Death Protein 5 and p53-Upregulated Modulator of Apoptosis Mediate the Endoplasmic Reticulum Stress–Mitochondrial Dialog Triggering Lipotoxic Rodent and Human β-Cell Apoptosis

The culture of INS-1E and primary fluorescence activated cell sorter–purified rat β-cells is described in the Supplementary Data.

Human islets (from 4 donors aged 63 ± 7 years, BMI 25 ± 1 kg/m², cause of death cerebral hemorrhage) were isolated by collagenase digestion and density gradient purification. The islets were cultured, dispersed, and transfected as described previously (24). The percentage of β-cells, examined by insulin immunofluorescence (9), was 69 ± 11%.

For FFA exposure, primary β-cells were cultured in medium with 1% BSA without FBS, and INS-1E cells were cultured in medium with 1% FBS and 1% BSA (5). Oleate and palmitate (Sigma, Schnelldorf, Germany) were dissolved in 90% ethanol (6) and used at a final concentration of 0.5 mmol/L, which in the presence of 1% BSA results in unbound FFA concentrations in the nanomolar range (5). The chemical JNK inhibitor SP600125 (Sigma) and the peptide JNK inhibitor L-TAT-JNKi were used at 10 μmol/L for 2 h before and during FFA exposure (9). The IRE1 inhibitor 4μ8C was used at 25 μmol/L during FFA exposure (25).

Details on the microarray analysis, real-time PCR primers, RNA interference, chromatin immunoprecipitation (ChIP), promoter reporter assay, and antibodies are provided in the Supplementary Data.

Apoptotic β-cells and INS-1E cells were counted in fluorescence microscopy after staining with the DNA-binding dyes propidium iodide (5 μg/mL) and Hoechst 33342 (10 μg/mL) (15). Apoptosis was confirmed by additional methods, including Bax translocation, caspase 9 and 3 cleavage, and cytochrome c release, measured as described elsewhere (13).

Male DP5^−/− and wild type (WT) C57BL/6 J mice were given standard chow (fat 10% of caloric intake, D12450B) or HFD (fat 60% of caloric intake, D12492; Research Diets). After 12 weeks, intraperitoneal glucose tolerance tests were done in 16-h fasted mice by injecting glucose (2 mg/g body weight). Blood was collected from the tail at times 0, 15, 30, 60, 90, and 120 min, and glucose was measured with a FreeStyle Lite glucometer (Abbott Diabetes Care). Plasma was separated by centrifugation (10 min, 5000g) and stored at −20°C. Plasma insulin was measured with the Ultrasensitive Mouse ELISA Kit (Crystal Chem, Downers Grove, IL). For the intraperitoneal insulin tolerance test, 0.75 mU/g body weight Actrapid (Novo Nordisk) was injected after a 4-h fast, and glucose was measured at the time points listed previously.

Pancreases were taken from mice killed after 25 weeks on the different diets. Three nonconsecutive 5-µm thick pancreatic sections 150 µm apart were labeled with a standard immunoperoxidase method for paraffin sections with a mouse anti-insulin antibody (1/4000; Sigma). The β-cell mass was quantified blindly as β-cell volume density multiplied by pancreas weight (26). To assess β-cell size, nuclear crowding was determined as the number of β-cell nuclei per 100 μm² β-cell cytoplasm (27). The β-cell proliferation was examined as described in the Supplementary Data.

After isolation, DP5^−/− and WT mouse islets were cultured for 30 min in M199 medium containing 5.6 mmol/L glucose and washed with modified Krebs-Ringer bicarbonate HEPES solution (150 mmol/L NaCl, 3.5 mmol/L KCl, 0.5 mmol/L MgCl₂, 1.5 mmol/L CaCl₂, 5 mmol/L NaHCO₃, 0.5 mmol/L NaH₂PO₄, 10 mmol/L HEPES, and 0.1% BSA at pH 7.4). Insulin secretion was induced by 1 h of incubation with Krebs-Ringer bicarbonate HEPES solution containing 1.7 or 16.7 mmol/L glucose. Insulin was measured by ELISA in cell-free supernatants and acid-ethanol extracted cell lysates.

Data are presented as means ± SE of the indicated number (n) of independent experiments. Comparisons were performed by ANOVA, followed by paired t test with the Bonferroni correction for multiple comparisons. P < 0.05 was considered statistically significant.

