Acute Insulin Signaling in Pancreatic Beta-Cells Is Mediated by Multiple Raf-1 Dependent Pathways

Abstract

Insulin enhances the proliferation and survival of pancreatic β-cells, but its mechanisms remain unclear. We hypothesized that Raf-1, a kinase upstream of both ERK and Bad, might be a critical target of insulin in β-cells. To test this hypothesis, we treated human and mouse islets as well as MIN6 β-cells with multiple insulin concentrations and examined putative downstream targets using immunoblotting, immunoprecipitation, quantitative fluorescent imaging, and cell death assays. Low doses of insulin rapidly activated Raf-1 by dephosphorylating serine 259 and phosphorylating serine 338 in human islets, mouse islets, and MIN6 cells. The phosphorylation of ERK by insulin was eliminated by exposure to a Raf inhibitor (GW5074) or transfection with a dominant-negative Raf-1 mutant. Insulin also enhanced the interaction between mitochondrial Raf-1 and Bcl-2 agonist of cell death (Bad), promoting Bad inactivation via its phosphorylation on serine 112. Insulin-stimulated ERK phosphorylation was abrogated by calcium chelation, calcineurin and calmodulin-dependent protein kinase II inhibitors, and Ned-19, a nicotinic acid adenine dinucleotide phosphate receptor (NAADPR) antagonist. Blocking Raf-1 and Ca²⁺ signaling resulted in nonadditive β-cell death. Autocrine insulin signaling partly accounted for the effects of glucose on ERK phosphorylation. Our results demonstrate that Raf-1 is a critical target of insulin in primary β-cells. Activation of Raf-1 leads to both an ERK-dependent pathway that involves nicotinic acid adenine dinucleotide phosphate-sensitive Ca²⁺ stores and Ca²⁺-dependent phosphorylation events, and an ERK-independent pathway that involves Bad inactivation at the mitochondria. Together our findings identify a novel insulin signaling pathway in β-cells and shed light on insulin’s antiapoptotic and mitogenic mechanisms.

Insulin acts through a network of proteins including Raf-1 to promote the survival of insulin-secreting pancreatic beta-cells.

Insulin supports sufficient pancreatic β-cell mass by increasing the proliferation and enhancing the survival of β-cells (1,2,3). However, the signal transduction mechanisms downstream of β-cell insulin receptors are not well understood and remain controversial. Upon insulin binding to its receptors in other tissues, two major signaling pathways can be activated, the phosphoinositide 3-kinase/phosphoinositide-dependent kinase 1/Akt pathway and the Ras/Raf-1/ERK cascade. To date, the majority of studies on β-cell survival have focused on upstream regulators and downstream targets of Akt kinase (4,5). Studies performed with cultured insulinoma cells and transgenic overexpression systems initially suggested that insulin may promote β-cell survival via Akt (6). However, transgenic mice lacking virtually all β-cell Akt activity have normal β-cell mass and do not exhibit increased β-cell apoptosis (4), suggesting that another arm of the insulin signaling pathway might be more important for the regulation of β-cell fate.

Raf-1 kinase has only recently been investigated in β-cells and upstream factors that activate Raf-1 have not been identified. We demonstrated that endogenous Raf-1 signaling is critical for suppressing basal β-cell apoptosis (7). Raf-1 also appears to participate in β-cell proliferation (8,9). Raf-1 is ubiquitously expressed and tightly regulated at the posttranslational level by phosphorylation, interactions with adaptor/scaffolding proteins and by its subcellular localization (10). Raf-1 is localized in the cytoplasm, mitochondria, and the nucleus in islet β-cells and MIN6 mouse insulinoma β-cells (7). Full Raf-1 activation involves the dephosphorylation of an inhibitory site at serine 259 (11) and phosphorylation of an activation site at serine 338 (12). Active Raf-1 can then phosphorylate MAPK kinase, an upstream kinase activator of ERK. Additionally, a novel ERK-independent mechanism involving Bcl-2-mediated targeting of Raf-1 to the mitochondria has been described (13). Raf-1 phosphorylates Bad on serine 112 at the outer mitochondrial membrane, thereby causing the inactivation and sequestration of Bad in the cytoplasm by 14-3-3 scaffolding proteins.

Ca²⁺ stores sensitive to the second messenger nicotinic acid adenine dinucleotide phosphate (NAADP) are present in human β-cells and β-cell lines (14,15,16). NAADP is essential for the initiation of Ca²⁺ signals by insulin (14). Pancreatic β-cells from mice lacking CD38, one of the enzymes capable of generating NAADP, have reduced Ca²⁺ signals in response to insulin but not glucose (17). CD38-null islets also display increased apoptosis (17). In T cells, increased CD38 activity within lipid rafts leads to the activation of prosurvival ERK pathways (18). Whether NAADP-sensitive Ca²⁺ stores play a role in ERK activation in β-cells remains untested.

In the present study, we investigated the mechanisms by which insulin acts on β-cells. We evaluated if insulin activates Raf-1 and examined downstream targets of Raf-1, including both ERK activation and Bad inactivation at the mitochondria. Moreover, we tested whether insulin stimulates ERK via release of Ca²⁺ from NAADP-sensitive intracellular stores.

