Recent years have witnessed extensive progress in our understanding of the physiology and pharmacology of ATP-sensitive potassium (KATP) channels. The cloning of the sulphonylurea receptor (SUR1) in 1995 [1], and soon afterwards of its binding partner, the inwardly rectifying potassium channel Kir6.2 [2, 3], opened the way for an integrated investigation into KATP channel structure and function, combining techniques drawn from molecular biology, electrophysiology, biochemistry and transgenic/knockout technology (reviewed in [4, 5, 6]). As a result, we are now beginning to recognise the physiological roles of KATP channels in a range of tissues, and the molecular basis for the observed variation in their pharmacological properties.

Therapeutic modulation of beta cell KATP channel activity has been used for many years to enhance insulin release in subjects with Type 2 diabetes [4, 5, 6]. Sulphonylureas and the newer class of 'glinides' stimulate insulin secretion by interacting with SUR1, thereby bringing about KATP channel closure and increased electrical activity in the beta cells. However, KATP channels have been described in a number of extrapancreatic tissues, including cardiac, skeletal and smooth muscle and some neurones [4, 5, 6, 7, 8, 9, 10, 11, 12, 13], and are blocked to varying degrees by the different sulphonylureas and glinides used in clinical practice. beta cell-selective and non-selective agents are generally used interchangeably in the treatment of diabetes, because studies have not conclusively shown that blocking extrapancreatic KATP channels has a deleterious effect. As we are now beginning to understand the physiological roles of non-beta cell KATP channels and the molecular basis for tissue-selective KATP channel inhibition, perhaps the time has come to revisit this issue. This review will therefore focus on the molecular mechanisms of action of sulphonylureas and related drugs, and the possible consequences of non-selective KATP channel inhibition, based on the knowledge we have gained from studies on recombinant KATP channels and KATP channel-deficient mice.

The unrivalled role of KATP channels in beta cell physiology

Although KATP channels have been identified in a range of muscular, neuronal and neuroendocrine tissues, their function is best understood in pancreatic beta cells where they are involved in coupling membrane electrical events and insulin release to changes in the plasma glucose concentration [14, 15]. The release of insulin-containing secretory granules is triggered by an increase in the local cytoplasmic Ca2+ concentration, and occurs when Ca2+ enters the beta cell through Ca2+ channels in the surface membrane. The opening of these channels is controlled by the membrane potential, which is itself determined by the balance between the inward and outward currents (Fig. 1). At low glucose concentrations outward currents are dominant, due to the high rate of K+ efflux through open KATP channels. The net efflux of positively charged ions sets up a membrane potential of about –70 mV (inside negative). At high glucose levels the KATP channels close, the magnitude of the K+ current falls below that of the inward currents, and the membrane potential drifts to less negative values (depolarisation). Voltage-gated Ca2+ channels open as the membrane depolarises, allowing Ca2+ entry and the stimulation of insulin secretion.

Fig. 1.
figure 1

KATP channel activity modulates insulin secretion in pancreatic beta cells. Glucose-dependent KATP channel closure results in membrane depolarisation and insulin release, which are further enhanced by the addition of arginine

Full size image

A principal determinant of KATP channel activity is believed to be the intracellular concentrations of the adenine nucleotides, ATP and ADP. These nucleotides have opposing actions on KATP channels, with ATP acting as a channel blocker, and MgADP as a channel opener [4, 5, 6]. In pancreatic beta cells, the metabolic rate is responsive to the plasma glucose concentration, resulting in enhanced glycolytic and mitochondrial generation of ATP from ADP at high glucose levels [16, 17, 18]. The increase in ATP and concomitant fall in ADP concentration therefore combine to close the KATP channels, triggering Ca2+ entry and insulin release.

The glucose-dependent reduction in K+ efflux also enables beta cells to respond to small inward currents, such as that generated by arginine uptake (Fig. 1). Transport of positively charged arginine molecules by the cationic amino acid transporter is electrogenic, i.e. there is a net charge movement across the cell membrane associated with arginine uptake, producing a small depolarising current [19]. Arginine-stimulated insulin release is glucose-dependent because the transporter current is very small and is drowned by the opposing KATP current at low glucose concentrations. It only becomes an important player when the KATP channels are largely closed (at raised glucose levels), and can then enhance depolarisation and insulin release. Although it is not known whether other electrogenic transporters play any role in beta cells, sodium glucose cotransporters in GLP-1 secreting cells exert a similar effect, providing a glucose-dependent inward current that enhances depolarisation and GLP-1 release following KATP channel closure [20, 21].

Molecular architecture of KATP channels

The K+ conducting pore of KATP channels is formed from four inwardly-rectifying potassium (Kir) channel subunits, either Kir6.1 or Kir6.2 [2, 3, 22, 23, 24]. Kir6.2 forms the KATP channel pore in the majority of tissues, but expression of the alternative pore subunit, Kir6.1, e.g. in vascular smooth muscle and astrocytes, generates KATP channels with different pore properties and nucleotide sensitivity [25, 26, 27]. Kir subunits have intracellular N and C termini, and two transmembrane helices separated by a pore loop (Fig. 2). Isolated Kir6.2 tetramers are retained in the endoplasmic reticulum by an Arg-Lys-Arg (RKR) motif located in the C-terminus, but are trafficked to the surface membrane on co-assembly with sulphonylurea receptors [28]. Kir6.2 channels can also be engineered to reach the surface membrane without SUR by truncation of the RKR-containing C-terminus, a feature that has been exploited to investigate the intrinsic properties of the Kir6.2 subunit [29]. One of the most notable properties of truncated Kir6.2 subunits, expressed without SUR, is that they are blocked by adenine nucleotides (ATP>ADP) [29, 30].

