Rapid Turnover of Phosphatidylinositol-4,5-Bisphosphate in Insulin-Secreting Cells Mediated by Ca²⁺ and the ATP-to-ADP Ratio

Plasmids encoding the fusion constructs between the pleckstrin homology (PH) domain of PLCδ1 and the green fluorescent protein (GFP) (28) and membrane-targeted GFP were provided by Professor Tobias Meyer (Stanford University, Stanford, CA).

MIN6 β-cells (29) of passages 17–32 were cultured in Dulbecco’s modified Eagle’s medium containing 25 mmol/l glucose and supplemented with 2 mmol/l glutamine, 70 μmol/l 2-mercaptoethanol, 100 units/ml penicillin, 100 μg/ml streptomycin, and 15% fetal calf serum. After plating onto 25-mm coverslips at a density of 1.5 × 10⁵/ml, the cells were transiently transfected with 2 μg of plasmid DNA and Lipofectamine 2000 (Invitrogen, Carlsbad, CA) in a 1:2.5 DNA:lipid ratio according to the manufacturer’s protocol and further cultured for 12–24 h.

Before experiments, the cells were transferred to a buffer containing 125 mmol/l NaCl, 4.8 mmol/l KCl, 1.3 mmol/l CaCl₂, 1.2 mmol/l MgCl₂, and 25 mmol/l HEPES with pH adjusted to 7.40 with NaOH. For [Ca²⁺]_i measurements, transfected cells were loaded with the Ca²⁺ indicator Fura Red by 30 min incubation at 37°C with 10 μmol/l of its acetoxymethyl ester (Molecular Probes, Eugene, OR).

In some experiments, the cells were permeabilized with Staphylococcal α-toxin (PhPlate Stockholm, Stockholm, Sweden). Before permeabilization the cells were superfused with an intracellular-like medium containing 140 mmol/l KCl, 6 mmol/l NaCl, 1 mmol/l MgCl₂, 2 mmol/l EGTA, 0.465 mmol/l CaCl₂, and 10 mmol/l HEPES with pH adjusted to 7.00 with KOH. Temporarily interrupting the perfusion, 5 μl α-toxin (0.46 mg/ml) was added directly into the 50-μl superfusion chamber. After permeabilization, 0.025–3 mmol/l MgATP was added to the medium. Free Ca²⁺ was buffered to 100 nmol/l using 2 mmol/l EGTA.

The coverslips with the attached cells were used as exchangeable bottoms of the 50-μl open chamber and superfused at a rate of 0.3 ml/min. All experiments were performed at 37°C. Plasma membrane concentrations of the fluorescent protein constructs were measured using an evanescent wave microscopy setup as previously described (16). Selection of excitation and emission wavelengths were made with the following filters (center wavelength/half-bandwidth nm): GFP, excitation 488/10 nm and emission 525/25 nm; Fura Red, excitation 488/10 nm and emission 630 nm long pass. Images or image pairs were acquired every 5 s. To minimize exposure to the potentially harmful laser light, the beam was blocked by an electronic shutter (Sutter Instruments, Novato, CA) between image captures.

Confocal imaging was performed with a Yokogawa CSU-10 spinning disk system (Andor Technology, Belfast, Northern Ireland) attached to a Diaphot 200 microscope (Nikon) equipped with a ×60 1.40 NA objective (Nikon). The 488-nm beam from an argon ion laser (Melles-Griot, Didam, the Netherlands) was coupled to the scan head through an optical fiber (Point-Source, Southampton, U.K.). Fluorescence was detected at 520/35 nm using an Orca-AG camera (Hamamatsu) under MetaFluor software control.

Image analysis was made with ImageJ software (W.S. Rasband, National Institutes of Health [http://rsb.info.nih.gov/ij]). Curve fitting was made with the Igor Pro software (Wavemetrics, Lake Oswego, OR). Fluorescent protein and Fura Red intensities are expressed as changes relative to initial fluorescence (ΔF/F₀) after subtraction of background. Statistical analysis was performed using Student’s t test. All data are presented as means ± SE.

