Membrane Phosphoinositides Control Insulin Secretion Through Their Effects on ATP-Sensitive K⁺ Channel Activity

Kir6.2 containing point mutations or the influenza A virus hemagglutinin (HA)–epitope tag (YPYDVPDYA inserted between amino acid 100 and 101 of Kir6.2; an extra nine amino acids, DLYAYMEKG, inserted between amino acid 98 and 99 to aid accessibility of the epitope) were prepared using the QuickChange site-directed mutagenesis kit (Stratagene) and subcloned into the pCMV6b vector. Recombinant adenoviruses were constructed using the AdEasy system (Stratagene). The cDNAs encoding wild-type PIP5K Iβ (PIP5K:WT) or a deletion mutant lacking the NH₂-terminal 238 amino acids (PIP5K:Δ1--238) (20) or the various rat Kir6.2 wild type and mutants were subcloned into the pCMV shuttle vector (pShuttle). They were then recombined with the pAdEasy vector in the BJ5183 strain of Escherichia Coli. Positive recombinants were selected, and pAdEasy plasmids containing the correct insert were used to transfect HEK293 cells for virus production. Single plaques were isolated for further amplification in HEK293 cells. Recombinant viruses were purified on a CsCl gradient and titers determined by spectrometry. All constructs were sequenced to verify the correct mutations.

For protein expression in INS-1 cells, recombinant adenoviruses with desired titers were used for infection as follows. INS-1 cells [clone 832/13, a gift from Dr. Chris Newgard (ref. 22)] were plated at 10⁶ cells/35-mm dish and cultured in RPMI-1640 with 11.1 mmol/l d-glucose (Invitrogen) supplemented with 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, 10 mmol/l HEPES, 2 mmol/l glutamine, 1 mmol/l sodium pyruvate, and 50 μmol/l β-mercaptoethanol. Twenty-four hours later, cells were infected with appropriate titer of each adenovirus diluted in 1 ml OptiMEM at 37°C for 1.5 h. One milliliter of RPMI-1640 with 2× supplement was then added to each dish for overnight incubation. The next day, the infection medium was replaced with regular medium and incubated for another day before the experiment.

INS-1 cells were lysed 24 h postinfection in 20 mmol/l HEPES, pH 7.0, 5 mmol/l EDTA,150 mmol/l NaCl, and 1% Nonidet P-40 (IGAPEL) with Complete protease inhibitors (Roche Applied Science, Indianapolis, IN). Proteins were separated by SDS/PAGE, transferred to nitrocellulose membrane, probed with appropriate primary antibodies followed by horseradish peroxidase–conjugated secondary antibodies (Amersham Biosciences, Piscataway, NJ), and visualized by enhanced chemiluminescence (Super Signal West Femto; Pierce, Rockford, IL). The primary antibodies used were rabbit polyclonal anti-Kir6.2 for Kir6.2 (Santa Cruz Biotechnology, Santa Cruz, CA) and rabbit polyclonal anti-PIP5K, which recognizes both the wild type and the Δ1-238 mutant (kindly provided by Dr. J. Nathan Davis), for PIP5K:WT and PIP5K:Δ1-238. To confirm expression of mutant Kir6.2 at the cell surface, living INS-1 cells infected with a virus carrying HA-tagged R206A Kir6.2 were immunostained for surface HA using protocols described previously (23).

INS-1 cells were grown on 35-mm dishes and infected with PIP5K:WT or PIP5K:Δ1-238 viruses. Twenty-four hours postinfection, cells were preincubated for 1 h in a phosphate-free medium before labeling with 20 μCi ³²P for 4 h. Cells were then washed twice with phosphate-free medium and scraped into 1 ml MeOH:HCl (10:1). After scraping, the dish was washed again with 1 ml of 2-mmol/l AlCl₃ in H₂O, and the wash was combined with the cell lysate. The lipids were extracted by 2 ml CHCl₃. The aqueous phase was transferred to a new tube and extraction repeated twice. The CHCl₃ phase from the three extractions was combined and extracted with 2 ml MeOH:HCl (1:1). The final lipid extract was dried and resuspended in 15 μl solvent (CHCl₃:MeOH:10 mmol/l HCl [20:10:1]) and analyzed by thin-layer chromatography using oxalate-coated plates. Approximately 30 μg of each lipid standard in the same solvent was run simultaneously. The plates were developed with CHCl₃:MeOH:H₂O:NH₄OH (90:90:20:7) and the labeled phospholipids quantified by phosphorimaging. Lipid standards were visualized by iodine staining.