Fig. 2.
figure 2

Membrane topology of Kir6.2 and SUR1 subunits. The transmembrane and intracellular domains of SUR1 and Kir6.2 are indicated. The cytoplasmic linkers are depicted as separate intracellular domains, by analogy with the structure of MsbA [163]. Regions implicated in the binding and action of sulphonylureas are shown in grey, and the position of residue S1237 is indicated by a white circle. TMD = transmembrane domain; NBD = nucleotide binding domain

Full size image

In fully-assembled KATP channels, each of the four Kir6 subunits is associated with a regulatory sulphonylurea receptor (Fig. 2) [23, 24]. SURs originate from two separate genes and occur in several alternatively-spliced isoforms. SUR1 has been identified in pancreatic beta and alpha cells, GLP-1 secreting intestinal L-cells, and neurones (e.g. in the ventromedial and mediobasal hypothalamus, substantia nigra and myenteric nervous system) [1, 12, 20, 31, 32, 33]. SUR2 is predominantly found in muscle, and exists in two important alternatively-spliced isoforms: SUR2A in skeletal and cardiac muscle, and SUR2B in smooth muscle [25, 34, 35]. Although additional splice variations of both SUR1 and SUR2 have been described, their physiological roles remain unclear [36, 37, 38, 39, 40]. SURs are members of the ATP binding cassette (ABC) transporter family. They have 17 transmembrane helices (TMs 1–17) arranged in three domains (TMD0, TMD1 and TMD2), containing five, six and six helices, respectively. They also possess two large intracellular nucleotide binding domains (NBDs), the first located between TMDs 1 and 2, and the second lying distal to TMD2 [41, 42].

The sulphonylurea receptors endow KATP channels with sensitivity to the stimulatory actions of Mg-nucleotides and KATP channel openers (e.g. diazoxide, nicorandil) and the inhibitory effects of sulphonylureas and glinides [4, 5, 29, 34, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56]. Their NBDs contain consensus nucleotide binding motifs—the Walker A and Walker B motifs—which, in other ABC transporters, are the sites of MgATP binding and hydrolysis [58, 59]. In SUR, binding of MgADP at NBD2 is believed to underlie nucleotide-dependent channel activation [60, 61, 62, 63], and it has been suggested that ADP shuttling from mitochondria to KATP channels might be facilitated by enzymes such as adenylate kinase and creatine kinase, which have been found to form physical associations with SUR2 [64, 65]. In biochemical studies, however, it has also been shown that the NBDs of SUR can hydrolyse MgATP [62, 63, 66, 67, 68]. This would generate MgADP at NBD2 and presumably therefore promote channel opening, raising the possibility that channel activity could also be regulated in vivo by physiological modulation of the ATPase activity of NBD2 [62].

Human KATP channel mutations associated with hyperinsulinism and diabetes

A number of polymorphisms in Kir6.2 and SUR1 subunits have been detected in the human population. Many have no effect on KATP channel function, but others have been associated with hyperinsulinism and diabetes. The first pathogenic mutations in SUR1 to be discovered were in subjects with congenital hyperinsulinism (CHI or persistent hyperinsulinaemic hypoglycaemia of infancy, PHHI) [56, 69]. This condition presents in early life with persistent life-threatening hypoglycaemia and raised insulin concentrations, and in the most severe cases is treatable only by sub-total pancreatectomy. Pancreatic beta cells from these subjects show a lack of KATP channel activity in the surface membrane, causing continuous membrane depolarisation, persistently raised intracellular calcium concentrations and unregulated insulin release [70]. Over 40 disease-causing mutations in SUR1, and 4 in Kir6.2 have been described, accounting for about 50% of the known cases of CHI [5, 6, 71, 72, 73]. Further characterisation of the SUR1 mutations has shown that they result in premature stop codons, trafficking defects or loss of KATP channel sensitivity to the endogenous activator, MgADP [56, 69, 71, 72, 74, 75, 76, 77]. The loss of MgADP sensitivity prevents KATP channels from opening during hypoglycaemia, and the association of such mutations with CHI supports the idea that the crucial signal opening KATP channels during metabolic stress is the nucleotide-dependent channel activation mediated via SUR1, rather than the falling ATP concentration [56].

A mutation in SUR1 (E1506K) that causes mild autosomal dominant CHI in infants, has also been found to cause autosomal dominant Type 2 diabetes in adult life [78, 79]. This mutation in NBD2 interferes with channel activation by MgADP, accounting for the CHI phenotype [78]. By puberty, however, a severe defect in glucose-stimulated insulin secretion could be measured in both normoglycaemic and diabetic carriers of the mutation [79]. The maintenance of normoglycaemia in many of these subjects indicates the physiological importance of KATP channel independent pathways in the beta cell such as incretin-induced secretion and vagal stimuli. The development of diabetes was associated with obesity in three out of four subjects, and probably resulted from an increased insulin requirement that could not be met by the failing beta cells. Although the mechanisms underlying beta cell failure in this group are not clear, an increased rate of apoptosis, perhaps as a result of the chronic membrane depolarisation, has been proposed. If this is true, it could have implications for chronic therapy with potent long-acting sulphonylureas, which might be predicted to induce a similar state of persistent beta cell depolarisation.

A number of recent studies have also demonstrated a strong association between Type 2 diabetes and the common KATP channel polymorphism, Kir6.2-E23K [80, 81, 82]. Functional studies revealed that recombinant KATP channels containing the Kir6.2-E23K mutation spend more time in the open state, and have reduced sensitivity to ATP inhibition and enhanced activation by MgGDP [83, 84]. In human physiological studies the E23K variant was associated with impaired insulin secretion and glucagon suppression during an OGTT and increased BMI, but normal insulin release in response to intravenous glucose [85, 86]. One possible explanation for the diabetic tendency might be that release of incretins in response to oral glucose is also impaired, because the KATP channels in GLP-1 secreting L-cells have the same molecular composition as those in beta cells [20]. Why the beta cells show apparently normal glucose sensitivity in vivo, despite the KATP channel defect, is unclear. Impaired secretion of the L-cell hormones, GLP-1 and Peptide YY, and alterations in hypothalamic KATP channel activity involved in feeding behaviour might also contribute to the development of obesity [86].

