Glucose-dependent Insulinotropic Polypeptide (GIP) Stimulation of Pancreatic (cid:1) -Cell Survival Is Dependent upon Phosphatidylinositol 3-Kinase (PI3K)/Protein Kinase B (PKB) Signaling, Inactivation of the Forkhead Transcription Factor Foxo1, and Down-regulation of bax Expression* □ S

The hormone glucose-dependent insulinotropic polypeptide (GIP) potently stimulates insulin secretion and promotes (cid:1) -cell proliferation and cell survival. In the present study we identified Forkhead (Foxo1)-medi- ated suppression of the bax gene as a critical component of the effects of GIP on cell survival. Treatment of INS- 1(832/13) (cid:1) -cells with GIP resulted in concentration-de-pendent activation of the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (PKB)/Foxo1 signaling module. In parallel studies, GIP decreased bax promoter activ- ity. Serial deletion analysis of the bax promoter demonstrated that the region (cid:2) 682 to (cid:2) 320, containing FHRE-II (5AAAACAAACA), was responsible for GIP-me- diated effects. Foxo1 bound to FHRE-II in gel mobility shift assays, and Foxo1-FHRE-II interactions conferred GIP responsiveness to the bax promoter. INS-1 cells incubated under proapoptotic and glucolipotoxic condi- tions demonstrated increased nuclear localization of Foxo1 and bax promoter activity and decreased cyto- plasmic phospho-PKB/Foxo1. GIP partially restored expression PKB/Foxo1 and bax promoter activity. Similar protective effects were found with dispersed islet cells from C57BL/6 mice, but not with those from GIP recep- tor knock-out (GIPR (cid:2) / (cid:2) ) mice. GIP treatment reduced glucolipotoxicity-induced cell death in C57 BL/6 and Bax (cid:2) / (cid:2) islets, but not GIPR (cid:2) / (cid:2) mouse islets. Chronic treatment of Vancouver diabetic fatty Zucker rats with GIP resulted in down-regulation of Bax and up-regula-tion of Bcl-2 in pancreatic (cid:1) -cells. The results show that PI3K/PKB/Foxo1 signaling mediates GIP suppression of bax gene expression and that this module is a key pathway by which GIP regulates (cid:1) -cell apoptosis in vivo

Recent studies have revealed that Foxo (FKHR) transcription factors (7) are important mediators of insulin and growth factor actions on gene expression downstream from PI3K and PKB (8). Members of the Foxo subfamily share several structural features; in addition to the highly conserved central DNA binding Foxo domain (Fox box), Foxo members contain a Cterminal transactivation domain and three consensus phosphorylation sites for PKB located at Thr 24 , Ser 256 , and Ser 316 (9). Foxo transcription factors regulate the expression of a wide variety of genes involved in the control of cell survival and apoptosis (6,8), cell cycle progression (10), protection against oxidative stress (11), and DNA repair (12). Phosphorylation by PKB causes redistribution of Foxo1 from the nucleus to the cytoplasm, and the resulting decrease in nuclear Foxo1 has been proposed as the possible mechanism for the inhibition of Foxo1-mediated transcription (13).
Regulation of insulin secretion involves the gut-derived peptide incretin hormones glucose-dependent insulinotropic polypeptide (GIP) and the proglucagon gene products glucagonlike peptide-1(GLP-1)-  and GLP-1-(7-36)-amide (14 -15). Recently, anti-apoptotic functions of these hormones in pancreatic ␤-cells have been identified (16 -19) and, because ␤-cell death is a major contributing factor to diabetes mellitus, an understanding of the underlying mechanisms that increase cell survival is clearly important. GIP has been shown to activate PKB in INS-1 ␤-cells (20), but neither the signaling pathway responsible nor the subsequent downstream events have been □ S The on-line version of this article (available at www.jbc.org) contains supplemental Fig. 1 ("Cytoprotective effect of GIP against glucolipotoxicity-induced apoptosis in INS-1 cells") and Fig. 2 ("Cytoprotective effect of GIP against glucolipotoxicity-induced apoptosis in mouse islets").
identified. In the present study, we demonstrated that GIP, acting via the PI3K/PKB signaling pathway, decreases nuclear Foxo1 interaction with the Foxo1 response element (FHRE) in INS-1 ␤-cells, resulting in down-regulation of the bax gene during glucolipotoxicity-induced apoptosis. Additionally, we have shown that these events are intimately involved in GIPinduced reduction of glucolipotoxicity-induced apoptosis in both dispersed wild-type C57BL/6 mouse islet cells and Vancouver diabetic fatty (VDF) Zucker rats in vivo. This is the first demonstration of PI3K/PKB/Foxo1-mediated transcriptional regulation of bax expression, and it adds the bax gene to the list of Foxo1-responsive genes.
