Regulation of insulin secretion: a matter of phase control and amplitude modulation
- First Online:
- Cite this article as:
- Henquin, J.C. Diabetologia (2009) 52: 739. doi:10.1007/s00125-009-1314-y
The consensus model of stimulus–secretion coupling in beta cells attributes glucose-induced insulin secretion to a sequence of events involving acceleration of metabolism, closure of ATP-sensitive K+ channels, depolarisation, influx of Ca2+ and a rise in cytosolic free Ca2+ concentration ([Ca2+]c). This triggering pathway is essential, but would not be very efficient if glucose did not also activate a metabolic amplifying pathway that does not raise [Ca2+]c further but augments the action of triggering Ca2+ on exocytosis. This review discusses how both pathways interact to achieve temporal control and amplitude modulation of biphasic insulin secretion. First-phase insulin secretion is triggered by the rise in [Ca2+]c that occurs synchronously in all beta cells of every islet in response to a sudden increase in the glucose concentration. Its time course and duration are shaped by those of the Ca2+ signal, and its amplitude is modulated by the magnitude of the [Ca2+]c rise and, substantially, by amplifying mechanisms. During the second phase, synchronous [Ca2+]c oscillations in all beta cells of an individual islet induce pulsatile insulin secretion, but these features of the signal and response are dampened in groups of intrinsically asynchronous islets. Glucose has hardly any influence on the amplitude of [Ca2+]c oscillations and mainly controls the time course of triggering signal. Amplitude modulation of insulin secretion pulses largely depends on the amplifying pathway. There are more similarities than differences between the two phases of glucose-induced insulin secretion. Both are subject to the same dual, hierarchical control over time and amplitude by triggering and amplifying pathways, suggesting that the second phase is a sequence of iterations of the first phase.
KeywordsAmplifying pathwayBeta cellBiphasic insulin secretionCytosolic calciumKATP channelsOscillationsPotentiationSulfonylureas
Cytosolic free Ca2+concentration
- KATP channel
ATP-sensitive K+ channel
Sulfonylurea receptor 1
Secreting insulin at the right moment and in appropriate amounts is the vital function of pancreatic beta cells. Any alteration of their functioning perturbs glucose homeostasis, leading to life-threatening hypoglycaemia when insulin secretion is excessive and to damaging hyperglycaemia whenever insulin secretion is insufficient. It is therefore not surprising that beta cells are under the tight control of an array of stimulatory and inhibitory factors, among which glucose plays the leading role through mechanisms that remain incompletely understood after more than 40 years of investigation [1–3].
The more recent identification of an amplifying pathway, originally referred to as the KATP channel-independent pathway [13, 14], has led to the proposal of a dual, hierarchical control of insulin secretion by glucose and other fuels [8, 15]. Although this concept has progressively gained credit over the years, it has not yet been unanimously accepted and incorporated in all published models of the beta cell. The purpose of this review is to present recent experimental evidence supporting the physiological importance of the amplifying pathway and to discuss how both pathways complement each other to ensure temporal control and amplitude modulation of insulin secretion.
The triggering pathway: essential for stimulating insulin secretion
In the presence of non-stimulatory concentrations of glucose, the rate of metabolism in beta cells is relatively low and enough KATP channels are open in the plasma membrane to counteract depolarising currents and maintain the membrane potential at values more negative than the threshold for opening the voltage-dependent Ca2+ channels. The influx of Ca2+ is minimal, [Ca2+]c is low, and insulin secretion is basal. When the concentration of glucose increases, beta cell metabolism accelerates, leading to changes in the concentration of cytosolic adenine nucleotides (in particular an increase in ATP and a decrease in MgADP) which induce closure of KATP channels [11, 16]. The resulting decrease in K+ efflux causes membrane depolarisation, followed by opening of Ca2+ channels and Ca2+ influx into the cell (as reflected by the electrical activity shown in the inset of Fig. 1). The ensuing increase in [Ca2+]c then activates an effector system that promotes exocytosis of insulin-containing granules [17–20]. The triggering Ca2+ signal is essential. All experimental conditions that interfere with the rise in [Ca2+]c impair glucose-induced insulin secretion, whereas all agents, physiological or pharmacological, that increase beta cell [Ca2+]c (regardless of the mechanism) induce insulin secretion [8, 21].
