Type 2 diabetes mellitus (T2DM) is a metabolic disorder characterized by hyperglycemia, insulin resistance and hyperinsulinemia in early disease stages but a relative insulin insufficiency in later stages. Insulin, a peptide hormone, is produced in and secreted from pancreatic β-cells following elevated blood glucose levels. Upon its release, insulin induces the removal of excessive exogenous glucose from the bloodstream primarily by stimulating glucose uptake into insulin-dependent tissues as well as promoting hepatic glycogenesis. Given the increasing prevalence of T2DM worldwide, elucidating the underlying mechanisms and identifying the various players involved in the synthesis and exocytosis of insulin from β-cells is of utmost importance. This review summarizes our current understanding of the route insulin takes through the cell after its synthesis in the endoplasmic reticulum as well as our knowledge of the highly elaborate network that controls insulin release from the β-cell. This network harbors potential targets for anti-diabetic drugs and is regulated by signaling cascades from several endocrine systems.
The human body relies on the tight control of its blood glucose levels to ensure normal body function. There are two principal hormones responsible for maintaining glucose homeostasis, namely insulin and glucagon. While insulin prevents postprandial hyperglycemia by various actions, including increasing insulin-dependent glucose uptake into adipose and muscle tissue [1–3] and the promotion of glycogenesis , glucagon prevents hypoglycemia by activating hepatic glucose production, i.e. hepatic glycogenolysis and gluconeogenesis . Both hormones are produced in and secreted from specific, yet distinct cell types clustered together within the islets of Langerhans: glucagon is released from α-cells, whereas insulin is secreted from β-cells. Together with other cell types, such as γ-, δ- and ε-cells releasing pancreatic polypeptide, somatostatin (SST) and ghrelin, respectively, these cell clusters comprise the endocrine portion of the pancreas . In contrast with the endocrine manner of action, i.e. the direct secretion of pancreatic hormones into the bloodstream, the exocrine portion, consisting of acinar cells, releases digestive enzymes, e.g. amylase, pancreatic lipase and trypsinogen, by means of ducts .
In type 2 diabetes mellitus (T2DM), the relative (i.e. relative to insulin requirements in early disease stages) and absolute (in advanced disease stages) insufficiency of insulin secretion results in elevated blood glucose levels, thereby causing various health complications such as cardiovascular disease (CVD) , diabetic neuro-, nephro- and retinopathy, as well as diabetic foot syndrome and cataracts . Due to the worldwide increasing prevalence of T2DM in developed and developing countries, there is an urgent and pressing need for more efficient medical intervention . Therefore, elucidating and understanding the molecular mechanisms underlying insulin biogenesis and secretion are of paramount importance for the development of novel therapeutic strategies. Here, we review established models and recent advances in our understanding of how insulin is synthesized, stored in granules and released via regulated exocytosis from pancreatic β-cells. Furthermore, we elucidate how the various processes involved in insulin biogenesis and secretion can be regulated by drugs and physiological stimuli.
Insulin granule biogenesis
The endocrine portion of the pancreas accounts for only 1–2% of the whole organ , with β-cells representing 65–80% of the total islet cell number . One single β-cell is estimated to contain about 10 000 insulin-containing granules with an average diameter of approximately 300 nm . However, of these 10 000 granules fewer than 100 complete exocytosis during the first 10–15 min of insulin secretion , with newly synthesized insulin secretory granules preferentially undergoing exocytosis over older granules . The estimated average half-life of insulin granules is three days, with older granules being cleared from the cell by various lysosomal degradation mechanisms . This overwhelming redundancy in insulin granule biogenesis possibly reflects the importance of insulin secretion in mammalian physiology.
A mature insulin molecule consists of 51 amino acids and arises from its precursors, preproinsulin and proinsulin, in a multistep process. The nascent product translated from insulin mRNA is the single chain preproinsulin [16,17], a biologically inactive precursor whose N-terminal signal peptide is removed by signal peptidases in the lumen of the rough endoplasmic reticulum (RER), thereby generating proinsulin. This molecule is then transported to the trans-Golgi network (TGN) where it is sorted into immature ISGs, which not only contain proinsulin but accumulate various proteins, divalent ions, such as Ca2+ and Zn2+ , as well as protons while traversing the Golgi apparatus before exiting the TGN . This process is summarized in Figure 1 (for reviews, see [20,21]). Advances in mass spectrometry have allowed extensive coverage of the ISG proteome , yet the functions of the various ISG-enriched proteins, peptides and lipids in the formation of ISGs have yet to be fully elucidated [23–25].
Schematic of insulin granule biogenesis.
