Neutralizing Interleukin-1β (IL-1β) Induces β-Cell Survival by Maintaining PDX1 Protein Nuclear Localization*

The transcription factor PDX1 plays a critical role during β-cell development and in glucose-induced insulin gene transcription in adult β-cells. Acute glucose exposure leads to translocalization of PDX1 to the nucleoplasm, whereas under conditions of oxidative stress, PDX1 shuttles from the nucleus to the cytosol. Here we show that cytosolic PDX1 expression correlated with β-cell failure in diabetes. In isolated islets from patients with type 2 diabetes and from diabetic mice, we found opposite regulation of insulin and PDX1 mRNA; insulin was decreased in diabetes, but PDX1 was increased. This suggests that elevated PDX1 mRNA levels may be insufficient to regulate insulin. In diabetic islets, PDX1 protein was localized in the cytosol, whereas in non-diabetic controls, PDX1 was in the nucleus. In contrast, overexpression of either IL-1 receptor antagonist or shuttling-deficient PDX1 restored β-cell survival and function and PDX1 nuclear localization. Our results show that nuclear localization of PDX1 is essential for a functional β-cell and provides a novel mechanism of the protective effect of IL-1 receptor antagonist on β-cell survival and function.

The transcription factor PDX1 plays a critical role during ␤-cell development and in glucose-induced insulin gene transcription in adult ␤-cells. Acute glucose exposure leads to translocalization of PDX1 to the nucleoplasm, whereas under conditions of oxidative stress, PDX1 shuttles from the nucleus to the cytosol. Here we show that cytosolic PDX1 expression correlated with ␤-cell failure in diabetes. In isolated islets from patients with type 2 diabetes and from diabetic mice, we found opposite regulation of insulin and PDX1 mRNA; insulin was decreased in diabetes, but PDX1 was increased. This suggests that elevated PDX1 mRNA levels may be insufficient to regulate insulin. In diabetic islets, PDX1 protein was localized in the cytosol, whereas in non-diabetic controls, PDX1 was in the nucleus. In contrast, overexpression of either IL-1 receptor antagonist or shuttling-deficient PDX1 restored ␤-cell survival and function and PDX1 nuclear localization. Our results show that nuclear localization of PDX1 is essential for a functional ␤-cell and provides a novel mechanism of the protective effect of IL-1 receptor antagonist on ␤-cell survival and function.
New therapies for diabetes that lead to protection of the insulin-producing ␤-cell are urgently needed. Only when the ␤-cell compensates for the higher insulin demand during insulin resistance can normoglycemia be maintained. Recent studies suggest that the low grade inflammation in type 2 diabetes mellitus (T2DM) 3 contributes to ␤-cell failure (1). Especially, interleukin-1␤ (IL-1␤), whose secretion has been postulated from intraislet macrophages (2) and from ␤-cells themselves when exposed to double-stranded RNA (3), to elevated glucose con-centrations (4 -6), or to free fatty acids (7), initiates ␤-cell destruction.
The receptor for IL-1␤, IL-1R1, is highly expressed in the ␤-cell, and more than 10-fold higher expression of IL-1R1 mRNA was observed in isolated islets than in total pancreas, which is attributed to the expression in ␤-cells (7). This may explain the high sensitivity of the ␤-cell to IL-1␤. A recent study shows that glucose-induced IL-1␤ secretion involves caspase 1 activation mediated by the NALP3 inflammasome. The inflammasome is activated by bacterial toxins or endogenous stress signals (e.g. ATP and ␤-amyloid (8 -10)) through the formation of reactive oxygen species (6,11). Glucose-induced IL-1␤ secretion by the ␤-cell is prevented in NALP3 Ϫ/Ϫ mice, indicating that IL-1␤ is generated through glucose-induced reactive oxygen species production and oxidative stress (6). The thioredoxin-interacting protein, which has been linked to insulin resistance (12), functions as an activator of NALP3. In line with these data, another recent study shows that thioredoxin-interacting protein is highly increased by elevated glucose in ␤-cells and that thioredoxin-interacting protein-deficient islets are protected against glucose toxicity (13). Although glucose-induced IL-1␤ production in the ␤-cell was not observed in all studies (14), IL-1␤ expression in islets from patients with T2DM has been confirmed (5,15), supporting the concept of blocking IL-1␤ signals as a target for diabetes treatment. Indeed, daily injection of IL-1Ra in mice fed a high fat diet (HFD) improves glycemia, glucose-stimulated insulin secretion, and survival (16); reduces hyperglycemia; and reverses the islet inflammatory phenotype in the GK rat (17). Treatment with an IL-1␤ antibody also improves glycemic control in dietinduced obesity in mice (18,19). Results from a clinical study in patients with T2DM showed that IL-1Ra improves glycemic control and ␤-cell function (20 -22). Blocking IL-1␤ signals reduces expression of inflammatory marker in fat tissue and in islets (16). None of the studies have studied the mechanisms of the protective effect of IL-1Ra at the level of ␤-cell gene regulation. IL-1␤ affects expression of the transcription factor PDX1 (pancreatic duodenal homeobox-1, previously called IPF1, IDX1, STF1, or IUF1) (23), a key factor in pancreas development and function. Reduced PDX1 expression levels negatively regulates insulin expression and secretion and predispose islets to apoptosis (24 -26). Its homozygous mutations result in pancreas agenesis associated with neonatal diabetes (27,28). PDX1 directly binds to the promoter and controls expression of important ␤-cell genes, which are vital for ␤-cell function, such as insulin, Glut2, and glucokinase (27). PDX1 deficiency contributes to impaired proliferation and enhanced apoptosis via transcriptional mechanisms in models of type 2 diabetes, such as Psammomys obesus and the leptin receptor-deficient (Lepr db/db ) db/db mice (29). Overexpression of PDX1 restores ␤-cell mass and function, thereby preventing the onset of diabetes in IRS2 knock-out mice, showing the critical role of PDX1 in ␤-cell survival (30).