Palmitate induced cytochrome c release from the mitochondria (Fig. 1A–C), Bax translocation from the cytosol to the mitochondria (Fig. 1B-D), and caspase 9 and 3 activation (Fig. 1E), demonstrating that palmitate engages the intrinsic pathway of apoptosis in pancreatic β-cells. Oleate induces less cell death (5,13) and does not stimulate cytochrome c release. In a time-course study, palmitate decreased Bcl-2 protein expression from 16 h on, independently of BCL2 mRNA expression (Supplementary Fig. 1). Knockdown of Bcl-2 induced apoptosis and sensitized cells to FFAs (Supplementary Fig. 1B). Palmitate also reduced Bcl-XL protein levels though at a late time-point of 24 h only (Supplementary Fig. 2A). Similar to Bcl-2, Bcl-XL knockdown induced β-cell apoptosis (Supplementary Fig. 2B). These results confirm the antiapoptotic role of Bcl-2 and Bcl-XL in β-cells under basal and lipotoxic conditions. The mitochondrial morphology of palmitate-treated cells was punctate, as opposed to the reticular adenosine 5´-triphosphate (ATP) synthase β staining in control cells (Supplementary Fig. 1C and D), and this coincided with the marked decrease in Bcl-2 protein expression, suggesting that Bcl-2 depletion by palmitate contributes to disruption of the mitochondrial network.

We next profiled the global gene expression of INS-1E cells exposed to palmitate for 6 and 14 h. In the array analysis, 20,405 probe sets corresponding to 10,524 genes were detected as present in control or palmitate-treated β-cells. Palmitate modified the expression of 6.7% (792 genes) and 8.2% (1,074 genes) of probe sets at 6 and 14 h, respectively. The top functional clusters included gene expression, lipid metabolism, cell growth, cellular function and maintenance, and cellular compromise at 6 h and cell death, cellular function, cellular compromise, protein synthesis, and lipid metabolism at 14 h. Two networks were enriched in proapoptotic signaling molecules. At 6 h, palmitate upregulated markers of ER stress, including ATF4, ATF3, CHOP, growth arrest and DNA damage protein 34, and TRB3 (Supplementary Fig. 3), which is in keeping with our previous findings on the proapoptotic signaling in the PERK branch of the UPR (9,15). At 14 h, we detected a network of mitochondrial cell death genes containing DP5 and PUMA, PERK, and Jun upregulations (Supplementary Fig. 4).

Genes were also classified by a previously described manual curation (28) according to their potential role in β-cell dysfunction and death (Supplementary Table 1). Palmitate decreased the expression of glucose metabolism genes involved in glycolysis and the citric acid cycle, and upregulated lipid metabolism genes such as Acsl, Cpt1a, and Fads1. The AP-1 family of transcription factors, including c-Jun and c-Fos, was induced at both time points. Genes mediating ER function and signal transduction in the different branches of the UPR were induced, including ATF3, ATF4, CHOP, and TRB3 and also Sec61a1 and Dnajc3 (Supplementary Table 1). The mRNA expressions of prodeath genes such as Bax, Bak, and Bim were not significantly modified; however, the BH3-only proteins DP5 and PUMA were markedly upregulated. On the basis of these gene expression analyses, these two Bcl-2 members were further studied for their role in mitochondrial β-cell death.

By real-time PCR, we confirmed the array finding that palmitate induced DP5 expression, with a maximum after 16 h (Fig. 2A). On the other hand, oleate did not induced DP5 expression (Fig. 2A). Efficient DP5 knockdown by RNA interference (Fig. 2B) reduced cytochrome c release, caspase 3 activation, and apoptosis in palmitate-treated cells (Fig. 2C and D) but did not prevent the mild oleate-induced apoptosis (Fig. 2D). Palmitate induced DP5 in primary rat β-cells (Fig. 2E), and DP5 knockdown partially protected β-cells against palmitate (Fig. 2E). In human islet cells, palmitate also induced DP5 (Fig. 2F), and DP5 knockdown protected cells from lipotoxicity (Fig. 2F).

