMA protein was demonstrated in INS-1E cells (Fig. 4D). SMARTpool (a combination of four different siRNAs) and a single sequence siRNA were used to silence PUMA, achieving a 70% reduction in mRNA expression as measured by real time RT PCR (Fig. 4E) and a 50% decrease in protein level as observed by Western blot analysis (Fig. 4F). PUMA knockdown protected INS-1E cells against TNF-␣ ϩ IFN-␥-induced apoptosis (Fig. 4G). Simultaneous knockdown of both DP5 and PUMA failed to fully protect INS-1E cells against cytokines (Fig. 5). Together, these data demonstrate that DP5 and PUMA contribute to the control of ␤-cell death induced by TNF-␣ ϩ IFN-␥. The partial protective effect obtained by knocking down these two proteins, however, sug-
The BH3-only Activator Bim Is Transcriptionally Activated by STAT1 in TNF-␣ ϩ IFN-␥-exposed ␤-Cells-Bim, together with PUMA, has been previously reported to play a key role in ␤-cell apoptosis induced by high glucose (33). On the other hand, Bim translation is not modulated by IL-1␤ ϩ IFN-␥ in ␤-cells (16). We next evaluated whether Bim is activated in the context of TNF-␣ ϩ IFN-␥. Bim is mainly localized at the mitochondria of the ␤-cells (supplemental Fig. S5A), and protein localization is not affected in INS-1E cells after TNF-␣ ϩ IFN-␥ exposure (data not shown). Treatment of islets derived from human donors, primary rat ␤-cells, and INS-1E cells with TNF-␣ ϩ IFN-␥ resulted in up-regulation of total Bim transcript level (Fig. 6, A-C). Thus, we analyzed by Western blot the different protein isoforms generated by alternative mRNA splicing of Bim (BimEL, BimL, and BimS). Among the three isoforms, BimEL and BimL were significantly induced from 8 to 24 h in INS-1E cells after cytokine treatment ( Fig. 6D; supplemental Fig. S5B). IFN-␥ alone up-regulated Bim mRNA expression to the same extent as TNF-␣ ϩ IFN-␥ (supplemental Fig.  S5C). The cytokine IFN-␥ activates signal transduction pathways that involve the tyrosine Janus kinases JAK1 and JAK2, which phosphorylate and induce the dimerization of the transcription factor STAT1 (34). To evaluate the role of STAT1 for Bim induction, we knocked down STAT1 by using two different siRNAs, achieving more than 60% silencing in INS-1E cells as demonstrated by mRNA and protein analysis (Fig. 6, E and G; supplemental  Fig. S6B). ChIP analysis confirmed STAT1 direct binding to the Bim promoter in positions Ϫ686 to Ϫ385; this region harbors two STAT1 binding sites, TTCtacGAA and TTCttgGAA (Fig. 6I). The phosphatase PTPN2, a candidate gene for T1D (35), was previously shown to down-regulate IFN-␥ signal transduction in ␤-cells (22). PTPN2 knockdown increased phosphorylation and activation of STAT1 in ␤-cells (22) and, in line with the above observations, increased cytokine-induced Bim expression (supplemental Fig. S6C). Interestingly, Bim protein expression does not change after IL-1␤ ϩ IFN-␥ exposure (supplemental Fig. S7A) (16), although Bim transcript is induced by this cytokine combination at a near similar level as with TNF-␣ ϩ IFN-␥ (supplemental Fig. S7B). IL-1␤ ϩ IFN-␥ trigger severe ER stress in ␤-cells (15), whereas TNF-␣ ϩ IFN-␥ did not increase expression of the ER stress marker XBP-1s (supplemental Fig. S7C) and only slightly up-regulated Chop transcript level (supplemental Fig. S7D). Because ER stress inhibits protein translation (36), we next used two different chemical ER stressors, namely cyclopiazonic acid and tunicamycin, to test whether ER stress affects Bim protein up-regulation. The two chemical ER stressors prevented TNF-␣ ϩ IFN-␥-induced Bim protein expression in INS-1E cells (supplemental Fig. S7E), sug-gesting that severe ER stress may explain why IL-1␤ ϩ IFN-␥ fail to induce Bim translation.
