Control Mechanisms of the Oscillations of Insulin Secretion In Vitro and In Vivo

The plasma insulin concentration oscillates even during postabsorptive periods (3–6). Two major frequencies characterize these oscillations. Ultradian oscillations (period of ∼120 min) probably result from a feedback loop between glucose production and insulin secretion and will not be further discussed here (7,8). There also exist faster oscillations that were initially reported to have a period of ∼15 min in peripheral blood (3–5,9), and turned out later to have a period between 5 and 10 min in portal blood (10,11) or in peripheral blood using deconvolution algorithms (6,12). The finding that the C-peptide concentration oscillates nearly in phase with insulin indicates that cyclic secretion rather than cyclic clearance produces the oscillations of insulinemia (13–15). Pulsatile insulin secretion has been reported to account for ∼70% of total insulin secretion in normal human subjects (6,10), and the major effect of glucose is to increase the amplitude of the oscillations (11,12,16,17).

The first convincing evidence for pulsatile insulin secretion in vitro was obtained with the isolated dog pancreas perfused with a constant concentration of glucose (18,19). The oscillations had a period of 5–10 min. Similar findings were made with the isolated pancreases of other species including monkeys (20) and humans (21). Oscillations of insulin secretion from groups of 12–200 islets have also been reported; their period was 5–10 min (22–24) or 15–30 min (25–28). Monitoring insulin secretion from single islets requires a sensitive radioimmunoassay (29) or enzyme-linked immunosorbent assay (30), or uses an indirect approach, based on the amperometric detection of 5-hydroxytryptamine (5-HT) that has accumulated in insulin granules during preincubation of the islet with the amine (31). Depending on the sampling frequency, variable periods were reported for pulsatile insulin secretion from the islet. Thus, oscillations with a period of 2–5 min were detected with a sampling interval of 10–20 s (29,30,32,33). These oscillations could be resolved into faster oscillations (period of 12–30 s) when the sampling interval was decreased to 3 s (32,34) or by using the amperometric real-time measurement of 5-HT release (35). Glucose increased the amplitude of the oscillations of the perfused pancreas (18) and of single islets (30,33,36) or groups of islets (24). At the single-cell level, several methods can monitor exocytosis of insulin granules with an excellent temporal resolution. They include the amperometric detection of 5-HT (37,38) or insulin itself (39), β-cell capacitance recording (38), and fluorimetric measurements of either quinacrine accumulated in secretory granules (40) or Zn²⁺ coreleased with insulin (41). Except in one study (37), no spontaneous oscillations of secretion were observed during glucose stimulation of single β-cells, probably because these techniques only detect large changes, cannot be used for long recordings, or put major constraints on the experimental protocols.

Glucose stimulates insulin secretion by activating two pathways that require metabolism of the sugar by β-cells (42). The triggering pathway, also referred to as the K_ATP channel-dependent pathway, depends on a rise in the ATP/ADP ratio that closes ATP-sensitive K⁺ channels (K_ATP channels) in the plasma membrane. The resulting decrease in K⁺ permeability leads to membrane depolarization, opening of voltage-dependent Ca²⁺ channels, Ca²⁺ influx, and eventual rise of the cytosolic Ca²⁺ concentration ([Ca²⁺]_c) that triggers exocytosis. The amplifying pathway, also referred to as the K_ATP channel-independent pathway, depends on an already elevated [Ca²⁺]_c but does not involve a further change in [Ca²⁺]_c. It increases the efficiency of Ca²⁺ on exocytosis. The pulsatility of insulin secretion might result from oscillations in either of these transduction pathways. Because metabolism and [Ca²⁺]_c play key roles in the control of insulin secretion and have been reported to oscillate, many efforts have been spent to investigate which of these two mechanisms is the primum movens of pulsatile insulin secretion.