PUMA was markedly induced by palmitate, peaking at 16 h, but not by oleate (Fig. 3A). Knockdown of PUMA (Fig. 3B) decreased palmitate-induced cytochrome c release, caspase 3 activation, and apoptosis (Fig. 3C and D), whereas it did not affect oleate toxicity (Fig. 3D). Palmitate also upregulated PUMA expression in fluorescence activated cell sorter–purified rat β-cells and human islet cells (Fig. 3E and F). PUMA RNA interference partially (Fig. 3E) and nearly completely (Fig. 3F) prevented palmitate toxicity in, respectively, primary rat β-cells and human islet cells. As a negative control, we studied a gene that was not detected as changed in the array analysis, namely Bim. Unlike DP5 and PUMA, Bim was not induced by palmitate (Supplementary Fig. 2C), and knockdown of its protein product did not modify apoptosis rates (Supplementary Fig. 2D), confirming that Bim does not play a major role in lipotoxic β-cell death.

To assess the mechanism by which palmitate induces DP5 expression, a DP5 promoter luciferase reporter was used. The promoter sequence between −125 to −85 is important for DP5 induction by palmitate, because its deletion reduced promoter activity by 42% compared with the full construct (Fig. 4A). This promoter region has a conserved ATF site where c-Jun binds after phosphorylation by JNK (23). Palmitate activated JNK and its target c-Jun from 30 min onward, peaking at 4–8 h (Supplementary Fig. 5A and B). JNK inhibition by SP600125 reduced c-Jun phosphorylation (Fig. 4B) and abrogated palmitate-induced DP5 promoter activation and mRNA expression both in INS-1E cells (Fig. 4C and D) and in primary β-cells (Fig. 4D). The small peptide JNK inhibitor L-TAT-JNKi, previously shown to inhibit JNK and protect β-cells against palmitate treatment (9), produced a similar reduction of c-Jun phosphorylation and decreased palmitate-induced DP5 expression (Supplementary Fig. 5C and D). The p-c-Jun binding to the DP5 promoter was increased by palmitate as observed in ChIP experiments (Fig. 4E), suggesting that p-c-Jun transcriptionally induces DP5. IRE1α knockdown (by 76 ± 3%) reduced XBP1 splicing (by 46 ± 6%) and decreased palmitate-induced JNK phosphorylation (Fig. 4F). Additional support for the role of IRE1 was provided by using the novel IRE1 inhibitor 4μ8C (25). This small molecule very effectively blocked IRE1, inhibiting spliced XBP1 expression by 96 ± 1%, and significantly decreased JNK phosphorylation and DP5 expression (Fig. 4G and H), confirming that IRE1-JNK mediates the DP5 induction. 4μ8C did not inhibit JNK phosphorylation by interleukin-1β and interferon-γ, suggesting it does not directly inactivate JNK (data not shown).

PUMA expression is controlled by different transcription factors, including p53 and nuclear factor-κB (NF-κB) (29,30). Time course experiments showed that p53 is not induced by palmitate in INS-1E cells, suggesting a p53-independent PUMA regulation in β-cells (Supplementary Fig. 5E). The activation of the PUMA promoter by palmitate was not affected by mutation of the NF-κB binding site (Supplementary Fig. 5F), previously shown to be important for cytokine-mediated PUMA upregulation (31). Furthermore, PUMA promoter activity and PUMA mRNA induction were not modified by JNK inhibitors (Supplementary Fig. 5G–I), suggesting a different role for JNK in the regulation of the DP5 and PUMA promoters.

The PERK pathway and its downstream effectors CHOP and TRB3 are proapoptotic in palmitate-treated β-cells (9,15,32). Knockdown of PERK by small interfering RNA (siRNA) decreased basal and palmitate-induced eIF2α phosphorylation (Supplementary Fig. 6A–C) but did not prevent JNK activation (Supplementary Fig. 6D). PERK knockdown reduced DP5 and PUMA inductions by palmitate (Fig. 5A and B). Knockdown of ATF4 or CHOP did not affect the induction of either BH3-only protein by palmitate (Supplementary Fig. 6E–H). ATF3 siRNA, however, mimicked the effect of PERK siRNA on DP5 and PUMA expressions (Fig. 5A and B), suggesting that PERK-ATF3 activation by palmitate leads to DP5 and PUMA expression. In primary β-cells, knockdown of PERK or ATF3 also decreased palmitate induction of DP5 and PUMA (Fig. 5C). This is consistent with the presence of a putative ATF3 binding site in the DP5 promoter region −125 to −85. Exposure of INS-1E cells to palmitate stimulated the association of ATF3 to the DP5 promoter, showing that this BH3-only protein is a direct target of the PERK-ATF3 pathway (Fig. 5D). The PUMA promoter, on the other hand, does not contain an ATF3 binding site, suggesting that ATF3 indirectly modulates PUMA.