Complicated signaling networks control PDX1 regulation, and nucleo-cytoplasmic shuttling plays a major role in the regulation of PDX1 function (31,32). Well characterized PDX1 nuclear import signal and nuclear export signal (NES) suggest that PDX1 might be regulated at the level of cellular localization (31,33). Post-translational modification of proteins is the most abundant form of cellular regulation, affecting many cellular signal pathways, including metabolism, growth, differentiation, and apoptosis. In response to acute elevation of glucose and survival factors, such as insulin, PDX1 is phosphorylated and translocates to the nucleus (34,35). By contrast, stimuli associated with diabetes, such as oxidative stress (31) and free fatty acids (36), cause nuclear exclusion of PDX1 (36). This suggests that cytoplasmic accumulation may represent a mechanism to reduce the nuclear action of PDX1 under pathologic conditions rather than to promote a specific cytoplasmic function.
We have shown previously that a diet enriched with fat and sucrose ("Surwit"; HFD) induces impaired glucose tolerance after 4 weeks of feeding, impaired fasting glucose after 8 weeks, and hyperglycemia after 12 weeks in C57Bl/6J mice (16,19,37). These changes in glycemia were accompanied by fluctuations in ␤-cell mass. Despite the reduction in ␤-cell proliferation and the increase in ␤-cell apoptosis, islets showed a compensatory increase in ␤-cell mass up to 12 weeks of diet. After 16 weeks, apoptosis was increased, and ␤-cell mass was reduced in the HFD-treated mice. IL-1␤ antagonism by anti-IL-1␤ antibody treatment, IL-1Ra injections, or overexpression restored normoglycemia and ␤-cell function and survival (16,19,37), but also changes in the inflammatory state of the fat tissue were involved in the protective effects of IL-1␤ antagonism. Therefore, we asked how IL-1␤ signals regulate gene transcription in the ␤-cell. We tested the effect of IL-1Ra on regulating glucose homeostasis in another animal model, the obese diabetic leptin receptor-deficient Lepr db/db mice (db/db). Commonly, we detected that IL-1Ra was able to maintain the cellular localization of PDX1. A diabetic milieu in vitro as well as T2DM in vivo induced a switch of PDX1 from the nucleus to the cytosol, which was accompanied by a loss in ␤-cell mass and function.
Whether PDX1 is altered in its localization in ␤-cells during the progression to diabetes and whether these changes may affect ␤-cell function in different levels was previously unknown. Because PDX1 regulates insulin and specific ␤-cell genes, altered localization may contribute to ␤-cell death and loss of function. In the short term, this stressful response may be tolerated, but under chronic situations in T2DM, prolonged stress conditions ultimately affect ␤-cell survival.

EXPERIMENTAL PROCEDURES
Animals-Transgenic mice overexpressing IL-1Ra (IL-1Ra-OE) were kindly provided by Dr. Emmet Hirsch (Northwestern University, Evanston, IL) (38). Beginning at 5 weeks of age and continuing for 12 weeks, transgenic animals as well as their wild type littermates were fed a normal diet (WT ND and IL-1Ra-OE ND) or a high fat/high sucrose diet (WT HFD and IL-1Ra-OE HFD, Surwit (Research Diets, New Brunswick, NJ), containing 58, 26, and 16% calories from fat, carbohydrate, and protein, respectively (39)) or were treated daily with IL-1Ra from the starting day of the diets (NDϩIL-1Ra and HFDϩIL-1Ra; Kineret, Amgen (Thousand Oaks, CA), intraperitoneal injection of 10 mg/kg body weight) as described previously (16). Four independent experiments with a total of 16 mice (4 mice/ cage) in each group were performed, respectively.
Heterozygous leptin receptor deficient mice on the C57BLKS/J background (Lepr db/ϩ , db/ ϩ ) were purchased from Jackson Laboratory. By cross-breeding these mice to C57BL/ 6J-IL-1Ra-overexpressing mice (OE), we obtained diabetic Lepr db/db (db/db) as well as non-diabetic Lepr ϩ/ϩ (WT) with endogenous overexpression of IL-1Ra (OE-db/db and OE) as well as littermates without IL-1Ra as appropriate negative controls (db/db and WT) with the same mixed background. In all experiments, littermates from the F2 breeding were used. All animals were housed in a temperature-controlled room with a 12-h light/dark cycle and were allowed free access to food and water in agreement with National Institutes of Health animal care guidelines and Section 8 of the German animal protection law.
Intraperitoneal Glucose and Insulin Tolerance Tests-After 4, 8, and 12 weeks of diet (HFD) or at the age of 6 and 10 weeks (db/db), all animals underwent in vivo studies (intraperitoneal glucose and insulin tolerance tests) as described before (16).
