Phosphatidylinositol 3-Kinase Suppresses Glucose-Stimulated Insulin Secretion by Affecting Post-Cytosolic [Ca²⁺] Elevation Signals

Insulin radioimmunoassay kit, enhanced chemiluminescence protein detection kit, and d-[U-¹⁴C]glucose were purchased from Amersham Pharmacia Biotech (Uppsala, Sweden); anti-p85α^PAN was from Upstate Biotechnology (Lake Placid, NY); collagenase, wortmannin, actinomycin D, LY294002, rapamycin, PD98059, diazoxide, and ATP bioluminescent assay kit were from Sigma (St. Louis, MO); and nitrocellulose membranes were from Schleicher & Schuell (Dassel, Germany).

p85α^−/− mice were maintained by a crossing of p85α^+/− mice on the original C57BL6/CBA hybrid background (36). The mice were housed on 12-h light/dark cycle with standard diet and water available ad libitum.

Islets were isolated by collagenase digestion and manual picking from the pancreata of p85α^−/− male mice and their wild-type male littermate mice. These islets were used for experiments immediately after isolation. The number of islet cells was microscopically counted after dispersing 10 islets per tube into single cells with trypsinization and pipetting through microtips.

For electron microscopic examination (37), islets were immersed in 2.5% glutaraldehyde in 0.1 mol/l phosphate buffer (pH 7.4). They were then subjected to double-fixation with 2.5% glutaraldehyde and 1% osmic acid, embedded in Epon 812, and cut into ultrathin sections. The sections were doubly stained with uranyl acetate and lead acetate and examined under a Hitachi HU-11B electron microscope. The number of insulin-containing dense core granules was counted in a 12.5-μm² rectangular area, except for the nuclear region, using ×4,000 photographs. The counting was performed on 150 β-cells from at least three mice in each group. For measurement of rough ER mass, their areas were traced and calculated on 150 β-cells using ×4,000 photographs.

For insulin content assay (38), islet homogenates were solubilized in 74% ethanol and 1.4% HCl solution at 4°C overnight, and insulin liberated into the solution was measured by radioimmunoassay.

Islets were homogenized in ice-cold buffer A (25 mmol/l Tris-HCl [pH 7.4], 10 mmol/l sodium orthovanadate, 10 mmol/l sodium pyrophosphate, 100 mmol/l sodium fluoride, 10 mmol/l EDTA, 10 mmol/l EGTA, and 1 mmol/l phenylmethylsulfonyl fluoride). For immunoblotting, the samples were boiled in Laemmli’s sample buffer for 5 min, separated on SDS-polyacrylamide gels, and transferred to nitrocellulose membranes, followed by detection with an ECL kit using anti-p85α^PAN antibody (36). For immunoprecipitation, islet homogenates were incubated in buffer A with anti-p85α^PAN antibody and protein G agarose for 2 h at 4°C. After the immunoprecipitates were washed three times with buffer A and three times with PI-3 kinase buffer (25 mmol/l Tris-HCl [pH 7.4], 100 mmol/l sodium chloride, 0.5 mmol/l EGTA), they were immunoblotted with anti-p85α mouse monoclonal antibody (Upstate Biotechnology).

After the islets had been incubated at 37°C for 60 min in the presence of the indicated concentrations of glucose, islet homogenates were immunoprecipitated as described above and used for a PI phosphorylation reaction (36). The reaction products were applied to thin-layer chromatography and quantified with an image analyzer (BAS 2000; Fuji Film, Tokyo, Japan), as described (39).

Insulin release from pancreatic islets was measured in static incubation or perifusion incubation with Krebs-Ringer bicarbonate (KRB) buffer composed of 129 mmol/l NaCl, 4.8 mmol/l KCl, 1.2 mmol/l MgSO₄, 1.2 mmol/l KH₂PO₄, 2.5 mmol/l CaCl₂, 5 mmol/l NaHCO₃, 0.2% bovine serum albumin, and 10 mmol/l HEPES (pH 7.4) (40). In static incubation, batches of 10 freshly isolated islets were preincubated at 37°C for 30 min in KRB buffer containing 2.8 mmol/l glucose. The preincubation solutions were replaced with KRB buffer containing test agents, and batches of islets were incubated at 37°C for 60 min. Insulin released into these supernatants was measured by radioimmunoassay. In perifusion incubation, 30 freshly isolated islets were suspended in 500 μl of Bio-Gel G-10 beads in each perifusion chamber and perifused with KRB buffer at 37°C at a flow rate of 0.6 ml/min. Islets were perifused for 30 min in the presence of 2.8 mmol/l glucose before higher glucose stimulation. Perifusate fractions were periodically collected, and insulin in these samples was measured by radioimmunoassay. When wortmannin was added to incubation media, it was also included in preincubation media if not otherwise stated.

