Transcription Factor FoxO1 Mediates Glucagon-Like Peptide-1 Effects on Pancreatic β-Cell Mass

LY294002 and AG1478 were purchased from Biomol (Plymouth Meeting, PA). Human GLP-1 fragment 7–36 amide and Exendin4 were obtained from Sigma (St. Louis, MO). RPMI 1640, FCS, and other culture media were purchased from Gibco BRL (Burlington, ON, Canada).

INS832/13 (22) cells (passage 36–70) were grown in RPMI 1640 medium supplemented with 10 mmol/l HEPES, 10% heat-inactivated FCS, 2 mmol/l l-glutamine, 1 mmol/l sodium pyruvate, 50 μmol/l β-mercaptoethanol, 100 IU/ml penicillin, and 100 μg/ml streptomycin at 37°C in a humidified 5% CO₂ atmosphere. Cells at 70% confluence were washed with PBS and preincubated in serum-free RPMI supplemented with 5 mmol/l glucose and 0.1% BSA (Fraction V; Sigma) for at least 4 h before treatment. LY294002 and AG1478 were applied 30 min before growth factor treatment.

Eight-week-old wild-type and FoxO1 transgenic mice were treated with daily intraperitoneal injections of Exendin4 (10 nmol/kg) or saline for 7 days (n = 4 each). Animals were killed and pancreas sections processed for insulin and Ki67 immunohistochemistry to assess β-cell proliferation. Cross-sectional islet area was measured using Image Pro Plus software (Media Cybernetics, Silver Spring, MD), and small β-cell clusters (<6 cells) were counted for each section. For each animal, at least three sections spaced 80 μm apart were studied (18).

Proteins were extracted and quantified by BCA (bicinchoninic acid) assay (Roche, Rockford, IL) before fractionation on 8 or 10% polyacrylamide gels. Primary antibodies (anti–forkhead transcription factor Foxo1 [anti-FKHR], anti–phospho-FKHR, anti–extracellular signal–related kinase 1/2 [anti-ERK1/2], and anti–phospho-ERK1/2) were purchased from Cell Signaling (Beverly, MA). Western blotting was performed as described previously (19).

Cells were transfected with FoxO1-GFP (23) using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) in LabTek chambered slides (Nalgene). In brief, 4 ×10⁵ cells were transduced with 1 μg plasmid DNA and 2 μl Lipofectamine 2000 in 0.5 ml RPMI medium. The following day, cells were treated for the indicated time periods and fixed in 4% paraformaldehyde in PBS for 15 min. Image acquisition was performed using a Zeiss confocal microscope (23).

The constitutively nuclear mutant FoxO1 carries single amino acid substitutions replacing the three main phosphorylation sites, Thr²⁴→Ala, Ser²⁵³→Asp, and Ser³¹⁶→Ala²⁴. Constitutively nuclear FoxO1 transduction was carried out as previously described (24). INS832/13 cells were seeded 2 days before use in 100-mm Petri dishes and cultured in RPMI. Cells were then infected with constitutively nuclear FoxO1 or β-galactosidase (β-Gal) adenoviral constructs at a multiplicity of infection of 50 plaque-forming units/cell for 1 h in 1 ml complete medium. The viral solution was then replaced with complete medium, and cells were allowed to recover for 24 h before being used.

Proliferation of INS832/13 cells was evaluated using an enzyme-linked immunosorbent assay–based bromodeoxyuridine (BrdU) incorporation kit (Roche). In brief, INS832/13 cells were transduced for 1 h in 100-mm Petri dishes, trypsinized, seeded in 96-well plates at 70% confluence, and incubated overnight in minimal RPMI medium before the experiment. BrdU was added to the culture medium for 1 h before harvesting cells. Cells were then fixed and incubated with a peroxidase-conjugated anti-BrdU antibody, and the immune complexes were quantified using a spectrophotometer to measure absorbance (Bio-Rad, Hercules, CA).