Currently, there are two hypotheses with respect to sorting, one being ‘sorting for entry’ taking place in the TGN and the other is ‘sorting by retention’, which occurs in the immature secretory granule (ISG) [26,27]. The former process is present in every cell type and describes the constitutive or continuous secretion of proteins, e.g. collagen release from fibroblasts, without the need of external stimuli to initiate exocytosis. In contrast, ‘sorting by retention’ refers to the regulated secretory pathway occurring only in exocrine, (neuro)endocrine and neural cells, and involves stimulus-secretion-coupling to evoke granule exocytosis [28–30] as found in β-cells, from which the large majority of insulin is released upon glucose stimulation . However, there is a constitutive-like process within the regulated pathway to remove redundant proteins and enzymes [32–34], and this purge is thought to contribute to the maturation of secretory granules (for reviews, see [29,35,36]). Within the regulated secretory pathway there are two models, namely ‘aggregation-mediated sorting’ and ‘receptor-mediated sorting’, describing how the secretory granule proteins are sorted into ISGs. While the first model proposes members of the granin protein family mediate the sorting, the latter model suggests carboxypeptidase E (CPE) is involved in granule protein sorting. Granins, namely chromogranin-A (CgA) and -B (CgB) as well as the secretogranins (Sg) II–IV, are a family of regulated secretory proteins [37–39], which are present in the ISGs of neuroendocrine  and pancreatic β-cells . These proteins are thought to be involved in large dense-core secretory granule biogenesis [42,43] due to their homo- and heterotypic aggregation, which is generally assumed to promote sorting and the coalescence of various ISG components, and is driven by high calcium but low pH levels in the lumen of the TGN and budding ISGs [44,45]. Additionally, CgA and secretogranin III were found to bind hormones in vitro  and secretogranin III has also been shown to bind cholesterol, which is an abundant component of secretory granules . However, although the loss of CgA was shown to result in dysregulated dense-core granule formation in adrenal chromaffin cells in vivo in one study , Chga gene ablation did not lead to abnormal neuroendocrine dense-core secretory granules in another study  and CgB-deficient mice display normal insulin granule biogenesis . This inconsistency hence challenges the essential role of granins in granule biogenesis. Along the same lines, there is some controversy regarding CPE acting as a/the receptor mediating proinsulin sorting into the ISGs, in addition to its role in converting proinsulin to insulin , as described later. CPE was found to be important in routing proinsulin to secretory granules in some studies [51–54] and, moreover, its disruption may account for familial hyperproinsulinemia . However, except for the latter, all the other studies were conducted using neuroendocrine cells, i.e. Neuro-2a [52,53] or Sf9 cells . Interestingly, two previous studies performed in pancreatic islets did not show an involvement of CPE in proinsulin sorting [56,57], whereas a more recent study reinvestigating the sorting capabilities of CPE reconfirmed its potential role as a sorting/retention receptor .
Following assembly, ISGs bud off from the TGN in a process mediated by the ADP-ribosylation factor (ARF) family of small guanosine triphosphate (GTP)ases, their GTPase-activating proteins (ARF-GAPs) and guanine nucleotide-exchange factors (ARF-GEFs) . In general, ARFs undergo consecutive rounds of GTP binding, which is catalyzed by ARF-GEFs, followed by GTP hydrolysis facilitated by ARF-GAPs. In the GTP-bound state, ARFs are activated, that is, they bind the membrane as well as downstream effector proteins to initiate the assembly of vesicle coat proteins which eventually mediate membrane scission . Recent reports have shown that BIG3, a defective ARF-GEF that is abundant in β-cells, acts as a negative regulator of insulin granule biogenesis . BIG3 knockout mice exhibit an enlarged pool of insulin secretory granules, leading to an enhanced insulin release frequency and hence increased total insulin secretion. The inhibitory function of BIG3 potentially arises from its non-catalytic nucleotide exchange domain, which could act as a non-functional competitor of active ARF-GEFs .
Several studies also highlight the importance of the arfaptin family in regulated exocytosis. Arfaptins contain a characteristic lipid-binding Bin/amphiphysin/Rvs (BAR) domain which is critical for insulin granule biogenesis. Arfaptin-1 binds to ARF and ARF-like proteins to inhibit their membrane scission activity by preventing the activation of phospholipase D (PLD) [63,64], which is involved in the regulation of vesicle budding from the TGN . Conversely, phosphorylation of arfaptin-1 on a serine residue near its BAR domain by protein kinase D-1 (PKD1) was shown to dissociate arfaptin-1 from ARF, thereby allowing vesicle generation in in vitro assays. Knockdown of arfaptin-1 in a cell line model caused accumulation of small non-functional ISGs in the cytosol, resulting in diminished insulin secretion .
Two other members of the arfaptin family, namely protein kinase C-binding protein 1 (PICK1) and islet cell antigen 69 kDa (ICA69), form heteromeric complexes associated with ISGs near the TGN. The PICK1 BAR domain was found to cause tubulation of liposomes in in vitro assays, suggesting that vesicle budding from the TGN might be defective in the absence of this protein . In support of this hypothesis, PICK1 and ICA69 knockout mice accumulate proinsulin and immature insulin granules, resulting in glucose intolerance possibly due to decreased insulin secretion . Additionally, a PDZ domain (an acronym made up of the initial letters of postsynaptic density protein (PSD95), Drosophila disc large tumor suppressor (Dlg1) and zonula occludens-1 protein (zo-1)), which is a common structural domain present in eukaryotes and eubacteria  mediating protein–protein interactions , on PICK1 indicates a potential role for this protein in the assembly of membrane proteins during the maturation of insulin granules .
