Signals and Pools Underlying Biphasic Insulin Secretion

Owing to the critical role of glucose metabolism for stimulus-secretion coupling, one could envisage that the biphasic pattern of insulin secretion is determined by an intrinsically biphasic change of β-cell metabolism elicited by the sudden rise in glucose concentration. Unfortunately, not many biochemical events lend themselves to monitoring with the adequate time resolution. Reduced pyridine nucleotides (NADH and NADPH) increase in glucose-stimulated islets. Biochemical measurements at selected time points in batches of rat islets have suggested that this increase is biphasic (20), and continuous recording of the autofluorescence of NAD(P)H in rat β-cells has shown that the response is often biphasic (21). In mouse islets cultured under conditions favoring their flattening, hence permitting studies in one cell layer, the NAD(P)H fluorescence increase produced by glucose was faster in cytoplasm than mitochondria (22); however, this did not result in a clearly biphasic change of the global signal. In intact single mouse islets stimulated by glucose, the increase has consistently been found to be monophasic (23–25). In perifused mouse islets stimulated by different glucose concentrations, oxygen consumption increased monophasically although simultaneously measured insulin secretion displayed its typical biphasic pattern (26). In single mouse islets, a biphasic increase in oxygen consumption has sometimes been detected, but this pattern is thought to be secondary to concomitant changes in [Ca²⁺]_i (27). Overall, there is no convincing evidence indicating that biphasic secretion can be attributed to a primary (i.e., Ca²⁺-independent) biphasic change of a metabolic signal.

Before [Ca²⁺]_i could be directly measured in β-cells, indirect methods were used to identify the mechanisms leading to the rise of Ca²⁺ that was assumed to trigger secretion. This led to contradictory proposals that first phase of glucose-induced insulin secretion was caused by Ca²⁺ influx from the extracellular medium (28) or by Ca²⁺ mobilization from intracellular pools (29). There is now general consensus that Ca²⁺ influx through voltage-dependent Ca²⁺ channels is necessary. In mouse islets, glucose-induced rise in [Ca²⁺]_i displays a biphasic time course: an initial sustained elevation is followed by oscillations or by a lower plateau, depending on the glucose concentration (8,24,30,31). This pattern is similar to that of the electrical activity induced by glucose in β-cells (17). The correspondence is not surprising because the electrical activity reflects Ca²⁺ influx through voltage-dependent Ca²⁺ channels.

The obvious question is thus whether the biphasic time course of glucose-induced insulin secretion could not simply be the consequence of a biphasic triggering signal. Figure 1 shows simultaneous changes in [Ca²⁺]_i and insulin secretion in the same mouse islet. Several stimulations with 20 mmol/l glucose for periods of 2.5 min every 5 min evoked similar [Ca²⁺]_i increases and simultaneous pulses of insulin secretion. When the period of stimulation by high glucose was extended to 25 min, [Ca²⁺]_i declined after the initial peak and rapid oscillations appeared, resulting in a lower mean [Ca²⁺]_i. Again, the initial [Ca²⁺]_i rise triggered a rapid and large secretory response, but insulin secretion subsequently decreased before stabilizing. This biphasic pattern with a lower, fairly stable second phase is typical for the mouse pancreas (32,33). The triggering signal [Ca²⁺] and the response (insulin secretion) thus show parallel time courses. However, the difference between the secretory rate at peak and plateau is clearly greater than the difference in [Ca²⁺]_i at the same times. This suggests that the correspondence between [Ca²⁺]_i and insulin is better temporally than quantitatively, in other words that other factors may be involved.

If the time-course of the glucose-induced [Ca²⁺]_i change in β-cells influences that of the secretory response, its alteration should perturb the two phases of insulin secretion. Stimulation of control islets with 15 mmol/l glucose caused a transient decrease in [Ca²⁺]_i followed by an increase whose biphasic pattern is here attenuated by the averaging of traces obtained in several islets (Fig. 2A). Mean [Ca²⁺]_i between 7 and 11 min was about 30 nmol/l higher than during the second phase. Test islets were pretreated with thapsigargin, a drug that blocks Ca²⁺ pumping into the endoplasmic reticulum. As shown previously (34), thapsigargin suppressed the small [Ca²⁺]_i decrease that normally precedes the first-phase rise (Fig. 2A). After thapsigargin pretreatment, glucose-induced [Ca²⁺] elevation was faster and larger than in control islets, and was monophasic. Insulin secretion, measured in separate islets, was increased by thapsigargin and no longer showed the characteristic biphasic pattern (Fig. 2B). The monophasic secretion of insulin induced by glucose thus followed the altered time course of the [Ca²⁺]_i change. It is possible, however, that the decline in secretion normally associated with second phase in the mouse was here prevented because [Ca²⁺]_i was increased to supranormal concentrations.