Cells were plated on chambers slide (Nalgene Nunc, Naperville, IL) and incubated for 24 h in complete RPMI. The following day, cells were washed and treated under different conditions in minimal RPMI medium for 4 h. Caspase activation was measured with the sulforhodamine multi-caspase activity kit and mitochondrial membrane potential with Mit-E-Ψ mitochondria permeability detection kit (Biomol). Briefly, SR-VAD-FMK, a sulforhodamine derivative of a potent caspase inhibitor, was added during the last 1 h of a 4-h incubation. SR-VAD-FMK is cell permeable and covalently binds to active caspases, allowing for detection by fluorescence microscopy. The lipophilic Mit-E-Ψ reagent was added for the last 15 min of a 90-min incubation. Mit-E-Ψ enters the mitochondria and aggregates to emit red fluorescence. In the presence of a loss of mitochondrial membrane potential (ΔΨ), red fluorescence is replaced by cytoplasmic green fluorescence.

RNA was isolated using Qiagen kits (Valencia, CA), and quantitative real-time PCR was performed as described previously (25).

Cells were fixed to isolate intact chromatin and treated as described previously. FoxO1 was immunoprecipitated with anti-FKHR (H-128) antibodies (Santa Cruz Biotechnology, Santa Cruz, CA), and bound DNA was amplified by PCR. Primers sequences are available on request (18,19).

Data are presented as means ± SE. Statistical analyses were performed with SPSS using ANOVA.

FoxO1 is a prominent effector of PI 3-kinase/Akt, a signaling cascade that is activated by GLP-1 in β-cells. We investigated whether GLP-1 affects FoxO1 phosphorylation and subcellular localization in pancreatic β (INS832/13) cells. In cultured β-cells, FoxO1 is constitutively phosphorylated and cytoplasmic, probably because of endogenously produced insulin (26). To measure the effects of GLP-1 on FoxO1 phosphorylation, we cultured INS832/13 cells in serum-free RPMI medium supplemented with 5 mmol/l glucose for 24 h. At the end of the incubation period, we added 10 nmol/l GLP-1 for various lengths of time (Fig. 1A). Under these conditions, GLP-1 induced FoxO1 phosphorylation in a time-dependent manner, with a maximal effect observed at 30 min. The action of GLP-1 was mimicked by the EGFR ligand BTC (not shown). The effect of GLP-1 was blocked by inhibitors of EGFR (AG1478, 250 nmol/l) and PI 3-kinase (LY294002, 50 μmol/l) signaling (Fig. 1B). Insulin also increased FoxO1 phosphorylation. Unlike the effect of GLP-1, that of insulin could be blocked by the PI 3-kinase inhibitor LY294002 but not by the EGFR inhibitor AG1478 (Fig. 1C).

We next studied the GLP-1 effect on the subcellular localization of FoxO1 by fluorescence microscopy. To this end, we transiently expressed a FoxO1-GFP fusion protein in INS832/13 cells (23). GLP-1 caused a time-dependent translocation of FoxO1 from the nucleus to the cytoplasm, consistent with FoxO1 phosphorylation (Fig. 2, first row). Insulin, BTC, and serum had comparable effects (Fig. 2, second row). Inhibition of PI 3-kinase by LY294002 preempted translocation induced by all ligands tested (Fig. 2, third row). In contrast, AG1478 was able to prevent translocation induced by GLP-1 and BTC but not that induced by insulin (Fig. 2, fourth row).

These data indicate that GLP-1 inhibits FoxO1 through phosphorylation-dependent nuclear exclusion. Given that GLP-1 has been shown to activate sequentially EGFR and PI 3-kinase (9,12,13,16), the observed changes in FoxO1 phosphorylation/subcellular localization suggest that FoxO1 mediates GLP-1 action on β-cells.

Because FoxO1 controls cell cycle progression (27), we examined whether its inhibition is required for GLP1-dependent β-cell proliferation. Incubation of serum-deprived cells with high glucose increased BrdU incorporation by ∼50%, as did GLP-1 and BTC. Addition of 10% serum to the medium resulted in a 3.5-fold increase in BrdU incorporation (Fig. 3A). We next transduced cells with a constitutively nuclear FoxO1 mutant (20) or with control β-gal adenovirus. Overexpression of constitutively nuclear FoxO1 inhibited β-cell proliferation induced by glucose, GLP-1, BTC, and serum, without significantly affecting basal proliferation, whereas control virus encoding β-gal had no effect (Fig. 3A). This result indicates that FoxO1 nuclear exclusion is necessary for GLP-1–and BTC-induced proliferation.

