The Oscillatory Behavior of Pancreatic Islets from Mice with Mitochondrial Glycerol-3-phosphate Dehydrogenase Knockout*

Glucose stimulation of pancreatic β cells induces oscillations of the membrane potential, cytosolic Ca2+ ([Ca2+] i ), and insulin secretion. Each of these events depends on glucose metabolism. Both intrinsic oscillations of metabolism and repetitive activation of mitochondrial dehydrogenases by Ca2+ have been suggested to be decisive for this oscillatory behavior. Among these dehydrogenases, mitochondrial glycerol-3-phosphate dehydrogenase (mGPDH), the key enzyme of the glycerol phosphate NADH shuttle, is activated by cytosolic [Ca2+] i . In the present study, we compared different types of oscillations in β cells from wild-type and mGPDH−/− mice. In clusters of 5–30 islet cells and in intact islets, 15 mm glucose induced an initial drop of [Ca2+] i , followed by an increase in three phases: a marked initial rise, a partial decrease with rapid oscillations and eventually large and slow oscillations. These changes, in particular the frequency of the oscillations and the magnitude of the [Ca2+] rise, were similar in wild-type and mGPDH−/− mice. Glucose-induced electrical activity (oscillations of the membrane potential with bursts of action potentials) was not altered in mGPDH−/− β cells. In single islets from either type of mouse, insulin secretion strictly followed the changes in [Ca2+] i during imposed oscillations induced by pulses of high K+ or glucose and during the biphasic elevation induced by sustained stimulation with glucose. An imposed and controlled rise of [Ca2+] i in β cells similarly increased NAD(P)H fluorescence in control and mGDPH−/− islets. Inhibition of the malate-aspartate NADH shuttle with aminooxyacetate only had minor effects in control islets but abolished the electrical, [Ca2+] i and secretory responses in mGPDH−/− islets. The results show that the two distinct NADH shuttles play an important but at least partially redundant role in glucose-induced insulin secretion. The oscillatory behavior of β cells does not depend on the functioning of mGPDH and on metabolic oscillations that would be generated by cyclic activation of this enzyme by Ca2+.

Glucose stimulation of pancreatic ␤ cells induces oscillations of the membrane potential, cytosolic Ca 2؉ ([Ca 2؉ ] i ), and insulin secretion. Each of these events depends on glucose metabolism. Both intrinsic oscillations of metabolism and repetitive activation of mitochondrial dehydrogenases by Ca 2؉ have been suggested to be decisive for this oscillatory behavior. Among these dehydrogenases, mitochondrial glycerol-3-phosphate dehydrogenase (mGPDH), the key enzyme of the glycerol phosphate NADH shuttle, is activated by cytosolic [Ca 2؉ ] i . In the present study, we compared different types of oscillations in ␤ cells from wild-type and mG-PDH ؊/؊ mice. In clusters of 5-30 islet cells and in intact islets, 15 mM glucose induced an initial drop of [Ca 2؉ ] i , followed by an increase in three phases: a marked initial rise, a partial decrease with rapid oscillations and eventually large and slow oscillations. These changes, in particular the frequency of the oscillations and the magnitude of the [Ca 2؉ ] rise, were similar in wild-type and mGPDH ؊/؊ mice. Glucose-induced electrical activity (oscillations of the membrane potential with bursts of action potentials) was not altered in mGPDH ؊/؊ ␤ cells. In single islets from either type of mouse, insulin secretion strictly followed the changes in [Ca 2؉ ] i during imposed oscillations induced by pulses of high K ؉ or glucose and during the biphasic elevation induced by sustained stimulation with glucose. An imposed and controlled rise of [Ca 2؉ ] i in ␤ cells similarly increased NAD(P)H fluorescence in control and mGDPH ؊/؊ islets. Inhibition of the malate-aspartate NADH shuttle with aminooxyacetate only had minor effects in control islets but abolished the electrical, [Ca 2؉ ] i and secretory responses in mGPDH ؊/؊ islets. The results show that the two distinct NADH shuttles play an important but at least partially redundant role in glucose-induced insulin secretion. The oscillatory behavior of ␤ cells does not depend on the functioning of mGPDH and on metabolic oscillations that would be generated by cyclic activation of this enzyme by Ca 2؉ .