Construction of Human Bax Promoter-Luciferase Plasmids-The human bax gene promoter (972 bp; GenBank TM accession number AB183034) was cloned into the pGL3 vector (Promega), and various deletion constructs were prepared by PCR with NheI and HindIII insertions for directed cloning. Site-directed mutant constructs were prepared using the QuikChange site-directed mutagenesis kit (Stratagene). All transfection plasmids were prepared using Qiagen plasmid midi kits (Qiagen).
Transient Transfection and Luciferase Assays-INS-1 cells were plated at a density of 1 ϫ 10 6 cells per 6-well plate. On the following day, transfection was performed with 2 g of the indicated bax promoter-luciferase constructs, 1 g of pCMV-␤-galactosidase plasmid (Clontech), and the indicated amounts of control vector, pCMV5 or PI3Kexpressing plasmids (⌬p85 and p110CAAX), or Foxo1 expressing plasmids (FLAG-FK-WT and FLAG-FK-Tri, wild-type or triple dead kinase mutant, respectively). PI3Kinase and Foxo1 constructs were kindly provided by Dr. G. A. Rutter (Bristol, UK) and Dr. T. G. Unterman (Dundee, UK), respectively. Transfections were performed using Lipofectamine 2000 TM reagent (Invitrogen) for 4 h according to the manufacturer's instructions. On the following day, cells were treated with GIP for the times indicated in the legends to Figs. 1-6, washed with phosphate-buffered saline (Invitrogen), and lysed in 400 l of reporter lysis buffer (Promega). Luciferase activities were measured using the Promega luciferase assay system and normalized with ␤-galactosidase activities to correct for transfection efficiency. Relative luciferase activity is expressed as the normalized luciferase activity per microgram of protein.
Cell Fractionation-To prepare cytoplasmic and nuclear fractions, cells were scraped into hypotonic lysis buffer (10 mM NaCl, 20 mM HEPES at pH 7.4, 3 mM MgCl 2 , and 0.05% Nonidet P-40) and incubated on ice for 2 min. Nuclei were then pelleted at 2000 ϫ g in a microfuge, and the supernatant (cytoplasmic fraction) was collected. The nuclear pellet was resuspended in SDS sample buffer, and the fractions were subjected to Western blot analysis.
Western Blot Analysis-For studies on the effect of GIP on PKB and Foxo1 phosphorylation, INS-1 cells (10 6 cells per well) were incubated in the presence or absence (control) of GIP at the concentrations indicated in legends to Figs. 1-6. Proteins (15 g of protein per well) from each sample were separated on a 13% SDS-polyacrylamide gel and transferred onto nitrocellulose (Bio-Rad) membranes. Probing of the membranes was performed with phospho-PKB (Ser 473 ), PKB, phospho-Foxo1 (Ser 256 ), Foxo1, and ␤-tubulin purchased from Cell Signaling Technology (Beverly, MA). Immunoreactive bands were visualized by enhanced chemiluminescence (Amersham Biosciences) using horseradish peroxidase-conjugated IgG secondary antibodies. For quantification of band density as a measurement of the phosphorylation state, films were analyzed using densitometric software (Eagle Eye; Stratagene).
Gel Mobility Shift Assays-For the electrophoretic mobility shift assay, oligonucleotides for FHRE-I (5Ј-CACAAACACAAACATTC-GAGT-3Ј) and FHRE-II (5Ј-AGGAAAAAACAAACAAACAGA-3Ј) were radiolabeled with [␥-32 P]ATP using the T4 polynucleotide kinase (New England Biolabs, Inc., Beverly, MA) and annealed with 5-fold excess of its complementary, unlabeled oligonucleotide. The annealed probe was separated from unincorporated nucleotides using a QIAquick nucleotide removal kit (Qiagen). Nuclear proteins (5 g) were incubated in a binding buffer for Foxo1 (20 mM HEPES, pH 7.9, 50 mM KCl, 2 mM MgCl 2 , 0.5 mM EDTA, 10% glycerol, 0.1 mg/ml BSA, 2 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 2 g of poly(dI-dC) per reaction) with 5 ϫ 10 5 cpm of 32 P-labeled oligonucleotides. To verify the specificity of the binding reaction, a 12.5-to 100-fold excess of unlabeled oligonucleotide was added to the reaction before adding the labeled probe. For supershift assays, an antibody against Foxo1 was preincubated with nuclear proteins for 30 min before the addition of a labeled probe. Samples were separated by electrophoresis on a 6% polyacrylamide gel in 1ϫ TES buffer (6.7 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 3.3 mM sodium acetate) at 130 V for 1.5 h. Gels were dried and exposed for the appropriate period at Ϫ70°C with intensifying screens.