Materials and Methods

Reagents

Reagents were purchased from Sigma (St. Louis, MO). Raf inhibitor (GW5074) and 1,2-bis-(o-aminophenoxy)-ethane-N,N, N′,N′-tetraacetic acid, tetra(acetoxymethyl)-ester (BAPTA-AM) were from Calbiochem/EMD (Gibbstown, NJ) and Biomol International (Plymouth Meeting, PA), respectively. Raf-1 (51-131)-green fluorescent protein (GFP) was provided by Dr. Tamas Balla (National Institutes of Health, Bethesda, MD). Ned-19 (19) was a gift from Dr. Grant Churchill (University of Oxford, Oxford, UK).

Cell culture

Mouse islets were isolated from 3-month-old C57BL6/J mice using a previously described protocol (7) approved by the University of British Columbia Animal Care Committee. Insulin knockout mice were provided by Professor Jacques Jami (Institut National de la Santé et de la Recherche Médicale, Paris, France) and have been described (20). Human islets were provided by the Centre Human Islet Transplant and β-cell Regeneration (Vancouver, British Columbia, Canada). Islets were cultured in RPMI 1640 media with 5 mm glucose, 10% fetal bovine serum, 100 IU/ml penicillin, and 100 μg/ml streptomycin. MIN6 cells were cultured in DMEM containing 25 mm glucose with 10% fetal bovine serum (Invitrogen, Burlington, Ontario, Canada). Islets or MIN6 cells were serum starved for 6 h before treatment and washed three times with serum-free media before culturing in 5 mm glucose RPMI 1640 or DMEM with insulin with or without inhibitors.

Protein detection and mitochondria isolation

Immunoblotting was performed as described (7). MIN6 cells and islets were washed after treatments with PBS before adding cell lysis buffer with protease inhibitor cocktail (Cell Signaling, Beverly, MA). Primary antibodies against phosphorylated Raf-1 (serine 259 and serine 338), ERK, phosphorylated ERK (Thr202/Try204), Bcl-2 antagonist of cell death (Bad), and phosphorylated Bad (serine 112) were from Cell Signaling. Antibodies against β-actin, cytochrome c oxidase IV, and Raf-1 were from Novus Biologicals (Littleton, CO), Abcam (Cambridge, MA), and BD Biosciences (San Jose, CA), respectively. Mitochondria isolation kit (MITOISO1) was from Sigma.

Dominant-negative Raf-1 overexpression

Raf-1 (51-131)-GFP dominant-negative mutant DNA (1 μg) was transiently transfected into MIN6 cells (7,21). As a control, 1 μg of DNA of enhanced-GFP was transfected. Twenty-four hours after transfection using Lipofectamine 2000 (Invitrogen, Burlington, Ontario, Canada), MIN6 cells were washed and incubated with serum-free 5 mm glucose DMEM for 24 h and then subjected to insulin treatment for 5 min before protein detection by immunoblotting.

Coimmunoprecipitation

Coimmunoprecipitations of Raf-1 and Bcl-2 family proteins (Bcl-2, Bad, and Bcl-xL antibodies were from Cell Signaling) were performed on MIN6 cells lysates. Briefly, 600 μg of total protein were incubated with antibodies overnight, after which 50 μl of protein A/G beads (Santa Cruz Biotechnology, Santa Cruz, CA) were incubated for 3 h in PBS with protease inhibitor cocktail at 4 C. The slurry of antibody, beads, and lysates was spun down for 30 sec at 8000 rpm using Ultrafree-MC filters (Millipore, Billerica, MA). The immunoprecipitation beads were then washed three times with PBS with protease inhibitors, and the immunoprecipitate was then electrophoresed on a sodium dodecyl sulfate gel and probed with anti-Raf-1 antibody overnight. Raf-1 antibodies were from BD Biosciences and Epitomics (Burlingame, CA) and detected using protein A conjugated with horseradish peroxidase.

Immunofluorescence imaging

Immunofluorescence analysis of pancreas sections and dispersed islet β-cells was performed as previously described (7). Insulin receptor and phosphorylated Raf-1 antibodies were from Cell Signaling and phosphorylated insulin receptor, Raf-1, and insulin antibodies were from Stressgen (Ann Arbor, MI), BD Biosciences, and Linco Research (St. Charles, MO), respectively. Secondary antibodies conjugated to Texas Red, fluorescein isothiocyanate, or 7-amino-4-methyl-coumarinylacetic acid were from Jackson ImmunoResearch (West Grove, PA). Pearson’s correlation r values were calculated using SlideBook software (Intelligent Imaging Innovations, Denver, CO).

