Nuclear Factor-κB Regulates β-Cell Death: A Critical Role for A20 in β-Cell Protection

BALB/c and C57BL/6 mice were purchased from ARC (Perth, WA, Australia). NOD mice were purchased from The Walter and Eliza Hall Institute (Melbourne, VIC, Australia). Procedures complied with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.

Assays were performed using a modified crystal violet method (23,24). Briefly, β-insulinoma (β-TC₃) cells were pretreated (30 min) with 0.01 μmol/l actinomycin D (Calbiochem, Kilsyth, Australia) before the addition of 100, 500, or 1000 units/ml of recombinant mouse tumor necrosis factor (TNF)-α (R&D Systems, Minneapolis, MN). Twenty-four hours later, cell death was determined as described (23).

Human primary islets, provided by Prof. Phillip O’Connell (Westmead Hospital, Sydney, NSW, Australia), were isolated from cadaveric organ donors and cultured as described (25). Mouse islets were isolated and cultured as we have described (13,26). For cytokine stimulation, islets (150–200 islets/treatment) were stimulated with 200 units/ml of the appropriate cytokine for the indicated times. Human and mouse recombinant T-helper (Th) 1 cytokines interleukin (IL)-1β, TNF-α, or γ-interferon (IFN), and the Th2 cytokines IL-4, IL-13, or IL-15 (R&D Systems). β-TC₃ and Min6 cells were cultured as described (27). Cells were stimulated with 200 units/ml of mouse recombinant IL-1β or TNF-α for the indicated time points. In some cases islets were first pretreated with 100 μmol/l of the pharmacological NF-κB inhibitor, pyrrolidine dithiocarbamate (Sigma, St. Louis, MO) as described (28).

For cell sorting, primary mouse islets were first stimulated with cytokines as above and then washed (4% EDTA-PBS), and dispersed into single cells by a brief incubation with 0.05% trypsin-EDTA (Invitrogen). Dispersed islets were then washed free of trypsin and resuspended in 1% BSA-PBS and sorted in a fluorescence-activated cell sorter (FACS) based upon autofluorescence (FL-1) as described (29). Sorted cells were then analyzed for A20 expression as described below. The percentage of insulin-positive β-cells in the FL-1^Hi and FL-1^Lo sorted populations was determined by insulin (mouse anti-insulin; Zymed Laboratories, South San Francisco, CA) and glucagon (rabbit anti-glucagon; ICN Immunologicals, Lisle, IL) specific labeling on cytospun cells.

mRNA was extracted as described (26); we obtained ∼3–5 μg mRNA per 200 islets. cDNA was produced from 0.5 μg of RNA using the SuperScript III Reverse Transcriptase Kit (Invitrogen, Australia). Reactions were performed on the RG300 Real Time PCR System (Rotorgene), using SYBR green chemistry (JumpStart, Sigma, Castle Hill, Australia). A20 primers were human (F: AAAGCCCTCATCGACAGAAA; R: CAGTTGCCAGCGGAATTTA) and mouse (F: TGGTTCCAATTTTGCTCCTT; R: CGTTGATCAGGTGAGTCGTG). Human and mouse samples were normalized with glyceraldehyde-3-phosphate dehydrogenase (F: CACATCAAGAAGGTGGTG; R: TGTCATACCAGGAAATGA) or CPH2, respectively (F: TGGACCAAACACAAACGGTTCC; R: ACATTGCGAGCAGATGGGGTAG). Fold change was calculated following the 2ΔΔ^CT method.

Before hybridization, total RNA was amplified and labeled using the MessageAmp II aRNA Amplification Kit (Ambion, Austin, TX). Amplified RNA quality was evaluated by capillary electrophoresis using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Samples were hybridized on a custom-designed death-CHIP, a CustomArray 12K-Microarray platform (CombiMatrix, Mukilteo, WA). Genes were selected for inclusion on the microarray based upon their reported role in the regulation of cell death (Table 1). The microarray probe design and feature layout, hybridization and scanning protocol, data normalization, and quality control data will be reported elsewhere (D. Liuwantara, S.T. Grey, unpublished data). All data were first normalized and background corrected, then log (base 2) transformed. For each experiment, islets were isolated from 10 mice, pooled, and treated with cytokines, and RNA was harvested as described. Each experiment was performed on three (NOD) or four (BALB/c) independent islet preparations. All samples were run on duplicate arrays; 42 arrays were run in total. The mean coefficients of variation (CVs) based on replicate features on the microarrays were ≤20%.