Induction of Apoptosis and Apoptosis Assay-INS-1 cells or dispersed mouse islets were incubated with a 2 mmol/liter long chain free fatty acid mixture (2:1 oleate to palmitate) (Sigma) containing high glucose (15 mM). After 24 h, GIP (100 nM) was added every 8 h to culture media and incubated for an additional 24 h. Detection of apoptotic cells was performed with the APOPercentage apoptosis assay kit (Biocolor Ltd., Belfast, Northern Ireland) based on the specific uptake of dye by an apoptotic "flip-flop" mechanism. Quantitative analysis was performed using a colorimetric assay following release of the dye from the apoptotic cells according to the manufacturer's instructions. Concentrations of released APOPercentage dye (M), which are proportional to the level of apoptosis, were determined from absorbance values using a standard curve and normalized to levels observed with control cells.
Confocal Microscopy-INS-1 cells were transfected with wild-type or mutant forms of vectors encoding FLAG-Foxo1. After overnight serum starvation, the cells were treated for 6 h with GIP (100 nM). Cells were then fixed in 4% paraformaldehyde in phosphate-buffered saline for 30 min at room temperature and immunostained successively with an ␣-FLAG antibody and a secondary Texas Red ® dye-conjugated antirabbit antibody. Cell nuclei were counterstained with 4Ј,6-diamino-2phenylindole and mounted with Hydromount (with 2.5% (v/v) 1,4-diazobicyclo[2, 2, 2]octane anti-fade reagent). The transfected cells were imaged using a Zeiss laser-scanning confocal microscope (Axioskop2), and all imaging data were analyzed using the Northern Eclipse program (version 6).
GIP Infusion Study-A 2-week continuous infusion of GIP (10 pmol/ kg/min synthetic porcine GIP1-42; Bachem, Torrence, CA) was administered to VDF Zucker rats and their lean litter mate controls (8 -12 weeks old) via an Alzet micro-osmotic pump (Alza Corp., Minneapolis, MN) implanted in the intraperitoneal region. Prior to surgical implantation of the pump, the rats were anesthetized by intraperitoneal injections of pentobarbital (40 mg/kg). Rats were sacrificed at the end of the infusion, and pancreata were harvested and fixed in paraffin for histological studies. All animal experiments were conducted in accordance with the guidelines put forth by the University of British Columbia Committee on Animal Care and the Canadian Council on Animal Care.
Immunohistochemistry-Pancreatic sections from GIP-treated and control rats were subjected to a double immunostaining for Bax and Bcl-2. After deparaffinization, the sections were incubated with Bax and Bcl-2 antibodies and visualized with Alexa Fluor ® 488-conjugated anti-mouse secondary antibody and Texas Red ® dye-conjugated antirabbit antibody. Cell nuclei were counterstained with 4Ј,6-diamino-2phenylindole and imaged using a Zeiss laser-scanning confocal microscope (Axioskop2). A TUNEL assay was performed in adjacent sections to detect DNA fragmentation in apoptotic cells using the Apoptag® apoptosis detection kit (Intergen Company, Purchase, NY) according to the manufacturer's instructions. Quantification of staining intensity was performed by computer-based densitometric analysis using the Northern Eclipse program (version 6). Values of "average gray" were first calculated using a formula of total gray/pixel area. Values were then compared with those of lean islets and expressed as relative fluorescence intensity (RFI).
Statistical Analysis-Data are expressed as means Ϯ S.E. with the number of individual experiments presented in the legends to Figs. 1-6. Data were analyzed using the non-linear regression analysis program PRISM (GraphPad, San Diego, CA), and significance was tested using ANOVA with Dunnett's multiple comparison test or the Newman-Keuls post hoc test as indicated in the legends to Figs. 2-6.

GIP Activates a PKB/Foxo1 Signaling Pathway in INS-1 ␤-Cells-Treatment of INS-1 cells with 100 nM GIP increased
phosphorylation of PKB at Ser 473 (Fig. 1A), but had no effect on total PKB levels. Increased phosphorylation of PKB was observed by 15 min after the start of treatment, was almost maximal by 1 h (Fig. 1A), and was sustained for 6 h. GIP stimulation was concentration-dependent with an EC 50 value of 11.38 Ϯ 0.23 nM (Fig. 1B). GIP treatment also increased phosphorylation of Foxo1 at Ser 256 over a similar time frame (Fig. 1C) with an EC 50 of 11.41 Ϯ 0.23 nM (Fig. 1D). To determine whether GIP-induced phosphorylation of Foxo1 was accompanied by a change in Foxo1 localization, nuclear proteins were prepared from GIP-treated INS-1 cells and subjected to immunoblotting with Foxo1 antibodies. Within 1 h of GIP treatment the expression level of nuclear Foxo1 was decreased, reflecting nuclear exclusion. This decrease was sustained for 6 h (Fig. 1E) and demonstrated a GIP concentration dependence with an EC 50 value of 56.83 Ϯ 0.20 pM (Fig. 1F). The EC 50 value for GIP-induced nuclear exclusion of Foxo1 was significantly lower than that for phosphorylation of PKB and cytoplasmic Foxo1, implying that additional signaling pathways generated by GIP are involved in the translocation process.