Islet Isolation and Culture-Islets from all groups were isolated as described previously (40). Human islets were isolated from eight pancreases of healthy organ donors and from eight with T2DM at the University of Illinois (Chicago, IL), Lille University, or Pisa University as described previously (41) and cultured in CMRL-1066 medium as described previously (42). Islet purity was greater than 95% as judged by dithizone staining (if this degree of purity was not achieved by routine isolation, islets were hand-picked). Islets were exposed to 5.5, 11.1, 22.2, or 33.3 mM glucose or 5.5 mM plus 2 ng/ml recombinant human IL-1␤ (R&D Systems, Minneapolis, MN), with or without 500 ng/ml recombinant human IL-1Ra (R&D Systems) or 10 M JNKi (kindly provided by Xigen S.A., Lausanne, Switzerland) for 72 h.
Glucose-stimulated Insulin Secretion and Insulin Content-Islets used to perform glucose-stimulated insulin secretion experiments were kept in culture medium on matrix-coated plates derived from bovine corneal endothelial cells (Novamed Ltd., Jerusalem, Israel) for 4 days, allowing the cells to attach to the dishes and spread (43). These experimental conditions allowed direct comparison with our previous studies in human and mouse islets pretreated with IL-1Ra in vitro (4,16). Thereafter, islets were washed with PBS and extracted with HCl (0.18 N) in 70% ethanol for 24 h at 4°C. The acid-ethanol extracts were collected for determination of insulin content. Insulin was determined using a human insulin ELISA kit (Alpco).
␤-Cell Mass and Histochemical Analyses-Pancreatic sections from seven healthy controls and from seven patients with T2DM were obtained from the National Disease Research Interchange (NDRI) and from Kangnam St. Mary's Hospital (Seoul, Korea) as described (44); approval for the studies was granted by the Kangnam St. Mary's Hospital and by the Ethical Commission of Bremen University.
Mouse pancreases and sections from isolated islets were obtained as described (16). To determine ␤-cell mass, 10 sections (spanning the width of the pancreas) were analyzed as described (16). For detection of ␤-cell apoptosis, insulin and TUNEL staining were performed (In Situ Cell Death Detection Kit, TMR Red; Roche Applied Science) (16). For PDX1 localization studies, after dehydration, sections were incubated in blocking buffer containing 0.2% Tween 20, 3% IgG-free bovine serum albumin (BSA), 2% Triton X-100 for 1 h at room temperature and overnight at 4°C with rabbit anti-PDX1 (kindly provided by Christopher Wright, Vanderbilt University Medical Center, Nashville, TN) in antibody buffer containing 0.2% Tween 20, 3% IgG-free BSA, and 0.5% Triton X-100. Subsequently, all sections were double-stained for insulin and detected by donkey anti-guinea pig FITC-conjugated antibody (Dako). Fluorescent slides were analyzed using a Nikon MEA53200 microscope (Nikon GmbH Dusseldorf, Germany), and images were acquired using NIS-Elements software (Nikon).
Nuclear Fractionation-Nuclear and cytoplasmic extractions of human islets were performed according to the instructions for the NE-PER nuclear and cytoplasmic extraction reagents (Pierce). The purity of fractions was analyzed by incubation of the membranes with anti-tubulin and anti-GAPDH for cytosolic and anti-PARP or anti-histone H3 for nuclear extracts.
Western Blot Analysis-At the end of the incubation periods, islets were washed in ice-cold PBS and lysed as described (42). Membranes were incubated with rabbit anti-phospho-JNK (Thr 183 /Tyr 185 ), rabbit anti-histone H3, rabbit anti-PARP, rabbit anti-tubulin, rabbit anti-GAPDH, and rabbit anti-␤actin (Cell Signaling Technology) antibodies, followed by horseradish peroxidase-linked anti-rabbit IgG. Density of the bands was analyzed using DocITLS image acquisition 6.6a (UVP BioImaging Systems, Upland, CA).
Statistical Analysis-Samples were evaluated in a randomized manner by a single investigator, who was blinded to the treatment conditions. Data are presented as means Ϯ S.E. and were analyzed by paired, Student's t test or by analysis of variance with a Bonferroni correction for multiple group comparisons.

High Fat Diet Induces PDX1 Translocalization to the Cytosol and Is
Prevented by IL-1Ra-Looking into detailed mechanisms of the loss of adaption in C57Bl/6J mice fed a diet enriched with fat and sucrose (HFD; Surwit), we isolated RNA from islets from mice after 12 weeks of diet and IL-1Ra treatment and performed RT-PCR for insulin and its transcription factor PDX1. Whereas insulin mRNA in islets from HFD animals was decreased to 27% of control islets (Fig. 1A), PDX1 levels were 2.8-fold increased (Fig. 1C). Also, insulin content from isolated islets was significantly decreased in the HFD-fed mice (Fig. 1B, p Ͻ 0.05). Such changes did not occur in IL-1Ra-treated animals. Because it has been reported that PDX1 activity is primarily regulated by its subcellular localization (45,46), we stained pancreatic tissue sections for PDX1 and insulin. Sections from both normal diet groups showed PDX1 immunoreactivity predominantly (95 and 94%) in the nucleus of ␤-cells (Fig. 1D, b and d). In contrast, the majority (67%) of the ␤-cells under the high fat diet expressed PDX1 in the cytoplasm (Fig. 1C, f). IL-1Ra treatment inhibited this translocalization to the cytoplasm (30% cytosolic PDX1), and the islets showed more prominent staining in the nucleus (54% nuclear PDX1; Fig. 1D, h).
