Insulin Activates ATP-Sensitive K⁺ Channels in Pancreatic β-Cells Through a Phosphatidylinositol 3-Kinase–Dependent Pathway

Mouse pancreases were isolated from 25- to 30-g male Swiss-Webster mice. After cannulation of the bile duct, 2 mg/ml collagenase was injected into the pancreas, which was then removed and incubated for 5–10 min at 37°C to isolate individual islets. Single-cell suspensions were obtained by gently shaking islets in a low-calcium medium. Cell suspensions were plated onto glass coverslips in 35-mm petri dishes containing RPMI-1640 medium supplemented with fetal bovine serum, l-glutamate, and penicillin-streptomycin and were kept in a 95/5% air/CO₂ incubator at 37°C. Single cells were fed every other day and were used for up to 1 week after isolation. Whole islets were cultured in a similar fashion and used within 48 h in culture. To record electrical activity from in situ β-cells, islets were held in place by a wide-tipped glass suction pipette, which was mounted on a micromanipulator (21) (Fig. 2D).

A small-volume recording chamber containing β-cells or islets was mounted on an inverted microscope (Olympus IM-T2; Tokyo), which in turn was placed on a TMC air isolation table (TMC, Peabody, MA). Recording solutions were maintained at 35°C using a TC-1 temperature controller, a preheater, and a second transparent heater attached to the bottom of the recording chamber (Cell Micro Controls, Virginia Beach, VA). The electrophysiological recording solution contained (in millimoles per liter) 140 NaCl, 3.6 KCl, 2 NaHCO₃, 0.5 NaH₂PO₄, 0.5 MgSO₄, 5 HEPES, and 10 glucose, pH 7.4, with NaOH. Extracellular solution was superfused at a rate of 1–2 ml/min. Recording pipettes had resistances ranging from 5 to 10 MΩ and were pulled from borosilicate glass capillaries using a Sutter micropuller (P-97; Sutter Instrument, Novato, CA). The pipette tips were fire-polished and then filled with internal solution containing (in millimoles per liter) 76 K₂SO₄, 10 NaCl, 10 KCl, 1 MgCl₂, and 5 HEPES, pH 7.35, with KOH. The pipettes were then back-filled with the same solution containing 0.1 mg/ml amphotericin-B. The pipette solution for cell-attached single K_ATP channel recordings contained (in millimoles per liter) 140 KCl, 1 CaCl₂, 1 MgCl₂, and 10 HEPES, pH 7.3, with KOH. Seal resistances were >1 GΩ. Wortmannin was obtained from Calbiochem (Calbiochem-Novabiochem, La Jolla, CA), whereas all other chemicals, including recombinant human insulin, were obtained from Sigma (Sigma Chemical, St. Louis, MO). Insulin solutions were prepared fresh daily from 1 mmol/l stock solutions. Cytochrome-C, at a concentration of 0.1 mg/ml, was added to insulin containing solutions to prevent insulin from adhering to the plastic tubing or the experimental chamber. Control solutions contained an identical concentration of cytochrome-C.

Using the perforated patch clamp technique (22), isolated β-cells were identified by their size and characteristic patterns of electrical activity in 10 mmol/l glucose (23), whereas in situ β-cells were identified by their medium bursting in 10 mmol/l glucose (21,24). Integrity of the gigaseals was checked periodically. Data were acquired using an Axopatch 1D amplifier (Axon Instruments, Foster City, CA) and a Macintosh Quadra G3 computer (Apple Computer, Cupertino, CA) using an ITC-16 computer interface (Instrutech, Elmont, NY) and Igor Pro software (Wavemetrics, Lake Oswego, OR). Data were also stored on videotapes for off-line analysis. Linear voltage ramps from −90 to +50 mV lasting 6.2 s were applied from a holding potential of −65 mV after switching the amplifier from current to voltage clamp mode. Whole-cell conductance was measured by averaging currents obtained using five voltage pulses of ±10 mV from a holding potential of −65 mV.

