Metabolic Oscillations in β-Cells

Theories describing the feedback and oscillatory mechanisms can be divided into two categories. In one model, the oscillation is driven by inherent oscillations in metabolism (49). According to this theory, glycolysis, which has been demonstrated to have oscillatory properties in several systems, serves as a pacemaker and inherently oscillates to give rise to oscillations in ATP/ADP, membrane potential, [Ca²⁺]_i, and insulin secretion. The second model that has been advanced is that Ca²⁺ entry into cells provides negative feedback on further Ca²⁺ entry by an action involving the K_ATP channel (50). This model was originally proposed based on the observation that Ca²⁺ regulation of the period of oscillations was mediated by K_ATP channel conductance (50). Several mechanisms of feedback can be envisioned including [Ca²⁺]_i-induced decreases in ATP production (51,52) or increases in ATP consumption (53), which would decrease ATP/ADP, reopen K_ATP channels, and begin a new cycle. In this article, we consider these theories in light of recent data generated by dynamic measurements of metabolic parameters. Our focus is first on summarizing data supporting the occurrence of metabolic oscillations, a requirement for many of the models. We then consider the hypothesis that metabolic oscillations occur by an interaction with Ca²⁺. We emphasize this model because of recent data and the mathematical modeling that supports it and because reviews of the inherent glycolytic oscillation model have already appeared (49).

Evidence that metabolism can oscillate in islets was first obtained by measurements of oxygen level from groups of perfused islets (54). These data demonstrated that oxygen levels oscillate with a period of ∼10 min, which is similar to that observed for Ca²⁺ and insulin secretion in the same system. Greater detail on the oxygen oscillations has been obtained by using sensors that can be implanted directly into single islets or positioned near single cells (55–59). For single islet measurements, the intimate contact of sensor and tissue allows better sensitivity and temporal resolution than measurements of bath levels. Sensors implanted into single islets have revealed that oxygen levels oscillate with patterns remarkably similar to the known [Ca²⁺]_i fluctuations. This is illustrated in Fig. 1, which shows that oxygen levels have both fast and slow oscillations and mixed oscillations in which the slow oscillations appear to be composed of burst of faster pulses. Measurements at single clonal β-cells have revealed both slow oscillations (2–4 min) (HIT cells) (58) and fast oscillations (INS-1 cells) (57) in oxygen consumption. Oscillations in oxygen consumption have been correlated with oscillations in insulin secretion, demonstrating the relevance of the metabolic oscillation to secretion (59).

Other metabolic parameters have also been observed to oscillate. NADH levels, measured by native fluorescence in single β-cells, have been shown to oscillate with a period of ∼50 s in a subpopulation of single rat β-cells in response to 8.3 mmol/l glucose (60), although this result has not always been reproduced. In addition, oscillations in whole-islet NADH levels have not been reported despite numerous measurements (61,62). Mitochondrial membrane potential (ΔΨ), measured by the fluorescence of rhodamine 123, has been shown to oscillate with a period of ∼50 s in groups of β-cells upon step changes from 0.5 to 15 mmol/l glucose (63). Finally, glucose consumption, taken as a measure of glycolytic rate because glucose is primarily used by glycolysis in β-cells, oscillates with both short and long periods in single mouse islets, as detected using a glucose microsensor (Fig. 2) (56).

An important consideration is the ATP level or the ATP/ADP ratio, since this is taken as the parameter that translates metabolic changes into electrical activity and secretion. Measurements of ATP by a bioluminescent method have not revealed consistent oscillatory levels (64,65) with glucose concentrations >16 mmol/l. On the other hand, the conductance of K_ATP channels has been shown to oscillate (48), which would suggest oscillations in ATP or ATP/ADP. Finally, ATP/ADP, measured by a direct biochemical assay of cells sampled from a suspension of dissociated ob/ob islets, was found to oscillate with a period of ∼5 min after stimulation with 20 mmol/l glucose and synchronization with clonidine and high (5 mmol/l) Ca²⁺; however, these oscillations were apparently not sustained over a long period (66), and the infrequent sampling time makes it difficult to correlate these oscillations with the [Ca²⁺]_i oscillations observed in intact islets. The discrepancy among different ATP and ATP/ADP measurements could be due to several factors. For example, the bioluminescent method has only been used with higher glucose concentrations, which tend to promote continuous elevation of [Ca²⁺]_i and membrane depolarization and not oscillations. It is also possible that ADP level is more important in changing activity of K_ATP channels, which would mean that a measure of ATP level may not reveal oscillations. Finally, factors other than ATP or ADP may regulate the K_ATP channel and, if they oscillate, could give rise to oscillations in K_ATP channel conductance.