Two drugs that interact with the SUR1 subunit of KATP channels [10, 22] have been particularly instrumental for the studies of stimulus–secretion coupling discussed below. Sulfonylureas, such as tolbutamide, mimic the action of glucose on membrane potential and Ca2+ influx, by closing KATP channels independently of changes in metabolism. Conversely, diazoxide reverses the effects of glucose on membrane potential and Ca2+ influx through the direct opening of KATP channels. Their respective stimulatory and inhibitory effects on insulin secretion are thus attributable to changes in the triggering Ca2+ signal (Fig. 1). Additional mechanisms have been proposed but are unlikely to contribute to the effects that sulfonylureas and diazoxide produce at therapeutically relevant concentrations . Thus, both types of drug are inactive in beta cells lacking KATP channels because of a knockout of either Sur1 (Abcc8) [23, 24] or Kir6.2 (also known as Kcnj11) [25, 26].
Revision of the model: evidence for an amplifying pathway
Hierarchy of the triggering and amplifying pathways
Glucose-induced insulin secretion is characterised by a sigmoidal concentration-dependence curve. The threshold (6–7 mmol/l in mouse islets) corresponds to the concentration at which enough KATP channels have been closed by glucose metabolism to permit depolarisation of the plasma membrane up to the electrical threshold where voltage-dependent Ca2+ channels open, Ca2+ influx accelerates and [Ca2+]c increases. Although glucose metabolism accelerates [5, 16] and amplifying signals are generated below this threshold [8, 29], insulin secretion remains basal until the triggering Ca2+ signal is produced. Increasing [Ca2+]c by other means (membrane depolarisation with KCl or sulfonylureas) is sufficient to unmask the amplification exerted by at low-glucose concentrations. The supremacy of the triggering pathway is thus obvious (see [8, 15] for further discussion), but some confusion has stemmed from a paper reporting that glucose induced insulin secretion via a KATP channel-independent mechanism when [Ca2+]c was artificially maintained at very low levels . Such an effect of glucose indeed exists in vitro, but the extent of insulin secretion is negligible compared with that in the presence of normal [Ca2+]c, and the mechanism requires maximal stimulation of protein kinases A and C by pharmacological means . It is of no physiological relevance and does not cast doubt on the hierarchy between triggering and amplifying pathways. Although switched on by lower glucose concentrations than those that produce depolarisation and Ca2+ influx, the amplifying pathway does not influence insulin secretion if [Ca2+]c is not increased.
The amplifying pathway under physiological conditions
Because the experimental paradigms (either KCI plus diazoxide or a high sulfonylurea concentration) initially used to study the amplifying pathway in fact aimed to prevent an action of glucose on KATP channels, concerns have been expressed that detection of the amplifying effects of glucose may require bypassing these channels. More specifically, doubts were raised concerning the physiological significance of the amplifying pathway. For these objections to be addressed, experiments were needed that did not involve the pharmacological clamping of beta cell [Ca2+]c, i.e. experiments that allowed [Ca2+]c to fluctuate under the sole influence of the tested stimulus and quantitatively compared [Ca2+]c with the rate of insulin secretion measured at the same time.
Insulin secretion and [Ca2+]c were thus monitored in parallel in mouse islets stimulated by different concentrations of two non-metabolised agents, arginine and tolbutamide, in the presence of various glucose concentrations . The results showed that, in agreement with many previous studies, glucose potentiated insulin secretion induced by these non-metabolised secretagogues, but, further, demonstrated that this potentiation involves two mechanisms: augmentation of [Ca2+]c and amplification of the response to Ca2+ . These observations provided direct evidence in support of the conclusions of an electrophysiological study suggesting that glucose augments arginine-induced Ca2+ influx in beta cells and, by an unknown mechanism, also augments the efficacy of Ca2+ to stimulate insulin secretion . This better understanding of the mechanisms by which glucose potentiates insulin secretion induced by non-metabolised secretagogues may help elucidate the pathogenesis of the alterations in insulin secretion in type 2 diabetic patients .
