Pathways in Beta-Cell Stimulus-Secretion Coupling as Targets for Therapeutic Insulin Secretagogues

Insulin secretion is subject to tight control by glucose, other nutrients, neurotransmitters, and hormones. Although numerous and complex, the mechanisms underlying this multifactorial regulation can be schematized by a hierarchical interaction between triggering and amplifying pathways (Fig. 1) (4,5).

When the concentration of glucose increases, β-cell metabolism accelerates, leading to closure of K_ATP channels in the plasma membrane. These channels are composed of the pore-forming K⁺_IR6.2 and the regulatory sulfonylurea receptor 1 (SUR1). Binding of intracellular ATP to K⁺_IR6.2 closes the channel, whereas binding of MgADP to SUR1 opens the channel. The increase in the ATP/ADP ratio resulting from the metabolism of glucose thus closes the channel (Fig. 1). The consequence is a depolarization of the plasma membrane, with opening of voltage-dependent Ca²⁺ channels, acceleration of Ca²⁺ influx, and increase in the concentration of cytosolic free Ca²⁺ ([Ca²⁺]_i) that is necessary and sufficient to trigger insulin secretion (6–10). However, this triggering signal alone is poorly effective. Its efficacy is augmented by an amplifying pathway also using signals issued from glucose metabolism. The nature of these signals and their intracellular targets are still unclear, but a role of ATP and ADP is plausible (11–13). The same dual regulation applies to all nutrients that are actively metabolized and increase the ATP/ADP ratio in β-cells (5). Although no direct evidence is as yet available, both clinical investigation of diabetic patients and experimental studies of animal models suggest that the two pathways, triggering and amplifying, may be impaired in β-cells affected by type 2 diabetes (4).

Importantly, the triggering signal (rise in [Ca²⁺]_i) can also be produced or augmented by mechanisms that are independent of K_ATP channels (Fig. 1). Some hormones and neurotransmitters mobilize Ca²⁺ from intracellular stores (14–16). Agents acting on various ionic channels (e.g., inhibitors of voltage-dependent K⁺ channels) can augment glucose-induced depolarization, thereby potentiating the [Ca²⁺]_i rise (17). Cationic amino acids, like arginine, are poorly metabolized but depolarize β-cells because of their entry in a positively charged form, thus without direct interaction with an ionic channel (5).

Stimulatory hormones and neurotransmitters, such as glucagon-like peptide 1 (GLP-1) and acetylcholine, usually potentiate insulin secretion by a dual action. They moderately increase the triggering signal (rise in [Ca²⁺]_i) through complex, variable, but largely glucose-dependent mechanisms. They also produce major amplifying signals, mainly through activation of protein kinases, in particular protein kinase A (PKA) and protein kinase C (PKC) (14–16,18–19). In addition to PKA, cAMP-regulated guanine nucleotide exchange factors (GEFs, or Epac) might mediate part of the effects of cAMP on insulin secretion (16,20). Activation of PKA or PKC, or of GEF/Epac, augments the efficacy of Ca²⁺ on exocytosis. The biochemical mechanisms of this type of amplification are, however, distinct from those implicated in the amplifying pathway of glucose and other nutrients (5).

Finally, inhibitory hormones and neurotransmitters also act via two pathways. They depress insulin secretion partly by decreasing the triggering signal (via membrane repolarization) and mainly by reducing the efficacy of Ca²⁺ on exocytosis (attenuating pathway operating via kinases or small guanosine triphosphate [GTP]-binding proteins) (21).

Hypoglycemia induced by excessive insulin secretion is a major complication of current pharmacological treatments of type 2 diabetes. Under physiological conditions, it is the triggering pathway that determines whether insulin is secreted. The amplifying pathway serves to optimize the secretory response induced by the triggering signal but does not induce secretion if [Ca²⁺]_i is not increased (5). This strict hierarchy between the two pathways can be perturbed by pathological defects or by drugs. Theoretically, actions on either the triggering or amplifying pathways could result in hypoglycemia. Excessive insulin secretion will occur if a drug increases the triggering signal even when the glucose concentration is low, and does so to such an extent that β-cell [Ca²⁺]_i remains effective on exocytosis in face of the decrease in amplification that automatically accompanies a fall in blood glucose. This is how potent and long-acting sulfonylureas sometimes provoke hypoglycemia. Excessive insulin secretion can also occur if a drug produces an amplifying signal that makes a rise in [Ca²⁺]_i unnecessary to trigger insulin secretion. This is how strong and combined activation of PKA and PKC increases insulin secretion at very low β-cell [Ca²⁺]_i in vitro (22), but it is unlikely that amplifying pathways can ever be stimulated to such an extent in vivo. The risk of a moderate stimulation of amplifying pathways is minimal because excessive decrease in blood glucose automatically stops generation of the triggering signal and thus the insulin-secreting effect of the amplifying drug. This safeguard would however be lost if glucose no longer could modulate the magnitude of the triggering signal because another drug has taken over control of the K_ATP channels. The combination of drugs acting potently on both triggering and amplifying pathways might thus be dangerous.