Bim Activation Is a Major Event for Apoptosis Induced by TNF-␣ ϩ IFN-␥ in ␤-Cells-Bim was successfully silenced at both mRNA and protein levels by using two different siRNAs in INS-1E cells (Fig. 7, A and B). Bim knockdown resulted in 70 -80% down-regulation of cleaved caspase-9 and -3 after cytokine treatment (Fig. 7C). Inactivation of Bim consistently prevented TNF-␣ ϩ IFN-␥-induced cell death in INS-1E cells ( Fig. 7D; supplemental Fig. S8A). The protective effect of Bim inactivation was significantly higher for the cytokine combination TNF-␣ ϩ IFN-␥ than for IL-1␤ ϩ IFN-␥ (supplemental Fig. S7F). Importantly, primary rat ␤-cells and dispersed human islets were also protected against TNF-␣ ϩ IFN-␥-induced apoptosis by Bim knockdown (Fig. 7, E and F; supplemental Fig. S8,  B-D). Parallel knockdown of DP5 and Bim did not increase the anti-apoptotic effects of Bim alone (supplemental Fig. S9). In other cell types, Bim binds to and is inhibited by the anti-apoptotic proteins Bcl-2 and Bcl-XL (11). We thus characterized the role of these pro-survival Bcl-2 members in the context of TNF-␣ ϩ IFN-␥-induced ␤-cell apoptosis. Bcl-2 mRNA and protein expression was not modulated by TNF-␣ ϩ IFN-␥ (supplemental Fig. S10, A and B). Bcl-XL, on the other hand, was decreased at both mRNA and protein levels (Fig. 8, A and B). This effect was TNF-␣-dependent because this cytokine alone down-regulated Bcl-XL (supplemental Fig. S11). To address the role of Bcl-XL inactivation, we silenced this protein by using two different siRNAs (Fig. 8C). Bcl-XL knockdown increased ␤-cell death in both basal and cytokine-treated conditions (Fig.  8D). Interestingly, simultaneous knockdown of Bcl-XL and Bim reversed the lethal effect of Bcl-XL silencing (Fig. 8E). These results suggest that both Bim up-regulation (by IFN-␥) and inactivation of Bcl-XL (by TNF-␣) contribute to Bim release and activation, a critical step for TNF-␣ ϩ IFN-␥-mediated ␤-cell death.

DISCUSSION
The composition of and interaction between pro-inflammatory cytokines probably vary during the evolution of insulitis in early T1D (3). This might explain the different level of protection achieved by the blockage of TNF-␣ or IL-1␤ in rodent models of autoimmune diabetes (1,37). Therefore, the understanding of apoptotic networks induced downstream of each  cytokine combination might contribute to future individualized therapies. We presently demonstrate that activation of the BH3-only protein Bim is critical for TNF-␣ ϩ IFN-␥-induced ␤-cell death.
TNF-␣ combined with IFN-␥ induces DP5 expression in human islets and in rat primary and clonal ␤-cells. DP5 participates in ␤-cell apoptosis, as demonstrated by the observation that its knockdown partially prevents cell death induced by TNF-␣ ϩ IFN-␥. Islets derived from DP5 knock-out mice are also protected against this cytokine combination. Importantly, DP5 inactivation does not affect insulin secretion, as was previously reported with the BH3-only sensitizer Bad (38). The group of BH3-only sensitizer proteins is composed of DP5, Bad, Noxa, Bik, and other Bcl-2 members (9, 11). Bad was described to be activated by calcineurin in TNF-␣ ϩ IFN-␥-treated MIN6N8 cells (39). We also evaluated the putative response of Noxa and Bik in ␤-cells, but these proteins were not induced by TNF-␣ ϩ IFN-␥ (data not shown). On the other hand, TNF-␣ ϩ IFN-␥ induce transcriptional down-regulation of the anti-apoptotic Bcl-XL (present data), which may contribute (together with DP5 overexpression and Bad dephosphorylation) to "sensitize" ␤-cells to apoptosis. Inactivation of Bcl-XL is dependent on TNF-␣ alone, and this phenomenon was also observed in other cell types (40). The anti-apoptotic protein Bcl-2 is expressed at low levels in human islets (41) and was not modulated in INS-1E cells by TNF-␣ ϩ IFN-␥ (present data).