In the microarray analysis, palmitate markedly upregulated TRB3 (Supplementary Fig. 3 and Supplementary Table 1), an ER stress–inducible and proapoptotic gene in the PERK pathway (33). Because knockdown of CHOP did not change TRB3 expression (Supplementary Fig. 6G), we examined the role of ATF3. ATF3 knockdown attenuated TRB3 induction by palmitate (Fig. 5E). TRB3 is a pseudokinase that inhibits AKT phosphorylation, leading to the activation of death pathways including GSK-3β. Because AKT inactivates the forkhead box O (FoxO) family of transcription factors, among which FoxO3a has been shown to induce PUMA-dependent apoptosis, we examined whether the TRB3-AKT-FoxO3a pathway plays a role in the induction of DP5 and PUMA by palmitate. The induction of TRB3 mRNA by palmitate was confirmed at the protein level (Fig. 5F), and the FFA decreased both AKT and FoxO3a phosphorylation (Fig. 5F and Fig. 6A). The dephosphorylation of FoxO3a (Fig. 6A) was accompanied by its nuclear translocation (Fig. 6B and C), confirming activation of the transcription factor. The knockdown of TRB3 partially prevented early changes in AKT and FoxO3a phosphorylation (Fig. 5F). In parallel, TRB3 knockdown decreased palmitate-induced PUMA expression (but not DP5 expression), and reduced lipotoxic β-cell apoptosis (Fig. 5G and H). Additional TRB3/AKT-independent mechanisms may regulate FoxO3a activation at later time points. In silico analysis of the DP5 and PUMA promoters identified putative FoxO3a binding sites in both BH3-only genes, and ChIP demonstrated that palmitate induces FoxO3a binding to both promoters (Fig. 6D). FoxO3a knockdown partially prevented induction of DP5 and PUMA by palmitate (Fig. 6E–F), showing that the AKT-FoxO3a pathway downstream of ER stress controls the BH3-only gene expression and contributes to β-cell death.

Because DP5 is upstream of PUMA (17) and DP5 but not PUMA knockdown provides complete β-cell protection from palmitate at early time points (data not shown), we selected DP5^−/− mice for in vivo studies. DP5^−/− and WT mice were given standard chow or HFD (fat 60% of caloric intake). After 12 weeks, the two HFD-fed genotypes had similar weight gain and insulin sensitivity (Fig. 7A and B). In line with this, similar weight gain, perirenal fat accumulation and hepatic insulin-induced AKT phosphorylation were observed in WT or DP5 knockout mice after 25 weeks of chow or HFD (data not shown). DP5 mRNA expression increased threefold in islets from HFD-fed WT mice compared with chow-fed WT mice (P = 0.17, n = 5–6). DP5^−/− mice were protected from HFD-induced diabetes (Fig. 7C) as a result of strikingly increased insulin secretion (Fig. 7D). Islets isolated from HFD-fed DP5^−/− mice contained and secreted twofold more insulin after glucose stimulation (Fig. 7E-F), suggesting a protective effect at the β-cell level. The lack of the DP5 gene also protected mouse islets in vitro from palmitate-induced cell death (Fig. 7G). This resistance to lipotoxicity was paralleled by a doubling of β-cell mass in HFD-fed DP5^−/− mice but not in WT mice (Fig. 7H), confirming the physiological role of DP5 in vivo. We could not detect apoptotic β-cells in vivo in HFD-fed WT mice (data not shown) and were therefore unable to assess the putative protective role of DP5 deficiency. Increased β-cell proliferation, assessed by Ki67 staining (Fig. 7I), and β-cell hypertrophy (Fig. 7J) likely contribute to the greater β-cell mass of DP5^−/− mice. Thus DP5 may not only play a role in β-cell apoptosis but also serve as a brake on β-cell growth and proliferation.