Figure 1 schematizes the control of insulin secretion in a normal β-cell and indicates six possible pathways or sites that could be targeted by drugs designed to increase secretion. Several strategies can theoretically be followed to reach these targets (Table 1), but we will see that they are not equivalent and that a number of shortcomings often preclude transposition of theory into practice.

Alterations of glucose metabolism often characterize β-cells from animal models of type 2 diabetes and most likely β-cells from type 2 diabetic patients (23). Correction of specific defects would obviously be optimal, but a global improvement of glucose/fuel metabolism in β-cells could also have several advantages: restoration of normal triggering and amplifying signals, stimulation of proinsulin biosynthesis with lesser danger of store exhaustion, and little risk of hypoglycemia if the control steps are not offset.

Theoretically, β-cell metabolism might be boosted by a selective action on an identified perturbed site (e.g., hypo- or hyperactive enzyme due to a mutation or a change in expression), by supply with alternative fuels that bypass a defective enzyme or pathway, and by less specific, indirect measures to accelerate global metabolism. Some of these approaches have already been tested experimentally.

At the entry of glycolysis, glucokinase plays a primary regulatory role in the control of glucose metabolism in β-cells (10). Loss of function and gain of function of glucokinase respectively cause deficient and excessive insulin secretion in type 2 maturity-onset diabetes of the young (MODY2) patients and in certain infants with persistent hyperinsulinemic hypoglycemia (24,25). An allosteric activator of glucokinase (compound RO-28-1675) has been developed recently (26). It activates glucose metabolism and lowers the threshold concentration for glucose-induced insulin secretion in rat islets. It also increases plasma insulin and decreases plasma glucose levels in normal mice and promotes glucose usage in the liver (26). This novel family of drugs opens interesting perspectives for the development of original antidiabetic agents. One should, however, be aware that patients taking excessive doses of such a compound will mimic the activating mutations of glucokinase and thus incur a risk of hypoglycemia.

Glucose-6-phosphatase is essential for glucose production by the liver, but its operation in β-cells creates a futile cycle of glucose phosphorylation and dephosphorylation, with reduction of glucose usage and insulin secretion as consequences (27). The activity of the enzyme is insignificant in normal β-cells, and the possibility that an increase in activity contributes to abnormal insulin secretion in animal models of diabetes remains controversial (28–30). Available drugs inhibiting hepatic glucose-6-phosphatase are inactive on the high enzyme activity in ob/ob mouse islets (31). For this approach to be successful, it should first be established that glucose-6-phosphatase is so overactive in β-cells from type 2 diabetic patients that it eventually impairs oxidative glucose metabolism and that the enzyme can be inhibited independently from its hepatic congener.

Esters of carboxylic metabolites are effective insulin secretagogues in vivo and in vitro (32). In contrast to succinic, glutamic, and pyruvic acid, their methyl or ethylesters enter β-cells, where the carboxylic moiety is metabolized in the Krebs cycle and thus bypasses the early and possibly defective steps of glucose metabolism. Whereas problems linked to the mode of administration in vivo (route, amount) and to the production of methanol from the methylesters can be circumvented (32), it remains to be established that undesirable effects on glucose usage/production by the liver and peripheral tissues do not outweigh favorable effects on diabetic β-cells.

Monosaccharide esters, such as the pentaacetate ester of α-d-glucose, enter β-cells independently of the glucose transporters and stimulate insulin secretion (33). Because it is unlikely that glucose transport ever becomes rate-limiting in human β-cells, the potential interest of these compounds is elsewhere. Only minute amounts (much less than of glucose itself) are indeed sufficient to increase plasma insulin levels in normal rats. Unexpectedly, in vitro stimulation of insulin secretion is also observed with low concentrations of pentaacetate esters of nonmetabolized hexoses such as l-glucose or 2-deoxyglucose. The effects of these ester compounds are not attributed to an increase in β-cell metabolism, which is rather inhibited, but to an interaction with a still unidentified receptor (33). Developments in this area await further studies on the mode of action and potential negative effects of these compounds, as well as demonstration that their insulin-releasing activity is retained in models of type 2 diabetes.