Glucose oxidation in islets was measured by generation of ¹⁴CO₂ from d-[U-¹⁴C]glucose (41). Batches of 10 islets were incubated at 37°C for 90 min in KRB buffer containing the isotope. When wortmannin was added to incubation media, it was also included in preincubation for 30 min. The ¹⁴CO₂ generated was made volatile by adding HCl, captured by NaOH, and measured with liquid scintillation counting.

For determining ATP contents in whole islets (40), batches of 10 islets were incubated at 37°C for 60 min in KRB buffer. When wortmannin was added to incubation media, it was also included in preincubation for 20 min. The incubation was stopped by the addition of ice-cold HClO₄, and islets were homogenized by sonication. The homogenates were neutralized by addition of NaOH. ATP contents in the supernatants were measured with ATP bioluminescent assay kit.

Islets were loaded with 15 μmol/l fura-2/acetoxymethylester at 37°C for 60 min in Sol AII buffer (150 mmol/l NaCl, 5 mmol/l KCl, 1 mmol/l MgCl₂, 2 mmol/l CaCl₂, and 10 mmol/l HEPES [pH 7.4]) (40). An islet on a glass coverslip was placed on the stage of an inverted fluorescence microscope (IX70; Olympus, Tokyo, Japan) equipped with a charged-coupled device camera (Photometrics). The islet was viewed under a water immersion objective lens (LUMPlanFL ×60, Olympus) with alternate excitation wavelengths of 355 nm and 380 nm. The images of 256 × 256 pixels (170 × 170 μm) were obtained with an exposure time of 100 ms (42). The image analysis was carried out using an IPLab program (Signal Analysis Corporation).

Statistical analysis was performed using Student’s t test for unpaired comparisons or analysis of variance with Tukey’s post hoc comparison. Values are presented as means ± SE.

At 8 weeks of age, body weights of p85α^−/− male mice were 22.4 ± 0.5 g, which were similar to those of wild-type control mice (22.2 ± 0.3 g). Pancreas weights were also similar in wild-type (0.22 ± 0.02 g) and p85α^−/− mice (0.25 ± 0.02 g). We next measured the insulin contents of isolated islets. The p85α^−/− islets that we used for the present studies were comparable in size (∼150 μm diameter) to those of wild-type islets and contained a number of cells (1,601 ± 104 cells/islet) similar to that of wild-type islets (1,549 ± 76 cells/islet). The insulin content of p85α^−/− islets was 33.7 ± 1.8 ng/islet, which was significantly decreased (P < 0.01) as compared with that of wild-type islets (53.1 ± 1.4 ng/islet). When insulin content was expressed per cell, the content in individual p85α^−/− cells was 21.4 ± 1.4 pg/cell and was decreased by 39% (P < 0.01) as compared with that of wild-type cells (34.9 ± 2.1 pg/cell).

To investigate further the cause of decreased insulin content in p85α^−/− β-cells, we examined these cells with an electron microscope (Fig. 1B) (37). The number of insulin-containing dense-core granules (2.8 ± 0.3/μm²) was decreased (P < 0.05) as compared with wild-type β-cells (4.3 ± 0.4/μm²) (Fig. 1A) across ultrathin sections. In addition, as compared with wild-type β-cells (2.39 ± 0.21% of whole cell area), p85α^−/− β-cells possessed markedly reduced rough ER (0.53 ± 0.11% of whole cell area; P < 0.01 versus wild-type β-cells) (Fig. 1C). These data suggested decreased insulin synthesis in p85α^−/− β-cells, which may reflect decreased demand for insulin secretion as a result of increased insulin sensitivity in p85α^−/− mice.