Single K_ATP channels were recorded in the cell-attached patch configuration at a pipette potential of 0 mV using electrodes containing 140 mmol/l KCl. Nifedipine (10 μmol/l) was added to the bath solution in this configuration to prevent capacitatively coupled action potentials. Data were acquired at 5 kHz after filtering at 2 kHz. Single K_ATP channel openings were analyzed using TAC software to generate an event list (Bruxton, Seattle, WA). Single-channel conductance was determined by stepping the pipette potential from −20 to +20 mV. The change in mean single channel current amplitude for the applied step voltage was used to calculate the single-channel conductance, using the equation g = ΔI/ΔV (g = single-channel conductance, ΔV = step voltage, and ΔI = change in mean single channel current amplitude). Mean channel activity (NP_o) was calculated using 10-s sample traces, using the equation NP_o = I_avg/i (I_avg = the average current of the analyzed trace and i = the mean single channel current amplitude). The fact that liquid junction potentials were not compensated and the resting membrane potential of the cell was unclamped may have influenced the amplitude of the unitary currents. K_ATP channels were identified in cell-attached patch recordings by their characteristic slope conductance and their sensitivity to changes in bath glucose concentration. To further establish that these events correspond to the opening of single K_ATP channels, we confirmed that they were strongly activated by the addition of 2 mmol/l sodium azide to bath solutions containing 10 mmol/l glucose and were promptly blocked by the addition of 1 μmol/l glyburide (data not shown).

Islets were loaded with the Ca²⁺-sensitive probe Fura-2/acetoxymethyl. This probe (2 μmol/l) and 2 μl of 2.5% pluronic acid solution were added to culture dishes containing 1 ml of medium and were incubated for 30 min in an air/CO² incubator at 37°C. After loading, islets were washed and incubated in recording solution for 20 min. [Ca²⁺]_i was measured using an IX-50 inverted fluorescence microscope (Olympus, Tokyo) in conjunction with a xenon arc excitation source, a photomultiplier, and a photon counter (Ionoptix, Milton, MA). Data were acquired and analyzed with Ionwizard software provided by the manufacturer. Epifluorescence was measured from single islets, and the values in [Ca²⁺]_i were determined as the fluorescence ratio (R) of Ca²⁺-bound Fura-2 (340 nm) to unbound Fura-2 (380 nm), with emission collected at 510 nm. R was converted to free Ca²⁺ concentration using an in vitro standard curve generated with a calibration kit (Molecular Probes, Eugene, OR).

Statistical analysis was done using paired t test or nonparametric one-way analysis of variance with Newman-Keuls Multiple Comparison Test to compare more than two values. P values <0.05 were considered statistically significant.

The electrical activity of single dispersed β-cells in 10 mmol/l glucose consists of repeated large spikes, fast (e.g., <5-s) bursts, or plateau depolarizations (23), whereas medium bursting having a duration of 10–20 s only occurs in intact islets (21,24).

Using the perforated patch clamp technique (22) to record from single mouse β-cells bathed in 10 mmol/l glucose, we found that the application of insulin (1–600 nmol/l) resulted in a reversible inhibition of β-cell electrical activity in 73% (63 of 86) of the cells tested (Fig. 1A), with insulin doses >1 nmol/l producing statistically significant changes in β-cell membrane potential. The dose-response relationship for exogenous insulin yielded a maximal change (mean ± SE) in membrane potential of −13.6 ± 2.0 mV and a 50% effective dose (ED₅₀) of 25.9 ± 0.1 nmol/l, with an R² value of 0.9959, n = 63 (Fig. 1B). Although insulin-induced hyperpolarization was reversible, the recovery time appeared to be proportional to the total duration of insulin application.

Insulin is known to bind to and activate IGF-1 and insulin receptors (8,25). IGF-1 receptors are structurally and functionally homologous to insulin receptors and belong to the same family of receptors having tyrosine kinase activity (25). To rule out the possibility that insulin-induced β-cell hyperpolarization resulted from binding to IGF-1 receptors, equimolar concentrations of IGF-1 and insulin were tested on the same β-cells. As shown in Fig. 1C, 100 nmol/l IGF-1 applied alone for up to 5 min failed to suppress β-cell electrical activity in 10 mmol/l glucose (n = 6). However, the subsequent addition of 100 nmol/l insulin resulted in β-cell hyperpolarization within 2 min. Thus, the mean membrane potential after IGF-1 application was −45.0 ± 2.6 mV, which was not significantly different from control (−45.0 ± 3.2 mV), whereas the subsequent addition of 100 nmol/l insulin to the bath hyperpolarized these β-cells to −56.3 ± 4.5 mV (P < 0.001, n = 4).