Plasma membrane PIP₂ concentration was measured using the PIP₂- and IP₃-binding GFP-tagged PH domain from PLCδ1 (PH_PLCδ1-GFP). Conventional epifluorescence imaging revealed that the construct was predominantly located to the plasma membrane of unstimulated MIN6 β-cells maintained in buffer containing 3 mmol/l glucose (Fig. 1A). Selective illumination of the plasma membrane with evanescent wave microscopy demonstrated a homogenous fluorescence in the membrane area adhering to the coverslip (Fig. 1A). As previously reported (16,30), carbachol-stimulated hydrolysis of PIP₂ and generation of IP₃ via activation of PLC promoted loss of evanescent wave-excited fluorescence (Fig. 1A and B). This effect reflects the PH_PLCδ1-GFP dissociation from the membrane and binding to IP₃ in the cytoplasm. The translocation was rapidly reversed on removal of the stimulus (Fig. 1B).

Elevation of the glucose concentration from 3 to 11 mmol/l induced a small increase in evanescent wave-excited PH_PLCδ1-GFP fluorescence, reaching 4.9 ± 0.1% (n = 27) above basal level within <2 min of stimulation. This increase was followed by a more pronounced decline in fluorescence (9.2 ± 0.8%, n = 23) of variable duration, typically succeeded by a new increase and gradual stabilization to a steady level only slightly above baseline (Fig. 1C). The initial rise of PH_PLCδ1-GFP fluorescence most likely involved an effect on cell adhesion or cell volume rather than purely reflecting the PIP₂ concentration, because a similar rise was observed in cells expressing GFP in the cytoplasm (not shown) or targeted to the plasma membrane (n = 13; Fig. 1D). However, the pronounced glucose-induced decline and subsequent increase of PH_PLCδ1-GFP fluorescence had no counterpart in the control experiments, indicating that glucose stimulates hydrolysis and synthesis of plasma membrane PIP₂ in insulin-secreting cells.

Previous studies have indicated that glucose may stimulate PIP₂ hydrolysis either directly (13–15) or indirectly via elevation of [Ca²⁺]_i (11,12). We therefore investigated the temporal relationship between changes in [Ca²⁺]_i and PIP₂ during glucose stimulation of the MIN6 β-cells. Simultaneous recording of [Ca²⁺]_i with Fura Red and PIP₂ with PH_PLCδ1-GFP during elevation of the glucose concentration from 3 to 11 mmol/l demonstrated that the early reduction of [Ca²⁺]_i, known to reflect stimulation of the sarco(endo)plasmic reticulum and plasma membrane Ca²⁺-ATPases (31), was paralleled by the modest increase of PH_PLCδ1-GFP fluorescence, whereas the glucose-induced [Ca²⁺]_i elevation coincided with the drop of PH_PLCδ1-GFP fluorescence (Fig. 2A), most likely reflecting Ca²⁺-mediated activation of PLC (15,16,30,32). From this experiment, it can also be seen that the PH_PLCδ1-GFP rapidly increases again when [Ca²⁺]_i drops (t_1/2 = 19 ± 5 s; n = 22), indicating an efficient resynthesis of PIP₂ and/or degradation of IP₃. Consistent with the idea that depolarization-induced elevation of [Ca²⁺]_i after closure of K_ATP channels in β-cells results in activation of PLC and a reduction of membrane PIP₂, it was found that 200 μmol/l of the K_ATP channel inhibitor tolbutamide induced a rapid elevation of [Ca²⁺]_i paralleled by a drop of membrane PH_PLCδ1-GFP fluorescence (8.0 ± 1.2%, n = 3; Fig. 2B). Both changes of [Ca²⁺]_i and PH_PLCδ1-GFP fluorescence were reversed on washout of the drug. Some cells responded to elevation of the glucose concentration with the β-cell characteristic oscillations of [Ca²⁺]_i, reflecting periodic plasma membrane depolarization and Ca²⁺ influx (33). In this case, membrane PIP₂ concentration also varied in a periodic manner, with the nadirs in lipid concentration coinciding with increases of [Ca²⁺]_i and vice versa (Fig. 2C).