To confirm abrogation of the p85α molecule in islets from p85α^−/−mice, we performed Western blot analysis using an anti-p85α^PAN antibody that recognized full-length p85α protein (36). We detected no p85α in islets from the knockout mice (Fig. 2A, left panel). When PI 3-kinase activities after incubation with 2.8 mmol/l glucose were assayed in islet homogenates immunoprecipitated with the anti-p85α^PAN antibody, the activity in p85α^−/− islets was not abolished, but rather remained at 21.2 ± 3.7% (n = 3) of that of wild-type islets (Fig. 2B). With 22.2 mmol/l glucose, the PI 3-kinase activity was increased by 24.6 ± 4.1% (n = 3) as compared with 2.8 mmol/l glucose in wild-type islets. A similar slight activity increase in response to 22.2 mmol/l glucose (20.2 ± 2.9%, n = 3) was also observed in p85α^−/− islets, maintaining the activity ratio between p85α^−/− and wild-type islets unchanged at 22.2 mmol/l glucose. In skeletal muscle and adipocytes of p85α^−/− mice, p50α (43), a truncated splice variant of p85α with higher PI 3-kinase activity, was overexpressed so as to play a role in overcompensating for the lack of p85α, leading to rather increased PI(3,4,5)P₃ levels and enhanced whole-body insulin sensitivity (36). Thus, we presumed that this variant contributed to the remaining PI 3-kinase activity in p85α^−/− mice. In fact, we observed p50α to be expressed in wild-type islets (Fig. 2A, left panel). The expression of p50α in wild-type islets was far lower than that of p85α, as was reported in skeletal muscle and adipocytes (36). In p85α^−/− islets, p50α was expressed to the same level as in wild-type islets (Fig. 2A, left panel). We confirmed that the immunoprecipitated sample from p85α^−/− islets for PI 3-kinase activity measurement contained p50α that was comparable in amount to that from wild-type islets (Fig. 2A, right panel). These results demonstrated that p50α contributed to the residual PI 3-kinase activity and that the mechanism of compensatory overexpression of p50α was not operative in islets that lacked p85α, unlike the situation in skeletal muscle and adipocytes. The failure of p50α overexpression in tissues from p85α^−/− mice was also observed in B-lymphocytes (44).

We next studied glucose-stimulated insulin release from p85α^−/− islets during static incubation (Fig. 3A). Glucose-stimulated insulin releases did not differ from those of wild-type islets at 2.8 and 5.6 mmol/l glucose. However, they were increased by 27% at 11.1 mmol/l glucose (P < 0.05, n = 4) and by 53% at 22.2 mmol/l glucose (P < 0.01, n = 4) as compared with wild-type islets. Because the average cell number that constituted the p85α^−/− islets that we isolated and used for experiments was not different from that of wild-type islets, glucose-stimulated insulin releases per cell were clearly increased at 11.1 and 22.2 mmol/l glucose as compared with wild-type islet cells. Insulin contents per islet for 2.8, 5.6, 11.1, and 22.2 mmol/l glucose conditions were 42.1 ± 3.9, 44.2 ± 4.2, 50.0 ± 3.5, and 48.5 ± 5.1 ng for wild-type islets, and 29.4 ± 4.7, 37.9 ± 3.6, 31.7 ± 4.0, and 30.1 ± 4.0 ng for p85α^−/− islets, respectively. When secretion was expressed per insulin content, fractional insulin releases from individual p85α^−/− β-cells were increased by 110% (P < 0.01, n = 4) and 130% (P < 0.01, n = 4) at 11.1 and 22.2 mmol/l glucose, respectively, as compared with wild-type controls (Fig. 3B). Perifusion experiments revealed that knockout of the p85α molecule failed to affect the kinetics of biphasic insulin release in response to 22.2 mmol/l glucose as compared with wild-type controls (data not shown), although glucose-responsive secretion from p85α^−/− islets was greater by 35% in the first phase (0–12 min after stimulation) and by 50% in the second phase (12–60 min after stimulation) of secretion as compared with that of wild-type islets (Table 1).