Stimulation of insulin secretion by glucose involves a rise in the cytoplasmic concentration of Ca 2ϩ ([Ca 2ϩ ] i ) 1 in ␤ cells (1,2). This rise essentially results from the following sequence of events: closure of ATP-sensitive K ϩ channels (K ϩ -ATP channels) in the plasma membrane, membrane depolarization, and influx of Ca 2ϩ through voltage-sensitive channels (3)(4)(5). A second, important effect of glucose is the amplification of the action of [Ca 2ϩ ] i on the exocytotic process (6 -8). Both pathways require glucose metabolism and appear to depend on a rise in the ATP/ADP ratio (9).
Glucose metabolism in ␤ cells essentially occurs through aerobic glycolysis (10 -12). An increase in the glucose concentration is followed by an acceleration of glycolysis and an even greater stimulation of mitochondrial oxidative events, in which Ca 2ϩ may play an important role. Thus, the elevation of [Ca 2ϩ ] i is paralleled by an increase in mitochondrial Ca 2ϩ (13) that may then activate the three Ca 2ϩ -sensitive intramitochondrial dehydrogenases and promote ATP synthesis (14). Another feature of the ␤ cell metabolic organization is a low activity of lactate dehydrogenase (12,15). Cytosolic NADH formed during glycolysis is re-oxidized (and ATP synthesis concomitantly stimulated) by transfer of the reducing equivalents into mitochondria through the glycerol phosphate shuttle and the malate-aspartate shuttle (16,17). The rate-limiting enzyme of the former shuttle, mitochondrial glycerol-3-phosphate dehydrogenase (mGPDH) is located on the outer face of the mitochondrial membrane and can be activated by increases in cytosolic [Ca 2ϩ ] i (18,19). The importance of the NADH shuttles for glucose-induced insulin secretion (20) has received strong support from a recent study using islets from mice with a targeted disruption of mGPDH (21).
Many ␤ cell responses to glucose are oscillatory. Oscillations of the membrane potential drive oscillations of [Ca 2ϩ ] i , leading to oscillations of insulin secretion that can be amplified by metabolic oscillations (4,(22)(23)(24)(25)(26). It is still unclear whether oscillations of glucose metabolism are intrinsic and initiate the whole chain of other pulsatile events or are entrained by the oscillations of [Ca 2ϩ ] i (27). The Ca 2ϩ sensitivity of mGPDH makes the enzyme a possible key site of the interplay between glucose metabolism and [Ca 2ϩ ] i oscillations. Thus, imposed oscillations of free Ca 2ϩ in islet mitochondrial extracts induced oscillations of mGPDH activity (28). If the hypothesis is correct, the oscillatory behavior of stimulus secretion coupling should be perturbed by mGPDH defects. The present study addressed this question with islets isolated from mGPDH knock-out (mG-PDH Ϫ/Ϫ ) mice (21). We compared the oscillations of ␤ cell membrane potential, [Ca 2ϩ ] i , and insulin secretion in islet cell clusters or single islets from wild-type (mGPDH ϩ/ϩ ) and mG-PDH Ϫ/Ϫ mice. The impact of an inhibition of the malate-aspartate shuttle by aminooxyacetate (AOA) (17,29) was also evaluated in the two groups.
Preparations Used to Study the Oscillatory Behavior of ␤ Cells-One wild-type and one mGPDH Ϫ/Ϫ mouse were usually killed on the same day. Their islets were isolated by collagenase digestion of the pancreas, followed by hand-picking (32). The medium used was a bicarbonatebuffered solution containing 120 mM NaCl, 4.8 mM KCl, 2.5 mM CaCl 2 , 1.2 mM MgCl 2 , 24 mM NaHCO 3 , 10 mM glucose, and 1 mg/ml bovine serum albumin. It was gassed with O 2 /CO 2 to maintain a pH of 7.4. The islets were then cultured for 1 or 2 days in RPMI 1640 medium containing 10 mM glucose, 10% heat-inactivated fetal bovine serum, 100 IU/ml penicillin, and 100 g/ml streptomycin. To obtain clusters of cells, some islets were incubated for 5 min in a Ca 2ϩ -free medium. After brief centrifugation, this solution was replaced by culture medium and the islets were disrupted by gentle pipetting through a siliconized glass pipette. The clusters were then cultured for 2 days on circular glass coverslips (33).