Sulphonylurea action

Drugs that bind to sulphonylurea receptors and close KATP channels have been used for many years to stimulate insulin release in subjects with Type 2 diabetes. Although the first and second generation agents (e.g. tolbutamide, chlorpropamide, glibenclamide, gliclazide, glimepiride) possess a sulphonylurea group, this is not essential for drug activity. Indeed, meglitinide and its derivatives, which are structurally related to the non-sulphonylurea half of glibenclamide, form a new class of therapeutic KATP channel inhibitors, the glinides. Drugs within this category (e.g. meglitinide, nateglinide, repaglinide and mitiglinide) do not, however, seem to interact with identical regions of the sulphonylurea receptor.

Studies comparing the binding and functional effects of different KATP channel inhibitors have been carried out using recombinant KATP channels expressed in Xenopus oocytes and mammalian cell lines. Channels formed by coexpression of Kir6.2 subunits with either SUR1, SUR2A or SUR2B, replicate the properties of native KATP channels from beta cells, cardiac and smooth muscle, respectively, and express at high levels. They are amenable to genetic manipulations, facilitating investigations into the molecular mechanisms of drug action. Measurement of the current flowing through patches of membrane taken from oocytes expressing Kir6.2/SUR complexes, provides a direct measure of the K+ flux and therefore of the number of open channels (typically thousands) in the patch of oocyte membrane. A reduction in the current is observed when agents close the KATP channels, mirroring the pharmacological action of the drugs in vivo [4].

With the first studies of tolbutamide action on Kir6.2/SUR1 currents in oocyte membranes, it became apparent that the dose inhibition curve reflected block at two independent binding sites with very different affinities (Fig. 3) [87]. Low concentrations of the drug, causing high-affinity inhibition, only reduced the current amplitude by 50 to 75%. A flat section of the curve was then followed by a second, low-affinity, component of block, occurring at much higher drug concentrations. The high-affinity component reflects drug binding to the sulphonylurea receptor and occurs in the range of concentrations used therapeutically. The low-affinity block is only observed at suprapharmacological drug levels, and has been attributed to direct interaction of the compound with the Kir6.2 subunit [87, 88]. Recognition of the low-affinity component has clarified the interpretation of dose response relationships that were previously fitted with a single shallow curve.

Fig. 3.
figure 3

Dose response relationships for inhibition of beta cell and cardiac-type KATP currents by sulphonylureas and glinides. Kir6.2/SUR1 and Kir6.2/SUR2A currents are inhibited by increasing concentrations of sulphonylureas and glinides. In the upper row are shown drugs that inhibit SUR1-type channels with higher potency than SUR2-type channels. In the lower row are shown drugs that inhibit SUR1-type and SUR2-type channels with similar potencies. The conductance in the presence of drug (G) is shown relative to that in control solution (Gc). Experiments were performed in membrane patches excised from Xenopus oocytes expressing Kir6.2 with the SUR shown, and solutions did not contain nucleotides. SUR1, closed squares and solid lines; SUR2A, open circles and dashed lines. Data are taken from references [44, 45, 46, 47, 48]

Full size image

Subsequent studies with a range of sulphonylureas and glinides confirmed that they all exhibit both high and low affinity components of inhibition, with a variable separation between the affinities of the two sites (Fig. 3) [44, 45, 46, 47, 48, 50]. Drug interaction with the SUR subunit always resulted in a maximum reduction in the current amplitude of 50 to 75%, when the agents were added on their own to the cytoplasmic surface of the membrane. The interpretation at a molecular level is that the channels remain able to open, albeit less readily, when drug is bound to the sulphonylurea receptor [89]. This behaviour is modified in the presence of cytoplasmic nucleotides, as discussed later.

Whereas all hypoglycaemic sulphonylureas and glinides were found to block SUR1-type channels with high affinity, however, they fell into two distinct groups when their effects on SUR2-type channels were compared (Fig. 3) [44, 45, 46, 47, 48, 50]. Agents in the first group (glibenclamide, glimepiride, repaglinide, meglitinide) showed high affinity block of Kir6.2/SUR2 currents, with potencies similar to those observed with SUR1 subunits. Interestingly, these agents all possessed a non-sulphonylurea moiety. Drugs in the second group (tolbutamide, gliclazide, nateglinide) did not exhibit high-affinity block of Kir6.2/SUR2 currents, and only closed the channels at concentrations that could interact directly with the Kir6.2 subunit. Chlorpropamide is also SUR1-selective, as it inhibited Kir6.2/SUR1 currents by 75 ± 3% and Kir6.2/SUR2A currents by only 17±5% (n=5 of each) at a concentration of 100 µmol/l (five- to ten-fold above the Kd for chlorpropamide displacement of glibenclamide binding to beta cell membranes [90]) (Fig. 4).

Fig. 4.
figure 4

Reversibility and structure of sulphonylureas and glinides. The reversibility of different sulphonylureas and glinides was measured on Kir6.2/SUR1 and Kir6.2/SUR2A currents expressed in Xenopus oocytes. Currents are shown in response to repeated voltage ramps from –110 to +100 mV, and the vertical bars represent 1 nA. Drugs were added as indicated, at the following concentrations: chlorpropamide (100 µmol/l), gliclazide (10 µmol/l), glibenclamide (100 nmol/l), glimepiride (100 nmol/l), meglitinide (10 µmol/l), repaglinide (1 µmol/l). The drug structures are indicated to the left. Data were taken from the following references: gliclazide [45], glibenclamide and meglitinide [44], repaglinide [47]

