NADH Shuttle System Regulates KATPChannel-dependent Pathway and Steps Distal to Cytosolic Ca2+ Concentration Elevation in Glucose-induced Insulin Secretion*

The NADH shuttle system is composed of the glycerol phosphate and malate-aspartate shuttles. We generated mice that lack mitochondrial glycerol-3-phosphate dehydrogenase (mGPDH), a rate-limiting enzyme of the glycerol phosphate shuttle. Application of aminooxyacetate, an inhibitor of the malate-aspartate shuttle, to mGPDH-deficient islets demonstrated that the NADH shuttle system was essential for coupling glycolysis with activation of mitochondrial ATP generation to trigger glucose-induced insulin secretion. The present study revealed that blocking the NADH shuttle system severely suppressed closure of the ATP-sensitive potassium (KATP) channel and depolarization of the plasma membrane in response to glucose in β cells, although properties of the KATP channel on the excised β cell membrane were unaffected. In mGPDH-deficient islets treated with aminooxyacetate, Ca2+ influx through the plasma membrane induced by a depolarizing concentration of KCl in the presence of the KATP channel opener diazoxide restored insulin secretion. However, the level of the secretion was only ∼40% of wild-type controls. Thus, glucose metabolism through the NADH shuttle system leading to efficient ATP generation is pivotal to activation of both the KATP channel-dependent pathway and steps distal to an elevation of cytosolic Ca2+ concentration in glucose-induced insulin secretion.

In pancreatic ␤ cells, glucose metabolism in glycolysis and in mitochondria is pivotal to glucose-induced insulin secretion. Thus, mutations in the glucokinase gene (1,2) and mitochondrial DNA (3) can cause type 2 diabetes mellitus, which is characterized by insufficient insulin secretion in response to glucose. Moreover, abrogation of ␤ cell-specific glucokinase in mice (4,5) and mitochondrial DNA in insulinoma cell lines (6 -9) was associated with defective insulin secretion in response to glucose. The following cascade has been generally accepted as the glucose-induced insulin secretory pathway (10 -13). Glucose-stimulated increase in cytosolic ATP or in the ratio of ATP to ADP closes ATP-sensitive potassium (K ATP ) 1 channels, which depolarizes the plasma membrane potential above a threshold, leading to Ca 2ϩ entry into the cytosol through activation of voltage-dependent Ca 2ϩ channels (VD-CCs). The rise in cytosolic Ca 2ϩ concentration ([Ca 2ϩ ] c ) is thought to finally trigger exocytosis of insulin from secretory vesicles.
It has been known that pyruvate, an end product of aerobic glycolysis, cannot stimulate insulin secretion although it is oxidized as efficiently as glucose in ␤ cells (14,15). Studies with an inhibitor of pyruvate transport into mitochondria or an inhibitor of the tricarboxylic acid cycle suggested that metabolism of glucose-derived pyruvate in mitochondria or in the tricarboxylic acid cycle was not well correlated with glucoseinduced insulin secretion (16,17). These results suggested that glycolysis-derived factor(s) other than pyruvate are required for activation of mitochondrial metabolism and for generation of the metabolic signals such as ATP. NADH, which is generated at a step catalyzed by a glycolytic enzyme glyceraldehyde-3 phosphate dehydrogenase, is one such candidate. This is because the electrons of the cytosolic NADH are transferred into the mitochondrial electron transfer chain directly through the NADH shuttle system and enhance the mitochondrial membrane potential. There are two known NADH shuttles, the glycerol phosphate shuttle and malate-aspartate shuttle (18,19). Recently, we have generated mice that lack mitochondrial glycerol-3-phosphate dehydrogenase (mGPDH), a rate-limiting enzyme of the glycerol phosphate shuttle (20). When aminooxyacetate (AOA), a well characterized inhibitor of aspartate aminotransferases in the malate-aspartate shuttle (19,(21)(22)(23), was applied to mGPDH-deficient islets, glucose-induced increases in glucose oxidation at the tricarboxylic acid cycle, production of NAD(P)H autofluorescence, hyperpolarization of mitochondrial membrane potential, calcium influx into mitochondria, and cellular ATP