The rise in β-cell [Ca²⁺]_i produced by glucose and other secretagogues is followed by an influx of Ca²⁺ into mitochondria. It has been proposed that, by activating matrix dehydrogenases, Ca²⁺ augments the production of ATP and other putative messenger molecules (34,35). The exit of Ca²⁺ from mitochondria is mediated by a Na⁺/Ca²⁺ exchanger, whose blockade might thus be expected to promote metabolism. Compound CGP 37157 was developed for this purpose and was indeed found to increase mitochondrial [Ca²⁺], ATP production, cytosolic [Ca²⁺]_i, and insulin secretion in INS-1 cells incubated in the presence of glucose (36). It also increased glucose-induced insulin secretion by rat islets and augmented plasma insulin levels at certain time points of a hyperglycemic clamp. Although encouraging, these preliminary results should be confirmed with related compounds, and the consequences of their application for longer periods should be evaluated in primary β-cells, before deciding whether this new approach is worth pursuing. The issue of tissue selectivity may also be difficult to solve.

The binding of sulfonylureas to SUR1 leads to closure of K_ATP channels with subsequent depolarization of the plasma membrane, activation of Ca²⁺ influx, and rise of [Ca²⁺]_i. These drugs mimic the effect of glucose in the generation of the triggering signal, but do so independently of changes in β-cell metabolism. The different sites of sulfonylurea binding to SUR1 and the molecular mechanisms of its transduction into closure of K⁺_IR6.2 have been reviewed recently (37–40). In brief, SUR1 possesses an “A” site to which binds tolbutamide and the half of the glibenclamide molecule containing the sulfonylurea group, and a “B” site to which binds the nonsulfonylurea half of glibenclamide. Binding to one of these two sites is sufficient to produce the same final effects: closure of K_ATP channels, membrane depolarization, and rise in [Ca²⁺]_i.

Whereas it is widely accepted that, unlike glucose, sulfonylureas do not promote proinsulin biosynthesis, it has been suggested that, like glucose, they also stimulate insulin secretion through an amplifying pathway (41,42). This is based on electrophysiological studies in which single β-cells are usually patch-clamped in the whole cell mode (permitting unrestricted exchange between cytoplasm and pipette milieu), and exocytosis of insulin granules is monitored as changes in membrane capacitance. In the presence of fixed [Ca²⁺]_i, intracellularly applied sulfonylureas increase exocytosis (41,42). It is therefore proposed that sulfonylureas penetrate β-cells and interact with SUR1 or a related protein in the membrane of the insulin granules to confer them release competence by facilitating their acidification. Surprisingly, this effect persists in Sur1 KO β-cells, implying that the intracellular binding protein is not SUR1 (43). It would be expected therefore that sulfonylureas retain an effect on insulin secretion—via the intracellular binding sites—in intact β-cells without K_ATP channels.

Figure 2 shows that unstimulated (1 mmol/l glucose) and stimulated (15 mmol/l glucose) insulin secretion was increased by 10 and 50 nmol/l glibenclamide in control mouse islets. In Sur1 KO islets, insulin secretion was already high in 1 mmol/l glucose and was further increased by 15 mmol/l glucose, which reflects the operation of the triggering pathway and the effectiveness of the amplifying pathway of glucose in the absence of SUR1 (44). Glibenclamide, however, was completely ineffective, even at 1 μmol/l, a concentration that largely exceeds those reached in vivo (Fig. 2). Extracellular tolbutamide was also without effect on insulin secretion in islets without K_ATP channels due to a knockout of Sur1 (44,45) or K⁺_IR6.2 (46) at both low and high glucose levels.

In conclusion, intracellularly applied sulfonylureas interact with a binding site probably on the insulin granule membrane (42). It is plausible that this site is somehow implicated in the response to physiological secretagogues, but it does not seem relevant to the therapeutic action of sulfonylureas.

The nonsulfonylurea (benzamido) part of glibenclamide, known as meglitinide or HB-699, possesses blood glucose–lowering properties that have been attributed to stimulation of insulin secretion (47). Twenty years ago it was shown that meglitinide mimics the sequence of events by which tolbutamide and glibenclamide trigger insulin secretion (48). A number of other nonsulfonylurea compounds have been developed more recently, some of which are already in clinical use (49). The best known are repaglinide (AGEE-623), nateglinide (A-4166), and mitiglinide (KAD-1229 or S-21403). They are functionally related but structurally different. The common family name of “glinides” derives from their trade name but does not refer to any specific chemical structure. Except for repaglinide, it is incorrect to call them “meglitinide analogs” or “benzamido compounds.” Meglitinide and repaglinide bind to the “B” site in SUR1, whereas nateglinide and mitiglinide bind to the “A” site (39,40). Whatever the binding site, their eventual effect is similar to that of sulfonylureas: depolarization of β-cells, with subsequent rise in [Ca²⁺]_i and triggering of insulin secretion.