ISGs are coated with clathrin  and the progressive clustering and budding of clathrin-coated vesicles from ISGs has been proposed as a potential mechanism for insulin granule maturation . However, a recent study showed that clathrin was not involved in sorting proinsulin within the TGN, its conversion to insulin, forming dense-core secretory granules or the regulated exocytosis of proinsulin and/or insulin but rather in the removal of proteases during maturation . Despite this discrepancy, it is generally accepted that the proinsulin-processing enzymes CPE and prohormone/proprotein convertases (PC1/3 and 2) have to be sorted into immature insulin granules together with proinsulin [72,73] to initiate maturation of the insulin granules. Activation of these pH-sensitive and calcium-dependent enzymes by mildly acidic pH is critical for insulin granule maturation  and the required acidification is accomplished by the action of proton pumps and chloride channels [74,75]. Both CPE and PCs are responsible for cleaving the connecting or C-peptide from proinsulin which consists of three chains, namely the A-, B- and C-chain, whereby the C-chain corresponds to the C-peptide connecting the A- and B-chain; thus, the sequence of the proinsulin molecule is ‘B-C-A’. Removal of the C-peptide by PC1/3 dissociating the B- and C-chain  and by PC2 cleaving the A- and C-chain  eventually forms insulin, now consisting of the A- and B-chains connected by disulfide bonds. In addition to the well-known proinsulin-processing enzymes described above, the high-temperature-induced dauer formation (HID-) 1 protein has recently been identified as playing a role in the conversion from proinsulin to insulin. Mice lacking the hid-1 gene in pancreatic β-cells show a highly increased proinsulin/insulin ratio and accumulation of proinsulin, resulting in an impaired glucose tolerance which is most probably due to the incorrect processing of proinsulin .
Glucose-stimulated, calcium-dependent insulin secretion
Although food components other than glucose also trigger insulin secretion [79–82], the classical external stimulus is an increase in blood glucose levels resulting from the breakdown of ingested carbohydrates, thereby initiating the insulin signaling cascade referred to as glucose-stimulated insulin secretion (GSIS) . The first step in GSIS is the entry of glucose into the β-cell mediated by the glucose transporter GLUT2. Within the cell, glucose is catabolized to produce ATP, leading to an elevated intracellular ATP/ADP ratio, which in turn inhibits ATP-sensitive potassium channels (KATP). Under unstimulated conditions, that is, when the ATP/ADP ratio is low, these KATP-channels are open to maintain the cell's negative resting potential as a consequence of the equilibrium potential of K+-ions across the plasma membrane (PM). However, inhibition of the KATP-channels causes membrane depolarization due to more positive potentials emerging from KATP-channel closure on the one hand and the concomitant opening of non-selective cation channels (NSCCs) on the other. NSCCs, in particular transmembrane receptor potential (TRP) melastatin 2 (TRPM2), were shown to mediate inward background cation currents evoked by glucose metabolism and/or GLP-1  initiating the cyclic adenosine monophosphate (cAMP)/exchange protein directly activated by the cAMP (Epac) 2 signaling pathway , described later in ‘Enteroendocrine regulation’. Hence, the concurrent closure of KATP-channels and opening of NSCCs facilitates and amplifies membrane depolarization by altering the respective cation currents which in turn triggers the opening of voltage-dependent calcium channels (VDCCs) on the PM, allowing extracellular calcium to enter the cell. These increased intracellular calcium levels eventually trigger exocytosis, culminating in insulin release .
Calcium influx is coupled to vesicle fusion and exocytosis by a family of proteins known as synaptotagmins (Syts). At least 15 synaptotagmins (Syt 1–15) have been identified, cloned and are loosely defined by their calcium-transducing functions as well as their protein domain organization . Synaptotagmins are characterized by a short N-terminal domain which precedes a single membrane-spanning domain followed by a linker region of variable length separating the transmembrane anchor from two C-terminally located subdomains, namely C2A and C2B. Both contain a conserved calcium-binding motif, generally known as the C2 domain, which typically consists of eight β-strands arranged as a so-called β-sandwich. These β-strand-rich motifs are capable of binding multiple calcium ions and hence account for the dependence of regulated exocytosis on calcium . The calcium-binding affinities of C2 domains from different synaptotagmins vary by over an order of magnitude , allowing synaptotagmins to couple calcium to exocytosis in a wide range of systems. The founding member of the synaptotagmin family, Syt1, is enriched in neuronal tissue and is the primary calcium sensor for the synaptic exocytosis of neurotransmitters , whereas Syt7 has been proposed to be the calcium-transducing molecule in pancreatic β-cells. Syt7 knockout mice were found to not only exhibit marked impairments in the first and second phase of insulin secretion  but also a defective glucagon secretory response , thereby supporting a physiological role for Syt7 in pancreatic endocrine exocytosis. The identity of other synaptotagmin(s) responsible for the residual GSIS in Syt7 knockout mice remains to be clarified [93–95]. Early biochemical  as well as more recent structural studies contributed considerably to the establishment of physical models elucidating how synaptotagmins may interact with the SNARE components, described in detail later, to modulate vesicle fusion and exocytosis. Studies using single molecule fluorescence resonance energy transfer (FRET)  and nuclear magnetic resonance (NMR)  revealed that the synaptotagmin-SNARE complex is conformationally flexible, which in turn allows the complex to interact with other binding partners and adapt to remodeling membrane surfaces over the course of membrane fusion. A recent X-ray free electron laser (XFEL) crystallography study confirmed the close interaction of the Syt1 C2B domain with the SNARE complex via multiple interfaces, which is predicted to enable the synaptotagmin-SNARE complex to move as a single unit upon calcium binding to Syt1 via remodeling and eventually driving fusion between membranes .