Glucose triggers biphasic insulin secretion only when its concentration is rapidly raised as a “square-wave.” A “slow-ramp” increase induces gradually larger secretion without a first phase (15,35). The experiments shown in Fig. 3 were performed to determine whether [Ca²⁺] is the signal that must change rapidly for glucose-stimulated β-cells to develop first-phase insulin secretion. In a first series (Fig. 3A and B), mouse islets were stimulated by a rapid rise of the glucose concentration from 3 to 15 mmol/l. In the control group, the concentration of extracellular Ca²⁺ was abruptly raised from 0.5 to 2.5 mmol/l concomitantly with the glucose stimulation. This resulted in a rapid increase in [Ca²⁺]_i, followed by a slight decrease (averaging attenuates the biphasic pattern), and in a biphasic secretion of insulin. In the test group, the increase in extracellular [Ca²⁺]_i was delayed and stepwise (0.5–1.0 to 1.75–2.5 mmol/l), which resulted in a retarded and slower, but eventually similar increase in [Ca²⁺]_i (Fig. 3A). However, insulin secretion induced by glucose was monophasic, without first phase (Fig. 3B).

A slightly different approach was followed in the experiments shown in Fig. 3C and D. The glucose concentration was maintained at 15 mmol/l throughout, but its effects on [Ca²⁺]_i and insulin secretion were prevented by diazoxide. The triggering signal of secretion was thus provided by 30 mmol/l K⁺. In the control group, the concentration of extracellular Ca²⁺ was increased from 0.5 to 2.5 mmol/l at the time of K⁺ stimulation. This resulted in a rapid increase in [Ca²⁺]_i, with an initial peak followed by a small decrease, then a progressive rise (Fig. 3C). Insulin secretion was rapidly stimulated following a peculiar biphasic time course, with a larger second than first phase (Fig. 3D). This pattern has previously been observed using a similar experimental paradigm (36). In the test group, the increase in extracellular Ca²⁺ was delayed and stepwise, which resulted in a slower increase in [Ca²⁺]_i (Fig. 3C). As a result, K⁺-induced insulin secretion was progressive, without the initial rapid phase. Altogether these data show that, whether the glucose concentration is rapidly increased or is constantly elevated, first-phase insulin secretion does not occur if [Ca²⁺]_i does not rapidly increase. The rapid change in glucose metabolism accompanying the glucose stimulation is not a sufficient signal.

In the absence of glucose and presence of diazoxide, continuous depolarization of mouse islets with 30 mmol/l K⁺ caused a sustained and steadily increasing elevation of [Ca²⁺]_i (Fig. 4A). It also induced a monophasic secretion of insulin, characterized by a large initial peak followed by a progressive and almost complete return to basal values (Fig. 4B). A similar pattern of insulin secretion was observed when permeabilized islet or HIT cells were stimulated by a constant elevation of free Ca²⁺ (37,38). When the stimulation of intact islets with high K⁺ was not continuous but intermittent (periods of 6 min every 12 min), similar increases in [Ca²⁺]_i were produced by each pulse of K⁺ (Fig. 4A), but the magnitude of the insulin response decreased with the repetition of the stimulation (Fig. 4B). Interestingly, the area under the two insulin curves was similar, suggesting that only a finite pool of granules can be secreted during stimulation by [Ca²⁺] in the absence of glucose.

When similar experiments were repeated in presence of 3 mmol/l glucose, continuous depolarization caused a sustained, only slowly increasing elevation of [Ca²⁺]_i, and repetitive stimulations caused similar peaks of [Ca²⁺]_i (Fig. 4C). The presence of this low glucose concentration was sufficient to prevent the loss of Ca²⁺ efficacy on insulin secretion. During continuous depolarization, the immediate secretory response was followed by a partial decrease and a sustained second phase (Fig. 4D). During repetitive stimulation, the amplitude of the insulin responses slightly decreased between the first and second or third [Ca²⁺]_i pulse, but then remained stable (Fig. 3D). Similar observations have been made with single or groups of islets when [Ca²⁺]_i was intermittently raised in the presence of higher glucose concentrations (39,40).