Measurements of β-cell capacitance have also been used to assess whether glinides amplify insulin secretion at an intracellular site. Nateglinide, but not repaglinide, produces such an effect (50), but, again, this conclusion is not supported by experiments using intact islets. Thus, nateglinide, repaglinide, and mitiglinide do not increase insulin secretion from islets depolarized with KCl (51). As for sulfonylureas, I conclude that the intracellular effect of some glinides, observed in dialysed single β-cells, is not therapeutically important.

In conclusion, the value of the glinides as insulin secretagogues does not reside in an original mode of action but in pharmacokinetic properties associated with a rapid onset of action and a rapid reversibility of action, at least for mitiglinide and nateglinide. Both also display an advantageous greater selectivity for SUR1 than SUR2A or SUR2B (38) and, hence, for K_ATP channels of the β-cell. In this context, it has been speculated that drugs closing K_ATP channels in β-cells (SUR1/K⁺_IR6.2) and opening K_ATP channels in vascular muscles (SUR2B/K⁺_IR6.2) would be interesting to treat type 2 diabetic patients with hypertension. Agents like minoxidil sulfate (52) and MCC-134 (53) have such a profile, but their opposite effects on the two isoforms of the channel do not occur within the same concentration range. Blockade of the β-cell K_ATP channel requires concentrations that largely exceed those needed to produce vasorelaxation.

Many structurally different drugs used for the treatment of diseases other than diabetes occasionally produce hypoglycemia. In vitro studies have shown that they increase insulin secretion by closing β-cell K_ATP channels through a direct interaction with K⁺_IR6.2. Among these drugs are antimalarial quinolines such as quinine and mefloquine (54,55), antibacterial fluoroquinolones such as norfloxacin and lomefloxacin (56), antiarrhythmic agents such as disopyramide and cibenzoline (57–59), and others (60).

More interest has been paid to drugs with an imidazoline structure (phentolamine, antazoline, midaglizole), which also inhibit K_ATP channels in β-cells (61–63) by interacting with K⁺_IR6.2 (64). This effect largely explains their stimulation of insulin secretion. However, efaroxan and novel imidazoline compounds also or exclusively act on an amplifying pathway, as will be discussed below.

From a mechanistic point of view, it is unimportant whether drugs close K_ATP channels directly by an interaction with the pore formed by K⁺_IR6.2 or indirectly by an interaction with the regulatory subunit SUR1. The net result is the same: membrane depolarization, influx of Ca²⁺, and generation of the triggering signal. The major difference, however, is the distribution of the two targets. The much more restricted distribution of SUR1 than K⁺_IR6.2 considerably increases the tissue specificity of the drugs acting through it. Moreover, drugs closing K⁺_IR6.2 directly usually also affect other channels (61).

In vitro, all depolarizing agents increase insulin secretion regardless of the mechanism underlying the depolarization. However, this approach has limited clinical applications notably because of the difficulty to achieve tissue selectivity. A further underestimated problem is that the depolarizing action of a substance may be markedly dependent on the electrical resistance of the β-cell membrane, which is essentially determined by the extent of K_ATP channel closure. One instructive example is that of arginine and other cationic amino acids that depolarize β-cells because of their entry in a positively charged form (Fig. 1). The effects of arginine on membrane potential and [Ca²⁺]_i increase with the glucose concentration (65), which partially accounts for the glucose dependency of arginine-induced insulin secretion. Defective closure of K_ATP channels in diabetic β-cells may be expected to impair the production of a triggering signal by agents that depolarize by activating an inward current. The second mechanism explaining the glucose dependency of insulin secretion induced by arginine and related agents is the amplifying action of glucose (5).

The major voltage-dependent K⁺ channel in β-cells, Kv 2.1, participates in the repolarization of the membrane during spike generation (17). Its blockade augments the influx of Ca²⁺ and has long been shown to increase insulin secretion (66). The wide tissue distribution of these channels is a major limitation in their use as a target for insulin-secreting drugs in vivo. However, because the activity of these channels may be subject to modulation by different subunits and to regulation by hormonal or metabolic signals, it has been speculated that the development of selective blockers of β-cell voltage-dependent K⁺ channels is not unrealistic (17).