Resting membrane potential (RMP) of INS-1 cells was measured at room temperature using whole-cell or perforated patch-clamp recording with the Axopatch 1D amplifier and Clampex 8.1 (Axon, Foster City, CA). For measuring the RMP in 3 or 12 mmol/l glucose, cells were preincubated in indicated glucose concentrations for 3 h before recording. During recording, cells were bathed in Tyrode’s solution consisting of (in mmol/l) 137 NaCl, 5.4 KCl, 1.8 CaCl₂, 0.5 MgCl₂, 5 HEPES, 3 NaHCO₃, and 0.16 NaH₂PO₄, with glucose as specified in the figure legends. Pipette solution contained (in mmol/l) 140 KCl, 10 HEPES, 1 EGTA, 1 EDTA (referred to as K-INT), and 1 ATP plus 1 MgCl₂ for experiments shown in Fig. 1. For measuring RMP at 12 mmol/l glucose using whole-cell patch-clamp recording, pipette solution contained (in mmol/l) 10 KCl, 130 Kgluconate, 10 HEPES, 1 EGTA, 3 MgCl₂, and 5 ATP. The pH of all solutions was adjusted to ∼7.2. To measure RMP without dialyzing internal cellular contents, the perforated patch was adopted. The pipette resistance ranged between 3 and 5 MΩ. The tip solution was first filled up to ∼0.5 mm with the clean pipette solution containing (in mmol/l) 76 K₂SO₄, 10 NaCl, 10 KCl, 1 MgCl₂, and 5 HEPES, pH∼7.2; the pipette was then backfilled with the pipette solution containing ∼160–250 μg/ml amphotericin B (Sigma). The stable perforated patch usually followed after 6–10 min of initial Giga seal. For ATP sensitivity measurement, inside-out patch-clamp recording was used as described previously (21). K-INT was used in both the bath and pipette solution. All currents were measured at a membrane potential of −50 mV (pipette voltage = +50 mV). ATP dose-response curves were fitted by the Hill equation: I_rel = 1/{1 + [(ATP)/K_i]^H}, where I_rel = current in [ATP]/current in 0 ATP, H = Hill coefficient, and K_1/2 = [ATP] causing half-maximal inhibition. All data are presented as means ± SE.

Membrane capacitance measurements were carried out in the standard whole-cell configuration using an EPC-9 amplifier and Pulse software (version 8.4; Heka Elektronik, Lambrecht, Germany). Patch electrodes were made from borosilicate glass capillaries, coated with dental wax (Kerr, Romulus, MI) at the tips, and fire polished. The cells were constantly perfused with prewarmed extracellular solution consisting of (in mmol/l) 138 NaCl, 5.6 KCl, 1.2 MgCl₂, 2.6 CaCl₂, and 5 HEPES (pH 7.4 with NaOH). The pipette solution contained (in mmol/l) 125 CsMeSO, 25 CsCl, 10 NaCl, 1 MgCl₂, 10 HEPES, 0.05 EGTA, 3 Mg ATP, and 0.01 GTP, pH 7.2. The temperature close to the cell was 32–34°C. Exocytosis was elicited by a train of six (at 1 Hz) 300-ms voltage-clamp depolarization from −70 mV to 0 and detected as changes in cell capacitance estimated by the Lindau-Neher technique (“Sine + DC” feature of the lock-in software module). The amplitude of the sine wave was 20 mV, and the frequency set as 500 Hz.