Transfection of β-TC₃ cells, luciferase, and β-galactosidase (β-gal) activity was performed as described (21). The A20 reporter constructs were a kind gift of Dr. R. Dikstein (The Weizmann Institute of Science, Rehovot, Israel) (30). The expression plasmids encoding Fas-associated death domain (FADD), TNF receptor–associated factor (TRAF) 2, TRAF6, NF-κB–inducing kinase (NIK), P65, and protein kinase D (PKD) have been described (22,31,32). All data are expressed as relative light activity (21).

Islet transplantation into diabetic (blood glucose ≥300 mg/dl) recipients was performed essentially as described (13). Islets (isolated from three donor mice per recipient) from BALB/c mice were transplanted either into diabetic BALB/c recipients (syngeneic model) or into diabetic C57BL/6 mice (allogeneic model). To harvest the transplanted islets, nephrectomy was performed on postoperative day 5. The islet graft was collected by carefully peeling off the kidney capsule; mRNA was then isolated as described above. In some cases the graft was prepared for histological assessment as described (13).

Despite the importance of cell death in pathological β-cell loss, little is known about the genetic control of apoptosis in these cells. Cell death is a highly regulated event, controlled in a proinflammatory setting by an inducible set of antiapoptotic genes rapidly upregulated in response to this same inflammatory stress and/or injury (16–19). We were particularly interested in determining whether β-cells could also generate a rapid and inducible antiapoptotic response after exposure to inflammatory cytokines as has been described for other cell types (16–18). We first tested this concept by stimulating β-TC₃ cells with TNF-α in the presence or absence of the general transcription inhibitor, actinomycin D. Blockade of de novo gene transcription in combination with TNF-α stimulation sensitized β-TC₃ cells to cell death, whereas either treatment alone had minimal effect upon cell survival (Fig. 1). Cell survival decreased from 100% in control cells to 53% when cells were treated with 1000 units of TNF-α and actinomycin D (P ≤ 0.005). These data demonstrate that blockade of de novo gene transcription sensitizes β-cells to TNF-α–induced cell death, consistent with the notion that TNF-α triggers a proapoptotic death cascade that can be prevented by gene induction (16–18). Thus, β-cells do have a cytokine-dependent regulated antiapoptotic response as has been demonstrated for other cell types.

Having established that β-cells do have an inducible protective response, we next determined the molecular basis of this response. We used a custom-designed microarray (death-CHIP) to analyze the immediate early antiapoptotic gene response of primary rodent islets. Antiapoptotic genes from all known and putative antiapoptotic gene families were included on the death-CHIP (Table 1). cRNA was prepared from islets isolated from nondiabetic-prone BALB/c mice or diabetic-prone NOD mice. For each islet preparation, islets were either left untreated or stimulated with IL-1β or TNF-α for 1 h. These cytokines were chosen based on their established roles in inducing β-cell apoptosis both in vitro (21,33,34) and in vivo (35–37).

Of the 63 antiapoptotic genes represented on the death-CHIP, we found that islets constitutively expressed a broad selection of antiapoptotic genes over a range of expression levels (Table 1). To confirm the findings of the death-CHIP, the basal expression of selected genes (Table 1) was established by Western blot and/or PCR analysis in primary human and rodent islets and β-insulinoma cell lines (data not shown). Of the genes confirmed to be expressed in all three tissue samples, highly expressed genes (≥10,000 mean intensity units) including heat shock protein 70, HSPA1A/HSP70, and SOD1/CuZnSOD; moderately expressed genes (1–10,000 mean intensity units) including Heme oxygenase 1 and SOD2/MnSOD; and genes including the cellular inhibitors of apoptosis, BIRC2/c-IAP1, BIRC3/c-IAP2, BIRC4/XIAP, and the BCL family genes BCL2, BCL2L1/BCLXL, BCL2A1A/Bfl-1/A1, and TNFAIP3/A20 were expressed at relatively low constitutive basal levels (≤1,000 mean intensity units). In contrast, examination of the expression profile after cytokine stimulation indicated that islets have a relatively sparse immediate early antiapoptotic gene response (Fig. 2A and B). Establishing a twofold cutoff for statistical significance, only two genes were determined to be upregulated, namely, TNFAIP3/A20 and BIRC3/c-IAP-2, of which A20 was consistently the most highly regulated antiapoptotic gene (Fig. 2C). These data contrast islets with other tissues such as endothelial cells, which express a broad range of antiapoptotic genes under similar conditions (19,38,39).