The main voltage-dependent Ca²⁺ channels in β-cells are of the l-type and are inhibited by dihydropyridines, which decrease insulin secretion by lowering the triggering signal [Ca²⁺]_i (67). Some dihydropyridines (Bay K8644 or CGP2832) instead act as agonists. Almost 20 years ago they were shown to increase Ca²⁺ influx in β-cells and insulin secretion in vitro (60). Since then, no progress has been made in the field and, because of insufficient tissue selectivity, the development of this class of compounds for clinical use has been discontinued.

β-cells are equipped with many other ionic channels that may subtly modulate the changes in membrane potential induced by glucose and other secretagogues. None appears to have a sufficiently important role in stimulus-secretion coupling or a sufficiently restricted tissue distribution to qualify as a target for insulin-secreting drugs.

The [Ca²⁺]_i rise induced by glucose and other secretagogues must be counterbalanced by Ca²⁺ sequestration in intracellular organelles and by Ca²⁺ extrusion to the extracellular space. These tasks are achieved by several Ca²⁺-ATPases and Na⁺/Ca²⁺ exchangers (68,69,70). Inhibition of one or several of these processes is followed by a variable rise of [Ca²⁺]_i and increase in insulin secretion (70,71). However, should effective blockers of the plasma membrane Ca²⁺-ATPase and Na⁺/Ca²⁺ exhanger become available, tissue selectivity will still remain a serious issue. Moreover, the endoplasmic reticulum stress provoked by blockade of intracellular Ca²⁺-ATPases is likely to have deleterious consequences (72) that preclude this approach.

All measures resulting in acceleration of β-cell metabolism (see above) increase insulin secretion by a dual action on the triggering and amplifying pathways. It is accepted that the ATP/ADP ratio serves as a second messenger in the generation of the triggering signal not only because its changes occur over a wide physiological range of glucose concentrations (73), but also because the transducing molecules (SUR1 and K⁺_IR6.2) are known. In contrast, the molecular mechanisms of the amplifying pathway remain unclear, and the possibility that several factors are involved cannot be ruled out (Fig. 1). Our proposal (4,73) that variations in the ATP/ADP ratio (or associated changes in AMP or the GTP/GDP ratio) are implicated suffers from one shortcoming: the effector molecules have not been identified, so that suggestions for the development of insulin-releasing drugs specifically acting through that pathway are premature. One candidate effector is the SUR1-like protein of the insulin granule membrane (42). Whereas this protein is unlikely to contribute to the insulin-releasing action of sulfonylureas (see above), it might participate in the amplifying action of glucose and could thus become the target of novel drugs. SUR1 itself is also present in the granule membrane (74), but the amplification of insulin secretion by nutrients is not altered in its absence (44). Another possible target is AMP-activated protein kinase.

Islet AMP kinase activity is inhibited by glucose, which raises the possibility that the enzyme participates in a long- or short-term control of β-cell function, including insulin secretion (75). Activation of AMP kinase by AICAR or overexpression of a constitutively active form of AMP kinase-α1 inhibited glucose-stimulated insulin secretion but also virtually abolished glucose metabolism and the [Ca²⁺]_i rise in β-cells (75), making it difficult to identify the site(s) of action. Conversely, a dominant negative form of AMP kinase increased basal insulin secretion, but this increase was Ca²⁺-independent (75), which is in sharp contrast with all effects of glucose (4). Thus, it is premature to suggest that β-cell selective inhibitors of AMP kinase might become insulin secretagogues.

On the other hand, because the widely used metformin can activate AMP kinase (76), it is important to discuss whether the drug exerts deleterious effects in β-cells. Culture of human islets for 16 h in the presence of 1 mmol/l metformin (∼100-fold the plasma concentration in patients) (77) virtually abolished glucose-induced insulin secretion (78), whereas no effect was observed after culture with 20 μmol/l metformin (79). Permeation of metformin into cells is admittedly slow, but the key unanswered question is whether in vitro exposure to a very high or to a therapeutic concentration of metformin for a few hours only is equivalent to chronic exposure to the drug in vivo. When patients are treated with metformin, their plasma insulin level often decreases. This is attributed to improvement of the metabolic state rather than to inhibition of insulin secretion by the drug (77). This interpretation is supported by a study showing that 2 weeks of metformin administration to subjects without glucose intolerance did not change their insulin response to a hyperglycemic clamp (80).