Full size image

The ability of tolbutamide to inhibit Kir6.2/SUR1, but not Kir6.2/SUR2, channels with high affinity provided a pharmacological tool to search for the sulphonylurea binding site. Chimeric sulphonylurea receptors were constructed in which isolated domains were swapped between SUR1 and SUR2, to identify regions of SUR1 that are obligatory for the high affinity action of tolbutamide. In this way, the critical site was narrowed down to a section within the last TM domain (TMD2) of SUR1 (Fig. 2) [91]. A more detailed analysis of this area resulted in the identification of a single serine residue (S1237) in the cytoplasmic loop (CL8) linking TM helices 15 and 16, which, when mutated to its SUR2 counterpart (tyrosine), abolished both high-affinity tolbutamide block and [3H]glibenclamide binding [91]. Subsequent studies have confirmed the importance of this cytoplasmic loop, and of residue S1237, for sulphonylurea binding to SUR1 [49, 52]. Consistent with these data, introduction of the reverse mutation into SUR2B (Y1206S) increased the [3H]glibenclamide binding affinity about ten-fold [92]. As tyrosine has a more bulky side chain than serine, it is suggested that this in some way obstructs drug binding to SUR2.

When carrying out patch clamp experiments, it is evident that agents that block SUR1-type KATP channels with high affinity have marked differences in their rates of reversal (Fig. 4) [44, 45, 46, 47]. At a structural level, there is a close relationship between drug reversibility and the number of potential binding interactions with SUR1. Most drugs that possess only one active group (either a sulphonylurea or a non-sulphonylurea) are rapidly reversible over a recording period of less than 30 s. This group includes tolbutamide, chlorpropamide, gliclazide and meglitinide. By contrast, drugs that possess both sulphonylurea and non-sulphonylurea moieties (glibenclamide and glimepiride) are effectively irreversible in patch clamp experiments (measured over several minutes). As often occurs, there is an exception to the rule, in this case repaglinide, which does not possess a sulphonylurea group yet is very difficult to wash off KATP channels in membrane patches.

A similar idea holds for Kir6.2/SUR2 currents, but in this case the lack of binding of the sulphonylurea group to SUR2 results in a different pattern of reversibility (Fig. 4) [44, 45, 46, 47]. Thus, tolbutamide, chlorpropamide and gliclazide do not exhibit high-affinity block of Kir6.2/SUR2 currents, inhibition by meglitinide, glibenclamide and glimepiride probably only involves binding of a non-sulphonylurea moiety and is rapidly reversible, and block by repaglinide is effectively irreversible, as with SUR1.

The binding site for sulphonylureas and glinides could therefore be envisaged as a pocket with at least two binding motifs, one (exclusive to SUR1 and abolished by the S1237Y mutation) favouring sulphonylurea groups, and the other (common to SUR1 and SUR2) preferring meglitinide-like molecules [4, 44, 91]. Drug binding to more than one motif produces a particularly strong interaction, and is effectively irreversible. The slow reversibility of repaglinide inhibition might reflect additional interactions of the hydrophobic groups in this molecule. Slow reversibility seems to be crucial for the biochemical detection of drug binding, as specific binding to SUR1 and beta cell membranes can be measured with [3H] derivatives of glibenclamide, glimepiride and repaglinide, but not with [3H]nateglinide [1, 49, 90, 93, 94]. The difficulties sometimes encountered in trying to measure [3H]glibenclamide binding to SUR2-type channels and subunits [34] might be attributable to its rapid reversibility. In line with this idea, mutation of S1237Y in SUR1 increased glibenclamide reversibility in patch clamp experiments and impaired the binding of [3H]glibenclamide [49, 91], whereas the reverse mutation (Y1206S) in SUR2B increased the [3H]glibenclamide binding affinity [92]. Neither repaglinide reversibility, nor binding of [3H]repaglinide, were affected by the S1237Y mutation in SUR1 [47, 49].

It is debatable whether rapidly-reversible drugs provide any advantages over slowly-reversible agents for the pancreatic beta cell. Although the trend in therapeutic practice is to treat Type 2 diabetes with "short-acting" KATP channel inhibitors in an attempt to reduce the incidence of hypoglycaemic side effects, drug reversibility in vivo occurs over a very different time scale from that measured in patch clamp experiments, and involves different underlying mechanisms, including drug transport, metabolism and excretion. In clinical practice, even an agent that takes up to an hour to wash off a sulphonylurea receptor is unlikely to cause hypoglycaemia if it is rapidly cleared from the plasma and if dosing is arranged so that the drug levels are sufficiently low between meals. It is of interest, however, that high pre-prandial levels of agents that reverse rapidly at the level of the KATP channel appear to produce fewer hypoglycaemic side effects than slowly-reversible agents. Chlorpropamide, for example, has been found to produce fewer hypoglycaemic side effects than glibenclamide, even though both drugs have prolonged plasma half lives [95, 96, 97]. Also, the production of a modified-release preparation of gliclazide suitable for once-daily administration did not, apparently, increase the risk of hypoglycaemia over the conventional formulation, despite the high levels of drug that remain in the circulation between meals [98]. There could therefore be a case for suggesting that rapidly reversible action at the level of the KATP channel in some way reduces the risk of hypoglycaemic side effects. Whether drug reversibility has any effect on the age-related deterioration of beta cell function or the secondary failure rate of sulphonylurea therapy, are other possibilities that require clarification.

Studies on two new agents, nateglinide and mitiglinide, refute the classical idea that the 'sulphonylurea-binding motif' on SUR1 is selective for sulphonylureas. Nateglinide and mitiglinide, although not containing sulphonylurea groups, exhibit functional properties similar to those of tolbutamide and gliclazide. Thus, they both inhibited Kir6.2/SUR1 with higher affinity than Kir6.2/SUR2 currents, and inhibition of SUR1-type channels was impaired by the S1237Y mutation [48, 49, 50, 97]. Structural comparisons confirm that the nateglinide molecule can be aligned with tolbutamide, a carboxyl group in nateglinide taking the place of the sulphonyl group in tolbutamide, consistent with the idea that these drugs interact with a similar region of SUR1 [49]. Repaglinide, by contrast, is more structurally related to meglitinide, consistent with its similar affinity for SUR1 and SUR2.