Regulated exocytosis is a highly controlled process that generally involves docking and tethering of vesicles to the PM, fusion of vesicles with the PM and eventually the release of vesicle contents (for reviews, see [100,101]). Several sets of specific proteins located on the vesicle and PM interact to evoke the final exocytosis. The traditional canonical model of exocytosis comprises the soluble N-ethylmaleimide-sensitive factor (NSF) attachment protein (SNAP-) receptor proteins (SNAREs), which are defined by a signature SNARE domain of about 60 amino acids and has a high propensity to form highly stable coiled-coil structures. Specifically, the target (t-)SNARES and syntaxin 1, which are located on the PM [102,103] interact with the vesicle (v-)SNARE synaptobrevin 2, or vesicle-associated membrane protein VAMP2, which is incorporated into the vesicle's membrane, thereby forming the so-called SNARE complex [104,105]. Trans-zippering of SNAREs from opposing plasma and vesicular membranes into a highly stable four-helix coiled-coil allows the energetically demanding membrane fusion reaction to occur [106,107]. Apart from the SNARE proteins mentioned above, other isoforms, such as VAMP3 , SNAP-23  as well as syntaxin 3  and syntaxin 4 , have also been implicated in GSIS.
SNARE protein levels are regulated by an adaptor protein, phosphotyrosine interaction, the PH domain and leucine zipper containing (APPL) 1, which is a BAR-domain adaptor molecule that has been identified previously as a node in the Akt signaling pathway and, moreover, has been implicated in endocytosis . The contribution of APPL1 to insulin secretion was demonstrated in APPL1 knockout mice expressing lower levels of all three SNARE proteins as well as exhibiting significantly fewer granules in close proximity to the membrane, which in turn resulted in an impaired first-phase insulin secretion. These defects could be rescued by overexpression of a constitutively active Akt2, indicating the insulin signaling pathway to be involved in regulated exocytosis . Strikingly, exosome sequencing has revealed loss-of-function mutations of APPL1 in patients suffering from familial diabetes, thereby highlighting the critical role of this protein in GSIS in humans .
The formation of the SNARE complex has been shown to be regulated by Sec1/Munc18-like (SM) adaptor/tethering proteins [114–116] and Ras-related in brain (Rab) proteins [117,118] (see Figure 2). Extensive studies using total internal reflection fluorescence (TIRF) microscopy revealed the dynamics of various granule components – including SM  and Rab proteins  – in high spatio-temporal resolution [121,122]. These adaptor proteins are proposed to function at distinct stages, such as when the granules attach to the PM (docking) or become ready for the subsequent fusion process (priming). A well-studied example of the SM protein family and its involvement in insulin exocytosis is munc18 whose binding to syntaxin 1 is obligatory not only for neurotransmitter [123,124] but also insulin release [125,126]. Munc18 binds syntaxin 1 in a clasp-like conformation , resulting in cluster formations at the PM that are essential for and precede the recruitment of SNAP-25 and other SM and Rab proteins . The munc18-syntaxin 1 complex is thus a critical component in the docking stage of canonical regulated exocytosis and is obligatory in both β-cells and neurons [125,129]. Munc13, which is another member of the SM protein family, binds to the munc18-syntaxin complex and hence may also be critical for granule exocytosis. The binding of munc18 as well as munc13 to syntaxin 1 is proposed to prime granules for fusion [128,130,131] as indicated by impairment in the sustained second phase of insulin secretion, namely inhibition of the average secretion rate on the one hand and a delay in replenishing the readily releasable pool on the other, in munc13-1 knockout mice .
Cellular process and molecular players in the regulated exocytosis of insulin granules.
Various small GTPases from the Rab family are expressed in β-cells. Rab3A is an insulin granule-associated protein that binds munc18 and has been implicated in secretory granule docking in various models of regulated exocytosis [133–135]. While mutations in the human Rab27A gene were identified as the cause for genetic hypopigmentation arising from a deficiency in the trafficking of lysosome-related organelles , murine Rab27A mutations result in a loss of granule docking, a decrease in insulin secretion and diminished replenishment of insulin granules following GSIS [137,138]. Rabs are known to associate with various effector molecules; one example is Rab3-interacting-molecule (Rim)2α, a key interacting partner of Rab3A which is required to accomplish the role of Rab3A in granule docking. Rim2α simultaneously interacts with munc13 to facilitate the latter's priming function by forming a tripartite complex that is reported to be critical for neurotransmitter release . In β-cells, the nested binding of Rab3A and munc13 to Rim2α is conserved and critical for GSIS, which is substantiated by an attenuated GSIS response in Rim2α null mice . Another member of the Rab family, Rab27A, was reported to not only interact with effector molecules, such as myosin Va and Slac2-a/melanophilin, in the melanosome  but also with its effector, granuphilin, to facilitate – together with syntaxin 1 – granule docking in β-cells .