Glucocorticoids impair glucose homeostasis mainly by opposing the effects of insulin on hepatic glucose production and peripheral glucose uptake. Although plasma insulin is elevated in glucocorticoid-induced insulin resistance, the direct effect of glucocorticoids on insulin secretion is inhibitory (81). Overexpression of the glucocorticoid receptor in β-cells leads to hyperglycemia and hypoinsulinemia after several months (82). The mechanisms of this inhibition of insulin secretion may be multifactorial (82), but interference with the amplifying action of glucose is probable (81). Local transformation of inactive cortisone into active cortisol by 11β-hydroxysteroid dehydrogenase type 1 could play a pathogenic role in the metabolic syndrome, and inhibition of the enzyme is envisaged as a potential therapy in type 2 diabetes and obesity (83,84). 11β-Hydroxysteroid dehydrogenase type 1 is present in mouse and human islets (85), and its activity is increased in proportion to hyperglycemia in the islets from diabetic ZDF fa/fa rats (86). Inhibition of the enzyme in mouse islets prevents precursors of active corticosteroids from decreasing insulin secretion in vitro (85). This approach deserves further investigation.

Drugs are being developed with the aim of increasing the action of Ca²⁺ on contractile proteins, in particular to improve myocardial performance (87). One of these, pimobendan, increases insulin secretion by isolated rat islets. Because it remains effective when β-cell [Ca²⁺]_i is clamped by high KCl, the drug indeed behaves like a Ca²⁺ sensitizer (88). However, this study did not investigate whether the effect could be mediated by the phosphodiesterase inhibition that pimobendan can also cause (87). Effects of levosimendan, another available Ca²⁺ sensitizer, on the endocrine pancreas have not been reported.

Cyclic AMP is a potent amplifier of insulin secretion (14,18,89). Inhibition of its degradation by methylxanthines has long been known to increase plasma insulin concentrations (90). In β-cells, this degradation is mainly achieved by phosphodiesterase 3B, the inhibition of which strongly augments insulin secretion in vitro (89). Conversely, overexpression of phosphodiesterase 3B in β-cells reduces in vivo and in vitro insulin secretion in response to glucose and glucagon-like peptide-1 (GLP-1) and impairs glucose tolerance (91). Unfortunately, phosphodiesterase 3B is present in many other tissues, including hepatocytes and adipocytes where its blockade exerts anti-insulin effects. There is presently no evidence that cAMP degradation can be selectively prevented in β-cells by pharmacological agents (89). Hormones acting on β-cell membrane receptors linked to adenylyl cyclase are more promising insulin secretagogues (see below).

As already mentioned, the increase in insulin secretion produced by cAMP is not exclusively mediated by PKA but also involves GEF/Epac (20). Novel cAMP analogs activate GEF/Epac selectively and increase insulin secretion in vitro (16,43). However GEF/Epac is operative in other tissues, and it is uncertain whether some of these analogs will show enough tissue selectivity to be of therapeutic interest.

The interest in imidazoline compounds (such as phentolamine) as potential insulin secretagogues started with the idea that their ability to block α₂-adrenoceptors could relieve diabetic β-cells from a tonic sympathetic inhibition (92,93). Subsequently, it was shown that the increase in insulin secretion that phentolamine, efaroxan, and other imidazolines produce in vitro and in vivo is largely mediated by a blockade of K_ATP channels in β-cells (see above) rather than of α₂-adrenoceptors (61–63,94).

However, different experimental approaches and the development of novel compounds have shown that the mode of action of imidazolines is complex. Compound KU14R, the imidazol analog of efaroxan, does not interfere with the action of the latter on K_ATP channels, yet it inhibits its action on insulin secretion (95). Compound RX871024 increases insulin release from permeabilized β-cells and from islets depolarized with KCl, two preparations in which [Ca²⁺]_i is clamped (96). A novel imidazoline (BL11282) has no effects on K_ATP channels but increases insulin secretion in vitro (97). Three compounds, RX871024 (98), efaroxan (N.G. Morgan, personal communication), and BL11282 (S. Efendic, personal communication) increase insulin secretion from β-cells deficient in K_ATP channels. All these observations point to an intracellular site of action of imidazolines. This site is still unidentified and has tentatively been designated “I₃-imidazoline binding site” (62). A number of experimental arguments suggest that it is involved in the amplification of insulin secretion by PKA and mainly PKC (62,97) and, thus, not in the amplifying pathway of glucose, which is independent of these kinases (4). Although structurally unrelated to imidazolines, β-carbolines might be endogenous ligands of the “imidazoline binding sites.” Interestingly, two β-carbolines, harmane and pinoline, possess insulin-releasing properties in vitro (99).