To evaluate whether FoxO1 inactivation was required for the antiapoptotic effect of GLP-1, we examined caspase activation in response to H₂O₂, an inducer of oxidative stress (28), using a fluorescent probe in cells transduced with constitutively nuclear FoxO1 or β-Gal. Addition of H₂O₂ increased the number of cells expressing activated caspase by ∼threefold (Fig. 3B). GLP-1 reversed the effects of H₂O₂. Transduction with constitutively nuclear FoxO1 blocked the ability of GLP-1 to decrease the number of caspase-positive cells (Fig. 3B). These data indicate that GLP-1 action on β-cell survival requires FoxO1 inhibition.

Our results indicate that constitutively nuclear FoxO1 inhibits GLP-1 action on proliferation and survival. To determine the mechanism by which FoxO mediates its effects in β-cells, we examined whether constitutively nuclear FoxO1 affects GLP-1–dependent increase in cAMP levels, preserves mitochondrial membrane potential, and prevents GLP-1–induced ERK1/2 activation, three metabolic actions of the hormone that do not require gene transcription. GLP-1 induced a rise in cAMP levels in β-Gal–transduced cells, as did treatment with 10 mmol/l nicotinamide (29). Basal cAMP levels were ∼50% lower in cell transduced with constitutively nuclear FoxO1. This decrease is similar to that observed in growth-arrested cells and is likely to reflect the FoxO1-induced growth arrest. However, the magnitude of the GLP-1–induced rise was unaffected by constitutively nuclear FoxO1 (Fig. 4A). Next, we used a fluorescent probe (Mit-E-Ψ) to evaluate changes in mitochondrial membrane potential in cells transduced with constitutively nuclear FoxO1 after H₂O₂ treatment in the absence or presence of GLP-1. H₂O₂ caused a disruption of mitochondrial membrane potential, indicated by the reduction in red fluorescence and the appearance of green fluorescence. GLP-1 preserved the integrity of mitochondrial membrane potential in the presence of H₂O₂ and constitutively nuclear FoxO1 (Fig. 4B). Identical observations were made in untransduced cells (not shown). These data indicate that FoxO1 does not affect GLP-1 action on mitochondrial membrane potential. The mechanism by which GLP-1 acts on the mitochondria to preserve the membrane potential remains elusive but may be related to stimulation of mitochondrial ATP synthesis (30). Next, we investigated whether constitutively nuclear FoxO1 interferes with the ability of GLP-1 to activate ERK1/2. In untransduced cells, GLP-1 caused a moderate increase in ERK1/2 phosphorylation by 177 ± 34% (P < 0.05), as previously reported (12). Transduction of INS cells with constitutively nuclear FoxO1 did not prevent the GLP-1–induced rise in phospho-ERK (Fig. 4C). These results indicate that constitutively nuclear FoxO1 does not affect actions of GLP-1 that do not require gene transcription.

To investigate the role of FoxO1 in the proliferative actions of GLP-1 in vivo, we determined whether GLP1 affects β-cell mass in transgenic mice expressing constitutively nuclear FoxO1 in β-cells. The mutant transgene carries a single amino acid substitution replacing the main Akt phosphorylation site and is therefore constitutively nuclear (line 305) (19). We administered either saline or the long-acting GLP-1 analog Exendin4 (10 nmol/kg) daily for 7 days to either wild-type or 305 transgenic mice. We then measured β-cell proliferation by Ki67 immunostaining and islet mass by morphometry of pancreatic sections stained with anti-insulin antiserum (representative images are shown in Fig. 5A). Exendin4 administration resulted in a fourfold increase in Ki67-positive β-cells in control animals (Fig. 5B) and a twofold increase in cross-sectional islet area (Fig. 5C). We also observed a small increase in the number of small clusters of insulin-positive cells (<6 cells) in wild-type mice treated with Exendin4 compared with saline-treated littermates (Fig. 5D). In contrast, 305 transgenic mice failed to increase the number of Ki67-positive cells, islet size, and small β-cell clusters (Fig. 5B–D). These observations indicate that FoxO1 inactivation is required for the proliferative action of GLP-1 in vivo.