As A20 was the most highly regulated immediate early antiapoptotic gene in islets, we focused on its regulation in more detail. Primary islets isolated from BALB/c or NOD mice were stimulated with the cytokines IL-1β or TNF-α for 1–8 h, and steady- state mRNA expression was analyzed by PCR. As shown in Fig. 3A, A20 was expression was markedly upregulated by ∼23-fold (P ≤ 0.005) and ∼16-fold (P ≤ 0.001) in BALB/c and NOD islets, respectively, within 1 h after IL-1β stimulation. Kinetic analysis of A20 expression revealed that although the expression declined thereafter, it was still maintained at significant levels. Thus, A20 is a cytokine-inducible immediate early response gene in islets, consistent with its regulation in other cell types (40). In comparison with IL-1β, TNF-α was a relatively poor inducer of A20 expression, stimulating only an approximately fourfold increase in A20 mRNA expression at 1 h for islets isolated from either BALB/c or NOD mice (P ≤ 0.05). This finding may relate to differential activation of NF-κB family members by IL-1β versus TNF-α, respectively, which can then lead to distinct patterns of gene expression (41). The poor induction of A20 by TNF-α is interesting, given the fact that TNF-α is one of the earliest β-cell–toxic cytokines detected in the islet infiltrate during the development of type 1 diabetes in NOD mice (36). However, Fig. 3B shows that there was no difference in A20 expression between the male and female NOD mice, demonstrating that the lower disease incidence in the male NOD mice is not associated with A20 expression.

We next examined A20 regulation in primary human islets. As shown in Fig. 3C, A20 mRNA expression was rapidly induced after IL-1β stimulation, reaching a maximal ∼25-fold induction at 1 h after stimulation and decreasing thereafter. Again IL-1β was a more potent inducer of A20 expression than TNF-α. In contrast with IL-1β and TNF-α, other Th1-type cytokines such as γ-IFN and IL-15 and the Th2-type cytokines IL-4 and IL-13 did not induce A20 expression in human or rodent islets (data not shown). Together, these data demonstrate that A20 is an immediate early response gene in both human and rodent islets, preferentially regulated by the Th1-type cytokines, IL-1β and TNF-α.

Islets of Langerhans are a heterogeneous tissue comprising not only insulin-secreting β-cells but also additional hormone-secreting cells including α- and γ-cells. β-Cells from freshly isolated islets are autofluorescent due to intracellular flavin adenine dinucleotide levels and can be FACS-purified on this basis (29). To determine the cellular source of A20 expression in intact primary islets, we next examined A20 expression in FACS-purified primary β-cells and non–β-cells (Fig. 4A). We found that inducible A20 expression was restricted to insulin-positive β-cells (approximately threefold induction P ≤ 0.05) as A20 expression was not induced by IL-1β in glucagon-positive cells (Fig. 4B). We also examined inducible A20 expression in the insulinoma cell line, Min6. Stimulation of Min6 cells with IL-1β or TNF-α resulted in a rapid and marked induction of A20, with kinetics similar to that seen for primary human and rodent islets (Fig. 4C). Thus, these data demonstrate β-cell–specific regulation of A20 expression within pancreatic islets.

We next began to address the mechanism(s) by which A20 was regulated in islets. Primary rodent islets were pretreated with actinomycin D to block de novo gene transcription and then stimulated with IL-1β for 1 h. Analysis of A20 steady-state mRNA revealed that induction of A20 was completely abolished (>90%, P ≤ 0.001) by actinomycin D treatment, indicating that in islets, cytokine-dependent induction of A20 is regulated at the level of de novo gene transcription (Fig. 5A). To confirm that A20 was regulated at the level of transcription, β-TC₃ cells were transfected with an A20 promoter sequence upstream of a luciferase reporter and stimulated with IL-1β or TNF-α. As shown in Fig. 5C, IL-1β and TNF-α induced approximately fourfold (P ≤ 0.001) and approximately twofold (P ≤ 0.01) increases in luciferase activity, respectively. Together these data demonstrate that A20 is regulated at the level of gene transcription in β-cells, and prevention of A20 expression by inhibition of gene transcription correlated with sensitization to TNF-α–induced apoptosis.