Despite the identification of a site involved in sulphonylurea binding, the location of the meglitinide binding site remains elusive. The chimeric approach, used to identify residues involved in sulphonylurea binding, has been of no help because the properties of meglitinide-inhibition of channels containing SUR1 and SUR2 are indistinguishable. Two pieces of evidence, however, point to the involvement of a separate region of the sulphonylurea receptor in the formation of the drug binding pocket. Firstly, low concentrations of [125I]glibenclamide were found to label a peptide fragment corresponding to the N-terminal third of SUR1 [1]. Secondly, expression studies in insect cells showed that proteins corresponding to the N- and C-terminal halves of SUR1 were individually unable to bind [3H]glibenclamide, but formed high-affinity binding sites when expressed together [100]. Further binding experiments in insect cells showed that although TMD0 could be deleted without affecting glibenclamide binding, additional deletion of the cytoplasmic linker (CL3) between TMD0 and TMD1 abolished [3H]glibenclamide binding [101]. Whether CL3 directly contributes to the meglitinide binding region remains, however, to be established.

Kir6.2 could also be labelled by low concentrations of [125I]azidoglibenclamide in the presence of SUR1 [24], suggesting that part of the Kir6.2 subunit might lie close to the sulphonylurea binding site on SUR1. In favour of this idea, some (but not all) studies showed that coexpression with Kir6.2 or Kir6.1 enhanced the [3H]glibenclamide binding affinity of SUR1 or SUR2B-type subunits by five- to ten-fold, indicating either an allosteric or direct effect of Kir6.x on the sulphonylurea binding site [92, 99, 100, 102, 103]. In patch clamp experiments, high-affinity tolbutamide block of Kir6.2/SUR1 currents was abolished by deletion of the N-terminus of Kir6.2, and was impaired by addition of peptides corresponding to the Kir6.2 N-terminus, suggesting that this region of Kir6.2 might be involved in coupling to SUR [104, 105, 106]. This idea is supported by the finding that truncation of the Kir6.2 N-terminus also prevented co-photolabelling of the Kir subunit by [125I]azidoglibenclamide [106].

Interactions between sulphonylureas and nucleotides

The simplistic interpretation of the data discussed above is that the sulphonylureas and glinides are a heterogeneous group of drugs that would exhibit different degrees of cross-reactivity with cardiovascular KATP channels in vivo. Under physiological conditions, however, a number of additional factors need to be taken into consideration, such as the effects of cytoplasmic nucleotides. The finding that sulphonylureas block KATP currents in excised membrane patches by only 50 to 70%, whereas similar drug concentrations almost completely inhibit native KATP channels in beta cells, is a signal that the excised patch does not exactly mimic the physiological situation. The difference is attributable to the presence, in intact cells, of cytoplasmic nucleotides that modulate the degree of sulphonylurea block [87].

Early experiments on native beta cell KATP channels showed that inhibition by tolbutamide and meglitinide was enhanced by MgADP, or by the simultaneous addition of an inhibitory and a stimulatory nucleotide [107, 108]. The explanation for these findings became apparent during experiments on Kir6.2/SUR1 currents in oocytes. 100 µmol/l MgADP causes net activation of Kir6.2/SUR1 currents because the nucleotide simultaneously activates the channels strongly via SUR, and inhibits them weakly via the ATP binding site on Kir6.2. Tolbutamide was found to prevent MgADP activation, resulting in exposure of the inhibitory effect of the nucleotide on Kir6.2 [87]. The overall effect of adding tolbutamide in the presence of MgADP is therefore that channels go from an activated state (net MgADP activation) to a more inhibited state (combined block by ADP and sulphonylurea), and the overall block by the drug appears to be enhanced. Similar findings have been observed using other sulphonylureas and glinides [44, 47, 50, 91, 107, 109].

Whereas cytoplasmic nucleotides enhance the block of SUR1-containing channels by sulphonylureas, however, this is not observed with SUR2-containing channels [44, 47, 109]. KATP channels in isolated cardiac myocytes, for example, have been shown to be relatively unresponsive to glibenclamide during ischaemia [110, 111]. The region responsible for the SUR1/SUR2 difference was localised, by chimera studies, to TMD1 and the first part of NBD1 (Fig. 2) [109]. Exactly how these regions of SUR1 and SUR2 confer differential responsiveness in the combined presence of nucleotides and sulphonylureas remains unclear and is a complicated issue to resolve because of the theoretical potential for TMD1 to interact with other TMDs or NBDs in the same SUR subunit, with other SUR subunits in the octameric complex, or with the Kir6.2 subunits. Whatever the mechanism, the implications are physiologically significant: although drugs like glibenclamide appear to close cardiac-type KATP channels in patch clamp studies carried out without nucleotides, they might be less effective when the channels are opened in vivo by raised ADP concentrations occurring during ischaemia.

Interferences between the binding of nucleotides and sulphonylureas have been detected in [3H]glibenclamide and [32P]azido-ATP binding studies [61, 112, 113, 114, 115, 116]. Glibenclamide was found to impair azido-ATP binding to SUR1 by preventing the stabilisation of nucleotide binding at NBD1 that normally occurs when MgADP or MgATP binds at NBD2 [61]. Such an effect of sulphonylureas on nucleotide binding might underlie the ability of the drugs to prevent MgADP-induced activation of Kir6.2/SUR1 currents. In this respect it would be interesting to know whether sulphonylureas exert similar effects on nucleotide binding to SUR2, but such data is not currently available.