Glucose-stimulated insulin secretion was increased at higher glucose concentrations in p85α^−/− β-cells. However, the p85α^−/− mice also exhibited systemic insulin sensitivity, and the supernormal secretory phenotype thus might be a change secondary to this sensitivity, rather than a primary result of decreased PI 3-kinase activity itself in β-cells. Moreover, p85α^−/− islets still had ∼20% of the residual PI 3-kinase activity, and this incomplete abrogation of the activity should be cautiously taken into account when those results from the animal model islets are interpreted. To address these issues, we performed complementary experiments to analyze the effect of pharmacological inhibition of PI 3-kinase with wortmannin on glucose-stimulated insulin secretion in wild-type islets, resulting in confirmation of previous reports (29,34,35). We first confirmed that wortmannin dose-dependently suppressed PI 3-kinase activities in wild-type islet homogenates (Fig. 4). Its IC₅₀ was below 10 nmol/l, and PI 3-kinase activity was progressively suppressed by up to 100 nmol/l wortmannin, at which concentration the residual activity was 6.8% of the original activity. Wortmannin dose-dependently increased 22.2 mmol/l glucose–stimulated insulin release (Fig. 5A). The secretion peaked at 100–300 nmol/l wortmannin but decreased slightly at 1 μmol/l wortmannin. Thus, the insulin secretion curve (Fig. 5A) and the PI 3-kinase inhibition curve (Fig. 4) in response to up to 100 nmol/l wortmannin were well correlated in a reciprocal manner. At 100 nmol/l, wortmannin’s potentiating effect on glucose–stimulated insulin secretion was also clearly observed at 8.3 and 11.1 mmol/l glucose but not at 2.8 or 5.6 mmol/l glucose (Fig. 5B). This potentiating effect took 20 min to reach maximum levels, as shown in perifusion experiments (Fig. 5C) in which an additional stimulation with 100 nmol/l wortmannin was applied to wild-type islets during the second-phase secretion attained with 22.2 mmol/l glucose. When islets were exposed to wortmannin also during the preincubation period for 30 min, the drug potentiated not only the second-phase secretion but also the first-phase secretion (Table 1). The acute onset of wortmannin’s effect (Fig. 5C) suggested that a transcriptional regulatory mechanism was not involved in the potentiation of glucose–stimulated insulin secretion. Supporting this, the transcriptional inhibitor actinomycin D (5 μg/ml) (45) did not affect insulin secretion in response to 22.2 mmol/l glucose or 100 nmol/l wortmannin–induced potentiation of the secretion by 22.2 mmol/l glucose (data not shown).

Essentially identical effects on glucose–stimulated insulin releases were observed with LY294002, another inhibitor of PI 3-kinase (46,47). The 22.2 mmol/l glucose-stimulated insulin secretion was potentiated by ∼100% with 25–50 μmol/l LY294002 (data not shown). With 50 μmol/l LY294002, the potentiating effect was not observed at 2.8 or 5.6 mmol/l glucose but was evident at 8.3 and 11.1 mmol/l glucose (Fig. 5D). The common results from p85α^−/− islets and wild-type islets treated with pharmacological inhibitors of PI 3-kinase that glucose–stimulated insulin secretion was increased at higher glucose concentrations under those conditions enabled us to propose that PI 3-kinase normally plays a suppressive role in glucose-stimulated insulin secretion. Neither rapamycin, an inhibitor of p70 ribosomal S6 kinase located downstream from PI 3-kinase (47,48,49), nor PD98059, an inhibitor of mitogen-activated protein (MAP) kinase kinase (50,51), showed such potentiating effects in static incubations with 22.2 mmol/l glucose (data not shown).

Moreover, we studied the effect of wortmannin on glucose–stimulated insulin secretion in p85α^−/− islets with perifusion experiments to evaluate the effect of further reducing the residual PI 3-kinase activity of the islets. Without wortmannin, 22.2 mmol/l glucose–stimulated insulin secretion from p85α^−/− islets was nearly 50% greater than that from wild-type islets during 60 min of incubation (Table 1). After treatment with 100 nmol/l wortmannin, the secretion from p85α^−/− islets was increased but became comparable to that of wild-type islets (Table 1). Thus, we assumed that wortmannin further reduced the residual PI 3-kinase activity in p85α^−/− islets, which led to an enhancement of glucose–stimulated insulin secretion even in the islets, and that the blunted potentiating effect of wortmannin on secretion in p85α^−/− islets was consistent with the idea that the increased insulin secretion in the islets may indeed be causally related to reduced PI 3-kinase activity.