In summary, the family of insulin-releasing imidazolines is functionally heterogeneous. Older members (phentolamine, antazoline) owe their property to blockade of K_ATP channels (triggering pathway), whereas newer ones (BL11282) increase the action of Ca²⁺ on exocytosis (amplifying action). Because these novel compounds do not produce a triggering signal, their effect on insulin secretion is more strongly glucose dependent (97). However, the advantage of this gain in safety could be cancelled by loss of tissue specificity. Identification of their intracellular site of action (I₃-site) and investigation of their possible effects on kinases in other tissues are necessary before deciding whether imidazolines have a future as insulin secretagogues.

Beta-Cells are equipped with a number of inhibitory receptors whose activation decreases insulin secretion. The major ones are the α₂-adrenergic, galanin, and somatostatin receptors (100). Their inhibitory effects are mediated by complex mechanisms that include partial repolarization with a small decrease in [Ca²⁺]_i, inhibition of adenylyl cyclase with suppression of the amplifying effects of cAMP, as well as a poorly explained but major interference with the action of Ca²⁺ on exocytosis (21).

The use of antagonists of inhibitory receptors could be justified if tonic activation of these receptors contributed to the impairment of insulin secretion in pathological states. Studies in the 1970s suggested that excessive adrenergic activity could impede insulin secretion in type 2 diabetic patients (92,93), but this was not confirmed (101). Moreover, the plasma insulin increase that some α₂-adrenoceptor blockers (e.g., phentolamine) produce is now ascribed to a direct inhibition of K_ATP channels (see above). Finally, because adrenergic inhibition of insulin secretion is a safeguard against hypoglycemia during exercise, blockade of this action may not be without risk.

Effects of antagonists of somatostatin or galanin receptors on insulin secretion have not been reported. Mice with a general knock-out of the somatostatin receptor type 5, the major type in β-cells, do not show significant alterations of in vivo insulin secretion and glucose homeostasis. However, when their pancreas is perfused in vitro, insulin secretion is increased compared with controls, particularly in the basal state (102). Inactivation of the galanin gene unexpectedly reduces insulin secretion in vivo and impairs glucose tolerance. The inhibition of insulin secretion persists in vitro (103). Overall, there are no convincing arguments suggesting that blockade of somatostatin or galanin receptors in β-cells could be useful for increasing insulin secretion in type 2 diabetic patients.

Beta-Cells possess numerous membrane receptors for stimulatory hormones or neurotransmitters, and several have already been considered as potential therapeutic targets, but only few may qualify.

Acetylcholine is a physiologically important and potent amplifyer of insulin secretion, but its effects in β-cells are mediated by muscarinic receptors of the M3 type, whose characteristics and tissue distribution are such that selectivity of an agonist for insulin secretion is most unlikely (15).

Extracellular glutamate increases insulin secretion from the perfused pancreas by acting on AMPA receptors, and improves insulin secretion and glucose tolerance in vivo in rats (104). However, this approach is unlikely to be useful because of rapid desensitization of the receptor and undesirable extrapancreatic effects of glutamate or analogs.

Activation of purinergic receptors of the P2Y type increases insulin secretion from rodent and human islets. It is unclear, however, whether agonists with sufficient β-cell selectivity can be designed (105).

Pituitary adenylate cyclase–activating polypeptide (PACAP) exerts its effects through a PACAP-preferring receptor (PAC1) and two receptors shared by vasoactive intestinal peptide (VPAC1 and VPAC2). VPAC1 mediates stimulation of glucose production by the liver, whereas VPAC2 in β-cells mediates stimulation of insulin secretion (106). A synthetic agonist selective for VPAC2 increased plasma insulin and improved glucose disposal in normal rats (107). Confirmation of these effects in models of type 2 diabetes and demonstration that no side effects occur remain to be seen.

Glucose-dependent insulinotropic polypetide (GIP) and GLP-1 are the two major intestinal hormones released during meals that strongly potentiate nutrient-induced insulin secretion (incretin effect) (108). In type 2 diabetic patients, secretion of GIP is normal or slightly increased, but the efficacy of the hormone on β-cells is inexplicably impaired, which renders GIP administration inefficient treatment (109,110).