content were severely impaired, and insulin secretion was completely abrogated (20). These results demonstrated that the NADH shuttle system is essential for coupling glycolysis with mitochondrial ATP generation to trigger glucose-induced insulin secretion. However, the mechanism by which deprivation of cellular ATP led to the downstream events that culminated in abrogation of insulin secretion was not addressed in the previous study. Under the conditions where the NADH shuttle system was stopped, the change in [Ca 2ϩ ] c in response to glucose was significantly impaired. An initial drop below a base line and the first phase peak of [Ca 2ϩ ] c failed to occur, although it was followed by a gradual elevation of [Ca 2ϩ ] c to the second phase plateau level in 10 to 15 min (20). These results suggested that Ca 2ϩ uptake of endoplasmic reticulum by Ca 2ϩ -ATPase and Ca 2ϩ influx through VDCCs from the extracellular compartment were severely impaired (24 -26). In ␤ cells, the activity of VDCCs is controlled by the plasma membrane potential, which in turn is mainly regulated by the activity of the K ATP channel (27)(28)(29). Thus, we hypothesized that glucose metabolism without the functional NADH shuttle system was not sufficient to close the K ATP channel due to inefficient generation of ATP and to elicit depolarization of the plasma membrane and generation of action potentials. Moreover, steps distal to an elevation of [Ca 2ϩ ] c are also energy-requiring (30 -33). These steps including mobilization of secretory granules toward the plasma membrane, priming of the granules, and a final fusion process may also have been affected. In this article, using electrophysiological techniques we provide evidence that glucose metabolism through the NADH shuttle system is essential to regulate the K ATP channel-dependent pathway in glucose-induced insulin secretion. In addition, with diazoxide and a depolarizing concentration of KCl we show that steps distal to an elevation of [Ca 2ϩ ] c are also regulated by glucose metabolism through the system.

EXPERIMENTAL PROCEDURES
Islet Isolation and ␤ Cell Preparation-Isolation of islets from mG-PDH-deficient mice (20) and their wild-type littermates was carried out as described previously (34). In brief, after clamping the common bile duct at a point close to the duodenum outlet, 2.5 ml of Krebs-Ringer bicarbonate buffer (118.4 mM NaCl, 4.7 mM KCl, 1.3 mM MgSO 4 , 1.2 mM KH 2 PO 4 , 2.0 mM CaCl 2 , 10 mM NaHCO 3 , and 10 mM HEPES at pH 7.4) containing collagenase (Sigma) was injected into the duct. The swollen pancreas was taken out and incubated at 37°C for 3 min. The pancreas was dispersed by pipetting and washed two times with Krebs-Ringer bicarbonate buffer. Islets were collected by manual picking and incubated in RPMI medium (Life Technologies, Inc.) supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, and 10% fetal calf serum. Isolation of single ␤ cells was performed using 1000 units/ml dispase (Godo Shusei, Japan), as described previously (35).
Analysis of Insulin Secretion-Insulin secretion from islets was measured with Krebs-Ringer bicarbonate buffer with a basal glucose concentration of 2.8 mM unless otherwise stated. Batch incubation was performed with 10 islets/tube at 37°C for 1 h after preincubation with the basal glucose concentration for 20 min. Perifusion experiments were performed with 30 islets/chamber at 37°C with a flow rate of 0.6 ml. AOA, when used, was added to both preincubation and incubation medium. AOA and diazoxide were purchased from Sigma.
Measurement of [Ca 2ϩ ] c -A single islet was placed under a microscope (IMT-2, Olympus, Japan) and perifused with Sol II buffer containing 150 mM NaCl, 5 mM KCl, 1.0 mM MgCl 2 , 2.0 mM CaCl 2 , and 10 mM HEPES (pH 7.4) at 37°C. Fluorescence was excited with light emitted from a xenon lamp (TILL Photonics, Germany), collected through interference filters (Olympus, Japan), and detected using a photomultiplier (NT5783, Hamamatsu Photonics, Japan). Islets were loaded with 15 M fura-2/acetoxymethylester (Molecular Probes, Eugene, OR) at 37°C for 1 h. The fluorescence was excited with a dualwavelength ratiometric mode at 340 and 380 nm. The emission wavelength was filtered at 500 nm.