In an attempt to identify sites at which wortmannin exerted its potentiating effect, we analyzed glucose metabolism in wild-type islets with [U-¹⁴C]glucose oxidation (41). Application of 100 nmol/l wortmannin did not change [U-¹⁴C]glucose oxidation at either 22.2 mmol/l or 2.8 mmol/l glucose (Fig. 6A). ATP contents were also comparable between control and wortmannin-treated islets at both basal and stimulating concentrations of glucose (Fig. 6B). We next studied the effect of wortmannin on cytosolic [Ca²⁺], which is regulated by cellular ATP levels through K_ATP channels (9). The cytosolic [Ca²⁺] was monitored with the fura-2 fluorescence ratio at 355: 380 nm (42). Stimulation by 22.2 mmol/l glucose triggered a sharp cytosolic [Ca²⁺] rise (the first phase) followed by a sustained plateau (the second phase) (Fig. 6C). When 100 nmol/l wortmannin was applied to islets 10 min after 22.2 mmol/l glucose stimulation during the second-phase plateau, this treatment did not change cytosolic [Ca²⁺] levels (Fig. 6C). At 20 min after wortmannin application, increases in the fura-2 fluorescence ratio (0.25 ± 0.03, n = 6) were not different from those of 22.2 mmol/l glucose stimulation without wortmannin at the corresponding time (0.27 ± 0.04, n = 6).

Finally, we studied the relationship between wortmannin’s effect and cytosolic [Ca²⁺] in β-cells. For this purpose, cytosolic [Ca²⁺] was modified with diazoxide, an opener of K_ATP channels, plus a depolarizing concentration of KCl (Fig. 7) (11,12). Insulin contents per islet just after secretion experiments under four conditions (from the left in Fig. 7A) were 56.3 ± 4.2, 53.8 ± 4.6, 43.1 ± 6.2, and 50.4 ± 4.6 ng for wild-type islets, and 47.7 ± 5.3, 32.2 ± 3.0, 30.5 ± 3.7, and 37.8 ± 4.5 ng for p85α^−/− islets, respectively. When p85α^−/− islets were stimulated by Ca²⁺ influx with 250 μmol/l diazoxide and 30 mmol/l KCl in the presence of 2.8 mmol/l glucose, secretion was significantly less (P < 0.05, n = 4) than that of wild-type islets (Fig. 7A, columns 3 and 4). However, when it was corrected by insulin content, fractional secretion did not differ between wild-type and p85α^−/− islets (Fig. 7B, columns 3 and 4). These results suggested that Ca²⁺ influx stimulation without glucose only induced insulin secretion proportional to stored amounts of insulin, not evoking a potentiation effect as seen in stimulation with high glucose concentrations under decreased PI 3-kinase activities. Next, the effects of combined application of diazoxide and KCl were studied on 22.2 mmol/l glucose–stimulated insulin secretion (Fig. 7A, columns 5 and 6). Although this treatment further enhanced the insulin secretion in both wild-type and p85α^−/− islets, the secretion from p85α^−/− islets was significantly higher (P < 0.05, n = 4) than that from wild-type islets (Fig. 7A, columns 7 and 8). When expressed as fractional insulin secretion, secretion by 22.2 mmol/l glucose from p85α^−/− islets was higher by 46% (P < 0.01) than that from wild-type islets in the presence of 250 μmol/l diazoxide and 30 mmol/l KCl (Fig. 7B, columns 7 and 8).

The same series of experiments was performed in wild-type islets using wortmannin. Pharmacological inhibition of PI 3-kinase activity with wortmannin failed to modify insulin secretion in response to Ca²⁺ influx stimulation induced by 250 μmol/l diazoxide and 30 mmol/l KCl at 2.8 mmol/l glucose (Fig. 7C, columns 3 and 4), as was consistent with the results in columns 3 and 4 of Fig. 7B. Next, 250 μmol/l diazoxide and 30 mmol/l KCl were added to incubation media containing 22.2 mmol/l glucose. By this addition, 22.2 mmol/l glucose–stimulated insulin secretion in either the absence or the presence of wortmannin (Fig. 7C, columns 5 and 6) was further augmented (Fig. 7C, columns 7 and 8). However, the secretion in the presence of wortmannin was significantly higher (P < 0.01, n = 4) than that in the absence of this drug, which was also consistent with the results in columns 7 and 8 of Fig. 7B. These results, together with the results of cytosolic [Ca²⁺] measurements, indicated the effect of PI 3-kinase inhibition on glucose–stimulated insulin secretion to be exerted largely at sites distal to the cytosolic [Ca²⁺] elevation.