While these studies provide evidence for the involvement of the amplifying pathway under physiological conditions, a mechanistic explanation of the phenomenon is still lacking. Although many possibilities have been tested and ruled out [8, 29, 43, 44], neither the second messenger nor the cellular effector has yet been unambiguously identified. Several approaches using intact or permeabilised cells have shown the involvement of ATP at steps of the secretory process distal to the [Ca2+]c rise [45–47]. Adenine nucleotides may thus serve as second messengers in both pathways (Fig. 1), but other messengers could also be involved [8, 34, 48]. Recent studies call for a (re)evaluation of the roles of cAMP , NADPH [48, 50, 51], citrate cycle intermediates exported to the cytosol [48, 51, 52], AMP-activated protein kinase  and granule translocation by the cytoskeleton [18, 54]. Mathematical modelling of the amplifying pathway may also help to identify the effector system [55, 56]. These issues will not be developed further in this review, to keep the focus on the interactions of the two pathways in the temporal control of insulin secretion.
The biphasic time course of glucose-induced insulin secretion
Upon stimulation by an abrupt and sustained increase in the concentration of glucose, the rate of insulin secretion rapidly accelerates before slowing down (first phase), and eventually stabilises or progressively increases again (second phase) (Fig. 2b). This biphasic time course is observed in vivo and in vitro [57, 58], its characteristics varying somewhat with the experimental model and the species . Notably, there is a discrepancy between in vivo and in vitro responses in mice and humans. Whereas plasma insulin levels increase during the second phase of a hyperglycaemic clamp [41, 60–62], second-phase insulin secretion is flat when islets of both species are studied in vitro [32, 41].
Interest in these peculiar biphasic kinetics has been fostered by numerous clinical studies showing that impaired first-phase insulin secretion in response to glucose is an early sign of beta cell dysfunction in type 2 diabetic patients [40, 57]. Two major, non-exclusive models have been proposed to explain the phenomenon at the islet/beta cell level (reviewed in [18, 34, 59, 61, 63]). One model is based on the distribution of insulin granules in different pools. It proposes that the first phase corresponds to the exocytosis of a small pool of 50–100 docked and primed (readily releasable) granules, whereas the second phase requires the translocation of granules from a reserve pool and their docking and priming. The second model suggests that the biphasic secretory response results from the biphasic time course of a single signal or of the sum of several stimulatory signals.
The triggering pathway during the two phases of insulin secretion
The triggering Ca2+ signal, produced via KATP channel-dependent depolarisation in normal beta cells, is essential for the first and second phases of glucose-induced insulin secretion. Both phases are abrogated when the [Ca2+]c rise is prevented by omission of extracellular Ca2+ or by inhibition of depolarisation with diazoxide or inhibition of Ca2+ influx by blockage of voltage-dependent Ca2+ channels. Without disputing the essential role of L-type Ca2+ channels, it has been proposed that distinct channels variably contribute to Ca2+ influx during the two phases, and that Ca2+ entry through R-type Ca2+ channels creates the optimal conditions for granule translocation and priming during the second phase . A contribution of intracellular stores to the rise in [Ca2+]c during glucose stimulation is also possible , but the predominance of Ca2+ influx is indisputable. Whenever this influx is prevented, glucose lowers [Ca2+]c in beta cells and insulin secretion is not stimulated.
The amplifying pathway during the two phases of insulin secretion
Temporal control of first-phase insulin secretion by the triggering signal
Comparison of the changes in [Ca2+]c and insulin secretion induced by different glucose concentrations reveals another aspect of this time control (Fig. 2a, b). First phases not only increase in amplitude, they also increase in duration, and the similarity of the changes in [Ca2+]c and secretion strongly suggests a causal relationship. Furthermore, imposed alterations in the kinetics of the initial elevation in [Ca2+]c produced by glucose distort first-phase insulin secretion . A striking example is illustrated in Fig. 3c. When mouse islets are stimulated simultaneously with 15 mmol/l glucose and 500 μmol/l tolbutamide, first-phase [Ca2+]c rise is rapidly but transiently interrupted by a lowering that is reminiscent of the initial decrease in basal [Ca2+]c produced by glucose alone (Fig. 3c). First-phase insulin secretion is brief, shorter than with either tolbutamide or glucose alone (Fig. 3d). This peculiar pattern was first observed in the perfused rat pancreas . In contrast, when the stimulation with tolbutamide is applied 100 s after the increase in glucose concentration, at the end of the decrease in basal [Ca2+]c, neither first-phase [Ca2+]c increase, nor first-phase insulin secretion is interrupted (Fig. 3c, d). Together, these results show that first-phase insulin secretion is shaped by the time course and duration of the triggering Ca2+ signal.