In [3H]glibenclamide binding studies, it was found that Mg-nucleotides decreased the number of high-affinity glibenclamide binding sites by about 75% in membranes expressing SUR2B-type subunits, whereas they marginally reduced the affinity, but not the number, of sulphonylurea binding sites on SUR1 [115, 116]. This led to the speculation that SUR tetramers might undergo a conformational rearrangement in the presence of MgATP and sulphonylurea, resulting, in SUR2B but not SUR1, in the loss of three out of four high-affinity sulphonylurea binding sites [115]. Whether this contributes to the functional differences observed between channels containing SUR1 and SUR2 remains unclear. If it does, the results of the chimera studies [109] would predict that the effect of nucleotides on [3H]glibenclamide binding would depend on the origin of TMD1 and the first part of NBD1.

Interactions between sulphonylureas and KATP channel openers

KATP channel openers such as diazoxide and nicorandil are used in the treatment of hypertension and angina [117]. Opening KATP channels in vascular smooth muscle results in membrane hyperpolarisation (as in pancreatic beta cells), thereby reducing Ca2+ entry and promoting muscle relaxation and vasodilation [9]. The group of KATP channel openers includes both selective and non-selective agents [43]. Nicorandil, pinacidil and cromakalim are relatively selective for SUR2-type compared with SUR1-type KATP channels. Diazoxide, by contrast, activates channels containing either SUR2 or SUR1, and its pancreatic action is utilised therapeutically to reduce insulin secretion in selected subjects with insulinoma or congenital hyperinsulinism [71, 72].

Studies into the molecular basis for KATP channel opener selectivity have identified regions within TMD2 of SUR2 that determine the responsiveness to nicorandil and analogues of cromakalim and pinacidil. Fine mapping of this domain further localised the critical regions to the cytoplasmic linker between TMs13 and 14 (CL7), and residues within the last TM helix—sites that do not exactly overlap with the areas implicated in sulphonylurea binding [43, 51, 52, 57, 118, 119]. The non-identity of sites involved in sulphonylurea and KATP channel opener action is consistent with the results of binding studies on native and recombinant KATP channel subunits, which suggested that the interference between binding of sulphonylureas and the pinacidil analogue, P1075, are governed allosterically, rather than competitively [115, 120].

Studies on native tissues have suggested that diazoxide opens KATP channels in beta cells and smooth muscle, but not those in cardiac muscle [43]. This apparent selectivity for SUR2B over SUR2A is surprising because the SUR2A/B splice variation only modifies the distal C-terminus of NBD2. The explanation could lie in the fact that diazoxide activation is dependent on the presence of Mg-nucleotides [54, 121], the action of which is found to be modified by the C-terminus of SUR2A/B [118, 123]. Addition of MgADP permitted Kir6.2/SUR2A currents to respond to diazoxide [118], and under conditions in which diazoxide activated the channels, its concentration dependence (EC50) was very similar for SUR1 and SUR2 [43]. The location of the diazoxide binding site, like that of the meglitinide binding site, remains unidentified because of the EC50 similarity between the different SUR subunits.

Nicorandil is increasingly used in the treatment of angina, and operates both as a KATP channel opener and a nitric acid donor [117, 124]. Animal studies have implicated roles for both these properties in its therapeutic action. The likelihood that nicorandil will be prescribed for diabetic subjects on sulphonylurea therapy introduces the potential for interactions to occur between the drugs. It is crucially important to understand whether co-administration of nicorandil with a sulphonylurea that can close channels containing SUR2 would result in impaired activity of the opener or reduced inhibitor action. In oocyte experiments, glibenclamide abolished nicorandil-induced activation of Kir6.2/SUR2A and Kir6.2/SUR2B currents, indicating that the effect of the sulphonylurea is dominant over that of the opener [51]. This is perhaps not surprising, as it is well recognised that glibenclamide inhibits native KATP channels in vascular smooth muscle, in an experimental setting that classically involves prior channel activation by cromakalim or diazoxide [9, 125, 126]. Interestingly, although glimepiride abolished the effect of nicorandil on SUR2A-type channels, it was less effective on SUR2B-type channels, consistent with reports that this agent has lower cross-reactivity with vascular KATP channels [125]. As expected, SUR1-selective sulphonylureas, such as gliclazide, did not affect nicorandil activation [51]. Studies are now required to investigate the extent to which the in vitro findings are replicated in a clinical setting. Loss of the KATP channel opening activity but not the nitric oxide donating property of nicorandil would be predicted during concomitant therapy with glibenclamide, and might, in part, explain why nicorandil was found to retain some anti-anginal activity in a small group of glibenclamide-treated patients [127]. A larger controlled study will be required to determine how much of the therapeutic effect of nicorandil is abolished by glibenclamide and other non-selective KATP channel inhibitors.

KATP channel inhibition under physiological conditions

It is clear that the extent of channel inhibition by sulphonylureas and glinides is determined by the presence (or not) of intracellular nucleotides or KATP channel openers, and that channels containing SUR2 subunits exhibit particularly marked differences in the extent of sulphonylurea block under different conditions. In vivo, KATP channels in the cardiovascular system could be modulated by a number of factors in addition to nucleotides, including phophorylation status, phospholipid concentrations and the function of SUR2 accessory subunits such as creatine kinase, adenylate kinase and lactate dehydrogenase [64, 65, 128]. Phosphoinositides, for example, activate Kir6.2/SUR1 and Kir6.2/SUR2A currents, and impair inhibition by sulphonylureas [129, 130, 131]. It remains to be established whether the accessory enzymes associated with SUR2 modify the response to sulphonylureas. In vitro systems are therefore of only limited value when trying to establish whether non-selective KATP channel inhibitors block cardiovascular KATP channels under physiological conditions. In vivo studies might be the most reliable method to detect cardiovascular effects of sulphonylureas and glinides, but must be carried out with therapeutic concentrations of sulphonylureas and without the use of KATP channel openers.