Electrical Recordings-Dispersed islet cells were kept in a 35-mm Petri dish on an inverted microscope. Identification of ␤ cells from other islet cells was carried out by detecting the excitatory electrical response of the cells to 15 mM glucose or to 0.5 mM tolbutamide. The patch clamp method was used to record membrane currents and membrane potential with an EPC-7 patch-clamp amplifier (List Electronic, Germany) (36,37). The resistance of electrodes made of borosillicate glass capillaries was 2-4 megaohms. For recording the plasma membrane potential, the nystatin-perforated whole-cell configuration was used (38,39). For recording the whole-cell membrane currents, the conventional whole-cell mode was used. The membrane capacitance ranged from 12 to 20 picofarads. The membrane potential was repeatedly changed from Ϫ90 to Ϫ50 mV by ramp pulses with a ramp pulse generator (SET-2100, Nihon Kohden, Japan). Both the rising and falling time of each ramp pulse was 2 s in duration. In the inside-out mode, single channel currents were recorded. The data were analyzed by using a single channel current analysis program (QP-129J, Nihon Kohden). All experiments were performed at room temperature. The standard extracellular solution contained 130 mM NaCl, 5 mM KCl, 1.2 mM MgCl 2 , 1 mM CaCl 2 , 2 mM glucose, and 10 mM HEPES (pH 7.3). When the cells were stimulated with glucose, sucrose was added to the control solution to adjust the osmolarity. In the nystatin-perforation recordings, the pipette solution contained 135 mM KCl, 1.2 mM MgCl 2 , 0.5 mM EGTA, 200 g/ml nystatin (Sigma), and 10 mM HEPES (pH 7.3). In the conventional whole-cell recordings, the pipette solution contained 100 mM potassium gluconate, 35 mM KCl, 1.2 mM MgCl 2 , 0.5 mM EGTA, 1 mM Na 2 ATP, 1 mM GTP, 2 mM glucose, and 10 mM HEPES (pH 7.2). In single channel recordings, both pipette (external) and internal solutions contained 135 mM KCl, 1.2 mM MgCl 2 , 0.5 mM EGTA, 5 mM tetraethylammonium chloride (Sigma), 10 mM glucose, and 10 mM HEPES. The pH was 7.3 in the external solution and 7.2 in the internal solution. Cells in the bath were superfused with a stream of an extracellular solution. Drugs were dissolved in the extracellular solution and applied to cells by switching flow tubes.
Measurement of Islet ATP Content-Ten islets were incubated with Krebs-Ringer bicarbonate buffer containing 2.8, 10.0, or 16.7 mM glucose at 37°C for 1 h. After aspiration of medium, the metabolism was snap-stopped by an addition of ice-cold 1 N perchloric acid and freezing with liquid nitrogen. Then, samples were thawed, sonicated, and neutralized with NaOH (40). ATP content was measured using a luciferaseluciferin system (Sigma).
Statistical Analysis-Data were expressed as means Ϯ S.E. Statistical significance was evaluated with the Student's t test or by analysis of variance.

Effect of AOA on Glucose-responsive Changes in Insulin Secretion and [Ca 2ϩ ] c in mGPDH-deficient Islets-
In static incubation for 1 h, the increase in insulin secretion stimulated by 16.7 mM glucose was inhibited by 94% in mGPDH-deficient mice islets treated with 5 mM AOA (0.11 Ϯ 0.03 ng/islet/h) compared with wild-type islets without AOA (1.83 Ϯ 0.19 ng/ islet/h). Perifusion experiments revealed that the first phase of the secretion was severely decreased and the second phase was totally abrogated (Fig. 1A, closed squares) compared with wildtype islets without AOA (Fig. 1A, closed circles). The inhibitory effect of AOA on glucose-induced insulin secretion from mG-PDH-deficient islets was dose-dependent, and the half-maximal inhibitory concentration of AOA was ϳ1 mM at 16.7 mM glucose (Fig. 1B). As previously reported (20), an initial small drop and the first phase peak of [Ca 2ϩ ] c , which were normally observed in wild-type islets without AOA in response to 22.2 mM glucose (Fig. 1C, left panel), disappeared when 5 mM AOA was applied to mGPDH-deficient islets (Fig. 1C, right panel). However, [Ca 2ϩ ] c gradually rose in 10 to 15 min to the second phase plateau level, which was comparable to that of wild-type islets without AOA. At 1 mM AOA, the initial drop and the subsequent first phase peak of [Ca 2ϩ ] c remained missing, whereas the second phase plateau was reached in 6 to 8 min (data not shown). These results suggested that both the appearance of the first phase peak of [Ca 2ϩ ] c and the time required to raise [Ca 2ϩ ] c to the second phase level were dependent on glucose metabolism through the NADH shuttle system and that loss of function of the shuttle system impaired Ca 2ϩ influx into the cytosol. Stimulation by 15 mM glucose instead of 16.7 or 22.2 mM glucose produced essentially similar results to the above experiments.