Western blot analysis from nuclear and cytosolic fractions from isolated islets from WT and IL-1Ra-OE mice fed for 12 weeks with normal chow or HFD confirmed the decrease in nuclear together with the increase in cytosolic PDX1 in islets under the HFD condition, which was prevented by IL-1Ra overexpression (Fig. 1E).
Heterozygous db/ ϩ and db/ ϩ IL-1Ra mice showed clear nuclear PDX1 localization independent of their age (84 and 87% nuclear PDX1; Fig. 2A, a-f). In contrast, 6-week-old db/db mice showed a clear switch of PDX1 from the nucleus into the cytosol ( Fig. 2A, g-i). Littermates, which overexpressed IL-1Ra, were clearly protected from such localization change ( Fig. 2A,  j-l). The switch in localization was predominant in the 6-weekold mice. With the progression of diabetes in 10-week-old mice, loss of PDX1 expression was observed (only 15% PDX1 ϩ ␤-cells; Fig. 2A, m-o). IL-1Ra-OE mice partly restored overall PDX1 expression and its nuclear localization (51% ␤-cells had nuclear and 12% cytosolic PDX1; Fig. 2A, p-r). The localization switch was again confirmed by Western blot analysis in 6-weekold mice. Although heterozygous db/ ϩ and db/ ϩ IL-1Ra as well as db/dbIL-1Ra mice showed major nuclear PDX1 localization, we found PDX1 predominantly in the cytosol in the db/db mice (Fig. 2B).
The switch in the PDX1 localization to the cytosol was accompanied by impaired glucose tolerance. Although db/ ϩ IL-1Ra and db/ ϩ mice showed normal glucose tolerance, db/db mice revealed increased fasting glucose levels and impaired glucose tolerance. IL-1Ra overexpression in the db/ ϩ IL-1Ra mice could partially but significantly restore glucose tolerance at all time points during the intraperitoneal glucose tolerance test and decrease fasting glucose levels (supplemental Fig. 1A). Glucose-stimulated insulin secretion was completely abolished in the db/db mice but was significantly restored in the db/dbIL-1Ra mice with the significant increase in stimulated insulin secretion and a restoration of the stimulatory index (supplemental Fig. 1, B and C). ␤-Cell failure seen by the impaired GSIS was also confirmed when we analyzed ␤-cell mass and survival. ␤-Cell apoptosis was 4.5-fold FIGURE 1. High fat diet induces PDX1 translocalization to the cytosol. C57Bl/6J mice treated daily with vehicle or 10 mg/kg IL-1Ra were fed a diet enriched with fat and sucrose (HFD) or a normal chow diet (ND) for 12 weeks. A-C, RT-PCR analysis of insulin (A) and PDX1 (C) expression relative to control normal diet conditions and insulin content (B) of mouse islets isolated from the four treatment groups. In the LightCycler System, mRNA levels were normalized to GAPDH and tubulin with the same result. Islets were isolated from four mice per treatment group. D, double immunostaining for insulin (green; a, c, e, and g) and PDX1 (red; b, d, f, and h) in mouse pancreatic tissue sections from all four treatment groups; staining was performed on four different pancreases per treatment group (magnification, ϫ250; inset magnification, ϫ2000). The percentage of nuclear and cytosolic PDX1 was calculated by counting 2000 insulin-positive ␤-cells from four mice in each condition. E, Western blot analysis of PDX1 from nuclear and cytosolic fractions from isolated islets from C57Bl/6J WT-or IL-1Ra-overexpressing mice fed an ND or HFD for 12 weeks. Histone H3 and tubulin were used as loading controls, and purity of fractions for the nuclear and cytosolic extracts from the same protein lysates was assessed. One representative blot of three experiments is shown. Densitometry analysis of bands normalized to histone H3 or tubulin shows the ratio of nuclear and cytosolic PDX1 expression. Data are shown as mean Ϯ S.E. *, p Ͻ 0.05 HFD compared to ND WT mice, **, p Ͻ 0.05 IL-1Ra treated HFD compared to nontreated HFD mice.
increased, and ␤-cell mass was 2.1-fold reduced in the db/db mice. In contrast, ␤-cell apoptosis was reduced, and ␤-cell mass was partially restored in the db/db IL-1Ra mice (supplemental Fig. 1, D and E).