Measurements of Insulin Secretion, [Ca 2ϩ ] i and NAD(P)H-
The system has previously been described (23,32), and the control medium was the same as that used for islet isolation. When the concentration of KCl was raised to 30 mM, that of NaCl was reduced to 94.8 mM. Cultured islets were loaded with fura-PE3 during 2 h of incubation at 37°C in control medium containing 2 M fura-PE3 acetoxymethyl ester. After loading, one islet was transferred into a 110-l perifusion chamber with a bottom made of a glass coverslip and mounted on the stage of a microscope. The islet was held in place by gentle suction with a micropipette. The preparation was perifused at a flow rate of 1.8 ml/min and the medium was collected, in fractions of 30 s, just downstream of the islet. The temperature within the chamber was 37°C. The [Ca 2ϩ ] i was measured by dual wavelength (340 and 380 nM) excitation spectrofluorimetry, using a CCD camera to capture images (510 nM) at 2.4-s intervals. From the ratio of the fluorescence at 340 and 380 nm, the concentration of [Ca 2ϩ ] i was calculated by comparison with a calibration curve (4). Insulin was measured, in duplicates, in 400-l aliquots of the effluent fractions. The characteristics of the radioimmunoassay, using rat insulin as a standard, have previously been described in detail (32). The insulin content of the islets was determined after extraction in acid ethanol (34). It was similar in wild-type (112 Ϯ 9 ng/islet) and mGPDH Ϫ/Ϫ mice (110 Ϯ 7 ng/islet, n ϭ 22).
For the experiments in which only [Ca 2ϩ ] i was measured, 2-3 islets loaded with fura-PE3 were studied simultaneously in a larger chamber (ϳ1 ml) perifused at a flow rate of 1.8 ml/min. When [Ca 2ϩ ] i was measured in clusters of islets cells, a coverslip with attached cells was first incubated in the medium containing fura-PE3/AM before being transferred into the perifusion chamber of which it formed the bottom. At the end of the experiment, the perifusion was stopped and the chamber filled with control solution containing 1 M bisbenzimide. After 30 min, the preparation was excited at 365 nm and the number of cells in the studied cluster was determined by counting the fluorescent nuclei (at 510 nm) (33). The size of the studied clusters was similar for wild-type (14 Ϯ 1 cells) and mGPDH Ϫ/Ϫ (16 Ϯ 1 cells) mice.
The same experimental setup was used to measure the reduced forms of NAD and NADP, referred to as NAD(P)H. The islets were excited at 360 nm and the emitted fluorescence was filtered at 470 nm (4). The changes in fluorescence were expressed as a percentage of basal values within the same islet.
Measurement of ␤ Cell Membrane Potential-The membrane potential of a single cell within an islet was measured with a high resistance intracellular microelectrode (35). The only difference from the described method was that a single isolated islet cultured for 1 or 2 days was used instead of a piece of pancreas. ␤ Cells were identified by the typical electrical activity that they display in the presence of 10 mM glucose. The medium was the same as that used for islet isolation but did not contain albumin.
Presentation of Results-All experiments have been performed with 1-3 islets from four to six different wild-type and mGPDH Ϫ/Ϫ mice. The results are illustrated by representative traces and/or presented as means (Ϯ S.E.). The statistical significance between means was assessed by unpaired or paired Student's t test as appropriate, and that of differences between percentages by Fisher's exact test. Differences were considered significant at p Ͻ 0.05.

RESULTS
mGPDH Expression and Activity in Islets-Reverse transcriptase-PCR analysis showed that mGPDH mRNA was not expressed in mGPDH Ϫ/Ϫ islets (Fig. 1A, upper panel), whereas the glyceraldehyde-3-phosphate dehydrogenase gene transcript was amplified as effectively as that in wild-type islets ( Fig. 1A, lower panel). The absence of mGPDH in mGPDH Ϫ/Ϫ islets was confirmed by Western blot analysis with anti-mG-PDH antibody (Fig. 1B), and by an assay of the enzymatic activity in islets homogenates (Fig. 1C). No mGPDH mRNA and protein was detected in muscle and liver tissues of mG-PDH Ϫ/Ϫ mice. 2 The mGPDH mRNA was easily detectable in islets of heterozygous (mGPDH ϩ/Ϫ ) mice, but the experimental conditions do not permit reliable quantification. The mGPDH protein and activity were decreased by about 40% in these heterozygous islets (Fig. 1).