PI3K Is a Component of the GIP Activation of PKB/Foxo1 Signaling-Because phosphorylation of both Thr 308 and Ser 473 by PI3K is essential for PKB activation (24), the relationship between PI3K and the PKB/Foxo1 signaling module was studied using the pharmacological inhibitors of PI3K, LY294002, and wortmannin. Inhibition of PI3K greatly decreased basal levels of phospho-PKB, ablated PKB responses to GIP ( Fig. 2A), increased levels of phosphorylated Foxo1 in the cytoplasm (Fig.  2B) and decreased levels of Foxo1 in the nucleus (Fig. 2C). To substantiate involvement of PI3K in the PKB/Foxo1 signaling module, transient transfection analysis was performed using dominant negative ⌬p85 PI3K and constitutively active p110CAAX constructs. Transfection of ⌬p85 into INS-1 cells slightly decreased basal phospho-PKB levels, partially attenuated GIP-stimulated levels (Fig. 2D), decreased phospho-Foxo1 levels in the cytoplasm, and increased nuclear Foxo1. GIPinduced increases in phospho-Foxo1 in the cytoplasm and decreases in nuclear Foxo1 were reduced in magnitude. On the other hand, p110CAAX transfection resulted in increased PKB phosphorylation with no significant increase over GIP treatment, suggesting saturation of phosphorylated PKB (Fig. 2D). Basal cytoplasmic phospho-Foxo1 increased, and nuclear Foxo1 decreased. However, there was no significant response to GIP treatment (Fig. 2, E and F). In summary, GIP was still capable of exerting a small effect on PKB/Foxo1 in cells transfected with ⌬p85, whereas PI3K inhibitors completely blocked responses to GIP, probably due to differences in site access. Nevertheless, the results strongly support a role for PI3K in GIP-stimulated activation of the PKB/Foxo1 signaling pathway.
GIP Decreases Transcriptional Activity of the bax Gene, and the PI3K/PKB/Foxo1 Module Regulates bax Promoter Activity-Foxo1 binding to DNA activates transcription under conditions of growth factor withdrawal that can lead to apoptosis (6). This finding prompted us to consider critical death genes involved in apoptosis as candidate targets in the PI3K/PKB/Foxo1 signaling pathway. One of the possible Foxo1 targets, whose protein product is a mediator of cell death, is the pro-apoptotic bax gene. Because regulation of the bax gene is poorly understood, we first analyzed the human bax promoter using the MatInspector (Abteling Genetik, Braunschweig, Germany) program to identify potential regulatory sequences. Examination of the human bax gene promoter sequence showed two FHREs, potential binding sites for Foxo1, in addition to nuclear factor Y, activator protein-1, stimulating protein-1, and c-Rel consensus sequences. To test the possible involvement of Foxo1 in GIP transcriptional regulation of the bax gene, promoter activity analyses were performed. Treatment of INS-1 cells with GIP (100 nM) for 12 h resulted in a 70.5 Ϯ 8.02% decrease (p Ͻ 0.01) in bax promoter activity compared with control (Fig. 3A); concentration-response studies revealed an EC 50 value of 16.36 Ϯ 0.72 pM for GIP treatment (Fig. 3B). Inhibition of PI3K signaling with LY294002 or wortmannin significantly increased basal bax promoter activity and ablated responses to GIP treatment (Fig. 3C). Similar to the results with Foxo1, cotransfection of ⌬p85 resulted in partial blockade of the effects of GIP on bax promoter activity, whereas co-transfection of p110CAAX resulted in decreased basal bax promoter activity and no further change with GIP treatment (Fig. 3D). Co-transfection of FLAG-FK-WT, an overexpressing mutant of Foxo1, resulted in significant increases in bax promoter activity, and GIP treatment decreased bax promoter activity almost to vector-treated control levels. Co-transfection of FLAG-FK-Tri, a triple kinase dead mutant of Foxo1 (Thr 24 , Ser 256 , and Ser 319 ), increased bax promoter activity as compared with the vector co-transfected control, and GIP was incapable of inhibiting activity (Fig. 3E). When the subcellular localization of transfected Foxo1 was determined using confocal microscopy, wildtype Foxo1 was found to be transported from the nucleus to the cytoplasm in response to GIP treatment, whereas the triple mutant Foxo1 was mainly retained in the nucleus (Fig. 3F). These results therefore establish that the PI3K/PKB/Foxo1 module is a key component in the regulation of bax promoter activity in the ␤-cell and that GIP activation of this pathway is involved in transcriptional down-regulation of the bax gene through the depletion of Foxo1 from the nucleus.
