Visualising insulin secretion. The Minkowski Lecture 2004
- First Online:
- Cite this article as:
- Rutter, G.A. Diabetologia (2004) 47: 1861. doi:10.1007/s00125-004-1541-1
Insulin secretion from pancreatic islet beta cells is a tightly regulated process, under the close control of blood glucose concentrations, neural inputs and circulating hormones. Defects in glucose-triggered insulin secretion, possibly exacerbated by a decrease in beta cell mass, are ultimately responsible for the development of type 2 diabetes. A full understanding of the mechanisms by which glucose and other nutrients trigger insulin secretion will probably be essential to allow for the development of new therapies of type 2 diabetes and for the derivation of “artificial” beta cells from embryonic stem cells as a treatment for type 1 diabetes. I focus here on recent developments in our understanding of beta cell glucose sensing, achieved in part through the development of recombinant targeted probes (luciferase, green fluorescent protein) that allow islet beta cell metabolism and Ca2+ handling to be imaged in situ in the intact islet with single cell resolution. Combined with classical biochemistry, these techniques show that the beta cell is uniquely poised, thanks to the expression of low levels of lactate dehydrogenase and plasma membrane lactate/monocarboxylate transporters, to channel glucose carbons towards oxidative metabolism, ATP synthesis and inhibition of AMP-activated protein kinase, a newly defined regulator of insulin release. I also discuss the molecular basis of the recruitment of secretory vesicles to the cell surface, analysed by the use of new imaging techniques including total internal reflection of fluorescence, as well as the “nanomechanics” of the exocytotic event itself.
KeywordsATP Beta cell Ca2+ Exocytosis GFP Imaging Islet Luciferase Metabolism Secretion
AMP-activated protein kinase
enhanced green fluorescent protein
ATP-sensitive K+ channel
monomeric red fluorescent protein
sterol regulatory element binding protein-1c
total internal reflection of fluorescence microscopy
tissue plasminogen activator
Tight regulation of insulin release is essential for normal blood glucose homeostasis. Were daily insulin release to exceed ~2% of the total pancreatic content (more than 2500 units) then fatal hypoglycaemia would ensue. Conversely, whilst decreases in beta cell mass are apparent in late stage type 2 diabetes , defective insulin release from existing beta cells seems likely to be the underlying cause of this disease in most cases. Thus, hyperglycaemia is not usually evident before the loss of ~70% of beta cell mass in type 1 diabetes , while 40 to 50% of the beta cell mass remains intact after baboons are rendered diabetic with streptozotocin .
The study of glucose-regulated insulin secretion began in earnest in the 1960s with the establishment of a radioimmunoassay for insulin . More recently, studies on individual beta cells were made possible with the development of electrophysiological approaches including amperometry  and measurements of membrane capacitance [6, 7]. Though the latter techniques provide remarkable temporal resolution, they are, however, limited with respect to the spatial aspects of insulin release. Overcoming this limitation, molecular approaches now allow the expression in living beta cells and islets of a range of recombinant fluorescent and bioluminescent probes, often derived from lower organisms . Combined with advances in imaging technologies, these tools enable the secretory event to be imaged in real time, allowing the intracellular signalling mechanisms that control it to be dissected.
Glucose sensing: adenosine-triphosphate-sensitive K+ channels
Closure of KATP channels , and a progressive depolarisation of the plasma membrane (from a resting potential of about −70 mV to 0 mV) , cause the beta cell to fire action potentials as glucose concentrations increase. These, in turn, open voltage-sensitive, L-type Ca2+ channels , and prompt the influx of Ca2+ ions. The resulting increase in intracellular free Ca2+ concentration [19, 20] is then the major stimulus that triggers secretory vesicle fusion with the plasma membrane [21, 22], whilst Ca2+ uptake by mitochondria also enhances mitochondrial ATP synthesis  to sustain glucose signalling [10, 11].