Glucose-induced [Ca 2ϩ ] i Changes in Clusters of Islet Cells-In the presence of a non-stimulatory concentration of glucose (3 mM), [Ca 2ϩ ] i was low and stable (Fig. 2, upper panels). Raising the glucose concentration to 15 mM first induced a small drop in [Ca 2ϩ ] i , that was rapidly followed by a marked increase in three phases: a long first phase, followed by a partial and progressive decrease with rapid oscillations, and eventually by large and slow oscillations. These changes were essentially similar in islet cell clusters from wild-type and mGPDH Ϫ/Ϫ mice.
The characteristics of [Ca 2ϩ ] i oscillations occurring during steady state glucose stimulation are illustrated by the lower panels of Fig. 2. These oscillations often displayed a mixed pattern of small and fast transients superimposed on slower but larger ones (Fig. 2, B and E). In other cases only the slow oscillations were detected (Fig. 2, C and F). In no preparation continuously stimulated with 15 mM glucose were fast oscillations observed in the absence of slow ones. No difference could be identified in the appearance of the [Ca 2ϩ ] i oscillations in clusters from wild-type and mGPDH Ϫ/Ϫ mice (Fig. 2) (Fig. 3). These oscillations were sometimes rapid (several per min) during the whole period of stimulation (not shown), became slower after a few minutes (Fig. 3, A and C), or were slow immediately after the first peak (not shown). Again these patterns were seen in both wild-type and mG-PDH Ϫ/Ϫ islets. The effect of AOA (used to inhibit the malateaspartate shuttle) was, however, very different. Six to 7 min after addition of 5 mM AOA to the medium, [Ca 2ϩ ] i stopped to oscillate and returned to close to basal values in all mGPDH Ϫ/Ϫ islets (Fig. 3, C and D). In wild-type islets, AOA only had a weak inhibitory effect, characterized by a decrease in the frequency of the oscillations. As shown in the inset of Fig. 3B, average [Ca 2ϩ ] i slowly but steadily increased with time in control islets continuously stimulated with 15 mM glucose alone. In contrast, [Ca 2ϩ ] i slightly decreased after addition of AOA and averaged 190 Ϯ 6 nM between 35 and 40 min, which was significantly different (p Ͻ 0.001) from the concentration measured in control islets without AOA (250 Ϯ 5 nM) and in mGPDH Ϫ/Ϫ islets treated with AOA (131 Ϯ 6 nM).
Glucose-induced [Ca 2ϩ ] i oscillations were characterized further in experiments during which the islets were continuously stimulated with 12 mM glucose (Fig. 4). Two types of patterns were observed in wild-type islets: large and slow oscillations usually superimposed with smaller and faster ones (mixed pattern) (85%) (Fig. 4B) and fast oscillations only (15%) (Fig.  4A). The same patterns were observed in mGPDH Ϫ/Ϫ islets 2 K. Eto and T. Kadowaki, unpublished data.  Fig. 4 also compares the effects of 1 mM AOA in both types of islets. A small and progressive decrease in the frequency of [Ca 2ϩ ] i oscillations was observed in wild-type islets (Fig. 4B). In mGPDH Ϫ/Ϫ islets with fast oscillations, mixed oscillations appeared and then stopped (Fig. 4C). In some cases, the inhibition by AOA was preceded by a transient phase of [Ca 2ϩ ] i increase (Fig. 4D). We have no explanation for this phenomenon that was also occasionally seen in wild-type islets (not shown). Fig. 4E shows that average [Ca 2ϩ ] i was similar in the two groups of islets during perifusion with 12 mM glucose alone. Upon addition of AOA, [Ca 2ϩ ] i decreased slightly in wild-type islets (p Ͻ 0.01 versus untreated controls) and considerably more in mGPDH Ϫ/Ϫ islets.
Glucose-and Aminooxyacetate-induced Changes in ␤ Cell Electrical Activity-In the presence of 3 mM glucose, the resting potential of ␤ cells was similar in wild-type (Ϫ64 Ϯ 2 mV) and mGPDH Ϫ/Ϫ islets (Ϫ66 Ϯ 2 mV). Upon stimulation with 12 mM glucose, the membrane depolarized to a plateau potential with continuous spike activity (Fig. 5, A and C). The membrane potential then started to oscillate with bursts of spikes on top of each oscillation. During steady-state stimulation of wild-type islets with 10 or 12 mM glucose, the oscillations of the membrane potential were either regular and rapid (Fig. 5A), or displayed a mixed pattern (Fig. 5B) (50% of each pattern in 10 mM glucose n ϭ 20) (Fig. 5A). Both patterns were also seen in mGPDH Ϫ/Ϫ islets (46% regular and 54% mixed in 10 mM glucose, n ϭ 13), and could sometimes be observed in the same cell (Fig. 5, C and D). The similarity of glucose-induced electrical activity in both types of islets is consistent with the unaltered properties of the K ϩ -ATP channels in the ␤ cell membrane of mGPDH Ϫ/Ϫ mice (30).