To efficiently cope with the glucose influx and subsequent rise in blood glucose levels, insulin release is biphasic with a rapid first phase peaking at around 5 min followed by a slowly developing and sustained second phase [141–143]. In order to ensure their prompt availability, some mature insulin-containing granules are already docked to the PM, thereby representing the so-called readily releasable pool (RRP) [12,144,145]. During the first phase, insulin is released from this RRP, whereas the second phase of insulin secretion is provided by newly recruited granules from the storage pool [12,145,146]. However, there has been a recent debate over whether  or not  vesicles undergoing exocytosis are docked to the PM, indicating that age [14,148,149] and/or mobility [121,150] of the secretory granules play an important role, with newly synthesized rather than long-term stored insulin granules undergoing exocytosis . The debate is reminiscent of several previous studies using green fluorescent protein (GFP)-labeled cargo (e.g. atrial natriuretic factor (ANF)-EGFP (enhance GFP), and neuropeptide Y (NPY)-EYFP (enhanced yellow fluorescent protein)) or vesicle-resident protein in combination with fluorescence microscopy in PC12 cells [152–154], adrenal chromaffin cells [155,156] and dissociated neurons , which proposed that freshly made and/or mobile granules are preferably released compared with docked and stationary granules [152–157]. Moreover, Dehghany et al. recently proposed a spatial model in which there is only one secretory pool accounting for both the first and second insulin secreting phase due to an increased number of docking sites on the PM . Another issue in the discussion of the docking-tethering-fusion process is represented by the recent finding that synaptobrevin 2 is not required to bridge the vesicle and PM , and that pre-assembled SNARE complexes are absent in the unstimulated state [128,147]. Thus, the necessity of docking for subsequent exocytosis remains a subject of debate with evidence suggesting that both physiological and non-physiological stimuli elicit distinct exocytotic pathways which involve different sets of regulatory proteins [134,138,160].
Vesicle fusion and insulin secretion are dynamic processes, hence SNARE assembly and insulin granule pool depletion is involved as well as the disassembly of the SNARE complex and insulin granule replenishment. The first-mentioned is mediated by NSF, an ATPase that binds to the SNARE assembly via the adaptor protein α-soluble NSF attachment protein (α-SNAP) and dissociates the SNARE complex after membrane fusion [161,162] using the released energy from ATP hydrolysis to enable recycling and reuse of the SNARE components . Interestingly, NSF also seems to be involved in ATP-dependent SNARE prefusion  or even membrane fusion steps . In order to replenish the depleted insulin storage pools, granule exocytosis triggers a positive feedback loop during prolonged demands for insulin by up-regulating the transcription of insulin and other granule genes. Islet cell autoantigen (ICA) 512 is an intrinsic single pass transmembrane protein that is cleaved by the cytoplasmic calcium-dependent protease µ-calpain during insulin granule fusion with the PM, thereby generating ICA512-CCF, a cleaved cytosolic fragment. ICA512-CCF then enters the nucleus where it acts as a transcription factor increasing the transcription of genes involved in granule biogenesis and insulin secretion [166,167]. Another protein which enhances insulin secretory granule biogenesis and hence restoration of the granule pool is polypyrimidine tract-binding protein (PTB) whose activation results in increased translation of secretory granule proteins . Interestingly, insulin self-regulates its biogenesis and secretion  by enhancing the transcription of its own gene in an autocrine manner: insulin binding to the insulin receptor A type (IR-A or Ex11-) causes initiation of the insulin receptor substrate 2/phosphatidylinositol-4,5-bisphosphate 3-kinase/p70s6k (IRS-2/PI-3 kinase/p70s6k) and calcium-modulated protein, or calmodulin (CaM), kinase signaling pathway, eventually resulting in increased insulin biogenesis and release [170,171].
Modulating glucose-stimulated insulin secretion
Given the urgent need to develop not necessarily new but rather more effective drugs and therapies to manage and treat T2DM, i.e. with fewer side effects, it is critical to understand the mechanisms underlying and regulating insulin secretion. This section will discuss drugs and physiological signals that modulate insulin granule exocytosis as well as the pathways on which they act.
Drug-dependent regulation of insulin secretion
Due to the critical role of the closure of KATP-channels in triggering insulin exocytosis, it is not surprising that a particular class of KATP-channel blockers, which are referred to as sulfonylureas belonging to the class of insulin secretagogues, are among the earliest drugs that have been developed to treat T2DM (for review, see ). Originally developed and used as antibiotics, sulfonamides nowadays provide the basis for sulfonylureas after Janbon and colleagues accidently discovered the hypoglycemic action of p-amino-sulfonamide-thiodiazole in 1942 . Well-known examples of sulfonylureas are tolbutamide, which is rarely prescribed these days due to its possible link to lethal cardiac events [174,175], and glibenclamide. Their mode of action is based on blocking the outward flow of K+-ions through inhibition of the KATP-channels in the PM [176,177], thereby effectively mimicking a high ATP/ADP ratio and hence promoting insulin secretion. At the same time, this glucose-independent stimulation of insulin secretion may result in hypoglycemia, which is the main side effect of sulfonylureas [178,179]. The more recently developed class of glinides, which are non-sulfonylurea insulin secretagogues, share the mode of action of sulfonylureas, that is, KATP-channel inhibition leading to an enhanced insulin secretion ; however, they reduce the likelihood of hypoglycemia due to their earlier onset of action on the one hand and a faster dissociation from the sulfonylurea receptor on the other [181,182]. Conversely, congenital hyperinsulinism is treated by the activation of KATP-channels by drugs such as diazoxide, which cause membrane hyperpolarization resulting in the closure of VDCCs. The subsequent decline in calcium influx consequently leads to diminished insulin exocytosis .