Recordings from β-cells within intact islets exposed to 200 nmol/l insulin produced similar results (Fig. 2A). Thus, insulin hyperpolarized islets from −47.7 ± 3.3 to −65.6 ± 3.7 mV (P < 0.0001, n = 11). The mean change in membrane potential for islets versus single dispersed β-cells was −17.9 ± 2.8 mV (n = 11) vs. −12.7 ± 1.2 mV (n = 14), respectively (P > 0.05). However, the mean time to maximum response was 9.4 ± 1.7 min for islets (n = 11) and 3.7 ± 0.5 min for isolated β-cells (n = 14, P < 0.05), indicating that insulin took a significantly longer time to produce its effect on in situ β-cells. This difference likely reflects the longer diffusion time required for insulin to activate a sufficient number of β-cells within the intact islet (R. Bertram and L. Satin, unpublished observations).

We next tested whether insulin was capable of abolishing the regular [Ca²⁺]_i oscillations of intact mouse islets bathed in 10 mmol/l glucose. Using the ratiometric Ca²⁺-sensing dye Fura-2 (see research design and methods), regular [Ca²⁺]_i oscillations were observed (Fig. 2B) between 251.0 ± 26.8 and 457.6 ± 42.9 nmol/l (P < 0.05, n = 5) (Fig. 2C), corresponding to the silent and peak phases of the [Ca²⁺]_i oscillations. As shown in Fig. 2B, the [Ca²⁺]_i oscillations were completely abolished within ∼9 min of exposure to 100 nmol/l insulin, resulting in a new steady-state [Ca²⁺]_i level of 300.4 ± 40.0 nmol/l (n = 5) (Figs. 2B and C). These data clearly demonstrate that insulin not only hyperpolarizes islets but abolishes the regular [Ca²⁺]_i oscillations, which are known be important for pulsatile insulin secretion.

Insulin has been shown to hyperpolarize hypothalamic neurons by activating their K_ATP channels (20). We therefore hypothesized that a similar mechanism might account for the insulin-induced hyperpolarization of β-cells. In support of this hypothesis, we found that tolbutamide, a specific blocker of K_ATP channels (26), completely reversed insulin-induced β-cell hyperpolarization (Fig. 3A). Thus, whereas 100 nmol/l insulin hyperpolarized β-cells from −46.7 ± 2.4 to −62.1 ± 1.9 mV (P < 0.001, n = 7), the subsequent addition of 100 μmol/l tolbutamide depolarized membrane potential back to control levels (−44.1 ± 2.1 mV). Whole-cell conductance measured using voltage pulses in voltage clamp was 0.6 ± 0.1, 1.7 ± 0.2 (P < 0.001), and 0.5 ± 0.1 nS for control, insulin, and insulin plus tolbutamide-containing solutions (n = 10), respectively. Thus, insulin-induced hyperpolarization was associated with a threefold increase in cell membrane conductance, which was completely reversed by tolbutamide (Fig. 3B). Whole-cell conductance corresponding to the period of insulin-induced hyperpolarization was within the reported range (e.g., 0.27–7.6 nS) for the calculated whole-cell K_ATP conductance of single rodent β-cells (1,27). Figure 3B shows cell conductance after normalization for cell capacitance under various conditions.

Application of voltage ramps to single β-cells voltage-clamped to −65 mV in 10 mmol/l glucose revealed an outward-rectifying whole-cell current that reversed at −25.7 ± 2.5 mV (n = 9). Insulin (100 nmol/l) increased the slope of the linear part of the current-voltage relationship between −90 to −40 mV, which is known to reflect the contribution of K_ATP channels to β-cell whole-cell current, and shifted the reversal potential to −63.7 ± 1.0 mV (P < 0.001 vs. control, n = 9), toward the calculated K⁺ equilibrium potential of −90 mV (Fig. 3C). The subsequent application of tolbutamide inhibited the insulin-activated current and shifted its reversal potential to near control values (−25.2 ± 2.3 mV). These data thus strongly suggest that insulin hyperpolarizes β-cells by activating K_ATP channels and increasing β-cell K⁺ conductance.

PI 3-kinase is a pivotal enzyme in insulin signal transduction both in classical insulin target tissues (28) and insulin-secreting β-cells (8,9). A PI 3-kinase–dependent pathway has also been implicated in leptin activation of K_ATP channels in the insulin-secreting CRI-G1 cell line (29) and in insulin activation of neuronal K_ATP channels (20). We thus tested whether wortmannin, a PI 3-kinase inhibitor, could reverse insulin-induced β-cell hyperpolarization. As expected, the application of 100 nmol/l insulin hyperpolarized β-cells from −53.6 ± 1.6 to −69.8 ± 2.2 mV (P < 0.001, n = 5) and terminated their electrical activity (Fig. 4A). The subsequent application of 100 nmol/l wortmannin to inhibit PI 3-kinase slowly depolarized these cells to −53.2 ± 2.1 mV (P < 0.05) over a mean period of 4.5 ± 1.1 min (n = 5). In separate experiments, we found that pretreating β-cells with wortmannin fully blocked the inhibitory effects of insulin (data not shown).