PDX1 Is Localized in the Cytosol in T2DM-To investigate whether our results in the two diabetic mouse models can be translated into human diabetes, we analyzed PDX1 localization and expression in human T2DM (Fig. 3, A-D) and the effect of IL-1Ra on PDX1 localization in human isolated islets (Fig. 3F) and in the rat ␤-cell line INS-1E (Fig. 3, G and  H). In most ␤-cells of the control patients investigated in pancreatic sections (Fig. 3, A (a and b) and B) or in sections from isolated islets (Fig. 3, A (e and f) and C), PDX1 was clearly nuclear (Fig. 3A, b and f). In contrast, PDX1 expression was much weaker in islets from patients with T2DM, and no nuclear PDX1 was observed in (Fig. 3, A (d and h), B, and C). Similar opposite effects in insulin and PDX1 mRNA expression as observed in the HFD mouse model could be observed in human diabetes. Whereas insulin mRNA in islets from patients with T2DM was reduced to 40% of the expression in islets isolated from non-diabetic control patients, PDX1 was 2.8-fold increased (Fig. 3D). This is in  IL-1Ra overexpression. A, insulin staining in green (a, d, g, j, m, and p) and PDX1 in red (b, e, h, k, n, and q) of control heterozygous db/ ϩ (a-c) and db/ ϩ IL-1Ra (d-f) mice and db/db (g-i and m-o) and db/dbIL-1Ra (j-l and p-r) littermates at the age of 6 (g-l) and 10 weeks (m-r). Staining was performed on three different pancreases per treatment group from three independent experiments, respectively (magnification, ϫ250; inset magnification, ϫ2000). Percentage of nuclear and cytosolic PDX1 was calculated by counting 2000 insulin-positive ␤-cells from three mice in each condition. B, Western blot analysis of islet lysates from 6-week-old heterozygous db/ ϩ , db/ ϩ IL-1Ra, db/db, and db/dbIL-1Ra mice. Histone H3 and tubulin were used as loading controls, and purity of fractions in the same membrane after stripping was assessed. One representative blot of three experiments is shown. Densitometry analysis of bands normalized to histone H3 or tubulin shows the ratio of nuclear and cytosolic PDX1 expression. Data are shown as mean Ϯ S.E. *, p Ͻ 0.05 db/db compared to db/ ϩ , **, p Ͻ 0.05 OE-db/db compared to db/db. line with previous data from Del Guerra et al. (47). Insulin content from diabetic islets was decreased when compared with islets from control patients (Fig. 3E, p Ͻ 0.01).
Isolated human islets were treated with increasing glucose concentrations (5.5, 11.1, and 33.3 mM glucose) or 5.5 mM glucose plus 2 ng/ml IL-1␤ for 72 h. As reported before and in our previous studies (4,48), under conditions of elevated glucose levels or IL-1␤ treatment, ␤-cells failed to adequately increase insulin secretion in response to a glucose challenge. As shown in Fig. 3F (b), PDX1 is localized in the nucleus in insulin-positive ␤-cells at control conditions at 5.5 mM glucose. Co-exposure of islets to IL-1Ra did not change PDX1 localization (Fig.   FIGURE 3. PDX1 is localized in the cytosol in T2DM. A, representative double immunostaining for insulin (green; a, c, e, and g) and PDX1 (red; b, d, f, and h) performed in human pancreatic (a-d) and human isolated islet (e-h) sections from seven (c and d) or three (g and h) poorly controlled patients with T2DM and seven (a and b) or three (e and f) healthy controls (magnification, ϫ250; inset magnification, ϫ2000). The percentage of nuclear and cytosolic PDX1 was calculated by counting 2000 insulin-positive ␤-cells from seven pancreatic sections (B) and three islet sections (C) in each condition. D, quantitative RT-PCR analysis of insulin and PDX1 expression from mRNA; E, insulin content from HCl-ethanol extracts from human islets isolated from control patients and patients with T2DM. In the LightCycler system, mRNA levels was normalized to cyclophilin and tubulin with the same result. Data are shown as mean Ϯ S.E. (error bars) from six (mRNA) or three (content) islet isolations from six (three) control patients and six (three) patients with T2DM. *, p Ͻ 0.05 T2DM compared with controls. F, human isolated islets were treated for 72 h with 5.5 mM glucose (control), 33.3 mM glucose, or 2 ng/ml IL-1␤ with or without 500 ng/ml recombinant human IL-1Ra. Fixed and paraffin-embedded islets sections were double-stained for insulin (green; a, c, e, g, i, and k) and PDX1 (red; b, d, f, h, j, and l) and analyzed under the confocal microscope. (magnification, ϫ1000; inset magnification, ϫ4000). Islets were isolated from three different donors, and 3 independent experiments were performed. The percentage of nuclear and cytosolic PDX1 was calculated by counting 2000 insulin-positive ␤-cells from three experiments in each condition. G and H, Western blot analysis of PDX1 of glucose-treated (3 days; 11.1-33.3; G) and IL-1␤-treated (2-3 days; 2 ng/ml; H) nuclear and cytosolic cell lysates of the ␤-cell line INS-1E. GAPDH was used as loading control, and purity of fractions for the cytosolic extracts and full-length PARP for the nuclear extracts in the same membrane after stripping was assessed. One representative blot of three experiments is shown. Densitometry analysis of bands normalized to PARP or GAPDH shows the ratio of nuclear and cytosolic PDX1 expression. Because there were no changes, the 2-day results were not analyzed in H. 3F (h)). Increasing glucose concentrations to 11.1 (not shown) and 33.3 mM (Fig. 3F (d)) or the addition of IL-1␤ (Fig. 3F (f)) shifted the PDX1 signal to the cytoplasm, resulting in co-localization of PDX1 with insulin (yellow merged color), whereas co-incubation with IL-1Ra protected from such PDX1 shift; PDX1 remained in the nucleus in the IL-1Ra-treated islets (Fig.  3F, j and l). In order to show a ␤-cell-specific mechanism, such changes were confirmed by Western blot analysis of glucoseand IL-1␤-treated cell lysates of the ␤-cell line INS-1E. Glucose induced a dose-dependent decline in nuclear PDX1 together with an increase in cytosolic PDX1 expression after 72 h (Fig.  3G). 2 ng/ml IL-1␤ reduced nuclear and induced cytosolic PDX1 expression after 3 days of exposure but not after 2 days, showing a time-dependent IL-1␤ effect (Fig. 3H).