We next evaluated the contribution of FoxO1 nuclear exclusion to the GLP-1 transcriptional response in β-cells. GLP-1 increased the levels of mRNAs encoding Pdx1, Foxa2, and Insulin2 and decreased the levels of Igfbp1 (Fig. 6A). In all instances, constitutively nuclear FoxO1 prevented the effect of GLP-1. To determine whether this effect was due to a direct inhibition of FoxO1-dependent transcription, we measured occupancy of the forkhead binding site in the Igfbp1 and Foxa2 promoters using chromatin immunoprecipitation assays. Under basal conditions, we detected FoxO1 bound to both promoters. GLP-1 inhibited FoxO1 binding. GLP-1 was similarly able to decrease FoxO1 binding in H₂O₂-treated cells, consistent with its effect to inhibit H₂O₂-induced caspase activation (Fig. 3). In contrast, GLP-1 had no effect in cells expressing constitutively nuclear FoxO1 (Fig. 6B). It should be pointed out that GLP-1 inhibits Igfbp1 and promotes Foxa2 expression. Because FoxO1 binds to both promoters, the data are consistent with the notion that it acts as a transcriptional activator for some genes and as a repressor for others (18).

GLP-1 is a potent glucoincretin hormone with valuable properties for diabetes treatment. Although its proliferative and antiapoptotic properties have been demonstrated in the β-cell, the molecular mechanism by which GLP-1 exerts its effect on β-cell mass is not fully understood. Pharmacological and biochemical evidence indicates that PI 3-kinase/Akt signaling plays a central role in GLP-1 action on β-cell growth and survival. In the present study, we investigated the role of the transcription factor FoxO1 in the action of GLP-1 on β-cell proliferation and survival. As recently hypothesized (31), we show that GLP-1 inhibits FoxO1 transcriptional activity through phosphorylation-dependent nuclear exclusion in INS832/13 cells. Moreover, the ability of the long-acting GLP-1 derivative Exendin4 to increase β-cell mass was blunted in transgenic mice expressing constitutively nuclear FoxO1 in β-cells. Finally, we demonstrate that FoxO1 inactivation plays an important role in the effect of GLP-1 on the expression of important β-cell transcription factors, such as Pdx1 and Foxa2. Our data indicate that FoxO1 inhibition is required for GLP1-dependent proliferation (Fig. 7).

Restoration of insulin secretion and β-cell mass is central to the treatment of diabetes (32–34). Through its pivotal role in PI 3-kinase/Akt signaling, FoxO1 has the potential to integrate the proliferative and antiapoptotic signals of β-cell growth factors. We have previously shown that FoxO1 plays an important role in metabolic control, but because of the complex interactions among different tissues, it remained unclear whether this effect is due to intrinsic actions of FoxO1 in the β-cell or to its regulation of peripheral insulin sensitivity (19). In this study, we show that expression of constitutively nuclear FoxO1 inhibits β-cell proliferation induced by glucose, GLP-1, BTC, and serum. Moreover, a recent publication demonstrates that FoxO1 mediates the action of glucose-dependent inhibitory peptide on β-cell survival by regulating Bax expression (35). Based on those results and on the present study, it appears that FoxO1 is a general regulator of β-cell mass in response to incretins.

The mechanism by which FoxO1 inhibition by GLP-1 increases β-cell mass is complex. We have previously shown that FoxO1 inhibits Pdx1 expression, thereby providing a mechanism by which FoxO1 controls β-cell proliferation and insulin secretion (18). Several publications indicate that Pdx1 plays a role in β-cell differentiation, replication, and regeneration. Thus, Pdx1 haploinsufficient mice fail to increase β-cell mass in response to insulin resistance (36), and increased Pdx1 expression is associated with an early burst of β-cell proliferation after subtotal pancreatectomy (37). It has also been shown that GLP-1 increases Pdx1 expression, but the mechanism is unclear (9). The present findings indicate that GLP1 increases Pdx1 expression via inhibition of FoxO1. Moreover, we show that GLP-1 increases Foxa2 expression, and this is associated with decreased FoxO1 binding to the Foxa2 promoter. Thus, it appears that the mechanism by which FoxO1 controls Foxa2 expression is similar to the mechanism by which it controls Pdx1. Further work will be required to understand the molecular mechanism of transcriptional inhibition by FoxO1.

In conclusion, we show that FoxO1 is in the GLP-1 signal transduction pathway. As GLP receptor agonists enter the clinical arena, FoxO1 represents a potential target to modulate its beneficial effects in diabetes.

J.B. is a Juvenile Diabetes Research Foundation Postdoctoral Fellow. D.A. received National Institutes of Health Grant DK64819.