To summarize, unequivocal evidence has been amassed to support the notion that metabolism oscillates in β-cells. The most robust measurements appear to be oxygen consumption; however, the limited ΔΨ and glucose consumption measurements are also strong. Direct measurements of ATP or ATP/ADP have not strongly demonstrated oscillations, although indirect measurements based on the K_ATP channel conductance do. Importantly, the recent measurements have demonstrated exquisite moment-to-moment modulation of metabolism to meet the needs and function of the cell.

The above discussion demonstrates that metabolism oscillates within islets and the oscillations are correlated with secretion. Although this evidence is important, it does not distinguish among the different theories that purport to describe oscillatory insulin secretion. Testing of these hypotheses has been aided by modeling and dynamic simultaneous measurements of multiple parameters in single islets and single cells. In this discussion, we consider the evidence that Ca²⁺ is a significant modulator of metabolic oscillations.

The concept that metabolic oscillations could be driven by an interaction with cytoplasmic Ca²⁺ oscillations has previously been dismissed based on the observation that such a scenario would contradict the basic model of stimulus-secretion coupling, which holds that increases in metabolism lead to Ca²⁺ influx and secretion (49). Indeed, when glucose is increased to stimulatory levels, ATP/ADP (53,67), NADH levels (68), and ΔΨ (69) all rise before increases in [Ca²⁺]_i. A limitation of using these observations in evaluating the role of Ca²⁺ in an oscillation is that they have not been performed during oscillatory conditions.

Simultaneous measurements of oxygen, glucose, or ΔΨ with [Ca²⁺]_i during glucose-stimulated oscillations have revealed that changes in metabolism do not always precede the rise in [Ca²⁺]_i. This is illustrated in Fig. 3, which shows that during a glucose challenge, oxygen consumption increases dramatically before the [Ca²⁺]_i increase in agreement with the accepted model of stimulus secretion coupling. However, after the Ca²⁺ increases, a new dynamic emerges. As shown in Fig. 3, once [Ca²⁺]_i rises, an acceleration of oxygen consumption occurs. Furthermore, once the oscillations begin, it appears that the [Ca²⁺]_i rises are in phase with oxygen oscillations with the [Ca²⁺]_i rise, possibly slightly preceding the increase in oxygen consumption (Fig. 3B). In contrast, with low Ca²⁺ present in the extracellular fluid, glucose gives rise to an increase in oxygen consumption but no apparent oscillatory behavior (Fig. 3C). Oscillations can be restored by addition of Ca²⁺ to the medium (56,57). Similarly, oxygen oscillations are greatly dampened or nonexistent in the presence of L-type Ca²⁺-channel blockers (56,57). Glucose oscillations also trailed [Ca²⁺]_i oscillations by a few seconds and had a similar Ca²⁺ dependency; however, Ca²⁺ appeared to evoke a decrease in glucose consumption (56).

Similar observations have been made in small groups of cells when simultaneously measuring mitochondrial membrane potential and [Ca²⁺]_i (63) (Fig. 4). In this case, the membrane was observed to hyperpolarize, indicative of an increase in the driving force for ATP production, before the rise in [Ca²⁺]_i; however, during subsequent oscillations, increases in [Ca²⁺]_i preceded decreases in membrane potential by about 4 s. Ca²⁺ entry into the cell was found to evoke similar decreases in the membrane potential. It was also demonstrated that Ca²⁺ entry alone could decrease ΔΨ. (In this study, the [Ca²⁺]_i rise evoked a decrease in ΔΨ, whereas in other studies, oxygen consumption was seen to increase with increasing [Ca²⁺]_i. This apparent paradox will be discussed below.)