In the present study, we investigated the role of PI 3-kinase in β-cell functions, such as insulin storage and insulin secretion in response to glucose, using p85α^−/− islets and wild-type islets treated with PI 3-kinase inhibitors.

Insulin content was significantly decreased in p85α^−/− β-cells possessing only ∼20% of the PI 3-kinase activity of wild-type β-cells. Hypersecretion of insulin and the resulting depletion of stored insulin seemed not to be responsible for the decrease because p85α^−/− mice were insulin-sensitive and their serum insulin levels were low in both fasting and glucose-loaded states (36). The decrease in cellular insulin content may be attributed to lowered biosynthesis of the hormone, which would be consistent with the markedly reduced proliferation of rough ER in the β-cells (Fig. 1B), although we have not directly addressed the rate of proinsulin biosynthesis and its conversion to insulin, or the rate of insulin degradation. This speculation was supported by the data that insulin gene transcription was markedly suppressed by the PI 3-kinase inhibitor LY294002 (31,32). However, it is unknown whether signals to stimulate insulin gene transcription via PI 3-kinase is mainly originated from ligand binding to insulin receptor because β-cell–specific knockout of insulin receptor did not lead to a reduction in pancreatic insulin content (52).

Rough ER changes in p85α^−/− β-cells were thought possibly to represent adaptation to the increased insulin sensitivity and hence the decreased insulin demand on β-cells. Thus, we observed that rough ER proliferation was very striking in β-cells obtained from insulin receptor substrate-1 (IRS-1) ^−/− and IRS-2^−/− mice showing insulin resistance and compensatory hyperinsulinemia (K.E. and T.K., unpublished observations). The whole-body insulin sensitivity or resistance may adjust capacity of insulin-synthesizing machinery represented by rough ER.

Insulin secretion was studied in the two β-cell models of decreased PI 3-kinase activities: the β-cells that were genetically devoid of p85α regulatory subunit and the β-cells that were treated with the pharmacological PI 3-kinase inhibitors. In the former, the decrease in PI 3-kinase activity was chronic and not complete, whereas it was acute and more complete in the latter (33). Glucose-stimulated insulin secretion from p85α^−/− β-cells with lower insulin contents was increased at higher glucose concentrations as compared with the wild-type controls (Fig. 3A), and the increase became more striking when it was normalized by cellular insulin content (Fig. 3B). Both the PI 3-kinase inhibitors wortmannin and LY294002 were found to enhance potently glucose–stimulated insulin secretion from intact mouse islets at higher glucose concentrations (Figs. 5B and D). Although these drugs at the doses currently used are believed to be specific to PI 3-kinase activity (30,33,46,47), the similar effect of the genetically downregulated model of PI 3-kinase activity with that of the pharmacological downregulation in wild-type controls further reinforced the idea that the potentiating effects seen in both models stemmed from specific suppression of the PI 3-kinase activity. It must be noted that the residual PI 3-kinase activity in p85α^−/− islets may partly be attributed to class II PI 3-kinase-C2β, which is also sensitive to low nanomolar levels of wortmannin (53), although its expression or activity in β-cells has not been addressed so far.

The acute onset of wortmannin’s effect on secretion (Fig. 5C) suggested that the effect might be mediated by inhibition of metabolite production by PI 3-kinase via its acute metabolic effects, rather than through inhibition of gene transcription or de novo protein synthesis. Inability of actinomycin D or rapamycin to affect glucose–stimulated insulin secretion supported this speculation. We assumed that time lag for ∼20 min to reach the maximum effect of wortmannin might be time required for attaining a new equilibrium of PI(3–5)P3 levels after the drug stopped synthesis of the lipid molecules. Similar time course of wortmannin’s effect on glucose–stimulated insulin secretion was also shown by Zawalich and Zawalich (35). Specifically, the importance of PI 3-kinase in the suppression of glucose–stimulated insulin secretion was reinforced by the observation that inhibition of the MAP kinase pathway by PD98059 (54,55,56) had no effect on secretion, as reported previously (50,51).