Identification of a Functional Foxo1 Binding Site in the Human bax Promoter-The functional FHRE involved in GIP responsiveness of the bax promoter was identified next. We termed the two potential FHREs residing between Ϫ904 and Ϫ895 and between Ϫ577 and Ϫ568 FHRE-I, and FHRE-II, respectively. FHRE-I (CAAACACAAA) and FHRE-II (AAAA-CAAACA) exhibit 60 and 90% homology to the consensus FHRE (AAAACAAACT) of insulin-like growth factor-binding protein 1 (IGFBP1), respectively. To identify the functional Foxo1 binding site, deletion mutants serially deleted from the 5Ј-end of the Ϫ972 to Ϫ1 fragment were constructed. The deletion from Ϫ682 to Ϫ320, containing FHRE-II, completely abolished GIP responsiveness (Fig. 4A), whereas deletion of FHRE-I had a minimal effect. In gel retardation assays, the formation of a specific shifted band was observed with the FHRE-II probe but not with the FHRE-I probe. When competition assays were performed to verify the specificity of this reaction, the shifted band gradually disappeared with the addition of a competitor at 12.5-to 100-fold molar excess (Fig. 4B). The addition of an anti-Foxo1 antibody disrupted the band between nuclear proteins and the FHRE-II probe and led to the appearance of a slower migrating supershifted band (Fig. 4C), indicating direct binding of Foxo1 to FHRE-II. To further evaluate the functional significance of the FHRE-II in transactivation of the bax promoter, site-specific mutations were introduced into the FHRE-II, and the GIP responsiveness of mutant constructs was determined using transient transfection analysis. Substitutions of three base pairs flanking FHRE-II (M1-M4 FHRE-II) decreased basal activity, resulting in decreased GIP responsiveness of these mutant constructs as compared with wild-type (WT FHRE-II) (Fig. 4D). These results  (Table I and (Table I). bax promoter activity and protein expression levels were significantly increased under glucolipotoxic conditions, with GIP treatment partially restoring levels toward control values (Fig. 5, A and B). Additionally, PKB and phosphorylation of Foxo1 in the cytoplasm, which decreased under glucolipotoxic conditions, were also partially restored by GIP treatment (Fig. 5, C and D). The expression level of Foxo1 in the nucleus, which increased under glucolipotoxic conditions, was also partially restored toward normal (Fig. 5E). When we examined the effect of GIP on dispersed wild-type C57 BL/6 mouse islets under glucolipotoxic conditions, there were GIP responses similar to those observed with INS-1 ␤-cells. Treatment with GIP reduced glucolipotoxicityinduced apoptotic cell death in dispersed wild-type C57 BL/6 mouse islets but not in those from GIPR Ϫ/Ϫ mice (Table I and (Table I). In contrast, dispersed Bax Ϫ/Ϫ mouse islets tended to show increased resistance to glucolipotoxicity, and GIP treatment reduced glucolipotoxicity-induced apoptotic cell death (Table I and supplemental Fig. 2). bax mRNA and protein expression levels were greatly increased in dispersed wild-type C57 BL/6 mouse islets exposed to glucolipotoxic conditions, and GIP treatment partially restored levels toward those observed with untreated islets (Fig. 6, A and B). Additionally, the levels of PKB and phosphorylated Foxo1 in the cytoplasm also decreased under glucolipotoxic conditions and were partially restored by GIP treatment, whereas nuclear Ϫ972 bax promoter reporter construct (2 g), pCMV ␤-galactosidase (1 g), and various PI3K constructs or a control vector, pCMV5 (2 g). After transfection, INS-1 cells were starved overnight, and 100 nM GIP was added for 12 h. E, effects of Foxo1 constructs on bax promoter activity. INS-1 cells were co-transfected with a Ϫ972 bax promoter reporter construct (2 g), pCMV ␤-galactosidase (1 g), and Foxo1 constructs (wild-type (WT) or triple kinase dead mutant (Tri)) or a control vector, pCMV5 (2 g). After transfection, INS-1 cells were treated as described for panel D. F, subcellular distribution of transfected Foxo1. INS-1 cells were transfected with various Foxo1 constructs and starved overnight. Cells were then stimulated for 6 h with 100 nM GIP, and immunocytochemical staining was performed using an antibody against FLAG. The scale bar indicates 10 m. All data represent three independent experiments, each carried out in triplicate. Significance was tested using ANOVA with a Newman -Keuls post hoc test (A, C, D, and E) 6 -10) oligonucleotides. An unlabeled oligonucleotide of consensus FHRE was added as competitor at 12.5-to 100-fold molar excess where indicated. C, super-shift electrophoretic mobility shift assay. 0.1 g (ϩ) or 1 g (ϩϩ) of the anti-Foxo1 antibody was added to each binding reaction. D, GIP responsiveness of wild-type and mutant FHRE-II constructs. The sequences of wild-type (WT) and mutant FHRE-II are presented. All sequences of mutant FHRE-II constructs were identical to the wild-type construct except for the indicated mutated sequences in FHRE-II. INS-1 cells were transfected with pCMV ␤-galactosidase (1 g) and wild-type construct (Ϫ972 bax promoter reporter construct; 2 g) or mutated FHRE-II constructs (M1-M4 FHRE-II; 2 g) and treated as described for panel A. Blots are representative of n ϭ 3; all data represent three independent experiments, each carried out in duplicate. Significance was tested using ANOVA with Newman-Keuls post hoc test. **, p Ͻ 0.01 versus basal level.