IL-1Ra Prevents Prolonged Glucose-and IL-1␤-induced JNK Activation-PDX1 nuclear export is induced by oxidative stress through a signaling pathway that involves JNK activity (31,49,50). Therefore, we investigated whether JNK also mediates glucose-and IL-1␤-induced effects on PDX1 localization by first evaluating whether IL-1Ra regulates JNK activation.
In isolated mouse islets, JNK remained in its activated state over a 3-day period in the presence of elevated glucose concentrations (Fig. 4A, left). In islets from mice that endogenously overexpress IL-1Ra (IL-1Ra-OE), glucose induced transient JNK phosphorylation after 1 day of culture, whereas after 3 days, JNK activation decreased (Fig. 4A, right). We confirmed these findings in isolated human islets. Phosphorylated JNK levels were increased by a 3-day culture in 33.3 mM glucose or in the presence of 2 ng/ml IL-1␤ (Fig. 4B) compared with control cultures at 5.5 mM glucose. In both cases, this increase was ameliorated when IL-1Ra was added to the culture medium.
Because we postulated that glucose-induced JNK activation leads to PDX1 export, JNK inhibition should prevent shuttling despite the presence of elevated glucose or IL-1␤. To test this hypothesis, we treated human islets with 22.2 mM glucose or 2 ng/ml IL-1␤, with or without the addition of 10 M JNKi, a small peptide that inhibits JNK activity, as described before (51,52). We again observed glucose-and cytokine-mediated PDX1 translocalization (Fig. 4C, top, a-c). We not only found cells that co-express insulin and PDX1 in the cytoplasm but found that some cells displayed PDX1 in the nuclear periphery, which has been observed previously in cell lines kept at low glucose concentrations (34). JNK inhibition prevented glucose-and IL-1␤-induced PDX1 translocalization and kept the transcription factor in the nucleus in most cells, indicating that the JNK pathway is involved in regulating PDX1 localization (Fig. 4C,  bottom, d-f).
␤-Cell apoptosis and function were analyzed in the same experiments. The switch of nuclear to cytosolic PDX1 was accompanied by increased ␤-cell apoptosis. 3-day incubation of human islets with elevated glucose (22.2 mM) or 2 ng/ml IL-1␤ resulted in a 1.8-and 2.1-fold increase in ␤-cell apoptosis, and glucose-stimulated insulin secretion was impaired by 75 and 52%, respectively. In contrast, 1 h of pre-exposure and prolonged culture of the islets with 10 M JNKi or 250 ng/ml IL-1Ra restored ␤-cell survival (Fig. 4D) and function (Fig. 4, E  and F).

Mutation of the PDX1 Nuclear Export Signal Restores ␤-Cell
Survival and Function-Because we have observed PDX1 nuclear exclusion in a prodiabetic milieu, we tested whether maintaining PDX1 localization in the nucleus may restore ␤-cell function and survival.
PDX1 WT plasmid and a mutant PDX1, in which the NESlike sequence was disrupted by substituting alanine for leucine at positions 91 and 93 and linked to GFP as described before (33,49), were expressed in human islets by transient plasmid Lipofectamine-mediated transfection. Plasmids were kindly provided by Dr. Ingo Leibiger. Transfection efficiency was analyzed imunocytochemically by GFP plasmid overexpression, which showed a ϳ60% transfection efficiency (supplemental Fig. S2). We also evaluated transfection efficiency by Western blot analysis. Linking PDX1 to GFP resulted in an additional band at 70 kDa, which gave an intensive band in the transfected but not in the GFP-transfected islets (Fig. 5D). Under these conditions, we analyzed ␤-cell apoptosis and function. Overexpression of WT or NES mutant PDX1 did not result in significant changes of ␤-cell apoptosis or function at 5.5 mM glucose. 3-day exposure to 22.2 mM glucose or 2 ng/ml IL-1␤ resulted in a 1.9-and 2.3-fold increase in ␤-cell apoptosis and to a 56 and 68% decrease in glucose-stimulated insulin secretion, respectively (Fig. 5, A-C, p Ͼ 0.05), compared with GFP-transfected control. Although WT PDX1 and NES mutant PDX1 improved ␤-cell survival significantly, glucose-stimulated insulin secretion was only improved by the NES mutant PDX1 and not by WT PDX1 overexpression, indicating 1) that PDX1 overexpression improved ␤-cell survival and 2) the necessity of nuclear PDX1 for restoration of ␤-cell function. When PDX1-WT was overexpressed in human islets, IL-1␤ and glucose induced shuttling of transfected PDX1 to the cytosol (Fig. 5D), which resulted in a clear decrease in the nuclear/ cytosolic PDX1 expression ratio (Fig. 5E). In contrast, PDX1-NES overexpression inhibited such PDX1 shuttling in IL-1␤and glucose-treated islets, and the ratio of nuclear/cytosolic PDX1 expression was significantly increased, compared with PDX1-WT (Fig. 5, D and E).

DISCUSSION
Human mutations of PDX1 as well as changes in its transactivation are strongly associated with diabetes (53,54). In the present study, we show that localization of PDX1 correlates with T2DM, ␤-cell function, and survival. PDX1 shuttling is one of the mechanisms that may explain the dual role of glucose on ␤-cell function and survival. Although acute elevated glucose induces ␤-cell proliferation and insulin secretion, chronically elevated glucose concentrations impair ␤-cell function, induce apoptosis (40), and thus may accelerate diabetes. Upon acute exposure of ␤-cells to glucose, PDX1 translocates to the nucleus, leading to insulin gene transcription (34). In contrast, oxidative stress induces PDX1 shuttling to the cytosol (49) and impairs insulin secretion (55). We show here that in diabetic islets or under conditions of chronic hyperglycemia or IL-1␤ exposure for 3 days in vitro, conditions of impaired ␤-cell survival and function shifted PDX1 expression to the cytoplasm. These data show the impact of PDX1 localization in regulating ␤-cell turnover and function.