In an effort to further explore the connection between [Ca²⁺]_i and oscillations in metabolism, we recently performed experiments in which [Ca²⁺]_i was clamped at high and low levels. These experiments were performed because it was apparent that prior work demonstrated strong Ca²⁺ dependency for oxygen consumption oscillations but did not prove that [Ca²⁺]_i oscillations are driving the oscillations, since such oscillations may require the presence of Ca²⁺. In these experiments, we first demonstrated that addition of diazoxide eliminated any oscillation in glucose or oxygen consumption (Fig. 5), which illustrates that the metabolic oscillations were mediated by changes in the activity of the K_ATP channel and not the other way around.

When a similar experiment is performed, in which 30 mmol/l K⁺ is pulsed to depolarize the membrane after diazoxide has been added to cease the oscillations, then a pulse of metabolic change is observed (Fig. 6). For oxygen, we observe that depolarization results in a strong increase in oxygen consumption that eventually lessens, even though the Ca²⁺ remains elevated. Upon returning to 5 mmol/l K⁺, the oxygen returns to the predepolarization level. For glucose, we observed that glucose level increases briefly during the depolarization pulse, but this is followed by a return to baseline or even a decrease to below baseline, as shown in the Fig. 6. When K⁺ is returned to basal level, a small sustained increase in glucose consumption is observed. Oscillations are insignificant for both glucose and oxygen if the 30 mmol/l K⁺ is kept on the islet for longer periods (data not shown). Thus, in the presence of 10 mmol/l glucose, spontaneous oscillations in [Ca²⁺]_i, oxygen consumption, and glucose consumption are observed. However, if [Ca²⁺]_i is kept low by treatment with diazoxide or if [Ca²⁺]_i is clamped high, with K⁺/diazoxide treatment, no oxygen or glucose oscillations are observed. In contrast, if Ca²⁺ is pulsed in by a K⁺ depolarization in the presence of 10 mmol/l glucose/diazoxide, then a pulse of metabolic changes will consist initially of a decrease in glucose uptake, which is tentatively attributed to an inhibition of glycolysis, and an increase in oxygen consumption.

These results demonstrate that the K_ATP channel–independent pathway does not contribute to metabolic oscillations. More important to the current discussion, however, these results demonstrate that metabolism, measured as oxygen or glucose consumption, exhibits little or no oscillation in single islets without [Ca²⁺]_i oscillations whether [Ca²⁺]_i is held low or high; however, metabolic changes will accompany pulses of [Ca²⁺]_i. These data lend evidence to the argument that [Ca²⁺]_i changes substantially modulate metabolic changes seen during oscillatory behavior.

The conclusion from these observations is that initially metabolism accelerates independently of Ca²⁺ influx, but once Ca²⁺ enters the cell, it begins to modulate metabolism by increasing oxygen consumption, decreasing ΔΨ, and altering glucose consumption. Several other observations support the idea that Ca²⁺ can affect energy metabolism in the β-cell. Increases in NADH in response to Ca²⁺ entry induced by K⁺ or glucose (60,61,68) have also been observed. Measurements of intracellular ATP by bioluminescence in MIN6 β-cells during glucose stimulation showed that ATP levels increased before [Ca²⁺]_i but then increased more after Ca²⁺ entry, suggesting that ATP production was augmented by increased [Ca²⁺]_i (64). Similar measurements in INS-1 cells showed that treatment with depolarizing concentrations of K⁺ had a net effect of decreasing cellular ATP by 10% (70).