Failure of the pancreatic β-cells to compensate for high insulin needs is central to the pathogenesis of T2D, and several studies have reported a significant decrease in β-cell mass in T2D (2,26). Chronic exposure to FFAs causes loss of functional β-cell mass (34,35) and may contribute to T2D. Saturated lipids are harmful to β-cells, and pronounced ER stress signaling (especially in the PERK pathway) has been proposed to mediate lipotoxic apoptosis (7–10,15). Execution of FFA-induced apoptosis occurs through the mitochondrial pathway, as indicated by cytochrome c release from the mitochondria after palmitate treatment (13,36,37), but it remained to be clarified how palmitate-induced ER stress crosstalks with the mitochondria to culminate in β-cell death. We have now answered this question, showing that palmitate transcriptionally induces the BH3-only proteins DP5 and PUMA through PERK-dependent ATF3 expression (Fig. 6G).

In the hierarchical model for Bax/Bak activation, the BH3-only “sensitizer” proteins DP5, Bad, Bik, Bnip3, Bmf, and Noxa neutralize the prosurvival Bcl-2 proteins, allowing the BH3-only “activators” PUMA, Bim, and tBid to be released to activate Bax/Bak (38,39). The use of the different pro- and antiapoptotic proteins after a proapoptotic stimulus is cell and context dependent.

Palmitate decreases Bcl-2 expression (13); however its contribution to lipotoxic β-cell death had not been examined. We now show that the decreased Bcl-2 protein expression in palmitate-exposed β-cells occurs in conjunction with mitochondrial fragmentation and apoptosis. A reduction in Bcl-XL protein was seen only after 24 h of palmitate exposure, suggesting that decreased Bcl-XL levels are not pivotal in palmitate-induced apoptosis. A role for Mcl-1 protein downregulation has also been reported in palmitate- and ER stress–induced β-cell apoptosis (40).

The BH3-only prodeath sensitizers and activators in lipotoxicity are essentially unknown. Gene expression profiling of palmitate-treated β-cells indicated that palmitate induced a gene expression signature of ER stress at the early time-point (6 h). At a later time-point (14 h), palmitate upregulated apoptosis-related genes, shifting from an adaptive response to one geared for cell death. As part of the proapoptotic signal network, there was upregulation of DP5 and PUMA, genes for two BH3-only proteins previously shown to play a role in apoptosis of β-cells exposed to cytokines (31,41). Knockdown of the BH3-only sensitizer DP5 and the activator PUMA protected β-cells from palmitate, establishing their contribution to saturated FFA-induced apoptosis. Because the protection is partial in rat β-cells, other BH3-only proteins or alternative death mechanisms may play additional roles. Of note, oleate did not increase DP5 or PUMA expression, in keeping with the observation that this FFA is much less toxic (5,9).

We next set out to identify the mechanism of induction of these prodeath proteins by palmitate. Although the upstream UPR has been characterized in detail, the downstream transduction mechanisms through which ER stress connects to mitochondrial apoptosis have remained elusive. By promoter analysis, we identified a putative c-Jun binding site in the DP5 promoter, and confirmed p-c-Jun binding by ChIP. Inhibition of IRE1-dependent JNK activation reduced DP5 promoter activation and mRNA induction by palmitate. These results are consistent with our previous finding that JNK inhibition protects against lipotoxic β-cell death (9). JNK regulates the intrinsic death pathway through transcriptional as well as posttranscriptional modifications of Bcl-2 family members (42,43).

The BH3-only activator PUMA was not induced by JNK, p53, or NF-κB, previously shown to modulate PUMA expression (29–31,44). We therefore examined the role of ER stress in PUMA expression. Of the three branches of the UPR, sustained activation of the PERK pathway plays a large part in β-cell apoptosis (14,15). Downstream of PERK, CHOP contributes to palmitate-induced apoptosis (9). CHOP deletion also improves glycemic control and prevents β-cell loss in mouse models of T2D (45). ATF3 is also induced by PERK; its role in β-cell death is more controversial. ATF3 knockout confers minor protection against cytokine-induced islet cell apoptosis (46) which may involve the regulation of the insulin receptor substrate 2 (47). In HFD-fed mice, on the other hand, ATF3 is protective by regulating insulin synthesis (48). We have now demonstrated that PERK-induced ATF3, but not ATF4-CHOP, induces both PUMA and DP5. Interestingly, the mechanisms by which ATF3 regulates DP5 and PUMA expressions are not similar. ATF3 directly modulates DP5 mRNA expression, as suggested by its binding to the DP5 promoter, whereas PUMA regulation depends on a complex pathway involving TRB3, AKT, and FoxO3a (Fig. 6G).