A20 has been described previously to be an NF-κB target gene and contains two NF-κB binding sequences in its promoter region (42). To determine whether or not the induction of A20 transcription in β-cells is NF-κB dependent, we transfected β-TC₃ cells with an A20 promoter construct in which the two NF-κB binding sites were deleted (ΔNF-κB) (Fig. 5B) and examined its activation in response to either IL-1β or TNF-α. As shown in Fig. 5C, compared with the wild-type promoter, both the cytokine-dependent and constitutive A20 promoter activities were completely abrogated (P ≤ 0.001) in the NF-κB–deleted reporter.

We next utilized an alternative approach to confirm the importance of NF-κB in the regulation of de novo transcription of the A20 promoter. We achieved this by overexpressing the p65/RelA subunit of NF-κB (39) in β-TC₃ cells and examined whether this change would result in induction of the A20 promoter. As shown in Fig. 6B, forced expression of p65/RelA resulted in a marked dose-dependent activation of the A20 promoter, showing an ∼19-fold increase (P ≤ 0.001) at the highest concentration tested. Importantly, forced expression of p65/RelA also resulted in a fivefold induction (P ≤ 0.005) in endogenous A20 mRNA levels, indicating that NF-κB was sufficient to drive the endogenous A20 promoter (Fig. 6A). Finally, inhibition of NF-kB activity using the chemical inhibitor, pyrrolidine dithiocarbamate (28), abolished IL-1β–stimulated A20 mRNA (>90%, P ≤ 0.01) expression in islets (Fig. 6C.) Together these data demonstrate that NF-κB is both necessary and sufficient to initiate transcriptional activation of the A20 promoter and drive expression of A20 mRNA. These data are consistent with our finding that A20 is regulated by the NF-κB–activating cytokines IL-1β and TNF-α, but not cytokines such as γ-IFN, IL-4, IL-13, and IL-15, which do not activate NF-κB.

NF-κB constitutes a family of transcription factors that together act as major integrators of the cellular inflammatory response (39,43). Activation of discrete NF-κB pathways through the Toll-like receptors (TLRs) or TNR receptors (TNF-Rs) or via free radicals results in the induction of a distinct sets of genes (39,41,44). To determine the important NF-κB pathways regulating A20 expression in β-cells, we utilized a transient transfection approach, cotransfecting β-TC₃ cells with the A20 reporter and specific kinases or adapter proteins, known to activate distinct NF-κB signaling pathways. TRAF6 is an adaptor protein required for NF-κB signaling through the TLR family, whereas TRAF2 mediates NF-κB signaling through the TNF-R family. As shown in Fig. 7A and B, forced expression of either TRAF6 or TRAF2 dose dependently activated the A20 promoter with ∼9-fold (P ≤ 0.001) and ∼6.5-fold (P ≤ 0.001) increases in reporter activity at the highest concentrations used, respectively. Thus, we demonstrate that A20 is a downstream target gene of the canonical NF-κB pathway, activated via ligation of TLR and TNF-R families. These data are consistent with our experiments demonstrating that increased A20 expression in IL-1β–(TLR-activating) and TNF-α–(TNF-R–activating) stimulated β-cells.

We next addressed whether A20 would be regulated in response to the noncanonical and free radical–dependent pathways. NIK is critical for mediating signaling through the noncanonical TNF-R pathway (39), whereas free radicals activate NF-κB through a PKD-dependent signaling pathway (44). To determine whether A20 is a target gene of these pathways, we cotransfected β-TC₃ cells with the A20 promoter and either NIK or PKD (Fig. 7C and D). Forced expression of NIK resulted in an ∼3.6-fold (P ≤ 0.005) activation of the A20 promoter, whereas PKD induced a more modest ∼2.6-fold (P ≤ 0.001) induction of reporter activity. These data demonstrate that A20 transcription can be induced by TNF-Rs that activate the noncanonical pathway and by free radicals. Together our data identify A20 as a major NF-κB target gene regulated by diverse NF-κB signaling pathways.