In intact cells, sulphonylureas might also have effects on KATP channels in intracellular organelles. The existence of a mitochondrial KATP channel, for example, has been suggested for many years [132, 133, 134]. In pancreatic beta cells, there is convincing evidence that Kir6.2 and SUR1 subunits are predominantly located in the membranes of secretory granules [135, 136], although their role there remains largely speculative. It is interesting, for example, that SUR1 associates physically with a cAMP-dependent guanine nucleotide exchange factor, cAMP-GEFII (epac2), which is involved in cAMP-stimulated exocytosis (vesicle release) and calcium release from intracellular stores [137, 138, 139]. It has also been found that exocytosis is directly modulated by the ATP/ADP ratio [140], which could theoretically be detected by the granule SUR subunits. The function of the granule KATP channels, the nature of the interaction between cAMP-GEFII and SUR1, and the possibility that they might be modulated by sulphonylureas, will be interesting areas for future work.

Potential consequences of KATP channel blockade in different tissues

The possibility that non-selective KATP channel inhibitors might have extrapancreatic side effects in clinical practice remains highly controversial. Although the UK Prospective Diabetes Study did not demonstrate an increased risk of micro- or macro-vascular events attributable to glibenclamide therapy [141], this study was not designed to detect more subtle, but potentially significant side effects such as modest changes in exercise tolerance. Part of the explanation for the apparent failure to detect side effects could lie in our poor understanding (until recently) of the physiological roles of KATP channels in extra-pancreatic tissues.

Some clarity is now emerging in this area, however, as a result of the phenotypic descriptions of knockout mice lacking different KATP channel subunits (Kir6.2, Kir6.1, SUR1 and SUR2), (Fig. 5). Although mouse physiology is not necessarily equivalent to human physiology, studies on KATP-channel deficient mice might suggest potential consequences of long term non-selective KATP channel inhibition in humans. Whether therapeutic concentrations of agents like glibenclamide could produce such side effects will depend on a number of factors, including the similarities between human (diabetic) and mouse physiology, the efficacy of the drugs under physiological conditions and the consequences of short term or partial channel inhibition compared with a complete channel deficiency extending during development and the entire lifespan of the animal.

Fig. 5.
figure 5

Major phenotypes of KATP channel deficient mice. KATP channel deficiency in different organs has been produced in mice by knockout of Kir6.2, Kir6.1, SUR1 or SUR2. The figure shows the SUR subtypes expressed in the affected tissues, and the major phenotypes of the knockout in each tissue

Full size image

Effects of KATP channel deficiency on glucose homeostasis

The studies on knockout mice showed that glucose homeostasis is regulated by KATP channel activity in a number of tissues, including pancreatic islets, skeletal muscle and the hypothalamus [142, 143, 144, 145, 146, 147, 148, 149]. The best characterised defects in knockout mice are those affecting the function of pancreatic beta cells. Mice lacking either Kir6.2 (Kir6.2−/−) or SUR1 (SUR1−/−) suffered from transient neonatal hypoglycaemia [142, 144], mimicking, on a different time scale, the human condition of congenital hyperinsulinism. Pancreatic beta cells from these animals had no functional KATP channels, were chronically depolarised and had increased basal intracellular calcium concentrations [142, 144]. Acute secretory responses to glucose and sulphonylureas were impaired, consistent with the role of KATP channel closure in triggering first phase insulin release. The rate of insulin secretion declined with age, resulting in normoglycaemia followed, in Kir6.2−/− mice, by hyperglycaemia. The behaviour of Kir6.2−/− mice is therefore similar to that observed in human subjects with the mutation E1506 K in SUR1, which is associated with both congenital hyperinsulism in childhood and diabetes in adults [78, 79]. Morphological studies of pancreas from Kir6.2−/− mice revealed a loss of beta cells and an enhanced rate of apoptosis [143]. Whether this results directly from the chronic depolarisation and increased intracellular Ca2+, or some other mechanism, is unclear. Consistent with the involvement of KATP channels in cAMP-stimulated exocytosis, SUR1−/− mice also displayed impaired insulin release in response to incretins [145, 146].

Kir6.2−/− and SUR2−/−, but not SUR1−/−, mice were more insulin-sensitive than wild-type mice during insulin tolerance tests, reflecting a role of KATP channels in skeletal muscle [142, 144, 147, 148]. Studies on isolated skeletal muscle from Kir6.2−/− and SUR2−/− mice demonstrated increased rates of insulin-stimulated glucose uptake, but the pathway linking KATP channels to the glucose uptake mechanism has not yet been established [147, 148].

Delayed recovery from insulin-induced hypoglycaemia in Kir6.2−/− mice has been attributed to impaired release of glucagon in response to systemic hypoglycaemia or brain glucopaenia, despite the presence of functional pancreatic alpha cells [149]. Central detection of hypoglycaemia is believed to involve glucose-responsive neurones in the ventromedial hypothalamus, which were lacking in Kir6.2−/− mice [149].

KATP channels in the cardiovascular system

KATP channels in cardiac and vascular smooth muscle are believed to be comprised of Kir6.2/SUR2A and Kir6.2/SUR2B subunits, respectively [34, 35]. An additional class of vascular channels with a smaller single channel conductance and lower ATP sensitivity, can be accounted for by the association of the alternative pore subunit, Kir6.1, with SUR2B [25, 27]. Opening KATP channels in vascular smooth muscle causes vasodilation, and occurs physiologically in response to certain neurotransmitters and hypoxia, or pharmacologically during therapy with KATP channel opening drugs [9]. Activation of KATP channels in cardiac muscle, by contrast, is believed to protect the heart during metabolic stress by shortening the cardiac action potential, thereby reducing calcium entry and cardiac work [8]. Cardiac ischaemic preconditioning is a slightly different phenomenon, reflecting the protection against ischaemic damage by prior exposure to hypoxia. It has traditionally been attributed to the opening of KATP channels in mitochondria, resulting in changes in mitochondrial membrane potential and function [134, 150], but the molecular identity of the putative mitochondrial KATP channels has remained surprisingly elusive.