Amplitude modulation of first-phase insulin secretion
Needless to say, the first phase is diminished when severe depletion of insulin stores limits granule availability for release. In normal beta cells, the amplitude of first-phase insulin secretion appears to be controlled by both the magnitude of the triggering signal and the amplification of its action on exocytosis of insulin granules (Figs 2 and 3). This dual regulation and the temporal control that glucose also exerts (see preceding paragraph) indicate that the first phase is not attributable simply to the emptying of a limited pool of insulin granules. It is also noteworthy that peak first-phase insulin secretion in vitro is more sensitive to glucose than integrated first-phase secretion (area under the curve), the latter showing the same glucose dependence as second-phase secretion (Fig. 2). This may be important when comparing the two phases or when comparing the first phase between studies.
Temporal control of second-phase insulin secretion by the triggering signal
Second-phase insulin secretion, or total secretion in static measurements, is usually looked upon as a process subject to amplitude modulation, whereas its time regulation is often overlooked. We have seen above that the time dimension in the control of first-phase insulin secretion is not lost by averaging the responses from several islets. The situation is more complex during the second phase. Although the concentration of glucose is stable, the membrane potential of beta cells oscillates (Fig. 1 inset), generating oscillations of [Ca2+]c (Fig. 4). Different types of [Ca2+]c oscillations are known to occur in mouse islets stimulated by glucose. Their characteristics and possible mechanisms are discussed in other reviews [72–74]. In the present context, it is sufficient to emphasise that all these [Ca2+]c oscillations induce pulses of insulin secretion by individual islets [15, 72, 74].
In contrast to the first phase, most [Ca2+]c oscillations that occur during the second phase are not well synchronised between islets, even from the same mouse (Fig. 4a, lower two traces), so that average [Ca2+]c only shows dampened and irregular fluctuations (Fig. 4b). Pooling data from several islets thus largely masks the characteristic pulsatility of insulin secretion, which, nevertheless, has occasionally been measured  or disclosed by deconvolution  in groups of perifused islets, and persists in the intact pancreas through unidentified synchronising mechanisms [72, 74, 77].
Pulsatility of insulin secretion by individual islets requires synchronisation of the triggering signal in all beta cells. Whereas [Ca2+]c oscillations are asynchronous between isolated beta cells and between clusters, even when these originate from the same islet , they are well synchronised throughout each cluster or each islet, owing to the electrical coupling of beta cells  via gap junctions made of connexin36 . When gap junctions are disrupted by knockout of connexin36, the islets no longer display oscillations of [Ca2+]c and insulin secretion . Time (minute-to-minute) regulation of insulin secretion is mainly achieved by variations in the triggering Ca2+ signal, which is not incompatible with the possibility that metabolic oscillations occur simultaneously with, or even cause, [Ca2+]c oscillations [73, 74]. This question will be discussed further in the next section.
Temporal control of second-phase insulin secretion by oscillatory amplifying signals?
Oscillations of insulin secretion in the presence of a stable triggering Ca2+ signal would imply that amplifying signals oscillate independently of [Ca2+]c. Such oscillations of secretion have been reported in batches of rat islets . In single islets from ob/ob mice, small glucose-independent oscillations of insulin secretion (about 2% of the rate of secretion in 11 mmol/l glucose) have been detected in the presence of low and stable [Ca2+]c, and larger, glucose-dependent oscillations were observed in the presence of stably elevated [Ca2+]c . Although we could partly confirm the latter observations in ob/ob mouse islets , we never found evidence that the rate of insulin secretion significantly oscillates when [Ca2+]c does not in normal mouse islets [86, 87]. Moreover, imposed metabolic oscillations at stably elevated [Ca2+]c are poorly able to induce pulsatile insulin secretion compared with imposed [Ca2+]c oscillations . This does not mean that metabolic oscillations do not exist. In fact, oscillations of several metabolic events [89–92] and of potential second messengers [49, 92–94] have been detected in individual beta cells or single islets, but they are often associated with [Ca2+]c oscillations. While it is certain that metabolic amplification does not require an oscillatory [Ca2+]c signal (it was first detected when [Ca2+]c was held stable), amplifying signals might be oscillatory [49, 56]. Appropriate temporal coordination of these signals with [Ca2+]c oscillations could ensure optimal effects on insulin secretion.