Isolated pancreatic islets have widely been used to study the effects of glucose on β-cell [Ca²⁺]_c. Three types of [Ca²⁺]_c oscillations can be observed during stimulation by the sugar (Fig. 1E–G): regular and rapid (frequency of 2–3/min), regular and slow (frequency of ∼0.2/min), or mixed, characterized by rapid oscillations superimposed on slow ones (35,43,44). Oscillations of [Ca²⁺]_c are also detected in islets in vivo (45). Image analysis has demonstrated that the changes in [Ca²⁺]_c are synchronized between all β-cells within an islet (46,47). [Ca²⁺]_c also oscillates in glucose-stimulated single β-cells (2,48), but these oscillations are much less regular than those in islets and often occur at a low frequency only (Fig. 1A) (49). Mixed oscillations also occur in isolated cells (49) (Fig. 1B), suggesting that this peculiar pattern does not necessarily result from the sum of different signals produced in distinct β-cell populations of the islet (50). [Ca²⁺]_c oscillations are more regular in clusters of islet cells than in single β-cells, and this regularity (not the frequency) increases with the number of coupled cells (Fig. 1C and D) (49).

The essential role of Ca²⁺ influx in the generation of [Ca²⁺]_c oscillations by glucose, in either whole islets or single β-cells, is demonstrated by their abrogation upon omission of extracellular Ca²⁺ (2,47) or blockade of voltage-dependent Ca²⁺ channels (51). [Ca²⁺]_c oscillations are linked to oscillations of the membrane potential in β-cells (46,47), and it is assumed that mixed [Ca²⁺]_c oscillations result from an irregular (so-called “periodic”) electrical activity (35,52,53). Synchronization of the β-cell electrical activity (54) by gap junctions is likely to underlie the synchronization of [Ca²⁺]_c oscillations between β-cells within the islet (47,49,55). Synchronization by extracellular signals has been proposed (56,57), but our experiments do not support this hypothesis (58).

The endoplasmic reticulum also contributes to glucose-induced [Ca²⁺]_c oscillations. We have recently shown that it dampens the amplitude of the oscillations by rapidly taking up Ca²⁺ during the upstroke phase and releasing Ca²⁺ during the descending phase (59). Our model contrasts with a more widely received opinion that the endoplasmic reticulum amplifies glucose-induced [Ca²⁺]_c oscillations by releasing Ca²⁺ during the periods of depolarization in β-cells (50,60–63). Such a release could be triggered by depolarization (depolarization-induced Ca²⁺ release, mediated by type 1 ryanodine receptors), by inositol 1,4,5-trisphosphate (IP₃-induced Ca²⁺ release), or by Ca²⁺ itself (Ca²⁺-induced Ca²⁺ release, mediated by type 2 or type 3 ryanodine receptors or IP₃ receptors). Because amplification of the Ca²⁺ signal through some of these mechanisms usually requires the presence of caffeine or cAMP-producing agents (50,62,63), it is unclear whether it is operative during stimulation by glucose alone.

Because a rise in β-cell [Ca²⁺]_c is required for glucose to stimulate insulin secretion (64), and because glucose-induced [Ca²⁺]_c oscillations occur synchronously within all β-cells of an islet, it has been proposed that oscillations of insulin secretion are driven by [Ca²⁺]_c oscillations. This proposal was directly supported by simultaneous measurements of insulin secretion and [Ca²⁺]_c in the same islet. They showed that each [Ca²⁺]_c oscillation induced by glucose is accompanied by a synchronous oscillation of insulin secretion even when the frequency of the oscillations is modified (Fig. 2) (29,31,35,36,65). A similar temporal correlation was observed at the single-cell level (37). It also holds when the mitochondrial substrate, α-ketoisocaproate, or the hypoglycemic sulfonylurea tolbutamide are used as stimuli (29,66).