Glucose sensing: metabolic specialisation of the beta cell
It is generally accepted that the intracellular metabolism of glucose and ATP synthesis, rather than the binding of the sugar to a specific “receptor”, explains the triggering of insulin secretion (the “fuel hypothesis”) [24, 25, 26]. The mammalian beta cell enjoys an unusual metabolic configuration, which tunes its glucose sensing to the insulin requirements of the whole animal. Firstly, glucose uptake is catalysed by the high K m (i.e. low-affinity) liver-type glucose transporter, Glut2 . Secondly, the first committed step in glycolysis, glucose phosphorylation, is catalysed in the beta cell by the low-affinity type IV hexokinase, better known as glucokinase (GK) . Together, these “sensors” ensure that glucose phosphorylation increases sigmoidally as blood glucose concentrations rise over the physiological range (3.5–8 mmol/l). Correspondingly, inactivating mutations of GK in humans cause MODY2 , whilst beta-cell-specific inactivation of the GK gene causes decreased sensitivity to glucose  and impaired insulin release in vivo .
Whilst the proximal aspects of glucose metabolism, catalysed by Glut2 and Gk, are closely similar in the beta cell and liver (but different from the majority of cells), there are marked differences in the distal end of the glycolytic pathway in these two cell types. Firstly, beta cells express vanishingly low levels of lactate dehydrogenase (LDH) activity and of the plasma membrane monocarboxylate (lactate) transporter-1 (MCT-1) [32, 33, 34, 35]. These specialisations ensure that: (i) close to 100% of glucose-derived pyruvate enters the tricarboxylate/citrate (TCA) cycle and is either broken down to H2O and CO2 yielding ATP  (75%), or assimilated into newly synthesised proteins : this feature is important for the normal stimulation of insulin secretion by glucose since overexpression of LDH leads to a right shift in the dose response to the sugar [37, 38], and islet levels of LDH are increased in a model of type 2 diabetes, namely 85 to 95% pancreatectomy ; (ii) circulating lactate and pyruvate do not stimulate insulin secretion during exercise  (Fig. 1a). Correspondingly, glucose-induced increases in [ATP] are significantly smaller in pancreatic alpha than beta cells  (unpublished observations, M. Ravier and G.A. Rutter), which possess relatively high levels of LDH. In contrast, levels of mitochondrial glycerolphosphate dehydrogenase are elevated in beta cells [32, 36, 41], providing a mechanism for the re-oxidation of glycolytically derived NADH to NAD+.
Alternative beta cell glucose sensors: AMP-activated protein kinase
Glucose-stimulated insulin secretion is biphasic: visualisation in living beta cells
By labelling the secretory vesicle membrane with a chimeric construct encoding the membrane-resident protein phogrin (phosphatase on the granule of insulinoma, also called IA-2β ) fused to enhanced green fluorescent protein (EGFP), we aimed to image and quantify vesicle movement  in single living beta cells. Phogrin, a single transmembrane-spanning vesicle protein, is well suited to the task of localising EGFP to secretory granules, since it possesses multiple and partially redundant targeting sequences . Fusion of EGFP to the C-terminus of phogrin also allows the photoprotein to face the cell cytosol, rather than the lumen, thus avoiding fluorescence quenching at low intraluminal pH. In contrast, the fusion of insulin with EGFP  is more problematic , unless steps are taken to reduce expression levels (T. Tsuboi, G.A. Rutter, unpublished results). The use of a fluorescent marker combined with confocal microscopy [75, 78] provides advantages over the use of earlier microscopic techniques such as differential interference contrast [79, 80], since the latter may be complicated by interference from other organelles (e.g. mitochondria, lysosomes). In addition, “evanescent wave” microscopy (also called total internal reflection of fluorescence or TIRF microscopy) [53, 81, 82], in combination with fluorescent probes, allows selective analysis of fusion events at the cell surface (see below). Using the phogrin.EGFP chimera, it was possible to show that at low (sub-stimulatory) glucose concentrations vesicles displayed only short oscillatory (“jiggling”) movements about a fixed point . However, when the glucose concentration was raised, much longer (several µm) vesicle excursions occurred, frequently towards the plasma membrane. Using TIRF microscopy, others  have recently suggested that first-phase secretion involves predocked vesicles, whilst sustained release involves “newcomers” to the membrane.