After about 5 min, AOA (5 mM) abolished the electrical activity induced by 12 mM glucose and repolarized the ␤ cell membrane in mGPDH Ϫ/Ϫ islets (Fig. 5D). Fig. 5E is a quantification of the electrical activity induced by glucose (percentage of time at plateau potential with spikes) and of its inhibition by AOA. It first shows that glucose-induced electrical activity was quantitatively similar in the two types of islets. It then shows that AOA only had a small inhibitory effect in wild-type islets. After 15 min of AOA application, electrical activity was still present during 45 Ϯ 5% of the time versus 67 Ϯ 9% in control islets not treated with AOA (p Ͻ 0.05). This contrasts with the abrogation of electrical activity and the repolarization of ␤ cells to a potential (Ϫ61 Ϯ 3 mV) close to the resting potential in mGPDH Ϫ/Ϫ islets.
Correlations between [Ca 2ϩ ] i and Insulin Secretion Changes-In a first series of experiments, single wild-type and mG-PDH Ϫ/Ϫ islets were perifused with a medium containing 15 mM glucose; diazoxide (100 M) was also added to prevent glucose from depolarizing the membrane (36) and raising [Ca 2ϩ ] i (4). Oscillations of [Ca 2ϩ ] i were then imposed by repetitive 2-min depolarizations with 30 mM K ϩ (Fig. 6). Each of these triggered a peak of insulin secretion. The inactivation of mGPDH did not impair these responses.
In a second series of experiments, single islets were stimulated by 3 pulses (2.5 min) and a more sustained application (25 min) of 20 mM glucose in a control medium, without or with AOA. The upper panels of Fig. 7 illustrate individual responses,  5. Effects of glucose (G) and AOA on the membrane potential of ␤ cells within intact islets from wild-type (mGPDH ؉/؉ ) and mGPDH ؊/؊ mice. ␤ Cells were impaled with the microelectrode during perifusion with a medium containing 10 mM glucose. When the electrical activity was stable, the preparation was perifused with a medium containing 3 mM glucose for 10 min. The glucose concentration was then raised to 12 mM and 5 mM AOA was added 15 min later. The experiments were performed with islets cultured for 1 or 2 days. Records A and B were obtained in different mice. Records C and D show a continuous experiment without interruption. Complete experiments of this type were performed with 6 mGPDH Ϫ/Ϫ islets and 12 mGPDH ϩ/ϩ islets. The mean electrical activity (calculated as % of time at plateau with spikes) is shown in panel E, where the dotted line corresponds to the mean electrical activity in 6 control islets stimulated with glucose alone until the end of the experiment. and the middle panels show mean responses. In wild-type islets, each pulse of high glucose induced a large peak of [Ca 2ϩ ] i accompanied by a peak of insulin secretion. The sustained stimulation caused an initial peak of [Ca 2ϩ ] i followed by rapid oscillations, and a biphasic secretion of insulin (Fig. 7A). After the large first phase, the secretion rate dropped to lower values, which is typical for the mouse (37,38). Fluctuations of secretion were usually found to follow the slow trends behind the fast [Ca 2ϩ ] i transients, but no regular oscillations occurred. With collections every 30 s, the time resolution of the system is insufficient to monitor fast oscillations of secretion.
Several aspects of the responses of wild-type islets were altered when the experiments were performed in the presence of 5 mM AOA throughout (compare the middle and lower panels of Fig. 7, A and B). The rise in [Ca 2ϩ ] i evoked by 20 mM glucose was clearly delayed, reaching a maximum only at the end of the 2.5-min glucose pulses, thus resulting in shorter [Ca 2ϩ ] i oscillations. The reason of the delay is the presence of a marked drop in [Ca 2ϩ ] i immediately upon glucose stimulation, which is best seen between 20 and 25 min (Fig. 7B, lower panel). During the long glucose application in the presence of AOA the elevation of [Ca 2ϩ ] i displayed large oscillations (Fig. 7B, upper panel). Again, insulin secretion tightly followed the changes in [Ca 2ϩ ] i with delayed and shorter peaks during glucose pulses, and clear oscillations synchronous with those of [Ca 2ϩ ] i during sustained stimulation. Total insulin secretion was about 25% lower in the presence than absence of AOA, but this difference must be interpreted with caution because the experiments were performed with single islets whose individual insulin content could not always be determined.