INS-1 cells were seeded in 24-well tissue culture plates at ∼5 × 10⁵/well, cultured for ∼24 h, and infected with viruses as described above. Twenty-four hours postinfection, the culture medium was replaced by RPMI-1640 with 5 mmol/l glucose and cells incubated for at least 18 h. Insulin secretion was assayed in HEPES balanced salt solution (HBSS) consisting of (in mmol/l) 114 NaCl, 4.7 KCl, 1 MgCl₂, 1.2 KH₂PO₄, 1.16 MgSO₄, 20 HEPES, 2.5 CaCl₂, 25.5 NaHCO₃, and 0.2% BSA, pH ∼7.2 (22). Cells were washed twice with prewarmed (37°C) HBSS buffer with 3 mmol/l glucose followed by 2-h incubation in the same buffer before stimulation with 0.8 ml/well prewarmed HBSS buffer containing 3 or 12 mmol/l glucose for 2 h. The medium was harvested and insulin content determined using Immunochem coated-tube insulin radioimmunoassay from ICN Pharmaceuticals (Costa Mesa, CA). Insulin release in different concentrations of glucose was normalized to that observed at 3 mmol/l glucose and expressed as fold increase.

PIP₂ exerts its effect on K_ATP channel gating largely via the Kir6.2 subunit (11,12). There is evidence that binding of the negatively charged phosphate groups of membrane PIP₂ to the cytoplasmic domain of Kir6.2 leads to channel opening (11,12,24–27). Several positively charged Kir6.2 residues critical for channel gating by PIP₂ have been identified, including R176, R177, and R206 (27,11). Mutation of these residues to the neutral amino acid alanine either reduces or abolishes channel activity in inside-out membrane patches in the absence of inhibitory ATP (these mutants are referred to as PIP₂-insensitive mutants hereinafter). To elucidate the role of K_ATP channel–PIP₂ interactions in β-cell excitability and insulin secretion, we sought to disrupt channel-PIP₂ interactions by overexpressing PIP₂-insensitive Kir6.2 mutants in the rat β-cell line INS-1 clone 832/13, which exhibits robust insulin response to glucose stimulation (22). In COS cells, coexpression of the PIP₂-insensitive Kir6.2 mutant with wild-type Kir6.2 (WT Kir6.2) at 1:1 molar ratio in the presence of SUR1 led to a >50% reduction in total K_ATP currents (data not shown). Mutant Kir6.2 expressed in INS-1 cells is therefore expected to compete with endogenous WT Kir6.2 for incorporation into the channel complex and reduces overall channel sensitivity to PIP₂. Reduced K_ATP channel activity is in turn expected to cause membrane depolarization even in the absence of glucose stimulation (28). We have chosen the recombinant adenovirus approach for expressing mutant Kir6.2 in INS-1 cells because it is much more efficient than conventional transfection methods (29); infection with adenoviruses carrying the green fluorescence protein (GFP) gene yielded >95% GFP-positive cells 12 h after infection.

As shown in Fig. 1, in cells expressing the R176A, R177A, or the R206A Kir6.2 mutant, the RMP (as assessed by whole-cell patch-clamp recording with pipette solution containing 1 mmol/l ATP and 1 mmol/l MgCl₂) in low (3 mmol/l) glucose was indeed significantly more depolarized (approximately −35 to 40 mV) than that observed in uninfected control cells (approximately −72 mV) or cells expressing either GFP (approximately −74 mV) or WT Kir6.2 (approximately −69 mV). To further validate the membrane potential obtained using whole-cell patch-clamp recording, we also measured membrane potential using perforated patch-clamp recording, a procedure that better preserves the intracellular milieu. A substantially more depolarized membrane potential was again observed in cells infected with the PIP₂-insenstive R177A Kir6.2 mutant compared with uninfected cells or cells infected with WT Kir6.2 at low (3 mmol/l) glucose (Fig. 1C) but not at high (12 mmol/l) glucose (not shown). Note that the membrane potential obtained using the perforated patch-clamp technique was ∼20 mV more depolarized than that obtained by whole-cell recording after dialysis with 1 mmol/l MgATP. The difference likely reflects the higher intracellular MgATP concentration in intact INS-1 cells. The above results are consistent with the notion that K_ATP channel–membrane PIP₂ interactions are necessary to sustain K_ATP channel activity and maintain the membrane potential at the resting hyperpolarized state in the absence of glucose stimulation. The depolarized membrane potential in cells expressing mutant Kir6.2 could largely be hyperpolarized to the level seen in uninfected cells (R206A, for example) by adding the K_ATP channel opener diazoxide in the bath solution, demonstrating that the depolarized membrane potential is not due to adverse effects of virus infection or mutant Kir6.2 expression on the health of the cell. Although PIP₂-insenstive Kir6.2 mutants have been reported to render the channel less responsive to diazoxide (30,31), residual sensitivity of mutant channels to the channel opener as well as endogenous WT Kir6.2 subunits present in some channels may be sufficient to confer the diazoxide sensitivity we observed.