Essentially identical consequences that wortmannin potentiated glucose–stimulated insulin secretion at intermediate glucose concentrations can be extracted from data by us and others (29,34,35). Wortmannin at 100 nmol/l potentiated 15–25 mmol/l glucose–stimulated insulin secretion by ∼100% in MIN6 cells (29), which was in good agreement with our results. However, minor differences among them should be pointed out. Thus, Nunoi et al. (34) reported that the potentiating effect of 10 nmol/l wortmannin was not significant at 8.3 mmol/l glucose but was at 16.7 mmol/l glucose, whereas Zawalich and Zawalich (35) reported that the potentiating effect of 50 nmol/l wortmannin observed at 8.3 mmol/l glucose was no longer evident at 20 mmol/l glucose. These may originate from the differences in species used for experiments or in incubation conditions. It is widely known that progression of the second-phase insulin secretion after glucose challenge was substantially smaller in mouse islets than in rat islets (57). We might speculate that PI 3-kinase activity in mouse islets was high relative to rat islets, which led to seemingly stronger effects of PI 3-kinase downregulation in mouse islets.

Aspinwall et al. (58) proposed that insulin stimulated insulin secretion in single β-cells, using amperometric detection of exocytosis. This insulin-stimulated insulin exocytosis was completely abolished by an application of wortmannin (59). These results seemingly argued the opposite of our results. However, we assumed that these two findings were observing different effects of wortmannin on insulin secretion at different sites at different time points. Thus, Aspinwall and colleagues. observed exocytotic processes nearby the plasma membrane during the very short time span of ∼30 s, whereas we observed insulin secretion events that should include translocation of insulin-containing vesicles to the plasma membrane, conditioning of them to a readily releasable pool, and final exocytotic processes during a relatively long time span of ∼20 min. Molecular mechanisms of these dual effects of wortmannin, which presumably have distinct molecular targets, require further investigation.

The effect of wortmannin to potentiate glucose–stimulated insulin secretion was undoubtedly exerted distal to the cytosolic [Ca²⁺] elevation. Thus, under conditions in which K_ATP channels were fixed open with diazoxide and a depolarizing concentration of KCl was applied to wild-type islets, wortmannin further augmented this secretion (Fig. 7C, columns 7 and 8). This was also the case in the β-cell model of genetically downregulated PI 3-kinase activity (Fig. 7B, columns 7 and 8). It is noteworthy that Ca²⁺ influx–stimulated insulin secretion at basal glucose concentration was indeed decreased in p85α^−/− islets (Fig. 7A, columns 3 and 4) and became comparable to that of wild-type controls after calibration by cellular insulin content (Fig. 7B, columns 3 and 4). The pharmacological inhibition of PI 3-kinase was also unable to potentiate Ca²⁺ influx–stimulated insulin secretion in wild-type islets (Fig. 7C, columns 3 and 4). These results demonstrated the potentiating effect of PI 3-kinase inhibition on insulin secretion to be specific to glucose stimulation accompanied by Ca²⁺ influx but not relevant to “pure” Ca²⁺ influx stimulation without glucose. This mechanism should work in the first phase as well as the second phase of glucose-stimulated insulin secretion and actually enhance both of these phasic secretions (Table 1).

PI(4,5)P₂ has been shown to activate K_ATP channels by competing with ATP (60,61). Moreover, LY294002 completely blocked the activation of the K_ATP channels by PI(4,5)P₂, and application of PI(3,4,5)P₃ directly increased the channel currents in a rat insulinoma cell line (62). These results suggested PI(3,4,5)P₃ to be a candidate for direct channel activation and that wortmannin’s potentiating effect on glucose–stimulated insulin secretion might be mediated through the closure of these channels. Nevertheless, wortmannin did not further augment the cytosolic [Ca²⁺] elevation induced by glucose in our study (Fig. 6C). These results indicated that the increased K_ATP channel sensitivity to ATP contributed only slightly, if at all, to the mechanism of potentiation by wortmannin.