Live-cell Ca²⁺ imaging

Single-cell imaging was performed in HEPES-buffered Ringer’s solution containing 144 mm NaCl, 5.5 mm KCl, 1 mm MgCl₂, 2 mm CaCl₂, 20 mm HEPES, and 3 mm glucose. Cytosolic Ca²⁺ was imaged in fura-2-AM or fura-4F-AM-loaded cells or in D3cpv-expressing cells (22). Endoplasmic reticulum (ER) luminal Ca²⁺ was imaged using the fluorescence resonance energy transfer-based D1ER cameleon (23). MIN6 cells that exhibited a quiescent baseline Ca²⁺ level at the beginning of the experiment in a 3 mm glucose incubation (5–15 min) and a positive response to 30 mm KCl treatment at the end of the experiment were considered viable and analyzed. A cell was considered to have responded to insulin if its Ca²⁺ level differed significantly from baseline for at least five time points and if the fluorescent signal increased or oscillated above baseline levels by 5%.

Analysis of programmed cell death, proliferation, and insulin secretion

Primary mouse islets were dispersed (5,000–10,000 cells/well) 24 h after isolation, allowed to attach in a 96-well Viewplate (Packard, Meriden, CT), and left to grow overnight before adding treatments. For MIN6 cells, 10,000–15,000 cells/well were seeded. Treatments were prepared in serum-free media containing 2.5 ng/μl of propidium iodide (PI) and 5 mm glucose in the presence or without GW5074 and BAPTA-AM. PI incorporation was measured every hour up to 48 h by a high-throughput imaging system (Thermo/Cellomics, Pittsburgh, PA). Immunofluorescence analyses of proliferating MIN6 cells were performed as described (8). Primary antibody to 5-bromo-2′-deoxyuridine was from Roche (Laval, Québec, Canada). 4′,6′-diamino-2-phenylindole (Vector Labs, Burlingame, CA) was used as a nuclear counterstain. Insulin secretion was measured from conditioned-media by RIA (Rat insulin RIA kit; Linco Research).

Statistical analysis

Statistical significance was determined using unpaired, two-tailed Student’s t tests calculated in GraphPad Prism (La Jolla, CA) and Excel (Microsoft, Redmond, WA). ANOVA (Student-Newman-Keuls) was used where appropriate. Results are presented as mean ± sem and considered to be statistically significant when P < 0.05.

Results

Insulin induces Raf-1 and ERK activation in primary islets and MIN6 cells

Insulin receptors are highly expressed in pancreatic β-cells and can be found in the plasma membrane, cytoplasm, and endosomes (24) (supplemental Fig. S1, A–C, published as supplemental data on The Endocrine Society’s Journals Online web site at http://endo.endojournals. org). We have previously shown that a low dose of insulin (0.2 nm) was more effective at preventing β-cell apoptosis than higher doses (1,25). Unfortunately, little is known about the signaling pathways activated by low doses of insulin, specifically those that would not activate IGF-I receptors (26). Given our previous observations suggesting that low doses of insulin do not activate Akt (1), we hypothesized that insulin might activate Raf-1 in β-cells. In primary mouse islets, we found that insulin caused Raf-1 dephosphorylation at serine 259 (Fig. 1A), an event that is known to initiate activation of this kinase (27). Subsequent to the dephosphorylation of Raf-1 at serine 259, we observed a significant increase in phosphorylation of the stimulatory site of Raf-1 at serine 338 (Fig. 1B and supplemental Fig. S2A), suggesting activation of this kinase (12). To determine the outcome of Raf-1 activation, we investigated the phosphorylation status of ERK, the canonical downstream target of Raf-1 (7,28). Insulin increased ERK phosphorylation in mouse islets (Fig. 1C, supplemental Fig. S2B). We observed a similar activation of Raf-1 and ERK by insulin in human islets (supplemental Fig. S2, C and D). These are the first data demonstrating that insulin rapidly activates both Raf-1 and ERK in primary islets. In each of these cases, the effects of 0.2 nm insulin were robust and reproducible. Higher doses of insulin showed more variable responses and were not consistently more effective than lower doses. These observations mirror previous findings that β-cell survival and proliferation are selectively enhanced by 0.2 nm insulin relative to higher doses (1,8).

Figure 1.

Insulin promotes Raf-1 and ERK activation in mouse islets. A, Acute insulin signaling stimulation for 15 min in primary mouse islets with 0.2 and 200 nm insulin resulted in a loss of inhibitory phosphorylation of Raf-1 (pRaf-1) at serine 259. B, Mouse islets treated with 0.2 and 200 nm insulin for 30 min also caused an increase of the stimulatory phosphorylation of Raf-1 at serine 338 and a prosurvival phosphorylation of ERK (pErk; C). Bar graphs are quantification of Western blots using densitometry (n = 4 for mouse islets). *, Significant difference between treatment and serum-free control.