When [Ca²⁺]_c was forced to oscillate by imposed, repetitive depolarizations of β-cells with 30 mmol/l K⁺ (in the presence of diazoxide and 15 mmol/l glucose), insulin secretion also oscillated. In contrast, when [Ca²⁺]_c was stably elevated by 30 mmol/l K⁺, no oscillations of insulin secretion were detected (Fig. 3) (67). Although these data suggest that Ca²⁺ is the oscillophore and the moment-to-moment driver of insulin secretion, the hypothesis was challenged by reports of dissociations between [Ca²⁺]_c changes and insulin secretion. In ob/ob mouse islets, insulin secretion was found to oscillate at basal and stable [Ca²⁺]_c when glucose was omitted from the medium, in the absence or presence of clonidine or diazoxide (68). Oscillations of insulin secretion were also observed during sustained and stable elevation of [Ca²⁺]_c with high tolbutamide, K⁺, or glucose concentrations in ob/ob and rat islets (24,33,69). Because many of the experiments reporting dissociations between [Ca²⁺]_c and insulin secretion were performed with islets from ob/ob mice, we recently investigated the possibility that the islets from hypoleptinemic, hyperglycemic, and hyperinsulinemic ob/ob mice differ from those of normal mice (70). We indeed found that ob/ob mouse islets are more prone than islets from young normal mice to display fluctuations of insulin secretion at stably elevated [Ca²⁺]_c. However, even in these islets, the fraction of total insulin secretion that occurred in a pulsatile manner did not exceed 10–13%, which is far less than the proportion of pulsatile secretion when [Ca²⁺]_c oscillates in β-cells (70).

Oscillations of different metabolic variables, such as O₂ consumption, ATP/ADP ratio, NAD(P)H, dihydroxyacetone-phosphate, glucose 6-phosphate, and lactate release, have been observed in intact or permeabilized normal or clonal insulin-secreting cells (22,27,71–77). These oscillations have frequencies compatible with those of insulin secretion. The identification of these metabolic oscillations and the observation of pulsatile insulin secretion independent from [Ca²⁺]_c changes (24,33,68,69) prompted the suggestion of metabolism-driven oscillations of insulin secretion (1,78).

Oscillations of a metabolic signal in β-cells may result from intrinsic properties of various pathways (for example glycolysis, see [78]) or be secondary to oscillations of [Ca²⁺]_c (79–81). In the first case only can the metabolic oscillations be regarded as the true regulators of the oscillations of secretion. This metabolic regulation can be achieved in two ways. Through an action on K_ATP channels, intrinsic metabolic oscillations may induce oscillations of the membrane potential and Ca²⁺ influx, but [Ca²⁺]_c oscillations remain the ultimate effector of the insulin pulsatility. Alternatively, intrinsic metabolic oscillations may modulate the action of Ca²⁺ on insulin secretion through the amplifying pathway. The proof of this second possibility requires the demonstration that oscillations of a metabolic signal do occur and drive large oscillations of insulin secretion in the absence of [Ca²⁺]_c changes. However, only few experiments performed with intact cells support the existence of metabolic oscillations at stable [Ca²⁺]_c. Thus, oscillations of the K_ATP channel activity were observed in intact single β-cells perifused with 3 mmol/l glucose, when [Ca²⁺]_c was presumably low and stable. Oscillations of O₂ consumption nearly in phase with oscillations of insulin secretion were detected in the presence of 3 mmol/l glucose or of a combination of 11 mmol/l glucose and a high concentration of tolbutamide (1 mmol/l) (77). Strangely, increasing the glucose concentration in the medium did not change the amplitude of the oscillations of O₂ consumption but markedly increased that of the oscillations of insulin secretion. In single clonal β-cells (HIT), oscillations of O₂ consumption were detected in the presence, not in the absence, of glucose. They had a period of ∼3 min, similar to that of insulin secretion oscillations, and persisted at the same frequency in a Ca²⁺-free medium (76). It is important to keep in mind that clonal cells might behave differently from normal β-cells, because their glycolysis is preferentially anaerobic whereas glycolysis in normal β-cells is preferentially aerobic (82). Different results were obtained by simultaneous measurements of [Ca²⁺]_c and metabolic parameters in single intact mouse islets. Oscillations of glucose and O₂ consumption were detected only in the presence of stimulating concentrations of glucose (74). These oscillations were synchronous to those of [Ca²⁺]_c and abolished by omission of Ca²⁺ or damped by addition of nifedipine (74,75). They thus seem to correspond to Ca²⁺-induced metabolic oscillations.