When the cells were stimulated with 11 mmol/l glucose in Ca²⁺-deficient medium containing 2 mmol/l EGTA, there was an early reduction but no subsequent increase of [Ca²⁺]_i. Under this condition, glucose triggered a monophasic rise of PH_PLCδ1-GFP fluorescence to a level significantly above that obtained in the presence of extracellular Ca²⁺ (9.4 ± 2.1%, n = 6 vs. 2.5 ± 1%, n = 27; P < 0.01; data not shown). Similarly, when voltage-dependent Ca²⁺ influx was prevented by 250 μmol/l of the hyperpolarizing agent diazoxide, glucose triggered a modest decrease of [Ca²⁺]_i that was paralleled by a pronounced increase of PH_PLCδ1-GFP fluorescence reaching steady state at 8.3 ± 1.1% above baseline (n = 30; Fig. 2D). Subsequent omission of diazoxide resulted in elevation of [Ca²⁺]_i and a rapid drop of PH_PLCδ1-GFP fluorescence (Fig. 2D). Taken together, these results show that glucose promotes breakdown of PIP₂ in the plasma membrane of single insulin-secreting cells via elevation of [Ca²⁺]_i and that the lipid undergoes rapid resynthesis after termination of the [Ca²⁺]_i signal.

The regulation of plasma membrane PIP₂ concentration was further characterized in individual MIN6 β-cells permeabilized with α-toxin from Staphylococcus aureus. The pores formed in the plasma membrane by this toxin allow passage of molecules up to ∼1 kDa (34). Thus, small molecules, such as ions and nucleotides, can permeate, whereas most proteins, including the PH_PLCδ1-GFP probe, remain inside the cell. Permeabilization of PH_PLCδ1-GFP–expressing MIN6 β-cells in an ATP-free intracellular-like medium resulted in loss of evanescent wave-excited membrane PH_PLCδ1-GFP fluorescence. Confocal microscopy revealed that the probe was localized predominantly to the cytoplasm under this condition (Fig. 3A). Introduction of 3 mmol/l ATP resulted in a rapid (t_1/2 = 16 ± 1 s, n = 32) and prominent (131 ± 10%, n = 32) increase of evanescent wave-excited membrane PH_PLCδ1-GFP fluorescence (Fig. 3A and B). The effect was reversed on removal of ATP (t_1/2 = 39 ± 5 s, n = 32; Fig. 3B), likely reflecting lipid phosphatase activity. Reintroduction of ATP caused a new increase of fluorescence, albeit with somewhat lower amplitude (76 ± 5% of that of first pulse, n = 7; Fig. 3B). The effect of ATP on PH_PLCδ1-GFP fluorescence was completely abolished by 100 μmol/l phenylarsine oxide, a PI 4-kinase inhibitor (n = 7; Fig. 3C). ATP was also without effect in cells expressing GFP alone targeted to the plasma membrane (n = 4; Fig. 3D), indicating that the changes in PH_PLCδ1-GFP fluorescence reflected changes in membrane PIP₂ concentration.

We next investigated the ATP dependence of PIP₂ synthesis in permeabilized MIN6 β-cells. Application of increasing concentrations of ATP from 25 μmol/l to 3 mmol/l resulted in graded increases of membrane PIP₂ concentration (Fig. 4A). Half-maximal and maximal stimulatory effects were obtained at ∼300 μmol/l and ∼1 mmol/l of the nucleotide, respectively (Fig. 4B). The concentration dependence was strikingly steep with a Hill coefficient of 9.7. This value may be overestimated because of the gradual desensitization of the response during repeated or prolonged exposure to ATP. In contrast, the results were not skewed by PLC activation via purinergic receptors, because 5 μmol/l of the PLC inhibitor U73122 was without effect on the PH_PLCδ1-GFP fluorescence level reached with 300 μmol/l ATP (n = 6; Fig. 4C). At 300 μmol/l ATP the t_1/2 for PIP₂ synthesis averaged 49 ± 5 s (n = 29) and that for breakdown on removal of ATP 16 ± 2 s (n = 12), demonstrating high turnover of PIP₂ in the plasma membrane. Inositol availability was not rate limiting for PIP₂ synthesis on a short time scale as addition of 0.5 mmol/l myo-inositol was without effect on the steady-state level (n = 7; data not shown). Nevertheless, a fraction of the PIP₂ synthesis may require de novo generation of phosphatidylinositol, because 10 mmol/l Li⁺, which prevents inositol recycling from inositolmonophosphate (35), suppressed the steady-state PH_PLCδ1-GFP fluorescence level by 16.5 ± 2.9% (n = 9; Fig. 4D). Together, the data suggest that PIP₂ turnover essentially reflects a cycle of rapid phosphorylation of PI to PIP and PIP₂ and dephosphorylation back to PIP and PI.