Foxo1 was increased under glucolipotoxic conditions and also partially restored by GIP (Fig. 6, C-E). On the other hand, dispersed GIPR Ϫ/Ϫ mouse islets revealed similar responses to glucolipotoxicity but showed no responsiveness to GIP treatment (Fig. 6, A-E). Dispersed Bax Ϫ/Ϫ mouse islets demonstrated attenuated responses to glucolipotoxicity as compared with wild-type C57 BL/6 mouse islets, and GIP treatment restored protein expression levels almost to control (untreated) values (Fig. 6, C-E). These results strongly suggest that GIP protects against glucolipotoxicity-induced apoptotic cell death partially through down-regulation of bax gene transcription via PI3K/PKB-mediated depletion of Foxo1 from the nucleus.
Anti-apoptotic Actions of GIP in Vivo-The effect of in vivo treatment with GIP on ␤-cell apoptosis was next studied with long term (14 day) systemic administration of peptide in VDF Zucker rats. Densitometric quantification of immunohisto-chemical staining of pancreatic islet Bax protein expression was performed as described under "Experimental Procedures." Values were then compared with those of control lean islets and expressed as RFI units. Islets from obese diabetic (VDF) rats (Fig. 7) revealed significantly increased Bax levels compared with their lean litter mate controls (RFI values of 2.4 Ϯ 0.2 versus 1.0 Ϯ 0.3). GIP treatment decreased Bax levels in islets from both obese and lean animals (RFI values of 1.2 Ϯ 0.2 and 0.6 Ϯ 0.1, respectively) (Fig. 7). In contrast, expression levels of the anti-apoptotic protein Bcl-2 were decreased in obese VDF rats compared with their lean litter mates (RFI values of 0.4 Ϯ 0.2 versus 1.0 Ϯ 0.2), and GIP treatment increased RFI levels to 1.9 Ϯ 0.4 with islets from lean animals and 0.7 Ϯ 0.2 with those from the obese animals. Dual labeling revealed that the increase in Bcl-2 was mainly restricted to ␤-cells. These in vivo results substantiate a physiological role for GIP in the regula-TABLE I Quantification of apoptosis on INS-1 cells and dispersed islet cells from C57BL/6, GIPR Ϫ/Ϫ , and Bax Ϫ/Ϫ mice Apoptosis was determined using the APOPercentage kit. Concentrations of released APOPercentage dye (M) were determined from absorbance values using a standard curve. For INS-1 cells, values were normalized to those observed with control (untreated) INS-1 cells, whereas islet cell determinations were normalized to those obtained with control C57BL/6 islets. Significance was tested using ANOVA with Newman-Keuls post hoc test.
a p Ͻ 0.05 versus control. b p Ͻ 0.05 versus glucose plus FFA.

FIG. 5. GIP reduces glucolipotoxicity-induced apoptosis INS-1 ␤-cells.
A, decreased bax promoter activity in response to treatment with GIP under glucolipotoxic conditions. INS-1 cells were co-transfected with a Ϫ972 bax promoter reporter construct (2 g) and pCMV ␤-galactosidase (1 g) and exposed to mixtures of glucose (15 mM) and FFAs (2 mM) (glucolipotoxic conditions) for 24 h. GIP (100 nM) was added every 8 h to culture media and incubated for an additional 24 h. Promoter activities are shown as relative luciferase activity normalized to ␤-galactosidase activity. B-E, INS-1 cells were treated as described above, and Western blot analyses were performed using antibodies against Bax (B), phospho-PKB (P-PKB) and PKB (C), or phospho-Foxo1 (P-Foxo1), Foxo1, and ␤-tubulin (D and E). Treatment with GIP under glucolipotoxic conditions resulted in decreased Bax protein expression (B), increased PKB phosphorylation (C), increased cytoplasmic Foxo1 phosphorylation (D), and decreased nuclear Foxo1 (E). Western blot analyses were performed using antibodies against Foxo1 and ␤-tubulin. Western blots are representative of n ϭ 3, and all data represent three independent experiments, each carried out in triplicate. Significance was tested using ANOVA with Newman-Keuls post hoc test. **, p Ͻ 0.01 versus control; ##, p Ͻ 0.01 versus glucose plus FFA.
tion of ␤-cell apoptosis through the combined down-regulation of Bax and the up-regulation of Bcl-2.
In sections stained for the detection of DNA fragmentation and analyzed by densitometry, islets from lean rats showed extremely low levels of apoptosis (RFI value of 1 Ϯ 0.1; Fig. 7). In contrast, islets from VDF rats showed greatly elevated levels (RFI value of 31.5 Ϯ 3.1), whereas in animals treated with GIP, islet cell apoptosis was greatly reduced (RFI value of 2.5 Ϯ 0.6).