Ca²⁺ could potentially have several competing effects on metabolism at the level of the mitochondria. Metabolism could be accelerated by activation of mitochondrial dehydrogenases. Activation of mammalian mitochondrial dehydrogenase by increases in cytoplasmic Ca²⁺ is a well-known phenomenon (71) that has been demonstrated in β-cells (72–74). This idea has been further supported by the observation (75–78) that increases in cytosolic Ca²⁺ are followed by increases in mitochondrial Ca²⁺ to the micromolar range, a concentration that is sufficient to act on dehydrogenases within the mitochondria. This activation would explain the Ca²⁺-induced oxygen consumption, ATP production, and NADH production discussed above. Ca²⁺ could also accelerate metabolism by activating the glycerol phosphate NADH shuttle to increase the flux of reducing equivalents from the cytosol to the mitochondria (79). However, a role for this process in oscillations seems unlikely given that knockout of glycerol-3-phosphate dehydrogenase, the key Ca²⁺-sensitive enzyme in the shuttle, had little effect on oscillatory [Ca²⁺]_i (80). It has also been proposed (51,52) and demonstrated (63) that Ca²⁺ entering the mitochondria could exert negative effects on the metabolic rate by collapsing the ΔΨ, which would decrease the driving force for ATP production.

These competing effects have been taken into account in a quantitative model, which includes, among several mitochondrial variables, the possibility that as Ca²⁺ enters the mitochondria, it initially increases dehydrogenase activity; this increases ATP production and oxygen consumption, but as it continues to rise in the mitochondria, it begins to have a suppressive effect by decreasing ΔΨ (52). A variety of experimental observations support the possibility of a biphasic effect. For example, the model produced predictions of oxygen consumption that are similar to those observed, as shown in Fig. 7. Although the model does not match the data exactly, several key observations are predicted, such as fast pulses of oxygen consumption with an increasing period during the slow wave. Some discrepancies, such as the time between slow waves and frequency of fast waves, may result from the slow recovery of oxygen into the islet by diffusion. Also, it is intriguing that whereas oxygen consumption appears to follow very closely the increases in [Ca²⁺]_i during an oscillation (Fig. 3B), the mitochondrial membrane depolarization lags by several seconds (Fig. 4). It is reasonable to speculate that this reflects the biphasic nature of the effect of Ca²⁺ on the mitochondria. In the K⁺ pulse experiments illustrated in Fig. 6, oxygen consumption increased but eventually returned to a lower level even though [Ca²⁺]_i remained elevated. The initial respiratory burst may reflect Ca²⁺ activation of dehydrogenases, whereas the subsequent inhibition may reflect the effects of Ca²⁺ on ΔΨ. Finally, in mitochondria from β-cells of ob/ob mice, ATP production is activated at low Ca²⁺ levels, but inhibited at higher Ca²⁺ levels (81, 82). The evidence supports the hypothesis that Ca²⁺ entry into the cell is rapidly reflected in the mitochondria, resulting initially in an activation of respiration followed closely by inhibition of ATP production by decreases in ΔΨ.

Most measurements of metabolic parameters in β-cells have focused on events that primarily reflect mitochondrial metabolism. This is reasonable because the vast majority of ATP production, oxygen consumption, and NADH production occurs in the mitochondria. However, because respiration may be substrate-limited under some conditions, it is important to consider glycolytic flux as well. Besides the observation of oscillations in glucose consumption mentioned above, little evidence exists for glycolytic oscillations in intact islet cells except for the observation that glucose-6-phosphate levels pulse under certain conditions (clonidine synchronization and 5 mmol/l Ca²⁺) in dispersed ob/ob islet cells (66).

Glucose consumption oscillations have a strong Ca²⁺ dependency and occur in parallel with Ca²⁺, such that increases in [Ca²⁺]_i correspond to decreases in glucose consumption during the oscillation (56). In addition, pulses of Ca²⁺ induced by K⁺ depolarization (Fig. 6) briefly suppress glucose consumption. Suppression of glycolysis by Ca²⁺ could provide negative feedback to further Ca²⁺ entry via a decrease in substrate supply to the mitochondria, which would decrease ATP/ADP; however, it is not clear if the decrease in glycolysis is sufficient to decrease the substrate supply. For example, whereas glycolysis is briefly inhibited, oxygen consumption increases, suggesting no limitation of substrate (Fig. 6). Thus, it remains to be determined if Ca²⁺ effects on glucose consumption are significant with regard to generating oscillations in secretion or if they are simply a response to Ca²⁺ oscillations.