Ca²⁺-dependent release of neurotransmitters from synaptic vesicles requires an ATP-dependent priming step (23,63). The PI(4) 5-kinase requirement of this step suggested that PI(4,5)P₂ plays a role in the process. Synaptotagmins and rabphilin-3A, which contain Ca²⁺/phospholipid binding domains, are among the candidates for PI(4,5)P₂ effector molecules (64,65). By analogy with synaptic vesicles, the insulin-containing secretory vesicles may require PI(4,5)P₂ for their ATP-dependent priming step. Inhibition of PI 3-kinase activity should lead to accumulation of this phosphoinositide form. Thus, an increase in PI(4,5)P₂, rather than a decrease in PI(3,4,5)P₃, at the site of exocytosis might be responsible for the potentiation of insulin secretion in response to glucose.

In adipocytes, Akt kinase, a downstream effector of PI 3-kinase (66,67), was demonstrated to activate phosphodiesterase (PDE) 3B directly (68), which facilitates degradation of cAMP (69,70). If this cascade is operative in β-cells, then wortmannin may cause an increase in cellular cAMP content, thereby activating protein kinase A and resulting in the potentiation of glucose–stimulated insulin secretion. In fact, we observed that 100 nmol/l wortmannin augmented cAMP content by ∼30% in islets incubated for 60 min in the presence of 22.2 mmol/l glucose. However, the increase in cAMP content was too small to account solely for wortmannin’s potentiating effect on glucose–stimulated insulin secretion, as compared with a ∼10-fold increase in cAMP content induced by 1 mmol/l of the nonspecific PDE inhibitor 3-isobutyl-1-methylxanthine, which potentiated as much secretion as 100 nmol/l wortmannin in the presence of 22.2 mmol/l glucose (K.E. and T.K., unpublished observations). These results suggested that increased cellular cAMP is not the principal mediator of wortmannin’s effect, although it is possible that wortmannin changes sensitivity of cAMP-dependent steps to a small increase in cAMP.

It has been shown that leptin and insulin-like growth factor-1 suppress glucose plus cAMP-stimulated insulin secretion (71,72). One of the responsible mechanisms seems to involve PI 3-kinase, Akt kinase, and PDE3B activation and a subsequent reduction in cAMP content. These changes in response to leptin were reversed by treatment with wortmannin or LY294002 and actually resulted in recovery of insulin secretion. Thus, wortmannin’s effect of sustaining higher levels of cAMP through deactivation of PDE3B may contribute considerably to increased insulin secretion under conditions in which the cAMP concentration becomes high with adenylyl cyclase agonists or PDE inhibitors (71,72), in contrast to the situation of glucose stimulation alone, which is accompanied by a minimal, at most, increase in the cAMP concentration.

Mouse knockout technology has provided us with several model islets that were defective in components of insulin signal transduction pathways. In β-cell–specific insulin receptor knockout, a selective loss of acute-phase insulin secretion in response to glucose was exhibited. It is interesting that insulin content was not impaired in those β-cells. Roles of IRS-1 and IRS-2, immediately downstream molecules of the insulin receptor, in the proliferation and function of β-cells differed with a marked contrast (5,6,73). Comprehensive understanding of the relationship between insulin secretion/biosynthesis and insulin signal transduction pathways or the autocrine regulation by insulin of β-cell functions needs much more accumulation of experimental evidence. Our present results also should be reevaluated from a viewpoint of signal transduction cascades proximal and distal to PI 3-kinase.

In conclusion, decreased PI 3-kinase activities that were induced both genetically and pharmacologically were associated with increased insulin secretory responses to higher concentrations of glucose, indicating that PI 3-kinase normally plays a suppressive role in the regulation of glucose–stimulated insulin secretion. The suppressive effect on secretion seems to be exerted not through glucose-metabolizing steps that lead to ATP generation and Ca²⁺ influx, but through steps distal to the cytosolic [Ca²⁺] elevation. Clarification of the mechanism by which PI 3-kinase modulates β-cell secretory function will shed light on new aspects of glucose–stimulated insulin secretion and provide clues to the development of therapeutic agents that are capable of potentiating insulin secretion through a novel mode of action.

This work was supported by a grant-in-aid for creative basic research (10NP0201) from the Ministry of Education, Science, Sports, and Culture, Japan (T.K.).

We are grateful to Satoru Fukuda for assistance with the electron microscopy and Hiroshi Chiyonobu for care of the mice.