Whenever it was shifted to the cytosol, PDX1 showed a weaker staining and signal intensity, which suggests the possibility of PDX1 degradation after translocalization. This is also supported by the observation that overexpression of PDX1-NES, which remains in the nucleus, prevents PDX1 shuttling together with degradation at conditions of chronically elevated glucose. Although transfection could not be achieved in all islet cells, nuclear PDX1 overexpression prevented the deleterious effects of glucose and IL-1␤ on ␤-cell survival and function. Previously, we have observed such quantitative decrease in PDX1 induced by elevated glucose concentrations in human and rat islets. At this time, whole islet lysates were used for the analysis, and we did not investigate PDX1 localization. PDX1 decrease was dose-dependent on glucose concentrations (5.5-33.3 mM) (56). In the same previous study, we also show an age-dependent PDX1 decrease in human and rat islets, which was confirmed in pancreatic biopsy samples (57). Such PDX1 decrease correlated with the increased susceptibility to glucose-induced apoptosis and with a decline in ␤-cell proliferation at an older age. Neither Reers et al. (57) nor we were able to show nuclear PDX1 localization. It was suggested that the human pancreas embedding method prevented detection of nuclear PDX1. Here we improved tissue permeabilization and clearly detected nuclear PDX1 even in sections from human autopsy and biopsy, with strong PDX1 signals in the nucleus in non-diabetic control pancreases and a shift of PDX1 to the cytosol in diabetic conditions, a signal that appeared much weaker in sections from patients with T2DM. Together, the data support PDX1 shuttling in response to chronic glucose to the cytosol and its subsequent degradation as one deleterious factor contributing to ␤-cell failure.
Strategies to block these deleterious effects on the ␤-cell are needed for a successful diabetes therapy. Blocking IL-1␤ signals has been suggested as a novel treatment for diabetes. The antiinflammatory cytokine IL-1Ra prevents glucose-induced apoptosis by blocking proapoptotic IL-1␤ signaling in vitro (4) and FIGURE 5. Mutation of the PDX1 nuclear export signal restores ␤-cell survival and function. Human isolated islets were cultured on extracellular matrixcoated dishes and transfected with a GFP control plasmid, with WT PDX1, or with an NES mutant PDX1 (PDX1-NES), in which the NES-like sequence was disrupted. 24 h after transfection, islets were exposed to elevated glucose or IL-1␤ for 72 h. A, ␤-cell apoptosis expressed as a percentage of TUNEL-positive ␤-cells Ϯ S.E. The mean number of ␤-cells scored was 1545 Ϯ 112 for each treatment condition in three independent experiments from three different donors. B and C, glucose-stimulated insulin secretion of islets. Stimulatory index (B) denotes the ratio between stimulated (16.7 mM glucose) and basal (2.8 mM glucose) values normalized to insulin content (C) of insulin secretion during successive 1-h incubations. Results are means Ϯ S.E. from three independent experiments from three donors. **, p Ͻ 0.05 compared with control at 5.5 mM glucose. *, p Ͻ 0.05, PDX1-WT-or PDX1-NES-treated islets compared with untreated islets under the same conditions. D, Western blot analysis for transfected PDX1 fused to GFP was performed with nuclear and cytosolic fractions of islet lysates. PARP and GAPDH were used as loading control for nuclear (left) and cytosolic (right) extracts. One representative blot of three experiments is shown. E, densitometry analysis of bands normalized to PARP or GAPDH shows the ratio of nuclear and cytosolic transfected PDX1 expression. *, p Ͻ 0.05 compared with control at 5.5 mM glucose. Islets were isolated from three different donors, and 3 independent experiments were performed. F, fixed and paraffin-embedded islet sections were double-stained for PDX1 (green; a, c, and e and g, i, and k) and insulin (red; b, d, and f and h, j, and l) and analyzed under the fluorescent microscope (magnification, ϫ800).
In this study, we provide further mechanisms of the protective effect of IL-1Ra directly on the ␤-cell transcriptional regulation in C57BL/6J mice fed a high fat/high sucrose diet (Surwit) and in db/db mice, serving as two animal models of T2DM and in human islets exposed to a diabetic milieu. 12 weeks of high fat feeding induced impaired glucose tolerance, which was inhibited in the IL-1Ra-OE mice. Unexpectedly, insulin and PDX1 mRNA levels were oppositely regulated in the HFD-treated mice; insulin was reduced, and PDX1 was significantly increased in islets after 12 weeks of high fat/high sucrose diet. IL-1Ra prevented such changes. Many previous studies have observed no changes or increases in insulin mRNA in response to a high fat diet without increase in sucrose ("Western diet") in rat or mouse models (e.g. see Refs. 58 and 59). The addition of sucrose to the diet (Surwit) causes ␤-cell failure together with reduced insulin mRNA in the ␤-cell (60).