The initial, rapid [Ca²⁺]_i rise appears to trigger release of a limited pool of granules and glucose seems to ensure refilling of this pool. This conclusion is in full agreement with that reached by experiments measuring membrane capacitance as a marker of exocytosis in single β-cells (41). A step increase in [Ca²⁺]_i elicits a biphasic exocytotic response consisting in an initial, rapid, ATP-independent phase and a second, slower, sustained, ATP-dependent phase. The question is whether first-phase insulin secretion in intact cells corresponds to the emptying of this pool of readily releasable granules. The size of “first-phase pool” can be estimated by measuring the dynamics of glucose-induced insulin secretion and the insulin content in the same islets. In perifused mouse islets, about 0.45% of total insulin is secreted during the first 5 min of stimulation by 15 mmol/l glucose (42). A similar size of the readily releasable pool was calculated by capacitance measurements in single cells stimulated by a train of depolarizations, but the release was completed over a considerably shorter time (10 s) (41). In contrast, strong stimulations (2 μmol/l Ca²⁺) of single β-cells in the whole cell mode may release up to 25% of the insulin content in about 2 min (43). Such an enormous rate of secretion never occurs in intact β-cells and is irrelevant to studies of the pools of granules involved in glucose-induced biphasic insulin secretion.

The observation that inhibitors of protein synthesis selectively depressed the second phase of glucose-induced insulin secretion by the perfused rat pancreas (1) originally led to the proposal that this phase requires synthesis of new insulin granules. This interpretation was subsequently shown not to be correct (44,45). The consequences of protein synthesis inhibition in β-cells are time-dependent, the first mechanism causing inhibition of insulin secretion being the disappearance of short half-life proteins involved in the amplifying action of glucose (46).

According to the signal-limited model, two opposite effects of glucose might interact to shape the second phase: a time-dependent inhibition and a time-dependent potentiation (18). The concept of a time-dependent inhibitory signal is partly based on the following observations. When the perfused rat pancreas is stimulated by two brief (5 min every 10 min) elevations of the glucose concentration the second burst of insulin secretion is smaller than the first one as if an inhibitory signal was attenuating the response, and causing the trough between the two phases during sustained stimulation (18). The concept is not challenged by the experiments shown in Fig. 1, because the [Ca²⁺] and insulin changes evoked by the first application of high glucose were not measured. In other experiments in which the two changes were measured in distinct islets, successive stimulations with 20 mmol/l glucose (4 min every 8 min) induced similar [Ca²⁺] rises but the second insulin response was smaller than the first one (40). A similar behavior was observed upon repetitive stimulation with high K⁺ in the presence of glucose and diazoxide (Fig. 4C and D) (39,40). If such an inhibitory signal exists, it is not an impairment of the glucose-induced or K⁺-induced [Ca²⁺]_i rise. An alternative but disputed (47) interpretation of time-dependent inhibition would be that provision of insulin granules to the releasable pool is not rapid enough to refill the pool to the same extent as before the first stimulation.

According to the store-limited model, transfer of insulin granules from a reserve to a releasable pool is necessary for sustained second phase. Figure 4A and B clearly shows that an elevation of [Ca²⁺]_i in β-cells is not a sufficient signal to induce sustained insulin secretion in the absence of glucose (48) or other fuels (36). As little as 3 mmol/l glucose is sufficient to permit development of a second phase during depolarization with high K⁺ (Fig. 4C and D). The magnitude of this second phase increases with the concentration of glucose, and may even become larger than the first one above 10 mmol/l glucose (Fig. 3D), although the concentration of [Ca²⁺]_i does not increase (36). This phenomenon corresponds to the amplifying action of glucose, i.e., the increase by the sugar of the efficacy of Ca²⁺ on the secretory process (19). It is probably equivalent to the time-dependent potentiating action of glucose (18). Whereas its essential contribution to the second phase of glucose-induced insulin secretion is widely accepted (19,36,49,50), its underlying mechanisms are still incompletely understood. Mobilization of insulin granules from a reserve to a releasable pool (41) is plausible. This replenishment may involve physical movement thanks to the cytoskeleton (51–54) or change in granule properties (16).

Physiologically, the second phase of glucose-induced insulin secretion is determined by the amplitude of the [Ca²⁺]_i rise in β-cells and the amplification of the action of [Ca²⁺] on the secretory machinery (19). It is important to bear in mind that these two effects of glucose can normally not be dissociated during steady-state stimulation. Changes in the glucose concentration affect second phase by changing both signals.