Effect of AOA on the Membrane Potential in mGPDH-deficient ␤ Cells-We next directly measured the plasma membrane potential of ␤ cells with patch clamp techniques. Under the nystatin-perforated whole-cell mode, the resting membrane potentials in the presence of 2 mM glucose were Ϫ58.4 Ϯ 1.3 mV (n ϭ 24) in wild-type ␤ cells, Ϫ61.5 Ϯ 1.6 mV (n ϭ 24) in wild-type ␤ cells with 5 mM AOA, Ϫ61.3 Ϯ 1.9 mV (n ϭ 23) in mGPDH-deficient ␤ cells, and Ϫ62.5 Ϯ 1.7 mV (n ϭ 22) in mGPDH-deficient ␤ cells with 5 mM AOA. There were no significant differences among the four groups. In wild-type ␤ cells treated with 5 mM AOA, stimulation by 15 mM glucose normally depolarized the plasma membrane potential to ϳ50 mV in 70 -90 s, where action potentials were generated ( Fig. 2A). These normal responses were also observed in mGPDH-deficient ␤ cells without AOA (Fig. 2B). However, in mGPDHdeficient ␤ cells treated with 5 mM AOA, the membrane potential was only minimally depolarized in response to 15 mM glucose and failed to reach the threshold and to generate action potentials within the measurement period of 4 min (Fig. 2C).
Application of 0.5 mM tolbutamide to extracellular solution caused the depolarization of the plasma membrane and generated action potentials even in mGPDH-deficient ␤ cells treated with AOA (Fig. 2D). This suggested that sensitivity of the K ATP channel to sulfonylurea was preserved under these conditions.
Because the second phase plateau of [Ca 2ϩ ] c was attained within 10 to 15 min after glucose stimulation in mGPDHdeficient islets treated with 5 mM AOA (Fig. 1C, right panel), we presumed that the membrane potential of ␤ cells might take longer to reach the threshold. The measurement period was then extended to 15 min. Under these conditions, membrane depolarization and generation of action potentials were observed in 2 of 6 experiments; one at 6 min (Fig. 2E), and the other at 7 min after the glucose stimulation. Thus, small number of mGPDH-deficient ␤ cells treated with AOA retained electrical excitability to glucose stimulation, although the initiation of the response was markedly delayed. Effect of AOA on the K ATP Channel Current in mGPDHdeficient ␤ Cells-We then studied the K ATP channel current in response to glucose in a conventional whole-cell mode. Ramp voltage pulses from Ϫ90 mV to Ϫ50 mV were given every 4 s. In mGPDH-deficient ␤ cells, the K ATP channel current in response to the 40-mV voltage pulse was decreased from ϳ40 to ϳ15 pA by elevation of the extracellular glucose concentration from 2 to 15 mM (Fig. 3A, upper panel). At several time points during the trace, the mode of recording was transiently changed to a current clamp mode, where a stepwise depolarization of the membrane potential culminating in the generation of action potentials was clearly observed (Fig. 3A, lower   panel). However, in mGPDH-deficient ␤ cells treated with 5 mM AOA, no decrease in the K ATP channel current (Fig. 3B, upper  panel) or depolarization of the membrane potential (Fig. 3B, lower panel) in response to 15 mM glucose was observed. In contrast, application of 0.5 mM tolbutamide to the cells reversibly abolished the K ATP channel current (Fig. 3B, lower panel).