Other observations also cast doubts on the existence of metabolic oscillations independent from [Ca²⁺]_c changes. We did not detect oscillations of the K_ATP channel activity in single metabolically intact β-cells under conditions of low and stable [Ca²⁺]_c (83). No oscillations of ATP concentration were detected in the mitochondrial and cytosolic compartments of glucose-stimulated rat islets transfected with adenoviruses and expressing targeted luciferase (84). With the exception of one group that reported oscillations of NAD(P)H fluorescence in single islet cells (40), most authors do not detect NAD(P)H oscillations in islets that display [Ca²⁺]_c oscillations under the same conditions (47,85).

If Ca²⁺-independent metabolic oscillations really exist in β-cells and play a role in the control of insulin secretion, models other than glycolytic oscillations will be required to explain the observation that the mitochondrial substrate α-ketoisocaproate can also induce oscillations of insulin secretion (66).

To establish whether metabolic or [Ca²⁺]_c oscillations are the main driver of the oscillations of insulin secretion, each parameter was forced to oscillate while the other was stabilized as much as possible. In the presence of diazoxide, to prevent the effect of glucose on the membrane potential, alternating between 10 and 20 mmol/l glucose in the continuous presence of 30 mmol/l K⁺ was without effect on [Ca²⁺]_c but induced oscillations of metabolism [NAD(P)H fluorescence]. This resulted in small oscillations of insulin secretion (Fig. 4A). Conversely, when the islets were alternatively perifused with 4.8 and 30 mmol/l K⁺ in the continuous presence of 10 mmol/l glucose and diazoxide, large [Ca²⁺]_c oscillations occurred, accompanied by only minor oscillations of NAD(P)H fluorescence. However, the resulting oscillations of insulin secretion had an amplitude ∼100 times larger than that of the oscillations of secretion induced by imposed metabolic oscillations (Fig. 4B). In a model of permeabilized cells, oscillations of the Ca²⁺ concentration in the medium (hence within the cells) triggered oscillations of insulin secretion in the presence of a stable concentration of ATP and without possible glucose metabolism (86). Altogether, these data demonstrate that metabolic oscillations are poor effectors compared with [Ca²⁺]_c oscillations and strongly suggest that [Ca²⁺]_c oscillations are the direct drivers of the oscillations of insulin secretion at the islet level.

In vivo, insulinemia, glucagonemia, and glycemia oscillate at the same frequency. These oscillations are unaffected by agents modifying the function of the central nervous system (87). Initially, plasma insulin was reported to cycle nearly in and glucagon nearly out of phase with glucose (3,9,88,89), suggesting that pulses of insulin/glucagon secretion might induce pulses of glucose production by the liver; the resulting pulses of glycemia might in turn induce pulses of hormone release by the pancreas. Although very attractive, this hypothesis was challenged by subsequent studies that failed to find a consistent relationship between the periodic fluctuations in insulin, glucagon, and glucose (4,13,90). Moreover, suppression of glucose oscillations by a glucose clamp did not prevent the oscillations of insulinemia (91), and pulses of glucose or insulin failed to reset the cycles (17).