Amplitude modulation of second-phase insulin secretion
Reducing the magnitude of the [Ca2+]c rise during second phase decreases insulin secretion . Augmenting the amplitude of imposed [Ca2+]c oscillations in the presence of stable glucose increases insulin secretion . Amplitude modulation of second phase can thus be achieved by changes in [Ca2+]c. But how does amplitude modulation of insulin secretion manifest itself during second phase, when the triggering [Ca2+]c signal is mainly subject to temporal control by glucose, as shown above?
Heterogeneity of beta cells and amplification of insulin secretion
Individual beta cells are heterogeneous [42, 80, 97–99]. Dose–response curves such as those shown in Fig. 2 could reflect recruitment of beta cells displaying distinct sensitivities to glucose and responding in an all-or-none fashion once activated. Single beta cells are indeed progressively recruited to produce a triggering Ca2+ signal when stimulated by increasing glucose concentrations [42, 99, 100]. However, this recruitment is restricted to a narrow range of low concentrations (4–10 mmol/l), and the individual beta cell response is not all-or-none either with glucose or tolbutamide [99, 100]. Moreover, beta cell association and coupling within clusters or intact islets markedly decreases the heterogeneity [72, 74, 99]. Within an intact islet, virtually all beta cells respond simultaneously [15, 81, 83]. The temporal control of [Ca2+]c exerted by glucose (Fig. 5) is also incompatible with models attributing the increase in the average triggering Ca2+ signal to the recruitment of individual beta cells with distinct sensitivities to glucose.
Nevertheless, a similar triggering signal in all beta cells of an islet does not necessarily imply homogeneity of the secretory responses. Even though there is less metabolic heterogeneity between beta cells associated within an islet than between dispersed beta cells , differences in the production of amplifying signals remain possible. Two beta cells within the same islet could thus secrete insulin at different rates despite a similar triggering signal. This raises the intriguing possibility that the phenomenon of amplification corresponds to the recruitment of poorly secreting beta cells to produce amplifying signals. However, the gradual metabolic amplification observed in single beta cells [46, 47] does not support this hypothesis.
Overall quantitative contribution of triggering and amplifying pathways
Whereas determination of the hierarchy between both pathways is straightforward, estimation of the proportion of glucose-induced insulin secretion that is attributable to each pathway is more problematic and necessarily indirect. It is based on quantitative comparisons of insulin secretion in response to a similar rise in [Ca2+]c in the absence and presence of amplification. It should be noted, however, that the amplification of insulin secretion by glucose is relatively weak when [Ca2+]c is only moderately increased above basal values, and becomes more prominent when the elevation of [Ca2+]c, is larger, be it sustained or oscillatory [87, 95]. Nevertheless, from comparisons such as those in Fig. 3, and others using several distinct protocols [38, 44, 87, 95], it appears that amplification accounts for at least 50% of insulin secretion during the second phase. Figure 3 suggests that the contribution of amplification to the first phase may also be about 50%. However, further experiments comparing the first phases of elevations in [Ca2+]c and insulin secretion in response to different tolbutamide and glucose concentrations are necessary to verify this estimation. Nevertheless, although the metabolic amplifying pathway is second to the triggering pathway in the hierarchy, its quantitative contribution to both phases is by no means trivial.
There are more similarities than differences between the first and second phases of glucose-induced insulin secretion. Both are subject to the same dual hierarchical control. The essential triggering Ca2+ signal exerts both temporal and amplitude control of the response, but the modulation of amplitude is mainly achieved by the metabolic amplifying pathway. When biphasic insulin secretion is considered over its physiological duration of several minutes, the time course of [Ca2+]c changes in beta cells is certainly as important as the distribution of insulin granules in different pools to shape the response. Common views of the second phase also need revision. It is not a sustained secretory response to a monotonic stimulus whose amplitude increases with the glucose concentration. It is caused by a succession of large increases in [Ca2+]c that are subject to a dual regulation (time and efficacy) by glucose. The second phase could be viewed as iterations of the first phase.
I am grateful to all students, post-docs and technicians, and to the many colleagues and collaborators who inspired and contributed to my studies of the islets of Langerhans over the years. I particularly wish to thank M. Nenquin, P. Gilon, J. C. Jonas and J. Rahier. My research work has been made possible by the continuous financial support of the Université Catholique de Louvain, the Fonds National de la Recherche Scientifique in Brussels, the Direction de la Recherche Scientifique de la Communauté Française de Belgique, and the Politique Scientifique Fédérale belge.
Duality of interest
The author declares that there is no duality of interest associated with this manuscript.