Many processes in β-cells are regulated by the ratio of ATP to ADP rather than by the ATP concentration itself (36). We therefore investigated the influence of ADP on PIP₂ synthesis in the permeabilized MIN6 cells. Unexpectedly, application of ADP alone resulted in prompt stimulation of PIP₂ synthesis (Fig. 5A). This effect probably reflects adenylate kinase–catalyzed formation of ATP from ADP (37), because it was prevented by 25 μmol/l P¹,P⁵-di(adenosine-5′) pentaphosphate, an adenylate kinase inhibitor (Fig. 5A). To circumvent this complication, we made use of the ADP analog adenosine-5′-O-2-thiodiphosphate (ADPβS), which is not a substrate for adenylate kinase (38). In the absence of ATP, 0.1–1 mmol/l ADPβS was without effect on PH_PLCδ1-GFP fluorescence. However, the ADP analog counteracted ATP-stimulated PIP₂ synthesis. Thus, in the presence of 1 mmol/l ADPβS, the increase of PH_PLCδ1-GFP fluorescence after stimulation with 500 μmol/l ATP reached only 49.5 ± 6.3% of that in controls (n = 9; P < 0.001; Fig. 5B and C). PIP₂ synthesis was significantly stimulated by subsequent reduction of the ADPβS concentration to 100 μmol/l with PH_PLCδ1-GFP fluorescence reaching 80.2 ± 7.3% of that obtained in the absence of ADP analog (n = 9, P < 0.02; Fig. 5C and D). Similarly, introduction of increasing concentrations of ADPβS in the continuous presence of 300 μmol/l ATP caused a dose-dependent loss of fluorescence (Fig. 5D and E). When the relative PH_PLCδ1-GFP fluorescence was plotted against the ATP-to-ADPβS ratio, there was a clear sigmoidal relationship with half-maximum at a ratio of 0.74 and a Hill coefficient of 0.90 (Fig. 5F). These data indicate that changes of the ATP-to-ADP ratio in the physiological range (∼2–10) (39) can regulate the membrane PIP₂ concentration in insulin-secreting cells from ∼70 to 90% of maximum.

In the present study, we characterized the regulation of the plasma membrane PIP₂ concentration in individual MIN6 β-cells using the PH domain from PLCδ, a well-established tool that has provided important insights into the regulation of PLC activity in various types of cells (16,17,28,30,40). Glucose was found to trigger Ca²⁺-dependent hydrolysis of PIP₂ via PLC. We also determined the requirements for intracellular adenine nucleotides for the synthesis of the lipid. The data indicate that PIP₂ undergoes rapid turnover in the plasma membrane and that its concentration is controlled by the intracellular ATP-to-ADP ratio.

Glucose stimulation of pancreatic β-cells is associated with voltage-gated influx of Ca²⁺, which triggers exocytosis of insulin granules. We now observed that the glucose-induced rise of [Ca²⁺]_i is associated with loss of membrane PIP₂, an effect that can be attributed to Ca²⁺-dependent activation of PLC. It has previously been shown (16,17,30) that this enzyme is tightly and dynamically regulated by [Ca²⁺]_i and that depolarization and influx of Ca²⁺ independently trigger PLC activation in insulin-secreting cells (16). Accordingly, depolarization with the K_ATP channel inhibitor tolbutamide was found to cause both rise of [Ca²⁺]_i and loss of membrane PIP₂ and glucose-induced oscillations of [Ca²⁺]_i paralleled by oscillations of PIP₂. Furthermore, the glucose-induced loss of PIP₂ was abolished when Ca²⁺ influx was prevented by the hyperpolarizing agent diazoxide or by removal of extracellular Ca²⁺. These data reinforce the idea that glucose stimulation of β-cells leads to IP₃ formation via Ca²⁺-dependent activation of PLC. Periodic formation of IP₃ may contribute to the generation and shaping of the oscillations of membrane potential and [Ca²⁺]_i in insulin-secreting cells (33,41,42).