To ensure that the titer of each recombinant Kir6.2 adenovirus used for infection gives rise to similar protein expression level, we analyzed Kir6.2 protein level by immunoblotting. Furthermore, to confirm that the mutant Kir6.2 subunit is incorporated into the channel complex and expressed on the surface of INS-1 cells, we tagged one of the mutant R206A with a HA-epitope tag in the extracellular domain between amino acids 100 and 101 (HA-R206A-Kir6.2). This extracellular epitope tag allows for evaluation of surface expression of the mutant without permeabilizing the cell; the HA-tag does not interfere with the function of the channel (data not shown). As Fig. 2A illustrates, the titer of each of the WT-, R176A-, R177A-, or HA-R206A-Kir6.2 viruses could be adjusted to obtain similar protein expression levels (approximately fivefold of endogenous Kir6.2). Immunofluorescent staining of HA-R206A-Kir6.2 also confirmed that the mutant protein is indeed incorporated into the channel complex and expressed at the cell surface (Fig. 2B).

Next, we evaluated insulin secretion response in cells expressing the various PIP₂-insensitive Kir6.2 mutants at both low (3 mmol/l) and high (12 mmol/l) glucose concentrations. These glucose concentrations represent basal and maximal insulin response to glucose stimulation, respectively, based on a glucose dose-response curve (0–20 mmol/l glucose) established in INS-1 cells (not shown) (22). To exclude a direct effect of the Kir6.2 mutant on the insulin secretory machinery, we first measured membrane capacitance changes in response to 500 ms voltage-clamp depolarization. In these experiments, effects of channel-PIP₂ interactions on the membrane potential are excluded and the capacitance change represents fusion of insulin granules that are available for immediate release upon the evoked membrane depolarization. Figure 2C shows that no significant difference in capacitance increase is detected between GFP virus–infected control cells and cells expressing HA-R206A-Kir6.2. Insulin secretion assays demonstrated that at 3 mmol/l glucose, while cells infected with WT-Kir6.2 virus had insulin secretion response indistinguishable from that observed in uninfected control cells (1.1-fold), cells infected with each of the three PIP₂-insensitive Kir6.2 mutants all exhibited significantly elevated insulin secretion (1.9-, 2.4-, and 3.6-fold that of uninfected cells for R176A, R177A, and HA-R206A, respectively; Fig. 2D). These results are consistent with the electrophysiological data (Fig. 1) showing depolarized membrane potential in cells expressing the PIP₂-insensitive mutants. As expected, at a high glucose concentration (12 mmol/l) no significant difference in insulin secretion response could be discerned between cells expressing the various Kir6.2 or uninfected cells because the membrane depolarizing effect caused by the mutant Kir6.2 is now masked by the depolarizing effect caused by high glucose stimulation. Also consistent with the electrophysiological data, treatment of cells with diazoxide reversed the elevated basal level insulin secretion in cells expressing the PIP₂-insensitive Kir6.2 mutants to various extents (Fig. 2E). The results described above provide strong evidence that K_ATP channel–PIP₂ interactions are necessary to allow channel opening and to prevent insulin secretion when glucose concentrations are low.