Glucose has been shown to induce ERK phosphorylation in islets as well as β-cell lines (29,30). High glucose transiently induced ERK activation within 15 min (Fig. 2A), after which ERK phosphorylation returned to baseline (data not shown). However, because glucose is a potent insulin secretagogue, we tested whether these effects might be due in part to autocrine insulin signaling or whether they were mediated by an insulin-independent mechanism. We incubated MIN6 cells at 5 mm glucose or 25 mm glucose in the presence or absence of 1 μm somatostatin, which blocks glucose-induced insulin secretion (8). Glucose-induced ERK phosphorylation was completely blocked by somatostatin (Fig. 2A), consistent with a requirement for insulin secretion for ERK activation. Similarly, Ins1^−/−-Ins2^+/− mice with about 50% reduced insulin secretion in vivo (20,31) (Mehran, A., and J. Johnson, unpublished data) and in vitro (Fig. 2B) displayed an approximately 50% decrease on ERK activation induced by high glucose when compared with islets from phenotypically normal Ins1^−/−-Ins2^+/+ mice (Fig. 2C). While additional experiments are required to elucidate the mechanism of glucose-stimulated ERK activation, this finding implicates insulin feedback signaling as a factor in the mitogenic effects of acute hyperglycemia and provides the basis for detailed future studies.

Figure 2.

Insulin activates Raf-1 and ERK in transformed MIN6 cells. Panel A, ERK phosphorylation in MIN6 cells treated with low glucose (5 mm) or high glucose (25 mm) for 15 min with or without somatostatin (Soma, 1 μm; n = 3). Panel B, Insulin levels were measured in conditioned media of islets from Ins1^−/−-Ins2^+/+ mice treated with low and high glucose (n = 4). Panel C, ERK phosphorylation (pErk) in islets from Ins1^−/−-Ins2^+/− or Ins1^−/−-Ins2^+/+ mice or in response to 15 min treatment with 25 mm glucose or 5 mm glucose. Panels D and E, Insulin-induced ERK phosphorylation was blocked by a Raf inhibitor (E-GFP:GW5074; 10 μm; n = 4) or a dominant-negative Raf-1 (51-131)-GFP protein (n = 3). Panels F and G, Insulin (0.2 nm) caused an increase in Raf-1 and ERK phosphorylation in a time-dependent manner (n = 3). Panel H, Insulin induced phosphorylated ERK translocation in the nucleus of MIN6 cells (57,556 of 2,300,189 cells had nuclear ERK phosphorylation in control; 190,013 of 1,195,820 MIN6 cells had nuclear ERK phosphorylation with insulin). Bar graphs are quantification of Western blots using densitometry. *, Significant difference between treatment and control (serum free); ^, significant difference between insulin and insulin with GW5074 or dominant-negative Raf-1 (51-131)-GFP.

Next, we examined the requirement of Raf-1 for the activation of ERK by insulin as well as the dose dependence and time course of these events. We previously demonstrated that a Raf inhibitor (GW5074) and a dominant-negative Raf-1 mutant robustly inhibited basal ERK phosphorylation (7). In the present study, blocking Raf-1 kinase activity with GW5074 completely prevented insulin-induced ERK phosphorylation (Fig. 2D). However, this inhibitor has been shown to also block B-Raf and other protein kinases (32). We also used a dominant-negative Raf-1 mutant [Raf-1 (51-131)-GFP] and found that ERK phosphorylation induced by 0.2 nm insulin was similarly blocked (Fig. 2E). Together, these results are consistent with a critical role for Raf-1 in insulin-induced ERK activation.

We also examined the rapid kinetics of Raf-1 and ERK stimulation in MIN6 cells (Fig. 2, F and G). The temporal profile of Raf-1 activation was sustained, whereas ERK activation was more transient. In many cell types, ERK phosphorylation in the cytoplasm is immediately followed by its translocation into the nucleus in which it initiates the transcription of prosurvival proteins. MIN6 cells treated with insulin showed an increased number of cells with phosphorylated ERK in the nucleus (Fig. 2H). These data suggest that insulin rapidly activates Raf-1 and its canonical downstream target, the ERK pathway, in the β-cell.

Raf-1 and Bcl-2 protein families in pancreatic-β-cells

An ERK-independent mechanism involving Bcl-2-mediated targeting of Raf-1 to the mitochondria has recently been described (33). In this paradigm, Raf-1 is thought to act at the mitochondria to phosphorylate Bad at serine 112, which then inactivates this proapoptotic protein by tethering it to 14-3-3 proteins in the cytoplasm (34). We reported that a large fraction of Raf-1 localizes to the mitochondria in β-cells (7). In the present study, we observed a moderate but significant increase in mitochondrial Raf-1 localization in β-cells treated with 0.2 nm insulin compared with serum-free treated β-cells (Fig. 3, A and B). Insulin also had clear effects on mitochondrial morphology, inducing networks of rod and thread-shaped mitochondria, which contrasted with the clustered and rounded mitochondria seen in serum-free conditions (Fig. 3A). Furthermore, we detected increased phosphorylated Raf-1 at serine 338 in the mitochondrial fraction of MIN6 cells treated with insulin (Fig. 3C). Using immunofluorescence imaging, endogenous Raf-1 and Bad exhibited overlapping staining patterns in human β-cells, mouse β-cells, and MIN6 cells (Fig. 4A; data not shown). This association was specific for Raf-1 because the related kinase B-Raf showed little colocalization with Bad (supplemental Fig. S3A). Using immunoprecipitation, we confirmed the close physical proximity of these two proteins with Bcl-2, Bcl-xL and Bad antibodies pulling down endogenous Raf-1 (Fig. 4B). Moreover, insulin increased the protein-protein interaction between Raf-1 and Bad (Fig. 4, C and D). We previously demonstrated that blocking Raf-1 caused a dramatic loss of Bad phosphorylation at serine 112 (7), and thus, we would expect insulin to have the converse effect. Indeed, primary mouse islets treated with insulin showed an increase in Bad phosphorylation at serine 112 (Fig. 4E), a step that prevents Bad from performing its proapoptotic actions. Our findings strongly suggest that insulin modulates Bad-dependent apoptosis via Raf-1 in β-cells.