It is likely that glucose, apart from stimulating hydrolysis of PIP₂, also stimulates the synthesis of the lipid. It was recently suggested that glucose stimulates synthesis of PIP and PIP₂ via activation of the Ca²⁺-binding protein neuronal calcium sensor-1 and PI 4-kinase type IIIβ in INS1-E cells (43). Such an effect may escape detection in our system. The glucose-induced unspecific increase of plasma membrane fluorescence could mask a modest rise of PIP₂ concentration. Moreover, the higher affinity of the PLCδ PH domain for IP₃ compared with PIP₂ (44) may result in underestimation of the PIP₂ concentration in the presence of elevated [Ca²⁺]_i. To circumvent the potential problem of determining whether the observed kinetics represents changes in PIP₂ or IP₃ concentration, we now made use of permeabilized cells where any formed IP₃ rapidly will be washed away. The remaining PH_PLCδ1-GFP signal therefore reflects PIP₂ concentration changes in the plasma membrane. This conclusion is supported by the observation that ATP stimulated an increase of fluorescence that was abolished by inhibition of PI 4-kinases. Our novel experimental approach demonstrated that PIP₂ undergoes rapid turnover with a half-life of <40 s and provides direct support for the early observation that the plasma membrane lipid phosphatidylinositol undergoes futile cycles of continuous phosphorylation to PIP and PIP₂ and dephosphorylation to PIP and PI (1). Although metabolically expensive, a high turnover is a prerequisite for a messenger molecule to undergo rapid changes in concentration or to maintain concentration gradients in spatially restricted domains.

The synthesis of membrane PIP₂ was half-maximally stimulated at ∼300 μmol/l ATP, a concentration that compares favorably with previous reports based on biochemical characterization of phosphoinositide kinases (45). Although this sensitivity seems too high to enable regulation of PIP₂ synthesis by changes in the cytoplasmic concentration of ATP, which is in the low millimolar range, this possibility should not be excluded. The situation is reminiscent of that for the K_ATP channel, which is regulated by changes in cytoplasmic adenine nucleotides, although the channel is completely inhibited by micromolar concentrations of ATP in isolated membrane patches (46). In the case of the K_ATP channel, the action of ATP is counteracted by ADP, which may be formed locally by ATP hydrolysis (46) or phosphotransfer mediated by adenylate kinase (47). Interestingly, our data show that adenylate kinase has a functional effect also on adenine nucleotide–mediated PIP₂ synthesis. The glucose-induced formation of ATP occurs at the expense of ADP, which results in an increase of the ATP-to-ADP ratio in the β-cell. Changes in this ratio are important for regulating secretion, not only by modulating the conductance of K_ATP channels but also by altering amplifying pathways for insulin release (7). Our finding that the plasma membrane PIP₂ concentration has a very steep ATP dependence and is negatively regulated by an ADP analog supports the idea that PIP₂ synthesis in pancreatic β-cells, driven by stimulated metabolism, may be controlled by the ATP-to-ADP ratio rather than by ATP alone. A similar conclusion was reached using mouse pancreatic islets and an in vitro kinase assay (25). Such an effect may be mediated by direct nucleotide regulation of phosphoinositide kinase activity (48). Future studies will determine which phosphoinositide kinase isoforms and regulatory mechanisms account for PIP₂ synthesis in insulin-secreting cells.

Because PIP₂ is capable of regulating many targets in the vicinity of the plasma membrane, it is likely that the high turnover of the lipid in glucose-stimulated β-cells will influence proteins involved in the regulation of insulin secretion. Moreover, rapid turnover of PIP₂ allows generation of second messengers with maintenance of adequate levels of PIP₂ critical for sustained insulin granule exocytosis. It is interesting to note that impaired synthesis of PIP₂ with lower concentration of the lipid has been reported in an animal model of type 2 diabetes with insufficient insulin secretion (49).

This work was supported by grants from Åke Wiberg’s Foundation, the European Foundation for the Study of Diabetes/MSD, the Family Ernfors Foundation, the Harald Jeansson’s and Harald and Greta Jeansson’s Foundations, the Novo Nordisk Foundation, the Swedish Diabetes Association, the Swedish Research Council (32X-14643, 32BI-15333, and 32P-15439), and the Wenner-Gren Foundations.