Mechanisms involved in the regulation of vesicle recruitment to the plasma membrane
Our early studies  showed that vesicle movement was sensitive to microtubule disruption with nocodazole, but barely affected by the disruption of actin filaments with colchicine. Moreover, simultaneous imaging of vesicles and microtubules suggested that the vast majority of long excursions occurred along microtubules . Correspondingly, inhibition of vesicle recruitment to the cell surface through the expression of an inactivating (dominant-negative) mutant of kinesin lacking the motor domain  did not affect the initial phase of glucose-stimulated secretion in MIN6 cells (measured after 20 min), but completely blocked further release of the hormone (measured at 90 min). These findings demonstrate that whereas vesicle recruitment may not be the sole mechanism involved in the second phase of insulin secretion, it is an essential prerequisite for sustained release of the hormone. Interestingly, kinesin inhibition, achieved with either the dominant–negative mutant  or by RNA interference , led to an essentially complete cessation of vesicle movement, suggesting that anterograde movements of vesicles predominate in the beta cell.
Increases in ATP concentration activate kinesin-dependent vesicle movement in permeabilised cells  and may thus directly regulate the ATPase activity of kinesin by binding to the enzyme’s active site. The loss of an inhibitory phosphorylation event, catalysed by AMPK , is another potential mechanism. Supporting the latter hypothesis, activation of endogenous AMPK or overexpression of the activated enzyme markedly decrease vesicle movements, not only in intact cells (where the effects are largely attributable to the suppression of glucose metabolism, see above) and in permeabilised cells where ATP concentration could be altered at will. In future it will be important to determine whether either of kinesin’s subunits can be directly phosphorylated by AMPK (Fig. 3c). Another intriguing but untested possibility is whether inhibition of AMPK is involved in the second, sustained phase of glucose-stimulated insulin secretion, or in the “amplification” of insulin secretion by glucose when intracellular [Ca2+] is clamped .
Might changes in vesicle motility or recruitment play a role in the pathology of type 2 diabetes? We [93, 94] and others [95, 96] have recently shown that elevations of beta cell triglyceride content caused by overexpression of the lipogenic transcription factor sterol regulatory element binding protein (SREBP1c) leads to a substantial accumulation of intracellular lipid and the near-complete elimination of both phases of glucose-stimulated insulin secretion from beta cells and islets, consistent with reports of inhibitory effects on insulin secretion of lipid infusion in vivo  and the culture of islets with fatty acids in vitro . The effects of SREBP1c overexpression are associated with decreased glucose-induced increases in cytosolic [ATP], vesicle motility at both low and high glucose concentrations, and a reduction in the number of glucose-stimulated release events  (Fig. 5b). Whereas the effects of SREBP1c on glucose-stimulated vesicle excursions are probably due to the lowering of ATP concentrations in these conditions , the blockade of vesicle motility at low glucose concentrations, where there is no difference in ATP between control and SREBP-infected cells, may be due to the physical effects of numerous lipid droplets in the cell cytosol, as well as to changes in the expression of elements of the fusion machinery including Rim1 . Interestingly, depolarisation-stimulated release is unaffected by SREBP1c overexpression, despite a decrease in the number of “morphologically docked” vesicles (i.e. those within ~100 nm and so detectable within an evanescent field). This finding is consistent with the view that these short, Ca2+-dependent final movements of the vesicle towards the plasma membrane are independent of microtubules/kinesin, but rely instead on myoVa/actin interactions.
Mechanisms of vesicle release at the cell surface: full fusion or “kiss and run”?