Figure 3.

Insulin promotes Raf-1 translocation to the mitochondria. A, Insulin (0.2 nm) promoted Raf-1 mitochondrial localization compared with serum free in mouse pancreatic β-cell (n = 3). Scale bars, 10 μm. B, A plot of Pearson correlation r values between Raf-1 and Red-mitochondria tracker in mouse pancreatic β-cells treated in serum-free (SF) or 0.2 nm insulin condition. C, Phosphorylated Raf-1 (pRaf-1; serine 338), Raf-1, and cytochrome c oxidase IV (CoxIV) levels in total cell lysates or mitochondrial fraction of MIN6 cells treated with insulin for 10 min. Bar graph is a densitometry quantification (n = 3). *, Significant difference between treatment and control (serum free).

Figure 4.

Endogenous Bcl-2 family members and Raf-1 form protein-protein interactions in pancreatic β-cell. A, Immunofluorescence imaging of endogenous Raf-1 and Bad in primary human and mouse β-cells. Pearson correlation r values between Raf-1 and Bad in human and mouse β-cells were 0.7 and 0.86, respectively. Scale bars, 10 μm. B, Immunoprecipitation (IP) demonstrating Raf-1 ability to form protein-protein interaction with Bcl-2 family proteins (Bcl-2, Bcl-xL, and Bad). IB, Immunoblot. C, Insulin increased Raf-1 and Bad protein-protein interaction (n = 3). D, Mouse islets treated with 0.2 and 200 nm insulin for 30 min caused an increase in the inhibitory phosphorylation of Bad (pBad) at serine 112 (n = 3). Bar graphs are quantification of Western blots using densitometry. *, Significant difference between treatment and serum-free control.

Ca²⁺ signals are required for insulin-induced ERK phosphorylation

In human and mouse islet β-cells, acute treatment with insulin stimulates Ca²⁺ release from intracellular stores (14,17,35). We further confirmed the presence of these previously described Ca²⁺ signals (14) in MIN6 cells using both conventional Ca²⁺ dyes and fluorescent protein-based Ca²⁺ indicators. Small cytosolic Ca²⁺ signals were evident in human β-cells or MIN6 cells exposed to increasing doses of insulin (Fig. 5, A and B), as we have demonstrated previously with mouse β-cells (17). Using the response criteria described in Materials and Methods, 59 of 550 MIN6 cells showed a positive response to insulin. The average response rate per coverslip was 13.61 ± 3.3%. In control MIN6 cells not exposed to insulin, only 4 of 239 cells responded, a typical level of basal activity. The nature of these Ca²⁺ signals varied between cells, as we have observed previously (14,17). Most insulin-induced Ca²⁺ responses had short spikes with a negligible plateau phase and were smaller in amplitude compared with signals elicited by depolarization with 30 mm KCl. Next we addressed whether these Ca²⁺ signals were required for insulin-induced ERK phosphorylation by treating β-cells with the Ca²⁺ chelator BAPTA-AM. BAPTA-AM blocked the activation of ERK by insulin (Fig. 5C). We did not observe significant changes in total ERK in cells treated with BAPTA-AM for 10 min. Chronic treatment with BAPTA-AM mimicked the effects of the Raf inhibitor, reducing both ERK and Bad phosphorylation and increasing cell death in a manner that was not additive to Raf-1 inhibition alone (Fig. 5, D–H). These data confirm that insulin stimulates cytosolic Ca²⁺ signals and demonstrate that these antiapoptotic signals are required for the activation of ERK.

Figure 5.