Physiological studies on mice lacking Kir6.2, Kir6.1 or SUR2 are beginning to clarify the roles of cardiac and vascular KATP channels in cardiovascular pathophysiology [151, 152, 153, 154, 155, 156]. Kir6.2 deficient mice had a number of cardiovascular abnormalities during myocardial ischaemia. Although cardiac myocytes from Kir6.2−/− mice did not show the normal action potential shortening during metabolic inhibition, myocardial infarct size was similar to that in wild-type mice following left anterior descending coronary artery (LAD) ligation [151, 152]. LAD occlusion, however, elicited ST-elevation in the ECG of wild type, but not Kir6.2−/−, mice, supporting the idea that KATP channels are involved in the pathogenesis of ischaemic ST-elevation [152, 153]. Surprisingly, ischaemic preconditioning was also absent in Kir6.2−/− mice, implicating a critical involvement of surface membrane (sarcolemmal), rather than mitochondrial, KATP channels in the genesis of this phenomenon in mice [152].

Although a protective role for cardiac KATP channels during hypoxia therefore seems likely, it is less clear that this is their sole function. In this respect, it is interesting that cardiac KATP channels have now been implicated in the normal response to severe stress. Thus, whereas vigorous sympathetic stimulation in wild type mice resulted in adaptive responses in the heart that maintained myocardial function, in Kir6.2−/− mice it caused impaired cardiac performance and calcium handling, resulting in arrhythmias and sudden death [156]. It therefore seems that sympathetic stimulation normally triggers the opening of KATP channels, which enable the heart to maintain an appropriate level of function. Further studies on the knockout mice might lead to the identification of other physiological conditions that similarly require adaptive changes in myocardial KATP channel activity.

Impaired vascular smooth muscle function was a feature of Kir6.1−/− and SUR2−/− mice, manifesting as episodic coronary artery vasospasm and a high rate of sudden death [154, 155]. The pathophysiological picture was similar to the human condition of Prinzmetal angina. Raised systolic and diastolic blood pressure was also observed in SUR2−/−, but not in either Kir6.1−/− or Kir6.2−/−, mice [151, 154, 155]. This suggests that SUR2-containing KATP channels are involved in blood pressure regulation, but that this role cannot be exclusively attributed to channels with either Kir6.1 or Kir6.2 as the pore-forming subunit. Both Kir6.1/SUR2 and Kir6.2/SUR2 channels, or possibly SUR2 coupled to a third type of pore, might be involved.

The mouse studies suggest that adverse cardiovascular effects of non-selective KATP channel inhibition would be largely restricted to periods of ischaemia or severe stress, and might, for example, include impaired ischaemic preconditioning and attenuated electrocardiographic ST changes during myocardial infarction. The idea that sulphonylureas can inhibit cardiac KATP channels despite the high cytoplasmic nucleotide concentrations is supported by the finding that ischaemic preconditioning during coronary angioplasty was prevented by oral administration of 10 mg glibenclamide in human subjects [157]. Glibenclamide also impaired ischaemic ECG changes in dogs [158], leading to the suggestion that the drug might partially mask ST changes in diabetic human subjects with myocardial infarction [159]. The mouse studies support the involvement of KATP channels in ischaemic ST elevation, and should perhaps stimulate a discussion of whether biochemical cardiac markers should be more widely used in the clinical evaluation of diabetic subjects with non-specific symptoms. Despite the potential adverse consequences of cardiac KATP channel inhibition suggested above, however, there is also evidence that blocking sarcolemmal KATP channels might reduce the incidence of reperfusion arrhythmias – an idea that has actually led to the development of a more selective cardiac KATP channel inhibitor [160].

Musculoskeletal system

The role of KATP channels in skeletal muscle is not well understood, although the possibility that they could be involved in glucose uptake has been discussed above. It had been suggested that the opening of KATP channels in skeletal muscle might also underlie the decrease in muscle tension that occurs during fatigue [10]. Physiological studies on muscle function in wild type and Kir6.2−/− mice, however, demonstrated only a small increase in the resting tension in Kir6.2−/− mice following prolonged muscle stimulation, and enhanced age-related deterioration in tetanic force and the rate of recovery after fatigue [161].

Central nervous system

KATP channels are expressed in a number of regions in the brain. In addition to mediating counter-regulatory responses to hypoglycaemia [149], it has been suggested that KATP channel opening during hypoxia might protect neurones by reducing membrane excitability. In support of this idea, Kir6.2−/− mice were more susceptible to hypoxia-induced generalised seizure, perhaps as a result of the loss of KATP channels in the substantia nigra, a region that has been implicated in the control of seizure propagation [162].

Summary and conclusions

KATP channels are located throughout the body in neuroendocrine, muscle and neural tissues, and the list of their extra-pancreatic roles is still growing. The channels in beta cells have a different sulphonylurea receptor subunit from those in cardiac, skeletal and smooth muscle (SUR1 vs SUR2) and, as a result, show different sensitivities to pharmacological agents. The antidiabetic sulphonylureas and glinides fall into two groups, based on their selectivities for channels containing SUR1 or SUR2. Some drugs, such as tolbutamide, chlorpropamide, gliclazide, nateglinide and mitiglinide, seem to be relatively selective for SUR1-type channels, whereas others, such as glibenclamide, glimepiride and repaglinide, inhibit channels containing either SUR1 or SUR2. This is because the latter group can interact with a region of the drug binding site that is common to SUR1 and SUR2.

Adverse consequences of blocking KATP channels containing SUR2 have been notoriously difficult to demonstrate, leading to the continued use of non-selective agents in clinical practice. Studies on KATP channel deficient mice, however, support the idea that the cardiovascular and musculoskeletal consequences of KATP channel inhibition could be subtle, or restricted to rare situations such as myocardial ischaemia or severe stress. A final clinical decision on whether non-selective KATP channel inhibitors are safe to use in the treatment of Type 2 diabetes should depend on the results of studies designed specifically to look for the side effects suggested by the phenotypes of the KATP channel-deficient mice.