To assess how increased interactions between K_ATP channels and PIP₂ affect β-cell excitability and insulin secretion, we sought to upregulate PIP₂ levels by overexpressing lipid kinase PIP5K:WT (murine type I isoform β), which is involved in the synthesis of phosphatidylinositol 4,5-P₂ (10,20,21). Overexpression of PIP5K:WT in COSm6 cells has previously been shown to decrease the ATP sensitivity of K_ATP channels exogenously expressed in those cells (21). As a negative control, we infected cells with viruses carrying a mutant kinase whose NH₂-terminal 238 amino acids have been deleted to remove its enzymatic activity (referred to as PIP5K:Δ1-238) (20,21). Figure 3A is a Western blot showing similar expression level of PI4P5K:WT or PIP5K:Δ1-238 in INS-1 cells. To validate that expression of PIP5K:WT increases PIP₂ levels, we metabolically labeled cells with ³²P and analyzed the ³²P-labeled phospholipids by thin-layer chromatography (Fig. 3B). The amount of PIP₂ in cells expressing PIP5K:WT was dramatically increased (∼30-fold) compared with uninfected cells or cells expressing PIP5K:Δ1-238.

We next examined how the increased PIP₂ affects K_ATP channel activity by inside-out patch-clamp recording. In these experiments, patches were excised into K-INT solution with no ATP and then exposed quickly to K-INT containing various concentrations of ATP (all in the absence of Mg²⁺) before any significant current rundown occurred to ensure accurate measurement of ATP dose response (Fig. 4A) (21). As shown in Fig. 4B, K_ATP channels in cells infected with the PIP5K:WT virus exhibited a ∼26-fold decrease in ATP sensitivity compared with channels in uninfected cells or cells infected with the PIP5K:Δ1-238 virus. The large rightward shift in ATP dose response is predicted to cause K_ATP channels to remain active even when the glucose level rises, resulting in hyperpolarized membrane potential and reduced insulin response. To test these, we examined the membrane potential at high glucose using a perforated patch-clamp technique. As mentioned above, the perforated-patch recording permits measurement of membrane potential without disturbing intracellular factors that are involved in regulating K_ATP channel activity. Indeed, we found that at 12 mmol/l glucose, the averaged membrane potential is significantly more hyperpolarized in cells infected with the PIP5K:WT virus than in uninfected cells or cells infected with the PIP5K:Δ1-238 virus (Fig. 4C). Similar results were obtained using whole-cell recordings with 12 mmol/l glucose in the bath solution and 5 mmol/l ATP plus 3 mmol/l Mg²⁺ in the pipette solution (not shown). Accordingly, while in uninfected cells or cells infected with the PIP5K:Δ1-238 virus increasing the glucose concentration from 3 to 12 mmol/l led to a 9.5- and 8.4-fold increase in insulin release, respectively, in cells infected with the PIP5K:WT virus the same glucose concentration change only caused a 3.5-fold increase in insulin release (Fig. 5A). This reduced secretion is not due to an inability of insulin granules to fuse with the membrane because no difference in membrane capacitance increase was found between cells infected with the PIP5K:WT virus and uninfected cells either during the first depolarization or during an entire train of step voltage-clamp depolarizations (Fig. 5C). In addition, overexpression of PIP5K did not alter voltage-gated Ca²⁺ currents during these experiments (shown as the charge influx in Fig. 5D) (32). These results lead us to conclude that overexpression of PIP₂ reduces insulin secretion at high glucose by rendering K_ATP channels insensitive to ATP.