Electrophysiological Characteristics of the K ATP Channel on the Excised Plasma Membrane-We then performed electrophysiological analyses of the K ATP channel itself in an insideout patch of the plasma membrane excised from ␤ cells. At a recording voltage of Ϫ60 mV, the amplitude of the K ATP channel current in mGPDH-deficient ␤ cells was the same as in wild-type ␤ cells (Fig. 4A). When the K ATP channel currents were plotted against the recording voltage from Ϫ60 mV to 60 mV at an interval of 10 mV, the current-voltage relationship of mGPDH-deficient ␤ cells essentially overlapped that of wildtype ␤ cells, which show a weak inward rectification (Fig. 4B). The single K ATP channel conductance of mGPDH-deficient ␤ cells measured between Ϫ60 mV and 0 mV was 73.1 Ϯ 3.8 picosiemens (n ϭ 4) and was not different from that of wildtype ␤ cells (72.7 Ϯ 3.3 picosiemens, n ϭ 4).
Inhibition by ATP of the K ATP Channel Current on the Excised Plasma Membrane-We next studied the effect of ATP on the K ATP channel activity in the inside-out mode with the recording voltage at Ϫ60 mV. When 1 mM ATP was applied to the cytoplasmic surface of mGPDH-deficient ␤ cell plasma membranes, the K ATP channel current was immediately and completely abolished (Fig. 5A). The channel open probabilities at 0.1, 1, 10, 100, and 1000 M ATP in mGPDH-deficient ␤ cells were not different from those in wild-type ␤ cells (Fig. 5B). These values were well in accordance with the theoretical values calculated with the Hill equation (41) (Fig. 5B). Thus, the electrophysiological characteristics of the K ATP channel and its sensitivity to ATP per se were normal in mGPDH-deficient ␤ cells.

Role of the NADH Shuttle System in Steps Distal to an Elevation of [Ca 2ϩ
] c -When both the NADH shuttles were stopped by treating mGPDH-deficient ␤ cells with 5 mM AOA, the first phase peak of [Ca 2ϩ ] c in response to glucose disappeared (Fig. 1C, right panel). Moreover, it took 10 to 15 min for [Ca 2ϩ ] c to reach the second phase plateau. These findings indicated an impairment of Ca 2ϩ influx through the plasma membrane during this period. We then postulated that a restoration of Ca 2ϩ influx could recover insulin secretion. To test this, islets were treated with diazoxide, a K ATP channel opener, to hold the K ATP channel open in combination with a high concentration of KCl to depolarize the plasma membrane. Under these conditions, it has been shown that glucose does not further elevate [Ca 2ϩ ] c but amplifies the actions of Ca 2ϩ on the insulin-secreting processes (31,32,42,43). With application of 30 mM KCl in the presence of 250 M diazoxide, the insulin secretory response to glucose was indeed restored in mGPDHdeficient islets treated with 5 mM AOA (Fig. 6, A-C, closed  squares). The amounts of secreted insulin were increased by elevating glucose concentration from 2.8 to 10.0 mM and to 16.7 mM. Insulin released by this treatment was, however, still ϳ60% lower than that observed in wild-type islets treated with 250 M diazoxide and 30 mM KCl in the absence of AOA at each concentration of glucose (Fig. 6, A-C, closed circles).
To evaluate glucose metabolism leading to ATP generation in islets, we measured the ATP content of islets after glucose stimulation for 1 h in the presence of 250 M diazoxide and 30 mM KCl. At 2.8 mM glucose, the ATP content of mGPDHdeficient islets treated with 5 mM AOA was 10.5 Ϯ 0.3 pmol/ islet, which was not different from that of wild-type islets (10.1 Ϯ 0.7 pmol/islet). However, the ATP contents of mGPDHdeficient islets treated with AOA were significantly less than those of wild-type islets at both 10.0 mM glucose (11.2 Ϯ 0.2 versus 13.1 Ϯ 0.6 pmol/islet, respectively, p Ͻ 0.01) and 16.7 mM glucose (12.0 Ϯ 0.4 versus 15.4 Ϯ 0.05 pmol/islet, respectively, p Ͻ 0.01) (Fig. 7). These results indicated that efficient ATP generation by glucose metabolism through the NADH shuttles was important in the steps distal to elevation of [Ca 2ϩ ] c in glucose-induced insulin secretion. The ATP contents in wild-type islets incubated with diazoxide and KCl were comparable to those in wild-type islets incubated without diazoxide or KCl at 2.8, 10.0, and 16.7 mM glucose (data not shown). DISCUSSION This study is the first to provide direct evidence that the NADH shuttle system plays an essential role in regulating both the K ATP channel-dependent pathway and steps distal to Ca 2ϩ influx through the plasma membrane in glucose-induced insulin secretion. The K ATP channels are a weak inward rectifier that are expressed in a wide variety of tissues including pancreatic ␤ cells, cardiac myocytes, skeletal muscle, and brain, and serve to couple cellular metabolic state to electrical excitability (44). In ␤ cells, the K ATP channels are composed of the inward rectifier potassium channel (Kir) 6.2 (45) and the sulfonylurea receptor (SUR) 1, a member of the superfamily of ATP binding cassette proteins (46). Kir6.2 acts as the ATPsensitive pore-forming subunit of the K ATP channel complex (47,48), whereas SUR1 has been identified as the regulatory subunit, which confers sensitivity to sulfonylureas, channel openers, and Mg 2ϩ nucleotides (41,49,50). The single K ATP channel kinetics (Fig. 4) and its sensitivity to ATP (Fig. 5) on the excised membrane of mGPDH-deficient ␤ cells were preserved compared with wild-type ␤ cells, clearly discounting the possibility that the channel activity itself was impaired secondarily to the absence of mGPDH. When mGPDH-deficient ␤ cells were treated with AOA to stop the function of the NADH shuttle system, glucose-stimulated generation of ATP in mitochondria was severely decreased (20). The present study showed that under these conditions, glucose-stimulated closure of the K ATP channel was severely impaired and subsequent membrane depolarization and generation of action potentials were scarcely observed. These results demonstrated that activation of mitochondrial glucose metabolism through the NADH shuttle system and the resultant increase in cellular ATP content were pivotal to closure of the K ATP channel and initiation of the downstream events leading to insulin secretion.
The mGPDH-deficient ␤ cells treated with AOA completely preserved the response to sulfonylurea with respect to the K ATP channel current, the plasma membrane potential, and insulin secretion ( Fig. 3B and 2D) (20). This was in contrast to the observation that the effects of sulfonylurea on Ca 2ϩ influx and insulin secretion were impaired in insulinoma-derived ␤ cells that were treated for days with ethidium bromide to eradicate mitochondrial DNA, thereby abolishing oxidative phosphorylation (8). This difference is presumably dependent upon whether glucose-stimulated ATP generation was maintained at ϳ25% of the normal state (20) or was almost completely nullified (8). Although the mechanism of decreased sensitivity of the K ATP channel to the drug in the ATP-depleted state is unclear, it may involve decreased binding of sulfonylurea to the channel, alteration of the phosphorylation state of the channel, and defective function of accessory constituent(s) of the channel (8,51,52).
There was an evident dissociation between the elevation of [Ca 2ϩ ] c (Fig. 1C, right panel) and the inhibition of insulin secretion (Fig. 1A, closed squares) when the NADH shuttle system was blocked in islets. First, it should be considered that steps distal to an elevation of [Ca 2ϩ ] c such as mobilization of secretory granules toward the plasma membrane, energization of the granules, and a final fusion process have been shown to be ATP-requiring (30). Consistent with this notion, when inhibitors of oxidative phosphorylation were applied to ␤ cells, a similar dissociation between [Ca 2ϩ ] c increase and insulin secretion was observed (33). A marked dependence of insulin secretion on energy metabolism has also been postulated as the reason why insulin secretion was inhibited by cooling, although Ca 2ϩ influx was only slightly affected by this treatment (53). Moreover, the omission of ATP from the cytosol resulted in 89% inhibition of exocytosis elicited by Ca 2ϩ stimulation in permeabilized insulin-secreting cells (54). Recently, it has also been shown that a final step of Ca 2ϩ -stimulated exocytosis is dependent on post-priming actions of cytosolic ATP (55). Second, it should also be noted that the time course of the [Ca 2ϩ ] c rise to the second phase plateau in islets was considerably disordered (Fig. 1C, right panel). Disappearance of the initial [Ca 2ϩ ] c drop suggests that Ca 2ϩ -ATPase of the endoplasmic reticulum may not be able to sequester cytosolic Ca 2ϩ due to insufficient supply of ATP (26,56). Disappearance of the first phase peak of [Ca 2ϩ ] c and delay of its elevation to the second phase level indicated that the closure of K ATP channel and generation of action potentials were severely disturbed, which were confirmed in single ␤ cells (Fig. 3B and 2C). Furthermore, we postulate that the elevation of [Ca 2ϩ ] c may also be ascribed, at least in part, to diminished Ca 2ϩ uptake from cytosol by mitochondria (20) and to diminished Ca 2ϩ extrusion from cytosol by plasma membrane Ca 2ϩ -ATPase. Thus, the [Ca 2ϩ ] c rise may not take place at the critical time and site for the exocytotic processes. Collectively, the reduction of ATP supply accompanied by the nonphysiological [Ca 2ϩ ] c rise appears to be unable to induce insulin secretion.