We next tested the effects of insulin and wortmannin on unitary K_ATP channel activity measured in cell-attached patch recordings from single mouse β-cells. As shown in Fig. 4B, in the absence of glucose, single channels having a slope conductance of 44.8 ± 4.8 pS (n = 5) were observed. This conductance corresponds to the reported unitary conductance for β-cell K_ATP channels (∼50 pS) (1,2,3). NP_o in glucose-free solution was 0.49 ± 0.1 (n = 4) (Fig. 4C), whereas increasing the bath glucose concentration to 10 mmol/l decreased channel activity to 0.08 ± 0.04 (P < 0.005). The subsequent application of 100 nmol/l insulin reversed the action of glucose, resulting in an increase in mean K_ATP channel activity to 0.37 ± 0.07 (P < 0.005), supporting the hypothesis that glucose-inhibited K_ATP channels are activated by insulin. Because single K_ATP channels in this recording configuration were not directly exposed to insulin, the efficacy of bath-applied insulin suggests that insulin opens K_ATP channels through an intracellular second messenger. The subsequent addition of 100 nmol/l wortmannin resulted in the suppression of NP_o to 0.04 ± 0.01 (P < 0.005), which was not significantly different from that observed in solutions containing 10 mmol/l glucose (P > 0.05) (Fig. 4C).

We thus report that insulin activates the K_ATP channels of mouse pancreatic β-cells through a PI 3-kinase–dependent pathway, resulting in β-cell hyperpolarization, a cessation of glucose-induced electrical activity, and inhibition of islet [Ca²⁺]_i oscillations. The peptide hormones leptin (29,30,31), somatostatin (32), and galanin (33) and “K_ATP-opener” drugs, such as diazoxide (34) and minoxidil (35), hyperpolarize β-cells and inhibit insulin secretion by activating K_ATP channels, as we found for insulin. In the case of leptin, which activates K_ATP channels in ob/ob mouse β-cells (31) and rat CRI-G1 insulinoma cells, a PI 3-kinase-dependent pathway also appears to be involved (30), as we found for insulin activation of β-cell K_ATP channels.

The actions of insulin that we describe are likely to be mediated by insulin receptors because IGF-1 failed to hyperpolarize β-cells, and the EC₅₀ we obtained for insulin compared favorably with the reported EC₅₀ (∼4 nmol/l) of β-cell insulin receptors (8,16,17). Although we and others (8,11,16,17) have exposed β-cells to higher concentrations of insulin to observe acute effects (e.g. 100 nmol/l), even low nanomolar doses of insulin significantly activated β-cell K_ATP channels. Circulating insulin levels in this range have been reported in the plasma of insulin-resistant human subjects after a glucose load (36,37,38,39) and in the portal circulation (40). Additionally, we would expect that still higher insulin concentrations occur within islets because of the local accumulation of newly secreted insulin into their tight interstitial spaces. Thus, the mechanism that we describe is likely to functionally alter β-cell electrophysiology and modify insulin secretion.

β-cells began to hyperpolarize within tens of seconds of adding exogenous insulin, whereas maximal hyperpolarization occurred after several minutes. These kinetics are consistent with a large number of studies of insulin action in its classic target tissues, where insulin-induced insulin receptor phosphorylation, phosphorylation of IRS proteins, and activation of PI 3-kinase and Akt/protein kinase B occur tens of seconds to minutes after exposure to insulin (41,42,43,44,45), whereas the translocation of glucose transporters and glucose transport takes many minutes (46). In βTC-3 cells, the autocrine action of insulin to stimulate insulin receptor phosphorylation occurs within ∼2 minutes (8). It thus seems reasonable to conclude that the time lag we observed for insulin activation of β-cell K_ATP channels reflects the time course of the insulin receptor signal transduction cascade, as in muscle cells and adipocytes. We attribute the longer delays we observed using in situ versus isolated β-cells to the time required for insulin to diffuse within the intact islet. Although we do not yet know how insulin activates the K_ATP channels of β-cells, PI 3-kinase could conceivably activate K_ATP channels by altering membrane phosphoinositide levels, because these have been shown to reduce the ATP sensitivity of K_ATP channels (47).