Ca²⁺ signaling modulators abrogates insulin-induced ERK activation. A, Ca²⁺ signals generated by insulin in individual dispersed human islet cells loaded with fura-4F in response to a range of insulin doses. B, Ca²⁺ traces of MIN6 cells loaded with fura-2 exposed to a ramp of insulin doses (550 cells imaged; 59 cells responded to insulin). Baseline solutions contained 3 mm glucose. C, MIN6 cells treated with insulin showed an increase on ERK phosphorylation (pErk) in 10 min that was blocked in the presence of BAPTA-AM. FBS, Fetal bovine serum. D, Phosphorylated ERK levels in MIN6 cells treated with GW5074 (10 μm) and BAPTA-AM (50 μm) individually or both for 6 h (n = 3). E, Phosphorylated Bad (serine 112) levels in MIN6 cells treated with GW5074 and BAPTA-AM individually or both for 6 h (n = 3). Rapid increase in PI incorporation in dispersed mouse islet cells (F) and MIN6 cells (G) treated with GW5074, BAPTA-AM, or both (n = 6). Area under the curve (AUC) of PI incorporation from 1 to 48 h after treatment. H, Cleaved caspase-3 expression levels in MIN6 cells treated with GW5074 and BAPTA-AM for 6 h (n = 3). I–K, MIN6 cells treated with insulin show an increase in ERK phosphorylation that was blocked in the presence of Ned-19 (I, NAADP receptor antagonist), KN-93 (J, CaMKII inhibitor), and FK506 (K, calcineurin inhibitor) for 10 min treatment (n = 3). Bar graphs are quantification of Western blots using densitometry. *, Significant difference (P < 0.05) between the serum-free control and treatment; ^, significant difference between insulin and insulin with inhibitors: BAPTA-AM (50 μm), Ned-19 (100 μm), KN-93 (1 μm), and FK506 (50 nm).

We have shown previously that insulin-stimulated Ca²⁺ signals are initiated by NAADP-sensitive Ca²⁺ stores (14) and reduced in mice lacking CD38, an enzyme responsible for generating some of the cellular NAADP (17). Here we used a new NAADP receptor antagonist, Ned-19 (19), to assess the role of these Ca²⁺ stores in ERK phosphorylation. Ned-19 reduced the effects of insulin on ERK phosphorylation (Fig. 5I). These data are the first to directly implicate NAADP-sensitive Ca²⁺ stores on ERK activation in any cell type. NAADP-sensitive stores are thought to reside in acidic organelles, such as lysosomes (19) or vesicle-associated membrane protein-containing granules (15). Consistent with these findings, we did not observe acute, robust changes in ER Ca²⁺ levels in insulin-treated MIN6 β-cells (supplemental Fig. S4A) using the ER luminal cameleon D1ER (23). Direct imaging of Ca²⁺ dynamics in non-ER Ca²⁺ stores such as lysosomes awaits the development of pH-insensitive cameleons. These findings indicate that the initial Ca²⁺ mobilization induced by insulin is a non-ER Ca²⁺ store but do not rule out secondary effects on ER stores via Ca²⁺-induced Ca²⁺ release. Interestingly, preliminary experiments point to a potential role for thapsigargin-sensitive stores in β-cell proliferation (supplemental Fig. S5A). Thapsigargin-sensitive sarcoendoplasmic reticulum Ca²⁺-ATPase pumps have been implicated in both ER and Golgi Ca²⁺ homeostasis (36).

Our results with BAPTA-AM and Ned-19 pointed to Ca²⁺-dependent regulation of Raf-1 and ERK phosphorylation status. However, BAPTA-AM alone can have effects on ERK phosphorylation and events associated with ER stress (37,38). Thus, we sought to further examine the role of Ca²⁺ on the regulation of ERK phosphorylation using inhibitors of a Ca²⁺ dependent kinase and a Ca²⁺ phosphatase. Calmodulin-dependent protein kinase II (CaMKII) is known to mediate the activation of Raf-1 and ERK by Ca²⁺ in other cell types (29,39). The effects of insulin on ERK phosphorylation were abrogated by inhibitors of CaMKII, KN93 (Fig. 5J) and KN62 (data not shown). KN93 did not change the amount of total ERK after 10 min of treatment. As noted above, the activation of ERK by insulin also required the dephosphorylation of serine 259 on Raf-1, pointing to a Ca²⁺-dependent phosphatase such as calcineurin. Indeed, the clinically relevant calcineurin inhibitor FK506 blocked insulin-induced ERK phosphorylation (Fig. 5K). These data further indicate the importance of Ca²⁺ signaling and point to CaMKII and calcineurin as critical upstream regulators of Raf-1/ERK signaling in β-cells.

Discussion

The goal of this study was to determine the signaling pathways activated by low doses of insulin, a growth factor that is now recognized as an important endogenous regulator of β-cell survival and proliferation (2,3,24). We found that insulin activated Raf-1 kinase, leading to both ERK activation and the inactivation of Bad. We identified novel components linking the insulin receptor to ERK activation, namely NAADP-sensitive Ca²⁺ stores, calcineurin, and CaMKII (supplemental Fig. S6A). These findings illuminate previously unidentified mechanisms of rapid autocrine insulin signaling in the β-cell.