Although our results are in contrast to previous data showing PDX1 down-regulation in response to hyperglycemia in ␤-cells in culture as well as in type 2 diabetic animal models (e.g. in ZDF rats (61), in P. obesus (29), and in partially pancreatectomized rats (62)), studies in isolated human islets from patients with T2DM show a similar opposite regulation, with reduced insulin mRNA and increased PDX1 mRNA levels compared with islets isolated from non-diabetic controls (47).
PDX1 expression seems to be important for the ␤-cell response to a higher insulin demand (e.g. in insulin resistance and ␤-cell compensation may occur through increased PDX1 mRNA). But suppression of PDX1 expression in MIN6 cells did not lead to a decrease of insulin or glucokinase mRNA (63). To activate insulin transcription, PDX1 translocates into the nucleus, where it binds to the insulin gene (46,64). Therefore, post-translational changes, which define PDX1 localization, rather than mRNA expression levels may play a more important role under diabetic conditions. For instance, oxidative stress induces PDX1 shuttling from the nucleus to the cytosol and thus causes a severe reduction of PDX1 activity (49). Also, 24-h exposure of rat islets to palmitic acid at elevated glucose concentrations causes PDX1 localization to the cytosol together with a decrease in MafA expression and inhibition of insulin expression (36). In line with these previous observations, we detected PDX1 predominantly expressed in the cytosol in HFD-treated and in hyperglycemic db/db mice. In contrast, in IL-1Ra-treated as well as the normal diet groups, PDX1 was localized in the nucleus. Because IL-1Ra protected from the prodiabetic effect of the diet, we propose that this is a result of maintaining PDX1 functionally in the nucleus.
Homozygous Lepr db/db mice (db/db) on the C57BLKS/J background with this depletion in the leptin receptor become obese, hyperglycemic, and hyperinsulinemic within the first month of age. In these mice, IL-1␤-mediated innate immunity is augmented, which results from a diabetes-associated loss of IL-1␤ counterregulation (65). To test the hypothesis that IL-1Ra would prevent diabetes progression in this model of T2DM, we injected db/db mice daily with IL-1Ra or with vehicle from 4 weeks of age on. db/db mice that received IL-1Ra showed improved glucose tolerance during intraperitoneal glu-cose tolerance test experiments after 2 weeks of treatment as compared with their vehicle-treated littermates (p Ͻ 0.05 at time points 30 and 60 min; data not shown). From 3 weeks of treatment on, we did not observe any differences in glucose levels between the two groups anymore. Considering the short half-life of IL-1Ra (6 -8 h) and the 10 -100-fold excess that is needed to block IL-1␤-mediated effects (66), we tested whether constitutive endogenous overexpression of IL-1Ra would improve the outcome of elevated IL-1Ra levels in db/db mice. We could confirm in the db/db mouse model that IL-1Ra overexpression was protective against the development of diabetes and ␤-cell failure, although the effects of IL-1Ra in db/db mice were quite modest. However, considering the db/db mouse as a model of severe diabetes with rapid development of hyperglycemia, any significant rescue in the model and a combined effect on glucose tolerance, insulin secretion, ␤-cell mass, and apoptosis confirms a protective effect on the ␤-cell. This was paralleled by IL-1Ra-induced PDX1 stabilization in the nucleus in the db/db mice. Already at 6 weeks of age, PDX1 was predominantly expressed in the cytosol in the db/db mice, whereas at the age of 10 weeks, PDX1 was strongly decreased. IL-1Ra restored nuclear PDX1 in 6-week-old mice, and also in 10-week-old mice, nuclear PDX1 expression could be detected. Previous cytochemistry analyses of db/db mouse pancreases show a nuclear loss of MafA rather than of PDX1 (67). To exclude differences in the background, we also analyzed PDX1 in male BKSdb/db mice and could confirm the switch of PDX1 localization to the nucleus already after 6 weeks of age.
Activation of JNK is a hallmark in glucose and IL-1␤ effect on the ␤-cell and is involved in PDX1 regulation (68,69). We used such well known results to confirm our study design and could identify JNK as cellular component also involved in glucoseand IL-1␤-regulated PDX1 localization. In wild type islets, JNK was phosphorylated after short term (30 min (data not shown) and 1 day) as well as long term (3 days) incubations with elevated glucose. Therefore, it seems that acute JNK activation does not necessarily cause impairment in ␤-cell function and survival, whereas chronic activation correlates with glucotoxicity. IL-1Ra protected islets from prolonged JNK activation. JNK activity has previously been linked to PDX1 shuttling under conditions of oxidative stress (49) and in prostaglandin E2-induced ␤-cell dysfunction (32), together with Foxo1 as a key player (31). Foxo1 cellular localization determines PDX1 localization, and both are reversely expressed. Foxo1 itself is regulated by JNK and AKT activity; JNK induces Foxo1 nuclear import, which leads to PDX1 export, whereas AKT-mediated Foxo1 phosphorylation results in Foxo1 cytoplasmic and PDX1 nuclear localization. Our findings suggest the JNK-PDX1 pathway as a critical signaling network that transduces short term as well as long term glucose stimulation and therefore might partly mediate the dual effect of glucose on ␤-cell function and survival.
The fact that IL-1Ra potentially prevented hyperglycemia and improved ␤-cell function favors the critical role of IL-1␤ signaling in the ␤-cell not only in a type 1 but also in a type 2 diabetic environment. Our data provide new insights into mechanisms of the protective effect of IL-1Ra on ␤-cell function and turnover, establish the important role of nuclear PDX1 localization, and support IL-1Ra as a potential therapy for diabetes.