Physiological regulation of insulin secretion
Insulin exocytosis is regulated by hormones secreted from various endocrine systems and tissues/organs, including, but not limited to, the pancreas itself, the brain, the liver, the gut as well as adipose tissues, allowing blood glucose levels to vary with physiological demands (for reviews, see [184–186]). In this review, we selected some well-known insulin secretion modulators, such as acetylcholine, GLP-1 and leptin, as well as some lesser-known neuropeptides and regulation via the circadian rhythm.
The pancreas utilizes paracrine signaling to self-modulate GSIS. One example of modulation is the inhibition of GSIS by somatostatin (SST), a peptide hormone secreted by the δ-cells of the pancreas . SST interacts with its G-protein-coupled receptor (GPCR) (somatostatin receptor or SSTR) whose five isoforms (SSTR1–5) co-localize with α- and β-cells to varying degrees . SSTR activation has dual GPCR-mediated signaling effects: first, the activation of G-protein-coupled inwardly-rectifying potassium (GIRK) channels which hyperpolarize the cell and second, the G-protein-mediated inhibition of calcium channel activation resulting in a subsequently decreased calcium influx . Additionally, SST exerts calcium-independent effects on insulin exocytosis which is likely to occur through the G-protein-dependent activation of the Ca2+-dependent phosphatase calcineurin .
Acetylcholine (ACh) and other cholinergic agonists, such as carbachol and pilocarpine, have been reported to stimulate insulin secretion by priming β-cells for insulin secretion [190,191]. In addition to its main release site, i.e. cholinergic neurons, ACh is secreted from human α-cells, thereby sensitizing β-cells to glucose as well as to other insulin secretagogues in a muscarinic receptor (m3)-dependent fashion [192,193]. ACh priming of insulin secretion is proposed to occur through multiple mechanisms: on the one hand, stimulation of the m3 receptor activates phospholipase C (PLC) and consequently the generation of inositol 1,4,5-trisphosphate (IP3) from phosphatidylinositol 4,5-bisphosphate (PIP2), which evokes the mobilization of Ca2+ from intracellular calcium stores [194,195,196,197]. On the other hand, the concomitant production of diacylglycerol during PIP2 breakdown stimulates protein kinase C (PKC), which in turn enhances the insulin response by affecting the pool of secretory vesicles , to be exact, by sensitizing and thereby increasing the number of calcium-sensitive secretory granules stored in a highly calcium-sensitive pool (HCSP) . Phosphorylation of the m3 receptor facilitates phosphorylation and hence activation of serine/threonine-protein kinase D1 (PKD), which is known to act on membrane fission, vesicle transport and fusion [65,200], and thus is necessary for insulin secretion . m3 activation also leads to the G-protein-independent activation of a sodium leak, non-selective channel (NALCN), resulting in membrane depolarization, calcium influx and subsequently increased cytosolic calcium concentrations .
Being the master regulator, the brain/central nervous system modulates pancreatic insulin secretion [208,209] through several neurotransmitters, amongst others NPY, gastrin-releasing peptide (GRP), melanin concentrating hormone (MCH), vasoactive intestinal peptide (VIP) and pituitary adenylate cyclase-activating polypeptide (PACAP). While NPY diminishes insulin secretion by inhibiting the stimulation of adenylyl cyclase and hence the conversion of ATP to cAMP , the remainder of the neuropeptides increase insulin release through various pathways: GRP activates the phosphatidylinositol-4,5-bisphosphate 3-kinase PI3K/PKC signaling cascade, resulting in mobilization of Ca2+ from intracellular stores [211,212], MCH modulates Ca2+-influx via an extracellular signal-regulated kinase ERK/Akt pathway , and VIP and PACAP by means of cAMP and PI3K signaling [214,215]. Along the same lines, the so-called cephalic phase – that is, the conditioned reflex of releasing hormones, in this case insulin – stimulates insulin secretion  through both, cholinergic and non-cholinergic mechanisms .
The hormone and neurotransmitter adrenaline, or epinephrine, which is produced in and released from the adrenal glands, as well as from certain neurons, has also been shown to be involved in modulating insulin secretion. By binding its α2-adrenergic receptor, adrenaline initiates the G-protein-coupled signaling pathway described earlier; the resulting activation of GIRK channels in turn is responsible for the inhibitory effect of adrenaline on insulin secretion [218,219], thereby coupling the fight-or-flight response to blood glucose control.