While Ca²⁺ entry under some conditions appears to evoke a decrease in glucose consumption (see Fig. 6 and discussion above), the interactions of Ca²⁺ and glucose consumption are complex. Most obviously, the glucose consumption suppression is only temporary with depolarization. In addition, when nifedipine is added to oscillating islets, oscillations in glucose consumption and [Ca²⁺]_i cease, but the glucose level rises, indicating less glucose consumption as [Ca²⁺]_i falls (56). In addition, switches from Ca²⁺-free conditions to 2.4 mmol/l Ca²⁺ at 10 mmol/l glucose causes an increase in glucose consumption followed by initiation of oscillations (56). These discrepancies have not been investigated, but they suggest the possibility for a biphasic effect of Ca²⁺ on glucose consumption, where low intracellular Ca²⁺ is required for glucose consumption but higher concentrations are inhibitory. The effects of Ca²⁺ on glucose consumption could be direct by effects of Ca²⁺ on glycolytic enzymes or could be mediated by ATP. In the latter case, Ca²⁺ induces changes in ATP, which then feed back to affect the glycolytic rate (83).

Thus far, we have shown how Ca²⁺ could modulate ATP production at the level of glycolysis and mitochondrial metabolism. It is also possible that Ca²⁺ could exert negative feedback via the ATP/ADP by activating numerous ATP-consuming processes such as ion pumping and secretion (53). Thus, it has been demonstrated that Ca²⁺ influx induced by K⁺ or tolbutamide evokes decreases in ATP level (53,70). This hypothesis is not mutually exclusive with the modulation of energy production discussed above.

The data discussed so far provide evidence that cytosolic Ca²⁺ can and does exert strong effects on both ATP production and consumption in β-cells and islets. As mentioned above, it has also been suggested that inherent oscillations in glycolysis drive the metabolic oscillations by a mechanism involving phosphofructokinase (PFK). In a glycolytic oscillation, a burst of PFK activity is generated once the ATP/ADP ratio decreases to a certain level. The burst is terminated once the supply of high-energy phosphate donors is exhausted, allowing the ATP/ADP ratio to decay (83). Supporting the idea that glycolysis may oscillate is the presence of PFK-M, the PFK isoform with known oscillatory behavior, in islets (84). Interestingly, mutations in this enzyme are associated with impairment of oscillatory insulin release in humans (85). A key feature of this model is that high ATP/ADP ratios inhibit glycolysis.

According to this model, Ca²⁺ may modulate oscillations; however, oscillatory glycolysis will proceed without a necessary interaction with Ca²⁺. Evidence for an inherent oscillation, without a requirement for Ca²⁺ interaction, has been reported. Oscillations in K_ATP channel activity (86) and oxygen tension (59) have been detected at nonstimulatory glucose levels. Slow oxygen oscillations were detected in single HIT cells in 15 mmol/l glucose without added Ca²⁺, although these oscillations were amplified by extracellular Ca²⁺ (58). In our own laboratory, we observed irregular pulsatile oxygen (but not glucose) levels under Ca²⁺-free or low-glucose conditions; however, these fluctuations did not rise above the signal-to-noise ratio. Because of these contradictory observations, it would be premature to conclude that all oscillations, especially slower metabolic oscillations, are entirely Ca²⁺-dependent. However, we emphasize that metabolic oscillations are greatly amplified under conditions where [Ca²⁺]_i can oscillate (see Fig. 5 and comparison of Fig. 3A and 3C). Furthermore, it is apparent that rises in [Ca²⁺]_i can create rapid changes in metabolism like those observed during oscillations (Fig. 6). It is also possible, as mentioned above, that the inhibition in glycolysis observed with Ca²⁺ entry is mediated by an increase in ATP/ADP brought about by the increased oxygen consumption (and presumably ATP/ADP) associated with increases in [Ca²⁺]_i. In this case, the glycolytic oscillation mechanism would come into play but would require an interaction with Ca²⁺.