When islets from mGPDH Ϫ/Ϫ mice were stimulated with 20 mM glucose alone, the [Ca 2ϩ ] i and insulin secretion responses were essentially similar to those observed in wild-type islets (Fig. 7C). The situation was very different in the presence of 5 mM AOA (Fig. 7D). Raising the glucose concentration from 3 to 20 mM caused a drop in [Ca 2ϩ ] i , whereas the return to the low-glucose medium was followed by an increase in [Ca 2ϩ ] i . This is most easily seen in the lower panel of Fig. 7D. During sustained stimulation with glucose, [Ca 2ϩ ] i slightly increased after the initial fall, but remained much lower than in the absence of AOA. The upper panel of Fig. 7D illustrates the largest response in an mGPDH Ϫ/Ϫ islet treated with AOA. Under these conditions, insulin secretion was not stimulated at all.
Effects of Glucose and [Ca 2ϩ ] i on NAD(P)H Autofluorescence-For these experiments 100 M diazoxide was added to the perifusion medium to hold [Ca 2ϩ ] i at basal levels except in the presence of 30 mM KCl. Raising [Ca 2ϩ ] i by high K ϩ increased the NAD(P)H fluorescence at low and high glucose, but these stimulations were much smaller than that following the increase in glucose concentration from 3 to 15 mM (Fig. 8). The results were essentially similar in islets from wild-type and mGPDH Ϫ/Ϫ mice.

DISCUSSION
The mGPDH is the rate-limiting enzyme of the glycerol phosphate shuttle which, together with the malate-aspartate shuttle, permits reoxidation of cytosolic NADH by transferring reducing equivalents produced during glycolysis to the mitochondria.
mGPDH is particularly abundant in pancreatic islets (16) and, in contrast to lactate dehydrogenase, much more so in ␤ than non-␤ cells (12,15). This peculiar biochemical organization (high mGPDH/lactate dehydrogenase ratio) is thought to be important for optimal coupling of glucose metabolism by the ␤ cell and insulin secretion. Observations of a decreased activity of mGPDH in islets from type 2 diabetic patients and several animal models of the disease (11) have lent support to this concept, although recent studies have challenged the hypothesis (39,40).
Mice with a targeted disruption of mGPDH have recently been generated and found to be grossly normal. In particular, they were not diabetic and their islets responded to glucose stimulation by similar increases in the ATP/ADP ratio and insulin secretion to those observed in control islets (21). The present study extends these findings in showing that the glucose-induced electrical activity and [Ca 2ϩ ] i rise are quantitatively equivalent in wild-type and mGPDH Ϫ/Ϫ islets, and that the oscillatory characteristics of the electrical, ionic, and secretory events induced by glucose are similar in both types of islets. Since no mGPDH could be detected immunologically or enzymatically in mGPDH Ϫ/Ϫ islets it appears that the malateaspartate shuttle can compensate for the absence of glycerolphosphate shuttle in all these biological functions.