Another major regulator of insulin secretion is the enteroendocrine system within the gastrointestinal tract. It has long been known that oral administration of glucose results in significantly higher plasma insulin levels compared with intravenous glucose administration, which is referred to as the incretin effect [220,221]. This phenomenon results from the release of the incretin hormones glucagon-like peptide (GLP)-1 and glucose-dependent insulinotropic peptide (GIP) from their respective enteroendocrine L-  and K-cells  upon food intake . GLP-1 and GIP act on their respective G-protein-coupled receptors (GLP-1R and GIPR), both of which are located on the PM of β-cells . Receptor binding initiates the cAMP signaling pathway, which through several steps, including the receptor's conformational change, exchange of GDP for GTP, activation of adenylyl cyclase and conversion of ATP to cAMP, eventually potentiates insulin secretion [226,227]. One of the well-known effects of the second messenger cAMP includes the activation of protein kinase A (PKA) and of Epac2. Mouse models featuring a constitutive PKA activity exhibit enhanced GSIS but at the same time forfeit the potentiating effect of exendin-4, a stable peptide analog of GLP-1 which is used as an anti-diabetic drug, thereby establishing PKA as a key player in the GLP-1-induced potentiation of GSIS [228,229]. Moreover, inactivating mutations of PKA inhibitory subunits in humans result in increased GSIS upon an oral glucose challenge . Another mode of action of GLP-1 enhancing insulin secretion by elevating intracellular Ca2+ concentrations is via the Ca2+-dependent stimulation of PLC which in turn activates PKC as described in ‘Pancreatic regulation’. While this mode of action may contribute more to basal insulin release , the PKC-dependent activation of Na+-permeable TRPM4 and TRPM5 may play a role in GSIS by mobilizing Ca2+ from thapsigargin-sensitive calcium stores .
It is well-established that the potentiating effect of GLP-1 on insulin secretion only occurs upon the stimulation of β-cells by glucose. A recent study proposed glutamate links glucose stimulation and incretin potentiation. Following the metabolization of glucose to glutamate, the latter is transported into insulin granules via a route that involves PKA as well as the vesicular glutamate transporter (VGLUT) 1 and this incorporation in turn was found to amplify GSIS. Glutamate transport into insulin granules only occurs upon incretin treatment and consequently, glutamate alone has no effect on GSIS. However, the mechanism by which glutamate incorporated into insulin granules promotes exocytosis remains to be elucidated .
Synaptotagmin-7, the main calcium sensor involved in insulin exocytosis, has been proposed to be another PKA target and hence to modulate GLP-1-potentiated GSIS. PKA-dependent phosphorylation of the serine residue 103 in the linker region enhances exocytotic events in murine β-cells, whereas loss of Syt7 consequently results in a diminished GLP-1-potentiated insulin release. Interestingly, the phosphorylation status of the Syt7 linker does not affect the calcium sensitivity of the insulin secretion process and hence, this Syt7-dependent but calcium-independent potentiation of insulin release by GLP-1 might depend on an unidentified interacting partner of phosphorylated Syt7 .
Another key PKA phosphorylation target downstream of incretin stimulation is snapin, a SNAP-25-binding protein originally studied for its role in synaptic exocytosis . Snapin knockdown decreases both the first and second phase of insulin secretion, suggesting a role for snapin in regulated exocytosis. Overexpression of a phosphomimetic mutant restores the potentiating effects of exendin-4, suggesting that phosphorylated snapin mediates the potentiation effect induced by incretins .
Although not activated by PKA, VAMP8, or endobrevin, has also been implicated in GLP-1-potentiated GSIS by mediating insulin release from secretory granules (SGs) that are newly recruited to the PM. As these newcomer SGs undergo minimal docking, the VAMP8-SM-SNARE complex mediating exocytosis has to form quite fast. Supporting this suggestion, loss of VAMP8 was found to cause a less stable assembly of syntaxin-2 and syntaxin-3 with SNAP-25 and SNAP-23, resulting in a diminished GLP-1-dependent secretory insulin response .
Being the second cAMP effector, Epac2 has been shown to induce the opening of ryanodine-sensitive calcium stores (RyRs) in a Ca2+-induced Ca2+-release (CICR) manner with the subsequent increase in cytosolic calcium concentrations facilitating insulin secretion [236,237]. Despite a rather low abundance of RyRs in pancreatic β-cells  they may be considered key players in the insulin secretory response due to their presence on secretory vesicles  mediating a context-dependent signal for CICR-facilitated insulin release  as well as their large conductance . Furthermore, Epac2 exerts CICR-independent effects by modulating the density of secretory vesicles in close proximity to the PM, possibly via an Epac2/Rap1 (Ras-related protein) signaling pathway [160,242]. Interestingly, Epac2 is a direct target of sulfonylurea anti-diabetic drugs, as revealed in Epac2-deficient mice displaying a diminished sulfonylurea-induced insulin secretion, and shows an attenuated response to tolbutamide, exhibiting its blood glucose-lowering properties .