Despite the effects of membrane phosphoinositides, especially PIP₂, on K_ATP channel activity having been well documented in vitro studies for some time (11,12,33,34), the relevance of channel regulation by PIP₂ to insulin secretion remains controversial (15,35). To address this issue, it is necessary to directly correlate the interactions between K_ATP channels and PIP₂ in β-cells to insulin secretion. INS-1 clone 832/13 is a rat insulinoma cell line widely used to study β-cell function; it exhibits robust glucose-stimulated insulin secretion, mediated by both K_ATP-dependent and independent pathways (22). Our studies demonstrate that alterations in K_ATP channel–PIP₂ interactions have profound effects on insulin secretion, establishing a physiological role of PIP₂ in controlling insulin secretion through K_ATP channel regulation.

In altering channel–PIP₂ interactions, we took two approaches: one by manipulating the channel subunit Kir6.2, the other by manipulating cellular PIP₂ levels. While interpretation of experiments in which Kir6.2 subunit is the subject of manipulation is straightforward, interpretation of experiments involving the manipulation of PIP₂ concentrations is more complicated due to its role in diverse cellular processes. For example, recent studies indicate that PIP₂ is required for recruitment of insulin granules to the release sites during the secondary phase of insulin secretion: a decrease in PIP₂ levels reduces, whereas an increase in PIP₂ levels enhances insulin secretion (18,19). In addition, PIP₂ has been shown to control the rate of endocytosis (36–38). Curiously, unlike that reported in primary β-cells and chromaffin cells (18,39), we did not detect a significantly larger increase in membrane capacitance change upon direct voltage-clamp membrane depolarization in INS-1 cells with elevated PIP₂ (Fig. 5B and C). The discrepancy might be due to differences in the level of PIP₂ overexpression; the prolonged high level of PIP₂ seen in PIP5K:WT virus–infected INS-1 cells might lead to increased membrane endocytosis, masking the capacitance increase expected from enhanced exocytosis. Regardless, even if there is an effect of PIP₂ that acts directly on the exocytotic machinery, this effect is likely to be small compared with the effect of PIP₂ via K_ATP channels, as the net result of PIP₂ overexpression in our experiments is a decrease in glucose-induced insulin secretion (Fig. 5A). Increased endocytosis due to elevated PIP₂ might also affect the level of membrane proteins involved in stimulation-secretion coupling, such as K_ATP channels. If true, this would lead to fewer K_ATP channels at the cell surface and increased insulin secretion, making our observed impact of enhanced channel-PIP₂ interaction on insulin secretion an underestimate. Taken together, despite the multiple mechanisms by which PIP₂ may affect insulin secretion, our observation that elevated PIP₂ dramatically reduces the ATP sensitivity of endogenous K_ATP channels (Fig. 4) provides strong consistency with a critical role of membrane PIP₂ in controlling β-cell K_ATP channel activity and hence β-cell excitability and insulin secretion.

In conclusion, our study demonstrates that membrane phosphoinositides play an integral role in determining the activity of K_ATP channels in β-cells. Many hormones, neurotransmitters, and nutrients affect the metabolism of phosphoinositides in β-cells (40–44). Our finding raises the possibility that variation in membrane phosphoinositides under normal and pathological conditions may affect insulin secretion by altering K_ATP channel activity. Mutations in other Kir channels such as Kir2.1 that impair channel sensitivity to PIP₂ have been shown to underlie Anderson syndromes (45). Our study also raises the possibility that naturally occurring mutations in Kir6.2 that alter channel-PIP₂ interactions may cause insulin secretion diseases.

This work was supported by National Institutes of Health (NIH) Grant DK57699 (to S.-L.S.) and the European Foundation for the Study of Diabetes (to S.B.). S.B. was a recipient of a European Molecular Biology Organization long-term fellowship during the course of this study.

We thank Dr. Carol A. Vandenberg for providing the rat Kir6.2 clone, Dr. Christopher Newgard for the 832/13 clone of INS-1 cells, Dr. Peter Rotwein for the GFP adenovirus, and Dr. J. Nathan Davis for the antiserum against PIP5K. We are grateful to Dr. Wolfhard Almers (Portland, OR; supported by NIH MH60600) for the use of capacitance measurement equipment and to Drs. Elizabeth Wilson and Zhengfeng Zhou for advice on constructing adenoviruses.