AOA, an inhibitor of various aminotransferases, is widely used to block the malate-aspartate shuttle in islets and other tissues (17,29,41,42). At the concentration of 5 mM, AOA inhibed glucose-induced insulin secretion by 50 -60% in normal rat islets (17,29). Smaller effects were observed in wild-type mouse islets. Glucose-induced electrical activity and [Ca 2ϩ ] i rise were only attenuated and insulin secretion was slightly impaired and delayed. If the effects of AOA solely result from an inhibition of the malate-aspartate shuttle, which is uncertain owing to the drug action on several aminotransferases, our results would suggest that the malate-aspartate shuttle might exert functions that cannot be compensated for by the glycerol phosphate shuttle. However, the minor effects of AOA in control islets strikingly contrast with the abrogation by the drug of the electrical, [Ca 2ϩ ] i and secretory responses to glucose in mGPDH Ϫ/Ϫ islets. It has been reported that the rise in the ATP/ADP ratio that glucose produces in ␤ cells (43) of normal islets (34) is markedly attenuated by AOA in mGPDH Ϫ/Ϫ islets (21). This reflects a major alteration of glucose metabolism (21) and may explain why the ␤ cell membrane was no longer depolarized and Ca 2ϩ influx, manifested by the electrical activity, was no longer stimulated by glucose. At variance with a previous report (21), [Ca 2ϩ ] i remained low when mGPDH Ϫ/Ϫ islets were challenged with glucose in the presence of AOA. There was thus no paradoxical dissociation between [Ca 2ϩ ] i and the suppression of insulin secretion. We, therefore, conclude that the NADH shuttles play an important role in glucose-induced insulin secretion and that the normal functioning of one of the two shuttles can largely compensate for an impairment of the other. This at least partial redundancy of the two NADH shuttles supports the importance of the system for the ␤ cell functioning.
The major goal of this study was to evaluate whether mG-PDH is involved in the oscillatory behavior of ␤ cells. A number of events regularly oscillate in ␤ cells during stimulation with a constant concentration of glucose. The origin of these oscillations is still incompletely understood (27), but a subtle interplay between metabolic and [Ca 2ϩ ] i changes may be involved. In this respect, the sensitivity of mGPDH to cytosolic [Ca 2ϩ ] i changes potentially confers a central position to the enzyme (28). The present results, however, do not support this hypothesis. Thus, no differences could be identified between glucoseinduced oscillations of ␤ cell membrane potential and [Ca 2ϩ ] i in wild-type and mGPDH Ϫ/Ϫ ␤ cells. The different phases of the changes (initial fall, large increase and oscillations) and the distinct patterns of the oscillations (regular fast or slow, and mixed fast and slow) were all observed in both types of islets. Their quantitative characteristics were also similar.
It has been proposed that a rise in ␤ cell cytosolic [Ca 2ϩ ] i stimulates mitochondrial metabolism at the levels of mGPDH and intramitochondrial dehydrogenases (13, 14, 18, 19, 44). This mechanism is viewed as a feed-forward process promoting FIG. 7. Effects of intermittent or sustained stimulations with glucose on cytoplasmic [Ca 2؉ ] i and insulin secretion measured simultaneously in single islets from wild-type (mGPDH ؉/؉ ) and mGPDH ؊/؊ mice. The concentration of glucose (G) was increased from 3 to 20 mM for 2.5 min every 5 or for 25 min (shaded periods). When indicated, 5 mM AOA was present throughout the experiments. All experiments were performed with islets cultured for 1 day. The upper panels show individual experiments, the middle panels show means (ϮS.E.) for four to five islets from different mice, and the lower panels show details of the temporal correlations between mean [Ca 2ϩ ] i and insulin secretion changes. ATP synthesis to sustain the secretory response. The present study, however, shows that repetitive stimulations with high glucose or high K ϩ induced pulses of insulin secretion from mGPDH Ϫ/Ϫ islets, which did not differ from the control ones. No decrease in the response was observed, nor was there any tendency to a fall during the second phase of insulin secretion induced by sustained glucose stimulation. These results do not support the idea that mGPDH is a critical site where changes in cytosolic [Ca 2ϩ ] i play a regulatory role in glucose metabolism and subsequent functional events. Similar doubts have been raised by studies of the effects of Ca 2ϩ on ATP production by islet mitochondria incubated with glycerol 3-phosphate (45). One should also keep in mind that the increase in NAD(P)H fluorescence, that follows glucose stimulation, is largely independent of an intracellular [Ca 2ϩ ] i rise (Fig. 7). In addition and even more importantly, the increase in NAD(P)H fluorescence brought about by Ca 2ϩ in low or high glucose was not affected by the inactivation of mGPDH. [Ca 2ϩ ] i stimulation of glucose metabolism through activation of intramitochondrial dehydrogenases thus appears to be independent of the functioning of the glycerol phosphate shuttle and its activation by [Ca 2ϩ ] i .
In conclusion, NADH shuttles play an important role in the regulation of insulin secretion by glucose but seem to be at least partially redundant. Despite its [Ca 2ϩ ] i dependence and ability to display an oscillatory function in vitro, mGPDH is not the generator of metabolic signals that might in turn induce oscillatory biophysical and secretory responses in ␤ cells.