Numerous studies point to insulin as a key regulator of β-cell apoptosis and proliferation in vitro (1,8,40) and in vivo (2,3,24). In vivo experiments have established the requirement for insulin receptor signaling for β-cell mass homeostasis, but they do not distinguish between acute and chronic effects of insulin or provide information on the optimum concentration of insulin. While the exact concentration of dissolved insulin available for β-cell insulin receptors in vivo remains a subject of speculation and debate, our in vitro studies and others indicated that low nanomolar (0.2–20 nm) concentrations of insulin rescued human and rodent islets from serum withdrawal-induced apoptosis (1,40). Higher doses of insulin (200 nm) were relatively ineffective at reducing human β-cell apoptosis (1). Similarly, work from another group found that low insulin (0.1 nm) reduced caspase-3 activity in glucose-deprived β-cells, whereas 100 nm insulin increased caspase-3-mediated death (41), further suggesting that insulin signaling can be biphasic (14,42). These and other observations led us to postulate that the prosurvival and mitogenic autocrine effects of insulin might be lost outside a relatively narrow concentration range (1,8,25). It is also notable that concentrations of insulin in the picomolar to low nanomolar range have higher affinity for the insulin receptor than superphysiological doses (43). Both low (0.2 nm) and high (200 nm) doses of insulin stimulate β-cell proliferation (8), but the high doses of insulin readily cross-activate the IGF-I receptor (43). Indeed, the effects of 100 nm insulin and 100 nm IGF-I on MIN6 cell apoptosis were not additive (44), suggesting common mediators. In our preliminary experiments, 0.2 nm IGF-I did not induce statistically significant ERK phosphorylation in primary islets (n = 4; data not shown), suggesting that insulin may be more potent in activating ERK at this concentration. Collectively our new data point to ERK and Bad as downstream targets that can be significantly regulated by 0.2 nm insulin. Consistently, the effects of insulin at the high 200 nm dose were either nonsignificant or did not show a stronger effect than 0.2 nm insulin. Negative feedback in insulin signaling may be an adaptation to prevent the oncogenic effects of excess mitogenic and antiapoptotic signaling (45).

The present findings suggest that insulin stimulates Raf-1/ERK activation and Bad deactivation at concentrations expected in the systemic and portal circulation (40–1000 pm) (46). The in vivo significance of these findings remains to be determined. Specifically, it is not yet clear whether this means that insulin signaling is constitutively activated in islets in vivo or whether the local concentrations of insulin that β-cells are exposed to are lower than expected. Although it may seem intuitive that β-cells could be exposed to higher concentrations of insulin than those observed in the systemic or portal circulation, no direct in vivo measurements of insulin levels in the intact local microenvironment have been reported to date. Indeed, there are a few clues that point to the possibility that local concentrations of bioactive, monomeric insulin may be within the picomolar or low nanomolar range. Theoretically, the concentration of free monomeric insulin available for binding to insulin receptors is dictated by a number of factors, including the rate of local blood flow, the rate of insulin exocytosis, and the rate at which insoluble zinc-containing insulin crystals and hexamers dissolve into monomeric insulin. We have proposed that a mathematical model might be useful in estimating the contribution of each factor (47). It is also important to consider the complex architecture of the islet microvasculature, which allows for some β-cells to be perfused first (48,49,50). Regardless of the rates of insulin secretion, decrystalization, and blood flow, this would mean that a population of β-cells would be exposed to insulin concentrations in the picomolar range.

Insulin is known to activate multiple downstream pathways to suppress apoptosis in other cell types, but the specific pathways targeted by insulin in β-cells had not been elucidated. Previous evidence suggested that Akt was not likely to play a dominant role in β-cell insulin signaling (1) or the regulation of β-cell mass (4). Most of the protective effects of insulin were maintained in mice severely deficient in islet Akt activity (1). Moreover, insulin did not stimulate Akt phosphorylation at serine 473 in primary islets (1). Thus, instead of Akt, we focused on Raf-1, a kinase that has previously been established as an insulin target in other cell lines (51). Our studies indicate that treatment of islets and MIN6 cells with 0.2 nm insulin, which is likely to interact primarily with the insulin receptor rather than the IGF-I receptor, caused an increase in Raf-1 phosphorylation at serine 338. This is in accordance with earlier studies demonstrating that 0.1 or 7 nm insulin significantly increased Raf-1 activity and serine phosphorylation in cell lines expressing insulin receptors (51,52). We previously demonstrated that Raf inhibitor and dominant-negative Raf-1 specifically blocked the effects of 0.2 nm insulin on β-cell proliferation but did not decrease the proliferation stimulated by 200 nm insulin (8). Together our findings support the notion that the Raf/MAPK kinase/ERK arm of the insulin signaling pathway plays a critical role in both β-cell survival and proliferation (7,8). There are caveats that come with the use of chemical inhibitors and dominant-negative proteins. For example, Raf-1 (51-131)-GFP may bind to activated Ras, thereby potentially affecting multiple downstream pathways. Thus, it will be important in the future to examine the effects of specific deletion of the Raf-1 gene in β-cells on ERK activation and apoptosis.