FIG. 6. GIP reduces glucolipotoxicity-induced apoptosis in dispersed wild-type mouse islets. A, decreased bax mRNA expression levels in response to treatment with GIP under glucolipotoxic conditions. Dispersed wild-type C57 BL/6 mouse islets, GIPR Ϫ/Ϫ islets, and Bax Ϫ/Ϫ islets were exposed to mixtures of glucose (15 mM) and FFAs (2 mM) for 24 h. GIP (100 nM) was added every 8 h to culture media and incubated for an additional 24 h. Total RNA was isolated from each sample, and real time reverse transcription PCR was performed to quantify bax mRNA expression levels. bax expression levels are shown as the fold difference versus control normalized to glyceraldehyde-3-phosphate dehydrogenase expression levels. B-E, dispersed mouse islets were treated as indicated above, and Western blot analyses were performed using antibodies against Bax (B), phospho-PKB (P-PKB) and PKB (C), Foxo1 or phospho-Foxo1 (P-Foxo1), and ␤-tubulin (D--E). Treatment with GIP under glucolipotoxic conditions resulted in decreased Bax protein expression levels (B), increased PKB phosphorylation (C), increased cytoplasmic Foxo1 phosphorylation (D), and decreased nuclear Foxo1 (E). Western blots are representative of n ϭ 3, and all data represent three independent experiments, each carried out in triplicate. Significance was tested using ANOVA with Newman-Keuls post hoc test. **, p Ͻ 0.01 versus control, C57 BL/6; ##, p Ͻ 0.01 versus glucose plus FFA, C57 BL/6.

DISCUSSION
It has been proposed that GIP promotes cell survival at least in part by activating PI3K and its downstream target PKB (26). One function of PKB is to phosphorylate and, thereby, inhibit proapoptotic components of the intrinsic cell death machinery present within the cytoplasm, including BAD and caspase 9. However, the nuclear targets of the PI3K/PKB pathway and their role in cell survival have not, as yet, been clarified. In the current study, it has been demonstrated that activation of the PI3K/PKB pathway by GIP results in phosphorylation of Foxo1, a member of the Forkhead transcription factor family, leading to sequestration of Foxo1 in the cytoplasm away from its transcriptional targets, including the bax gene. To study the underlying mechanisms involved, we cloned the human bax promoter and, using the rodent ␤-INS-1 cell line, a functional promoter FHRE, FHRE-II (AAAACAAACA), was identified that is responsible for GIP-mediated effects. Foxo1 selectively bound to FHRE-II of the bax promoter, and deletion of the FHRE-II region resulted in significantly decreased basal activity as well as GIP responsiveness (Fig. 4A). Nucleotide substitutions in the FHRE-II region directly affected basal activity and GIP responsiveness of the promoter (Fig. 4D), demonstrating the importance of Foxo1/FHRE-II interactions for transactivation (Fig. 4D). Moreover, basal activity of the bax promoter was significantly increased when cells were co-transfected with an overexpressing mutant of Foxo1, FLAG-FK-WT, or a triple kinase dead mutant of Foxo1, FLAG-FK-Tri (Fig. 3E).
Recently, it has become evident that Foxo proteins play critical roles in cell survival, apoptosis, and cell cycle progression (6,8,10,11,12), and Foxo1 is believed to play a key role in the regulation of ␤-cell mass (27)(28)(29). Expression of a transgene encoding a gain-of-function Foxo1 in liver and pancreatic ␤-cells was shown to result in diabetes due to increased hepatic glucose production and the impaired ␤-cell compensation associated with reduced Pdx1 expression (27). In contrast, haploinsufficiency of the Foxo1 gene was shown to restore insulin sensitivity and rescue the diabetic phenotype in insulin-resis-tant mice. The underlying mechanisms involved reduced hepatic expression of genes regulating gluconeogenesis and increased expression of adipocyte genes associated with the promotion of sensitivity to insulin (27). Accili and co-workers have therefore proposed that Foxo1 is a major negative regulator of pancreatic ␤-cell growth (28) and that impairment in its signaling in both ␤-cells and other insulin sensitive tissues can account for many of the metabolic abnormalities associated with type 2 diabetes. There is currently little known regarding the up-stream regulators of ␤-cell Foxo1 activity or its target genes. In the current studies, it was shown that exposure of the INS-1 ␤-cell line or dispersed wild-type C57 BL/6 mouse islets to elevated levels of FFAs and glucose led to apoptosis accompanied by significant increases in nuclear Foxo1 and bax mRNA transcript levels. GIP treatment resulted in an alleviation of glucolipotoxicity-induced apoptosis and decreases in nuclear Foxo1 and bax expression levels in both dispersed wild-type mouse islets (Table I and Fig. 6, A, B, and E) and INS-1 ␤-cells (Table I and Fig. 5, A, B, and E). The most likely pathway by which GIP down-regulates bax gene expression and reduces glucolipotoxicity-induced apoptosis is therefore via inactivation and nuclear exclusion of Foxo1 through PKB-mediated phosphorylation. The importance of end-target bax gene expression in islet apoptosis was further supported by the results with Bax Ϫ/Ϫ mouse islets. Dispersed Bax Ϫ/Ϫ mouse islets tended to show increased resistance to glucolipotoxicity, reflecting the involvement of Bax in islet apoptosis, whereas GIP treatment reduced glucolipotoxicity-induced apoptotic cell death (Table I). Dispersed Bax Ϫ/Ϫ mouse islets showed attenuated responses to glucolipotoxicity in PKB and Foxo1 expression as compared with wild-type C57 BL/6 mouse islets, and GIP treatment partially restored protein expression levels toward normal, implicating multi-faceted effects of GIP on islet apoptosis (Fig. 6, C, D, and E).