Although this is the first report showing that insulin activates β-cell K_ATP channels, insulin activates the K_ATP channels of rat hypothalamic neurons through a similar mechanism involving PI 3-kinase (20). Thus, these authors found that 5–300 nmol/l insulin hyperpolarized a subpopulation of the hypothalamic neurons of lean Zucker rats and decreased their input resistance. Voltage clamp studies established that insulin activated an ionic current under these conditions that reversed at the K⁺ equilibrium potential and was blocked by 200 μmol/l tolbutamide. The authors extended these studies to the single-channel level and showed that insulin activated single K_ATP channels in cell-attached patches. As we found in β-cells, insulin-induced hyperpolarization of hypothalamic neurons required PI 3-kinase activation, as it was blocked by wortmannin or LY294002. Based on these findings, Spanswick et al. (20) proposed that insulin serves as a negative feedback signal in the hypothalamus to reduce food intake through the inhibition of glucose-responsive neurons.

However, in contrast to our findings, Aspinwall et al. (16,17) reported that 1–100 nmol/l insulin applied to single mouse β-cells increased rather than decreased insulin exocytosis, as assayed using 5–hydroxytryptamine amperometry, and did so independently of changes in β-cell membrane potential. These authors proposed that insulin stimulates insulin secretion by releasing Ca²⁺ from ER Ca²⁺ stores via a mechanism involving IRS-1 and PI 3-kinase, because the insulin-induced changes they observed were wortmannin-sensitive. However, because Aspinwall et al. (16,17) carried out their studies in 3 mmol/l glucose, it is reasonable to expect that most of the β-cell’s K_ATP channels were already open and were thus relatively insensitive to further modulation (1). Under basal conditions such as these, minimal amounts of insulin would be secreted, which further questions the relevance of these findings for glucose-stimulated insulin secretion. In another recent study, Harvey and Ashford (29) reported that whereas leptin activated CRI-G1 cell K_ATP channels, insulin did not. Although we do not know why insulin failed to open K_ATP channels in this preparation, CRI-G1 cells differ in several respects from the adult mouse β-cells we used and are glucose insensitive. Thus, it may be that metabolic differences between the two preparations might account for this discrepancy.

Insulin activation of β-cell K_ATP channels has a number of implications for understanding islet function. First, in general, this negative feedback mechanism would be expected to restrain excessive insulin exocytosis when intra-islet insulin is sufficiently high, preventing hyperinsulinemia. Second, in terms of islet electrophysiology, the mechanism may help explain why intact islets have regular 10- to 60-s electrical oscillations (21,24), compared with the fast bursting or spiking of isolated β-cells (23). Thus, it may be that insulin released within islets during glucose-stimulated electrical activity in turn inhibits islet electrical activity by activating β-cell K_ATP channels. Secretion would then decrease, as [Ca²⁺]_i decreases, and the cycle would be repeated as secreted insulin diffuses away from β-cell insulin receptors and K_ATP channels close. However, in isolated β-cells lacking tight intercellular spaces, insulin once secreted would be likely to dissipate quickly, precluding autocrine or possibly paracrine effects. Thus, a loss of intra-islet signaling might account for the lack of islet-like electrical activity displayed by isolated β-cells (21,23,24).

We have quantitatively tested the feasibility of this proposed mechanism using a recent paradigm of bursting, the phantom burster model (48). In this model, islet bursting results from the interaction of two slow processes, one (S1) with a time constant of a few seconds, and another (S2) with a time constant of a few minutes. The slow kinetics of activation of K_ATP current by insulin makes this mechanism an attractive candidate for the slower S2 process. In support of this hypothesis, we found that a variation of the phantom burster model incorporating insulin activation of K_ATP channels as the slow S2 process was able to simulate the medium bursting of intact islets (A. Sherman, L. Satin, R. Bertram, unpublished observations). Furthermore, removing the slow process, which might be akin to dispersing islets into single β-cells and thus eliminating the intercellular spaces needed for insulin to accumulate in, resulted in faster, S1-dominated bursting resembling that of dispersed β-cells (23). These computations thus support the general feasibility of this proposed mechanism of bursting, although the hypothesis remains to be tested by experiment.

Dysregulated insulin signaling in this novel negative feedback system, perhaps resulting from biochemical or molecular defects in the β-cell insulin receptor signaling cascade, could be significant for diabetes, because it may result in abnormal insulin secretion including defective pulsatility if insulin activation of K_ATP channels is required for periodic islet electrical activity.

This work was supported by National Institutes of Health grant DK-46409 to L.S.S.

We thank Drs. A. Sherman, R. Bertram, and C. Baumgarten for helpful discussions and H. Strange for technical assistance. We thank Dr. John Nestler for support and encouragement.