An apparent difference was also observed in Ca 2ϩ behavior in response to glucose between an islet and a single ␤ cell. Glucose-induced elevation of [Ca 2ϩ ] c to the second phase level did occur in all mGPDH-deficient islets treated with AOA, although it was considerably delayed. In contrast, glucoseinduced generation of the action potentials, which reflects Ca 2ϩ spikes, was observed in only a small portion of mGPDH-deficient ␤ cells treated with AOA in the same time range. The ␤ cells constituting an islet are heterogeneous with respect to glucose metabolism and glucose-induced insulin secretion (57,58). Although it is not certain how Ca 2ϩ behavior in individual ␤ cells is manifested in [Ca 2ϩ ] c change of an islet, it is possible that a small portion of ␤ cells that preserved electrical excitability in response to glucose may transmit the signal to adjacent cells. Thus, the glucose-induced change in [Ca 2ϩ ] c that we actually observed in an islet might be an integration of these localized electrical activities that took place at various time points after glucose stimulation.
Diazoxide, which directly and selectively opens the K ATP channel without interfering with ␤ cell metabolism (59, 60), did not affect glucose-stimulated ATP generation of islets in our study (Fig. 7). Because opening of the K ATP channel by diazoxide holds the plasma membrane potential at a resting level, VDCCs are not activated in response to glucose. However, application of 30 mM KCl in the presence of diazoxide depolarizes the membrane, which is followed by Ca 2ϩ influx through VDCCs and stimulation of insulin secretion. Under these conditions, glucose is still able to enhance insulin secretion dosedependently without significant alteration in [Ca 2ϩ ] c , which has been described as the K ATP channel-independent Ca 2ϩ -dependent pathway of glucose-induced insulin secretion (31,32). When glucose-induced insulin secretion was abolished by the inhibition of the NADH shuttle system, recovery of insulin secretion through the K ATP channel-independent Ca 2ϩ -dependent pathway was observed, and the extent of recovery was dependent on glucose concentration. However, the secretion was less than that observed in the presence of the functional NADH shuttle system (Fig. 6). Under these conditions, the increase in cellular ATP content of islets in response to glucose was suppressed when the shuttle system was blocked (Fig. 7). In a comprehensive study to identify regulators of insulin secretion through the K ATP channel-independent Ca 2ϩ -dependent pathway, adenine nucleotides emerged as the best candidate (61). Taken together, ATP generation by efficient glucose metabolism through the NADH shuttle system probably func- In conclusion, we have clarified the two mechanisms by which the shutdown of the NADH shuttle function abolished glucose-induced insulin secretion. The insufficient closure of the K ATP channel due to the depletion of cellular ATP was followed by impairment of the depolarization of the plasma membrane and Ca 2ϩ influx into ␤ cells. In addition, steps distal to Ca 2ϩ influx into the cytosol were also affected. These results demonstrate that mitochondrial metabolism through the NADH shuttle system plays a central role as a common regulator of both the K ATP channel-dependent pathway leading to [Ca 2ϩ ] c elevation and the pathway in which the elevated [Ca 2ϩ ] c promotes exocytosis of insulin from secretory vesicles. Recently, it has been reported that various steps in glucoseinduced insulin secretion are impaired in type 2 diabetes mellitus caused by mutations in the hepatocyte nuclear factor-1␣ gene (62,63). Defects in generation or transport of glycolysisderived NADH into mitochondria are implicated in the impaired insulin secretion (62). Genetic or acquired defects in the NADH shuttle system may therefore be candidates for the causes of type 2 diabetes mellitus.