Regulation by adipose tissue
In addition to the enteroendocrine and adrenal control of GSIS, the adipose tissue is also involved in modulating insulin secretion. While the binding of leptin, the so-called satiety hormone, to its pancreatic receptor (Ob-R)  reduces the release of insulin by inhibiting the transcription  and/or expression of insulin genes  and by stimulating KATP-channels and consequently diminishing Ca2+-influx , adiponectin promotes insulin gene expression and release [248,249].
Circadian regulation of GSIS
Dark/light cycles have also been implicated in modulating insulin secretion in a manner that is dependent on the circadian rhythm regulating genes Clock and Bmal . Pancreatic islets from Clock and pancreas-specific Bmal1 mutant mice showed reduced GSIS at high glucose levels as well as an attenuated response to the GSIS-potentiating effects of incretin mimics and cAMP signaling activators. The diminished GSIS was independent of both the calcium influx and content of insulin RNA and protein in the islets, indicating an unrevealed circadian-regulated aspect of the late-stage of exocytosis.
Modulation of blood insulin levels is the major paradigm in T2DM management. Thus, a better understanding of insulin granule biogenesis and subsequent insulin exocytosis will provide new opportunities for diabetes therapies. Among the various anti-diabetic drugs, i.e. insulin secretatogues such as sulfonylureas and incretin mimetics, insulin sensitizers, e.g. metformin or thiazolidinediones (TZDs) and α-glucosidase inhibitors, those enhancing insulin release by targeting specific players of the insulin signaling cascade are of particular interest as they initiate and/or amplify existing, physiological pathways. Furthermore, this may result in less severe side effects: hypoglycemia caused by sulfonylureas [178,179] versus weight gain caused by TZDs  versus nausea/vomiting caused by GLP-1 analogs . Yet, ‘insulin secretion enhancers' are predominately represented by GLP-1 analogs [253–255], receptor agonists [256–258] or dipeptidyl peptidase IV (DDP4) inhibitors [259–261] but future investigations may also focus on other PKA targets such as synaptotagmin-7 [91,233] and snapin , as well as on Ca2+-flux modulators such as Epac2 /ryanodine receptors  or various SNARE proteins such as syntaxin 1 , SNAP-25 [264,265], munc13  and munc18  as well as at Rab3A  and Rab27 . However, many players and their roles in this highly sophisticated network controlling GSIS remain to be elucidated, especially with respect to the identification and characterization of the specific roles and interactions of granular components involved in the early biogenesis of insulin granules. Conversely, the events that immediately precede exocytosis have been extensively studied. The molecular interactions that cause granule docking and fusion have been probed with ever-finer precision imaging approaches. Nowadays, we even have temporal assembly models and atomic resolution structures to investigate the large protein complexes involved in exocytosis. Signals from pancreatic, enteroendocrine, adrenal and adipose tissues have been shown to exert control over insulin granule exocytosis and future research should reveal whether these pathways may provide opportunities for therapeutic interventions.
Ach, Acetylcholine; APPL, adaptor protein, phosphotyrosine interaction, PH domain, and leucine zipper containing; ARF, ADP-ribosylation factor; BAR, Bin/amphiphysin/Rvs; cAMP, cyclic adenosine monophosphate; CgA, chromogranin-A; CgB, chromogranin-B; CICR, Ca2+-induced Ca2+-release; CPE, carboxypeptidase E; DPP4, dipeptidyl peptidase IV; Epac, exchange protein directly activated by cAMP; FRET, fluorescence resonance energy transfer; GAP, GTPase-activating protein; GEF, guanine nucleotide-exchange factor; GIP, glucose-dependent insulinotropic peptide; GIRK, G-protein-coupled inwardly-rectifying potassium; GLP, glucagon-like peptide; GRP, gastrin-releasing peptide; GSIS, glucose-stimulated insulin secretion; HID, high-temperature-induced dauer formation; ICA, Islet cell autoantigen; IP3, inositol 1,4,5-trisphosphate; ISG, Immature secretory granule; MCH, melanin concentrating hormone; NMR, nuclear magnetic resonance; NPY, neuropeptide Y; NSCCs, nonselective cation channels; NSF, N-ethylmaleimide-sensitive factor; PACAP, pituitary adenylate cyclase-activating polypeptide; PC, prohormone/proprotein convertases; PICK1, protein kinase C-binding protein 1; PIP2, phosphatidylinositol 4,5-bisphosphate; PKC, protein kinase C; PKD, protein kinase D; PLC, phospholipase C; PLD, phospholipase D; PM, plasma membrane; RER, rough endoplasmic reticulum; RRP, readily releasable pool; RyRs, ryanodine-sensitive calcium stores; Sg, secretogranin; SNAP, soluble NSF attachment protein; SNAREs, SNAP-receptor proteins; Syts, synaptotagmins; T2DM, Type 2 diabetes mellitus; TGN, trans-Golgi network; TIRF, total internal reflection fluorescence; TRP, transmembrane receptor potential; TZDs, thiazolidinediones; VGLUT, vesicular glutamate transporter; VIP, vasoactive intestinal peptide; XFEL, X-ray free electron laser.
The Authors declare that there are no competing interests associated with the manuscript.
Research in the authors' laboratories at SBIC and IMCB are supported by an intramural funding from A*STAR Biomedical Research Council.