PUMA and Bim belong to the BH3-only activator subgroup and must be released from anti-apoptotic Bcl-2 proteins to induce Bax and Bak conformational changes, mitochondrial pore formation, and apoptosis (9,10). We have previously shown that IL-1␤ combined with IFN-␥ activates an early induction of NF-B, which is responsible for PUMA up-regulation (16). TNF-␣ ϩ IFN-␥ induce PUMA mRNA and protein expression, and PUMA inactivation partially decreases cytokine-induced ␤-cell apoptosis. Nevertheless, double knockdown of DP5 and PUMA fails to fully protect ␤-cells against TNF-␣ ϩ IFN-␥, suggesting that an additional BH3-only activator protein is involved in the pathway of apoptosis. Herein we identified Bim as a key regulator of ␤-cell death induced by TNF-␣ ϩ IFN-␥. Thus, inactivation of Bim by siRNAs prevents to a large extent TNF-␣ ϩ IFN-␥-mediated ␤-cell apoptosis, as shown by caspase-9 and -3 cleavage, cytoplasmic histone-associated DNA fragment detection, and HO/PI staining. We observed that inactivation of Bcl-XL induced ␤-cell death, but this was reversed by concomitant knockdown of the downstream pro-apoptotic protein Bim. Furthermore, knockdown of both DP5 and Bim did not augment the protection provided by Bim knockdown alone. Together, these observations are compatible with the concept of a direct or hierarchical pathway of apoptosis (9 -11), where DP5 acts as an upstream sensitizer inhibiting Bcl-XL and thus releasing the BH3-only activator Bim. Bim has a less marked protective effect in ␤-cell death induced by IL-1␤ ϩ IFN-␥ and is not translationally induced by this cytokine combination, suggesting that different pro-inflammatory cytokines induce a context-dependent modulation of BH3-only proteins in ␤-cells. IL-1␤ combined with IFN-␥ triggers ER stress via NO formation, leading to inhibition of protein translation (42), whereas chemical ER stressors prevent Bim protein induction by TNF-␣ ϩ IFN-␥ (present data). It is conceivable that activation of severe ER stress by IL-1␤ ϩ IFN-␥, but not by TNF-␣ ϩ IFN-␥, explains why IL-1␤ ϩ IFN-␥ fail to induce Bim protein expression despite the fact that both cytokine combinations up-regulate Bim mRNA levels. In other words, TNF-␣ ϩ IFN-␥ induce a mild ER stress, which enables Bim translation and favors this particular pathway of apoptosis, whereas IL-1␤ ϩ IFN-␥ induce a more pronounced ER stress that inhibits the protein synthesis of Bim, allowing alternative pathways of cell death to prevail. It has been previously reported that ER stress induces apoptosis via up-regulation of Bim in thymocytes, macrophages, and kidney and breast epithelial cells (43). We did not observe, however, Bim protein induction following ␤-cell exposure to chemical ER stressors, indicating that the pathways by which ER stress leads to apoptosis might be cell-specific.
As described above, Bim is induced at the transcriptional level following cytokine exposure. The transcription factor STAT1 has a major role in the apoptotic mechanism induced by cytokines in ␤-cells (17,26,44,45). This transcription factor is activated by IFN-␥ downstream of JAK. Knockdown of STAT1 by two different siRNAs prevents Bim mRNA and protein induction, whereas silencing of PTPN2, a phosphatase that provides a negative feedback on STAT1 activation (22), enhances Bim expression. In line with these results, STAT1-dependent Bim activation was also suggested in chronic lymphocytic leukemia cells after IL-21 exposure (46). We propose, therefore, that STAT1-mediated Bim overexpression is a key mechanism by which IFN-␥ contributes to ␤-cell apoptosis. These intriguing observations provide a putative link between a candidate gene for T1D (PTPN2), a cytokine-induced transcription factor (STAT1), and an effector BH3-only activator protein (Bim).
In conclusion, we demonstrate that TNF-␣ combined with IFN-␥ induces Bim protein expression. This event, together with DP5 and PUMA up-regulation and Bcl-XL inactivation, is critical for pancreatic ␤-cell death following exposure to the pro-inflammatory cytokines TNF-␣ and IFN-␥. Our findings suggest the novel concept that induction of ␤-cell apoptosis by different cytokine combinations (i.e. IL-1␤ ϩ IFN-␥ or TNF-␣ ϩ IFN-␥) have a characteristic "signature" of ␤-cell death. This knowledge may contribute toward the future development of individualized therapies to protect ␤-cells in the early stages of T1D.