In an earlier study on the effect of GIP on apoptosis induced in INS-1 cells by serum and glucose deprivation (19), we provided evidence that GIP increased cell survival via cAMP-FIG. 7. Anti-apoptotic actions of GIP in vivo; decreased Bax and increased Bcl-2 expression in GIP-treated VDF rats. VDF rats and their lean litter mate (Lean) controls received a 2-week continuous infusion of GIP (10 pmol/kg/min), and pancreata were harvested. Immunohistochemical staining was performed using Bax and Bcl-2 antibody and imaged using a Zeiss laser-scanning confocal microscope (Axioskop2). A TUNEL assay was performed on adjacent sections to detect DNA fragmentation in apoptotic cells using the Apoptag ® apoptosis detection kit. All imaging data were analyzed using the Northern Eclipse program (version 6). Scale bar indicates 50 m. mediated inhibition of p38 mitogen-activated protein kinase. The addition of wortmannin potentiated the level of cell death resulting from glucose and serum withdrawal, implying that PI3K activation occurred in response to the dual stresses and making it difficult to establish conclusively whether PI3Kmediated pathways were also involved in the action of GIP. The mechanisms linking cellular stress and serum withdrawal to the activation of p38 mitogen-activated protein kinase in pancreatic ␤-cells are unclear. However, G␤␥ and G␣ q (30), components of dynein motors and functional microtubules (31), and a scaffold protein, the GTPase Rac (32), have all been implicated in the proximal signaling steps in different cell types. Downstream, activated caspase-3 and caspase-6 act as "executioner caspases" in serum deprivation-induced apoptosis (33), but caspase-9 does not appear to play a role and there is a lack of apoptosome formation (34), implying that cytochrome c is not involved. In the current study, wortmannin and LY294002 clearly inhibited GIP-mediated stimulation of PKB and Foxo1 phosphorylation and reductions in Bax expression. Responses to the dominant negative ⌬p85 PI3K also provided support for a role for PI3K. Lipotoxic effects on ␤-cells are believed to involve increases in cellular lipid peroxidation and oxidative stress as well as increased ceramide, NF-B, and nitric oxide production (35)(36). The final steps resulting in apoptosis involve both endoplasmic reticulum stress (37) and mitochondrial disruption (38). Glucose toxicity has also been associated with intrinsic pathways involving Bad and Bid that converge on mitochondrial disruption (39). Reductions in cytochrome c release and caspase 9 activation, key components of apoptosis associated with mitochondrial disruption, are probable consequences of the regulation of Bax expression by GIP. Therefore, GIP likely acts on glucolipotoxicity-induced cell death by inhibiting both the early events, such as the activation of p38 mitogen-activated protein kinase via a cAMP-mediated pathway, and the later mitochondrial events, including the currently described decrease in bax expression through depletion of Foxo1 in the nucleus via PI3K. The Bcl-2 family members, including Bax, act as integrators of survival and death signals and play critical roles in setting the threshold at which cells commit to apoptosis (40). Pro-and anti-apoptotic family members can heterodimerize and seemingly titrate one another's function, suggesting that their relative concentration may act as a "rheostat" for the suicide program (41). We have also found that GIP increases bcl-2 gene transcription through cAMPresponse element-binding protein (CREB) phosphorylation, 2 and this is likely to be an additional important pathway in the regulation of ␤-cell apoptosis. Taken together, it seems likely that GIP modulates the ratio between the pro-and anti-apoptotic proteins, Bax and Bcl-2, in the ␤-cell.
In summary, GIP-induced phosphorylation of Foxo1, resulting in inactivation of the bax gene, is likely to be an important pathway by which GIP regulates ␤-cell apoptosis. There is considerable evidence implicating apoptosis as the main mediator of islet ␤-cell death in diabetes, resulting in an absolute (type 1) or relative (type 2) deficiency of insulin-producing pancreatic ␤-cells. Because GIP powerfully reduces levels of the key pro-apoptotic protein Bax, while up-regulating Bcl-2 expression, it is probably a major contributor to the physiological maintenance of ␤-cell mass, and such actions have clear implications for therapeutic intervention.