What is the fate of the secretory granule membrane once it finally arrives at the plasma membrane? It is generally accepted that the molecular machinery of vesicle fusion involves the interaction between soluble N-ethyl-maleimide-sensitive fusion protein attachment receptors that are present on the vesicle and the plasma membrane , and such mechanisms appear to be operative in the beta cell .
For many years it was assumed that the release of insulin required the complete fusion of the secretory and plasma membranes, as was also believed to be the case during synaptic vesicle fusion . Moreover, the detection by electron microscopy of the vesicle core apparently leaving the open mouth of a fused vesicle en masse suggested a complete-collapse model . On the other hand, the uptake of relatively high molecular mass markers into dense core secretory vesicles has been reported both in insulin-secreting  and chromaffin cell-derived PC12 cells , suggesting that the fusion pore must subsequently close, recapturing extracellular material. As in the nerve terminal , it now seems likely that such transient events are predominant at physiological levels of stimulation, where the rate of exocytosis does not exceed the cell’s capacity for endocytosis . Indeed, when the fate of the insulin-containing vesicle is imaged simultaneously in living beta cells, using either the low-molecular-mass dye acridine orange  or the vesicle cargo protein neuropeptide Y (NPY) fused to monomeric red fluorescent protein (mRFP) or Venus  (where Venus is a highly fluorescent and relatively pH-insensitive derivative or GFP) , genuine peptide release events occur without the concomitant release of the vesicle membrane protein phogrin.EGFP into the plasma membrane. This is perhaps the most compelling evidence that complete fusion of the vesicle membrane and the plasmalemma do not occur during insulin release. Very similar findings have also been made in PC12 cells , and suggest a conserved role for kiss-and-run or “cavicapture” exocytosis in neurosecretion.
The fate of the transiently fused vesicle: nanomechanics of insulin vesicle recapture
By what mechanisms is the recapture of the vesicle membrane achieved without full fusion? Dynamin is a 100 Mr GTP-driven mechanochemical enzyme related to mammalian mx-proteins, the yeast vps 1 gene product and Drosophila melanogaster shibire . First identified as a microtubule-associated motor protein-like activity , dynamin is implicated in the recapture of intact secretory vesicles in PC12 cells [105, 116]. Dynamin-1, the principal isoform in neuronal and neuroendocrine cells, possesses GTPase, pleckstrin homology and C-terminal proline-rich domains, and is proposed to act either as a mechanochemical enzyme that cleaves the neck of an endocytosing vesicle (“pinchase”) , or as a molecular spring (“poppase”) that extends and eventually ruptures the tubule linking the vesicle and donor membrane . Dynamin-1 is required for “rapid endocytosis” detectable by capacitance measurements after mild (physiological) stimulation of chromaffin cells . Imaged by TIRF microscopy, dynamin-1.EGFP is recruited, in the absence of other components of the classical endocytic pathway (e.g. clathrin and epsin), to sites of NPY.Venus release in MIN6 beta cells (Fig. 6b). Furthermore, overexpression of mutant dynamin-1 bearing a defective pleckstrin homology domain permits the release of large peptide cargoes including phogrin.EGFP and tPA . Together, these findings indicate that dynamin-1 plays a key role in the endocytic mechanism by which the semi-fused secretory vesicle is recovered from the plasma membrane, essentially intact, following cargo release.
Imaging techniques have provided previously unsuspected information on the complex molecular machinery of insulin release, and demonstrated new layers of regulation. By identifying the important role of particular gene products (kinesin, myoVa, dynamin etc.), these approaches provide potential new therapeutic targets, as well as identifying potentially new diabetes genes that may be useful predictors of this disease.
Research in the author’s laboratory is supported by grants from the Wellcome Trust (Programme Grant 067081), the Juvenile Diabetes Research Fund International, the Biotechnology and Biological Sciences Research Council and Diabetes UK. G.A. Rutter is a Research Leave Fellow of the Wellcome Trust. I thank Drs Takashi Tsuboi, Aniko Varadi, Gabriela daSilvaXavier, Laura Parton and Isabelle Leclerc for useful discussion.