All experimental results presented in the figures of this article have been obtained with isolated mouse islets in which second-phase insulin response to glucose is usually flat and lower than the peak of the first phase. A similar pattern is also found with the perfused mouse pancreas (32,33). In contrast, in the rat pancreas and in isolated rat islets, the second phase of glucose-induced insulin secretion increases, often well above the peak of the first-phase, for up to 1–2 h of stimulation. This difference in the magnitude of second phase between the two species has variably been attributed to a greater production of cyclic AMP (55) or expression of distinct isoforms of protein kinase C (56) in the rat β-cell. These proposals implicitly mean that the response to the triggering signal is more strongly potentiated in the rat than the mouse. However, it should also be emphasized that the depolarization and subsequent rise in [Ca²⁺]_i induced by glucose in rat islets are different from those in the mouse. In single rat β-cells, glucose-induced [Ca²⁺]_i increase is extremely heterogenous: slowly rising, transient, monophasic or biphasic (57–59). In intact rat islets, a short lasting increase has been reported (60), but more recent experiments indicate that the [Ca²⁺] elevation is sustained, with a slightly higher initial response than the persistent plateau (31). The major difference with the mouse islet appears to be the absence of [Ca²⁺]_i oscillations under steady-state stimulation.

In normal human subjects, a rapid marked and sustained increase in plasma glucose concentration, achieved by glucose perfusion, induces a biphasic elevation of plasma insulin concentration (61–63). After the initial peak a delayed increase often occurs, a pattern resembling that observed in the perfused rat pancreas. However, when human islets are isolated and perifused, the two phases of glucose-induced insulin secretion occur following a pattern similar to that of mouse islets (64–68) (J.-C.H., unpublished data). Studies of glucose-induced [Ca²⁺]_i changes in human islets have also shown that the biphasic “mouse-like” pattern predominates over the “rat-like” pattern (69).

All conditions interfering with the rapid increase in [Ca²⁺]_i that glucose normally produces in β-cells may be expected to impair first-phase insulin secretion. This may be achieved experimentally, or occurs as a side-effect during treatment with inhibitors of voltage-dependent Ca²⁺ channels or activators of K⁺-ATP channels. Less obviously, this may also be produced by agents that maintain β-cells depolarized in a glucose-independent manner.

When islets are stimulated by a high concentration of K⁺ or sulfonylurea, [Ca²⁺]_i is elevated and insulin secretion is stimulated even in low glucose. A rise in the glucose concentration (e.g., from 3 to 15 mmol/l) does not elevate [Ca²⁺]_i further but strongly increases secretion. However, this increase in secretion is slightly delayed, progressive and monophasic, without typical first phase (40,42,70). It was anticipated that islets from mice without functional K⁺-ATP channels in their β-cells would respond to glucose as do islets from normal mice after blockade of these channels with sulfonylureas, i.e., essentially with a second phase mediated by the amplifying action of glucose (19). This was, at least qualitatively, the case after inactivation of the sulfonylurea receptor (SUR 1) subunit (71) but not after inactivation of the pore forming (Kir 6.2) subunit (72). In the latter model, glucose was practically ineffective on insulin secretion in vitro. Persistent hyperinsulinemic hypoglycemia of infancy is often caused by an inactivating mutation of SUR 1; unexpectedly, however, these infants may rapidly release insulin after intravenous glucose injection (73). Thus, no coherent picture emerges from these preliminary studies of artificial or spontaneous models of K⁺-ATP channel inactivation. Further studies are needed to understand stimulus-secretion coupling in these β-cells chronically subjected to high [Ca²⁺]_i.

Impaired first-phase insulin secretion in response to glucose is an early sign of β-cell dysfunction in type 2 diabetic patients (3,74). Because agents like tolbutamide or arginine remain effective (75), the hypothesis that the pool of releasable granules is empty can be rejected. It is more likely that glucose is unable to raise [Ca²⁺]_i rapidly in these diseased β-cells.

The first phase of glucose-induced insulin secretion requires a rapid and marked elevation of [Ca²⁺]_i in β-cells, and corresponds to the release of granules from a limited pool. Although this pool is not emptied within 5 min of glucose stimulation, sustained secretion (second phase) requires production of a signal other than the [Ca²⁺]_i rise. This second signal corresponds to the amplifying action of glucose and may serve to replenish the pool of granules that are releasable at the prevailing [Ca²⁺]_i. Whether physical translocation of granules, or change in granule properties is involved, remains open. The magnitude and the time-course of the triggering signal ([Ca²⁺]_i) are also important to shape the biphasic pattern of the secretory response. This is also clearly demonstrated by the synchrony between the oscillations of β-cell [Ca²⁺]_i and insulin secretion, an aspect of the second phase of glucose-induced insulin secretion that is developed in other contributions to this volume.