Transplanted islets are exposed to a hyperglycemic environment, nonspecific inflammatory reactions (i.e., including cytokines and free radicals), and immune-mediated rejection as well as stresses relating to hypoxia and lack of adequate nutrients (10,11). Many of these factors are able to induce NF-κB. Having identified A20 as a major NF-κB–regulated gene in β-cells, we hypothesized that A20 expression should be heavily regulated in islets in vivo after exposure to inflammatory stress. To assess this directly, we examined A20 expression levels in islets after transplantation into either diabetic syngeneic or diabetic allogeneic hosts. As shown in Fig. 8, already at day 5 after transplantation, there were significant approximately threefold (P ≤ 0.05) and approximately fourfold (P ≤ 0.05) increases in A20 steady-state mRNA expression in islets transplanted into either syngeneic or allogeneic recipients, respectively. These data demonstrate that transplanted islets do have a regulated antiapoptotic stress response and that A20 is a critical component of this stress response in vivo.

We demonstrated that blocking de novo gene transcription sensitizes β-cells to apoptosis and prevents A20 upregulation, indicating that loss of A20 expression may be partly responsible for the sensitization to TNF-α–induced apoptosis. To determine the importance of A20 in regulating β-cell apoptosis, we next asked whether A20 expression was sufficient to protect β-cells from TNF-α–induced apoptosis.

Activation of the TNF receptor by its cognate ligand results in the recruitment of FADD, which subsequently recruits and activates procaspase 8/FLICE, triggering apoptosis (1,39). For this experiment, β-TC₃ cells were transiently cotransfected with a FADD expression vector to induce apoptosis in the presence or absence of an A20 expression plasmid. Controls cells were transfected with a vector expressing a FADD dominant negative inhibitor (FADD-DN) or the empty vector pcDNA₃. All groups were transiently cotransfected with a cytomegalovirus-driven β-gal reporter, which was used to determine the percentage of viable cells for each group (22). As shown in Fig. 9, forced expression of FADD resulted in the rapid destruction of β-TC₃ cells as evidenced by the decrease in β-gal reporter activity (∼60%, P ≤ 0.005) compared with baseline levels observed in control groups (e.g., pcDNA₃ or FADD-DN), whereas expression of FADD-DN had little effect upon cell viability (P = 0.132). In contrast, A20 expressing cells were completely protected from the proapoptotic effect of FADD (P ≤ 0.05). These data demonstrate for the first time that A20 is sufficient to protect β-cells from TNF-α–induced apoptosis; thus, loss of A20 expression by transcriptional blockade sensitizes β-cells to apoptotic cell death.

Apoptotic β-cell death is an important pathological mechanism of β-cell loss in type 1 diabetes and islet graft rejection (5,6). Apoptosis is a highly regulated process, such that the decision to undergo apoptosis is dependent upon the balance between antiapoptotic and proapoptotic signals (19). In an inflammatory setting, cells are required to express a new or inducible set of antiapoptotic genes as a mechanism to counteract the physiological stresses that otherwise lead to cellular damage and apoptosis (19,39). Blockade of this regulated antiapoptotic response sensitizes many cell types to apoptotic death, underscoring the physiological relevance of this signaling network in cell biology (16–18).

In this present study we demonstrate that this paradigm is true for β-cells, as β-cells were sensitized to TNF-α–induced death by inhibition of de novo gene transcription. Our microarray-based approach revealed that TNFAIP3/A20 is one potential candidate gene providing a molecular basis for this protective response. A20 is a zinc finger–containing, immediate early-response gene with a potent antiapoptotic and anti-inflammatory function (20,40,45). This essential antiapoptotic and anti-inflammatory function of A20 is preserved in islets (13,21). Once expressed, A20 binds to and targets TNF receptor interacting protein for proteosomal degradation, thereby preventing TNF-α–induced NF-κB activation (46). The antiapoptotic mechanism of A20 is less well understood but may also involve inhibition of key proximal signaling events, as A20 inhibits activation of initiator caspases after death receptor ligation (22), consistent with our data demonstrating that A20 protects β-cells from FADD-induced apoptosis.