The observation that insulin and glucagon are secreted in a pulsatile manner by the perfused pancreas (18–21) has led to the proposal that a pacemaker system is present in the gland itself. An intrapancreatic network was suggested to pace and synchronize insulin secretion between all islets within the pancreas (92). Several observations are consistent with this proposal. Intrapancreatic neurons of the cat pancreas showed oscillations of their electrical activity with a period similar to that of the oscillations of insulin release (93). Pulsatile insulin secretion from the perfused dog (94) and rat (2) pancreas was altered by blockade of neurotransmission with tetrodotoxin or inhibition of postsynaptic nicotinic receptors (92). However, further identification of the nature of the neurotransmitter controlling the pulsatility of insulin secretion proved impossible because pharmacological blockade of endorphin, muscarinic, and α- or β-adrenergic receptors was without effect on the frequency of the oscillations in vivo (17,95) and in vitro (92). Several studies have reported oscillations of insulinemia after intrasplenic (96) or intrahepatic grafts (97,98) of isolated islets, or transplantation of a denervated pancreas (99,100). But the opinion that pulsatility of insulin secretion only reappears upon reinnervation of the grafted islets (98) is not shared by most groups (96,97,99,100). The role of intrapancreatic ganglia in synchronizing the signals between islets is also challenged by the observation of pulses of insulin secretion by pieces of pancreas containing several islets and electrically silent ganglia (101). Despite all these studies, it remains unclear whether intrapancreatic ganglia are really the pacemaker of the pulsatility of insulin secretion in vivo. The alternative proposal, that the in vivo pulsatility of secretion reflects an intrinsic property of each islet, also has weaknesses. Thus, the electrical activity that controls [Ca²⁺]_c has been reported not to be synchronized between islets in situ (102). Whether insulin secretion was pulsatile under these conditions is, however, unknown.

What is the reason for the oscillatory behavior of β-cells? [Ca²⁺]_c oscillations might be energetically less costly than a sustained [Ca²⁺]_c elevation and, by limiting the time during which [Ca²⁺]_c is elevated, might decrease the risk of Ca²⁺-cytotoxicity (103). This has never been documented for β-cells. [Ca²⁺]_c oscillations might be important for the activation of intracellular signals that control functions other than insulin secretion, such as gene expression (104). Changes in the frequency or duration of [Ca²⁺]_c oscillations might control insulin secretion more accurately than changes in the amplitude of a sustained [Ca²⁺]_c rise, and avoid inactivation or downregulation of transduction or effector pathways of stimulus-secretion coupling. However, our experiments have shown that, regardless of the oscillatory or steady pattern of the [Ca²⁺]_c elevation, insulin secretion at constant glucose is determined by the product mean [Ca²⁺]_c × time (67,105). No functional advantages of [Ca²⁺]_c oscillations have yet been identified in β-cells, but it should be emphasized that the long-term impact of a loss of oscillations has not been evaluated.

Oscillations of insulinemia might help in preventing insulin receptor downregulation (106) and development of insulin resistance (107). Indeed, insulin infused in a pulsatile fashion had a greater hypoglycemic effect than continuously infused insulin (108–111), being more potent in reducing hepatic glucose production and stimulating glucose utilization (112–115). This superior efficacy of pulsatile insulin is not unanimously accepted (116–118), however, perhaps because its disclosure critically depends on various in vivo factors, such as glucagonemia or insulinemia or the frequency of the pulsatility (5,109,119).

Some studies (16,17,120), but not all (12,121,122), have reported disorganized oscillations of insulinemia in patients with type 2 diabetes and their near relatives with mild glucose intolerance. It is unknown, however, whether these alterations are secondary to the prolonged hyperglycemia. Insulin resistance of obese patients also decreases the regularity of pulsatile insulin secretion (123). Because oscillations of insulinemia favor optimal glucose homeostasis, one may speculate that a disturbance of plasma insulin oscillations in type 2 diabetes contributes to the disease.

This work was supported by grant 3.4552.98 from the Fonds de la Recherche Scientifique Médicale (Brussels), grant ARC 00/05–260 from the General Direction of Scientific Research of the French Community of Belgium, and by the Interuniversity Poles of Attraction Program (P4/21), Belgian State, Federal Office for Scientific, Technical and Cultural Affairs. P.G and J.C.J. are Maître de Recherches and Chercheur Qualifié, respectively, from the Fonds National de la Recherche Scientifique, Brussels.