The situation is very different with GLP-1. Its secretion is impaired but its action is preserved in type 2 diabetic patients. Therapeutic substitution may thus be justified and has indeed already proved successful, in particular when the short half-life of the peptide is extended (109–111). Marked glucose dependency is an important feature of GLP-1 effects on insulin secretion. The increase in cAMP produced by GLP-1 activates a PKA- and GEF/Epac-dependent amplifying pathway (14,16,19) that augments the efficacy of [Ca²⁺]_i, the rise of which depends on glucose. Second, despite some inhibitory action on K_ATP channels, GLP-1 hardly produces a triggering signal at low glucose for two reasons. It mainly acts on ionic channels (voltage-dependent Ca²⁺ and K⁺ channels) that become operative only when the β-cell membrane has been depolarized by glucose (19). Moreover, the mobilization of intracellular Ca²⁺ that GLP-1 promotes results from a Ca²⁺-induced Ca²⁺ release, which itself depends on glucose-induced influx of Ca²⁺ from the extracellular medium (14,16). This glucose dependency of GLP-1 stimulation of insulin secretion might however be lost if the hormone were combined with a potent and long-acting sulfonylurea.

Several nuclear receptors, including the liver X receptor (LXR) and the peroxisome proliferator–activated receptors (PPARs), play a central role in lipid and carbohydrate metabolism in liver, muscles, and adipose tissue (112,113). Their importance for β-cell function is less clear. However, preliminary evidence suggests that activation of LXR in β-cells may influence insulin secretion (114).

PPAR-α and -γ are moderately expressed in rodent and human islets (30,115–117) and may exert a long-term influence on insulin secretion by controlling the expression of enzymes involved in fuel metabolism (118). Their role in β-cell adaptation to metabolic perturbations (e.g., insulin resistance) is under investigation, but it is known that PPAR-α is downregulated and PPAR-γ upregulated in islets from Zucker diabetic fatty rats (116). Similar changes are induced by exposure to hyperglycemia in vivo or in vitro (30,117). Two families of clinically useful compounds, fibrates and thiazolidinediones, are synthetic ligands of PPAR-α and -γ, respectively. Although they are not prescribed with the specific aim of changing insulin secretion, such changes do occur. It is not always easy to determine if these effects are indirect or direct.

An acute (<1 h) increase of insulin secretion by fibrates has been observed in vitro, but neither the mechanisms nor the specificity of the effect has been assessed (119). Culture of normal rat islets with fibrates increased PPAR-α expression and exerted variable effects on enzymes of pyruvate metabolism and fatty acid oxidation, as well as on insulin secretion (116,119). The interpretation of these experiments is problematic because supratherapeutic drug concentrations were used.

In vivo treatment of normal rats with a PPAR-α agonist for 24 h increased PPAR-α and pyruvate dehydrogenase kinase 4 expression in islets, but did not affect glucose-induced insulin secretion ex vivo (120). A similar treatment exerted distinct effects in two models of insulin resistance. Hypersecretion of insulin by islets from high-fat-fed rats was reversed (120), whereas that from pregnant rats was unaffected (121), suggesting an extra-islet site of action of the drug.

Treatment with thiazolidinediones is often accompanied by improved glucose-stimulated insulin secretion in type 2 diabetic patients and various animal models of the disease (118). In vitro studies are contradictory. Troglitazone and rosiglitazone have been reported to induce a glucose-dependent increase in insulin secretion by HIT cells, rat islets, or the perfused rat pancreas (122–124). These acute effects were too rapid (<1 h) for nuclear receptor involvement. They were variably attributed to binding to SUR1 without closure of the K_ATP channel, direct activation of Ca²⁺ channels, and stimulation of phosphatidylinositol (PI) 3-kinase (122–124). A more recent study, however, did not find any acute effect of rosiglitazone on insulin secretion by perifused rat islets (125). There also seems to be agreement that culture (24–48 h) of normal human or rat islets with thiazolidinediones did not influence the subsequent secretory response (115,126). Finally, insulin secretion was not altered in mice with a selective knockout of PPAR-γ in β-cells (127). When these mice were placed on a high-fat diet, the beneficial effects of rosiglitazone on insulin resistance and glucose tolerance was preserved despite the absence of PPAR-γ in β-cells (127). It is thus unlikely that PPAR-α or -γ qualify as primary targets for drugs destined to stimulate insulin secretion.

The experimental work performed in my laboratory was supported by the Université Catholique de Louvain, Fonds National de la Recherche Scientifique à Bruxelles, Direction Générale de la Recherche Scientifique de la Communauté Française de Belgique et Programme des Pôles d’attraction interuniversitaire–Politique scientifique Fédérale.

I am grateful to J. Bryan for providing Sur1KO mice and to J. Bryan, S. Efendic, N.G. Morgan, U. Quast, and M.C. Sugden for discussions and communication of unpublished data. I thank V. Lebec for editorial assistance.