Like insulin, glucose has been reported to regulate proliferation and survival of β-cells (30) and may act on the same set of genes as insulin (53,54). However, it is unclear whether the prosurvival effects of glucose occur via a mechanism independent of autocrine insulin signaling. In the present study, glucose-induced ERK phosphorylation was completely blocked in the presence of somatostatin in MIN6 cells. These data are in agreement with previous observations that somatostatin or neutralizing anti-insulin antibodies reduced β-cell proliferation in the presence of high glucose in dispersed primary mouse β-cells and MIN6 cells (8,44). In contrast to the effects of somatostatin on glucose-treated MIN6 cells, Wicksteed et al. (55) reported that somatostatin had no effect on high glucose-induced activation of ERK in rat islets. Differences in experimental procedures, species, or microenvironments (dispersed vs. intact islets) may explain this discrepancy. These experiments also come with the caveat that somatostatin might directly inhibit glucose-dependent ERK phosphorylation that has been proposed to occur via cAMP and protein kinase A (56). Therefore, we also directly assessed the role of secreted insulin and demonstrated that islets from mice with an approximately 50% reduction in glucose-stimulated insulin secretion have a proportional 50% loss of ERK activation in response to glucose. Ideally, these studies would be complemented with experiments using primary islets made acutely deficient in insulin receptors using conditional gene ablation. It was recently shown by others that cell lines with insulin receptor ablation (but IGF-I receptor up-regulation) had impaired activation of phosphatidylinositol 3-kinase and Akt in response to glucose, although ERK phosphorylation remained (57). Together these data suggest that both glucose and insulin can activate ERK, depending on the model. Whereas additional work is clearly required, studies from multiple groups point to an important role for autocrine insulin signaling in β-cells.

Our data presented here and work from other groups are in agreement that insulin and glucose stimulate the rapid translocation of ERK to the nucleus (30), an event likely to promote β-cell survival and growth via ERK-dependent transcription (30). Additional support for the hypothesis that insulin and glucose may have overlapping mechanisms comes from the observations that the activation of ERK by insulin and glucose share many common signaling requirements, including Ca²⁺ mobilization, FK506-sensitive calcineurin, and KN93-sensitive CaMKII (29,30,58). The present study provides more evidence that Ca²⁺ mobilization is important for insulin-induced ERK phosphorylation. Blocking Raf-1, ERK, or Ca²⁺ signals induced β-cell apoptosis (7). Furthermore, we have now identified the NAADP-sensitive Ca²⁺ store as a mediator of ERK activation in β-cells. Like insulin, NAADP-stimulated Ca²⁺ mobilization follows a bell-shaped dose-response profile in human β-cells (14). Together with our previous data on CD38-deficient islets (17), these observations point to the complex role of Ca²⁺ in the β-cell.

Although ERK is known as a prosurvival kinase in many cell types, the role of ERK in β-cell survival is enigmatic. Some investigations point toward an antiapoptotic role for ERK, whereas other studies suggest a requirement for sustained ERK phosphorylation in various forms of β-cell apoptosis (28,59). Multiple studies reported that cAMP/protein kinase A-mediated-ERK signaling regulates β-cell growth and survival via Bcl-2 and Irs-2 (28,60). Our own work suggests that basal Raf-1/ERK signaling is essential for survival in primary β-cells (7). For instance, a dominant-negative Raf-1 mutant or a Raf inhibitor caused rapid death that was associated with down-regulated ERK (7). In the present study, we observed increased death in β-cells treated with Raf and Ca²⁺ signaling inhibitors that was partly attributable to blunted ERK activation. On the other hand, sustained ERK activation of up to 4 d has been shown to be required for β-cell apoptosis induced by chronic high glucose (33.3 mm) or IL-1β (59). The mechanisms of a putative dual role for ERK remain to be worked out, but clearly the duration, strength, and subcellular localization of ERK signaling may be important in determining its net effect on cell fate (61).

Another key finding of the present study involves insulin’s effect on Raf-1 at the mitochondria. We have previously shown that overexpressed Raf-1-GFP fusion proteins were targeted to the mitochondria in the MIN6 β-cell line (7). In the present study, we confirmed that endogenous Raf-1 localized to the mitochondria of islet β-cells and MIN6 cells. In addition to inducing ERK activation, insulin tightened the localization of Raf-1 to mitochondria and promoted mitochondrial Raf-1 activity. These events allow Raf-1 to contribute to cellular survival by phosphorylating Bad. The possibility that Raf-1 might also function by interacting with other proteins in the vicinity remains to be addressed. The effects of insulin on Raf-1 at the mitochondria of islets and MIN6 cells are consistent with studies done on other cell types (13,62). In other models, Bcl-2 appears to be critical for Raf-1 translocation to the mitochondria (13). We detected relatively stable protein-protein interactions between Raf-1 and Bcl-2 in β-cells. Because Raf-1 physically associates with Bad, it is possible that Raf-1 may regulate the function of Bad in glucose metabolism and insulin secretion (63). Thus, it will be of interest in the future to unravel exactly how insulin regulates Raf-1 binding to Bcl-2 proteins, the mitochondrial translocation of this complex, and the phosphorylation of Bad.

In conclusion, the present investigation demonstrated that two targets downstream of Raf-1 kinase, namely ERK and Bad, both play critical roles in the acute response to low concentrations of insulin. Studies on the mechanisms of β-cell insulin signaling provide new information that improves our understanding of β-cell biology and may be useful for combating the loss of β-cell mass and function that occurs in almost all types of diabetes.