The discrepancies in metabolic oscillation detection mentioned above highlight the experimental difficulties that can hamper studies of oscillations. Artifactual oscillations can be created by temperature fluctuations or pulsatile flow created by pumps and plumbing used for perfusion. On the other hand, a system that is not sensitive enough may not detect weaker oscillations. Finally, oscillations require a delicate balance of electrical activity and biochemical messengers. Therefore, the type of β-cell tissue and its metabolic state could play an important role in the observation of oscillations under different conditions.

For a whole islet to exhibit an oscillation in [Ca²⁺]_i, metabolism, and secretion implies that the cells of the islet are synchronized. Imaging [Ca²⁺]_i within islets (41,61) and use of multiple oxygen sensors in one islet (56) have both indicated that oscillations occur in synchrony throughout the islet. The mechanism for synchronization is not known; however, electrical coupling through gap junctions and chemical coupling through diffusible factors have been suggested (87). Evidence for chemical coupling has mainly been that changes in flow rate appear to change the period of oscillations (36). Many compounds have been investigated as possible chemical couplers, including pyruvate, lactate, insulin, and ATP, without clear effects on oscillations, thus the mechanism of synchronization remains unknown.

Another observation that remains to be explained is oscillatory insulin secretion without oscillatory [Ca²⁺]_i. Thus, oscillations in insulin secretion have been observed in the absence of glucose (88), with [Ca²⁺]_i clamped at elevated levels by depolarizing K⁺ and diazoxide in the presence of glucose (36), and in the absence of oscillatory [Ca²⁺]_i (89). Such pulses apparently involve oscillations in pathways other than the K_ATP-dependent pathway emphasized here. The significance of these oscillations remains uncertain because they tend to be much weaker than those observed with oscillatory [Ca²⁺]_i. Furthermore, imposed metabolic oscillations give rise to small insulin secretion oscillations when compared with imposed [Ca²⁺]_i oscillations (90).

We have presented a review of the existence of metabolic oscillations in β-cells and their interaction with [Ca²⁺]_i changes. The picture that has emerged is summarized in Fig. 8: during a glucose challenge, metabolism accelerates independent of extracellular Ca²⁺, producing a rise in the ATP/ADP ratio, closure of the K_ATP channel, membrane depolarization, and entry of Ca²⁺ through L-type Ca²⁺ channels. Once entering the cell, cytosolic Ca²⁺ increases are mirrored in the mitochondria where further acceleration of metabolism (increased oxygen consumption) occurs followed by reduction of ATP production (reduced ΔΨ). The rise in cytosolic Ca²⁺ might also result in a temporary inhibition of glycolysis. While the mechanism for such an inhibition is unclear, it could involve a direct effect of Ca²⁺ on glycolytic enzymes or inhibition by the increased ATP evoked by Ca²⁺ entry. The rise in Ca²⁺ may also activate ATP-consuming processes to further decrease ATP/ADP. The net result of Ca²⁺ entry is negative feedback via ATP/ADP by a reduction of metabolism (less glucose consumption and lower ΔΨ) and an increase in ATP consumption. The relative importance of these effects remains to be determined. These combined effects decrease the ATP/ADP ratio, causing a reopening of the K_ATP channel, repolarization of the cell, and a reduction in [Ca²⁺]_i. With reduction in [Ca²⁺]_i, the process can begin again. This model is consistent with much of the obtained data on dynamic processes occurring in β-cells and is supported by a mathematical model that incorporates most of these features. The resulting changes in Ca²⁺ and ATP/ADP would seem to explain the glucose-stimulated insulin secretion oscillations seen in single islets under most conditions. This model diverges from the proposal that inherent glycolytic oscillations drive all processes in the β-cell; however, the possibility that changes in ATP/ADP underlie Ca²⁺-dependent changes of glycolytic rate provides an intriguing link between the models.