TRB3 was identified as a proapoptotic effector of CHOP and ATF4 (49). Increased TRB3 expression was observed in islets from T2D donors and HFD-fed mice (50). In the same study, interaction between TRB3 and ATF4 decreased insulin secretion by reducing the expression of exocytosis-related genes. We show here that TRB3 is induced by ATF3 but not CHOP and causes FoxO3a-dependent PUMA induction and cell death. In other cell types, FoxO3a has been shown to induce PUMA expression independently of p53 (51). We also demonstrated that DP5 upregulation is partially FoxO3a dependent, which is consistent with the presence of a putative FoxO3a binding site in its promoter. This DP5 induction by FoxO3a is apparently TRB3-independent, suggesting alternative FoxO3a activation mechanism.

The physiological relevance of these findings was examined in DP5^−/− mice exposed to HFD. Compared with WT mice, the DP5^−/− mice were resistant to lipotoxic loss of glucose tolerance as a result of significantly greater glucose-stimulated insulin secretion and an adaptive increase in β-cell mass. Interestingly, β-cell proliferation and hypertrophy were increased in HFD-fed DP5^−/− mice, pointing to a hitherto unknown inhibitory role of DP5 in cell growth and the cell cycle. Because the mice were whole-body knockouts, it cannot be excluded that DP5 deficiency also exerts beneficial effects in peripheral tissues.

In conclusion, lipotoxic ER stress causes JNK- and PERK-dependent ATF3 and TRB3-FOXO3a activation, and these transcription factors regulate expression of Bcl-2 family members. Palmitate engages the mitochondrial pathway of cell death through induction of the BH3-only sensitizer DP5, loss of antiapoptotic Bcl-2 and Mcl-1, and upregulation of the BH3-only activator PUMA, culminating in the activation of Bax/Bak and mitochondrial permeabilization (Fig. 6G). These results provide insight into the mechanisms of lipotoxic β-cell ER stress and identify hitherto unknown transcriptional regulation and signal transduction between ER stress and mitochondrial apoptosis. The findings also delineate potential new areas for β-cell therapy in the prevention and treatment of T2D.

This work was supported by the European Union (Collaborative Projects CEED3 and BetaBat in the Framework Program 7), a European Foundation for the Study of Diabetes (EFSD)/Lilly grant, the Fonds National de la Recherche Scientifique (FNRS), Fonds de la Recherche Scientifique Médicale (FRSM) and Actions de Recherche Concertées de la Communauté Française (ARC), Belgium. D.A.C. was supported by an FNRS post-doctoral fellowship, E.N.G. was supported by an EMBO long-term fellowship, and J.-M.V. is research director at the FNRS.

No potential conflicts of interest relevant to this article were reported.

D.A.C., E.N.G., D.L.E, and M.C. contributed to the experimental design of the study. D.A.C., M.I.-E., E.N.G., C.M.G., N.N., I.M., M.F., J.-M.V., C.G., and D.R. carried out experiments, helped with data analysis, or both. C.M., L.M., P.M., H.P.H., and D.R. contributed materials and data interpretation. D.A.C., M.I.-E., E.N.G., D.L.E, and M.C. analyzed and interpreted the data. D.A.C., D.L.E, and M.C. wrote the manuscript. M.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Mingtao Li, Sun Yat-sen University, for providing the DP5 promoter; Lin Zhang, University of Pittsburgh, for the PUMA promoter; Andrea Jurisicova, University of Toronto, and Gabriel Nuñez, University of Michigan, for the DP5^−/− mice; Miriam Hernangomez Herrero, Université Libre de Bruxelles, for help with the mouse experiments; and Gilbert Vandenbroeck, Marie-Anne Neef, Maryse Urbain, Anyishai Musuaya, Stephanie Mertens, and Rachid Makhnas, all from Université Libre de Bruxelles, for expert technical assistance.