The transcription factor NF-κB regulates multiple proinflammatory genes that can contribute to islet destruction. NF-κB can promote T-cell–mediated killing and the generation of β-cell toxins through the induction of molecules such as Fas (47,48), inducible nitric oxide synthase (48,49) and cyclooxygenase-2 (50), respectively. In addition, the promoters of other proinflammatory genes induced in β-cells, including chemokines (i.e., monocyte chemoattractant protein-1) and adhesion molecules (i.e., intercellular adhesion molecule-1), also possess binding elements for NF-κB (43). The importance of NF-κB in β-cell inflammatory responses is underscored by the fact that blockade of NF-κB in in vitro models, by means of an inhibitor of NFκB (IκBα) super-repressor or via A20 overexpression, prevents IL-1β–and γ-IFN–induced β-cell dysfunction and death (21,48,49,51). Collectively these data implicate NF-κB as a major player in cytokine-dependent β-cell dysfunction and indicate that targeting NF-κB may have therapeutic utility as a means to improve islet function in the face of inflammatory insults.

However, aside from its important proinflammatory role, NF-κB is a major regulator of cellular apoptosis through its ability to control the expression of multiple antiapoptotic genes including the c-IAPs, caspase-8-c-FLIP, Bfl-1/A1 (39), and A20 (20,42). In cells that lack the NF-κB family member p65/RelA, exposure to TNF-α will induce apoptosis (16). Furthermore, blockade of NF-κB activation, by nonspecifically inhibiting transcription with actinomycin D (52) or specifically via the use of an IκBα super-repressor (17,18), also sensitizes cells to TNF-α–mediated apoptosis. Importantly blockade of NF-κB in β-cells by means of an IκBα super-repressor (53) or via actinomycin D as we demonstrate here, also sensitizes cells to TNF-α–dependent apoptotic death. We propose that this sensitization was due to blockade of NF-κB activation and, at least in part, the subsequent loss of A20 expression. Consistent with this, prevention of A20 induction in vitro sensitizes endothelial cells to TNF-dependent apoptosis (20). Thus in β-cells, as has been demonstrated for other cell types, NF-κB activation governs both proinflammatory responses and protection from apoptosis (39). Therefore, targeting NF-κB activation in an in vivo setting in which multiple factors are at play may sensitize β-cells to apoptosis (53). In light of this, we propose that more sophisticated approaches than mere blanket blockade of NF-κB are warranted if one aims to prevent the unwanted proinflammatory effects of NF-κB activation without incurring the associated risk of sensitizing β-cells to apoptotic death. Alternative strategies might include expressing genes such as A20 (13,20) or targeting specific upstream regulators of NF-κB activation (54).

In conclusion, we demonstrate that primary islets upregulated a relatively small set of antiapoptotic genes in response to inflammatory stress, of which TNFAIP3/A20 was the most highly regulated. Here we demonstrate that A20 is regulated at the level of gene transcription in pancreatic β-cells, under the control of the transcription factor NF-κB; thus, tightly linking islet proinflammatory gene responses with protection from apoptosis. Together with our previous studies demonstrating an anti-inflammatory and antiapoptotic function for A20 in islets (13,21), these present data indicate that A20 is a critical component of the islet-regulated response to inflammatory stress and injury. Thus, pathologic loss of A20 expression may render β-cells susceptible to apoptotic death; conversely, enhancing A20 expression in β-cells may improve their survival in the face of inflammatory and autoimmune insults. Importantly, our data indicate that blockade of NF-κB, as a means to prevent islet inflammatory responses in vivo, may have the unwanted side effect of sensitizing β-cells to apoptosis by preventing the upregulation of antiapoptotic genes such as A20 (16,17,20).

This work was supported by the Juvenile Diabetes Foundation International (JDRF 2-2000-719) and a BioFirst Award from the New South Wales Ministry for Science and Medical Research (to S.T.G.).

The authors thank Drs. J. Gunton and R. Laybutt for critical reading of this manuscript and T. Shukri, L. Choi, J. Chan, Jerome Darakdjian, and the Biological Testing Facility staff for technical assistance.