Differential Interactions of Nateglinide and Repaglinide on the Human β-Cell Sulphonylurea Receptor 1

Human SUR1 cDNA (GenBank L78207) and human Kir6.2 (GenBank D50582) were cloned into pcDNA3.1(−) (Invitrogen). The point mutation SUR1[S1237Y] was constructed by standard molecular biology techniques and confirmed by DNA sequencing.

HEK293 cells were cultured at 37°C in a humidified atmosphere of 95% air and 5% CO₂ in Dulbecco’s modified Eagles medium with 4.5 g/l glucose (BioWhittaker) supplemented with 10% FCS, penicillin (100 units/ml), and streptomycin (0.1 mg/ml).

Transient transfections were performed using FuGene 6 Transfection Reagent (Roche), according to the manufacturer’s instruction. Cells were seeded at 50% confluency and transfected with Kir6.2 and SUR1[S1237Y] at a plasmid ratio of 1:3 on the next day. Cells to be used for electrophysiological experiments were also cotransfected with green fluorescent protein (GFP) to enable visual identification of transfected cells. Experiments were performed 1–3 days after transfection. HEK293 cells stably expressing hKir6.2 and hSUR1 were used for studies of wild-type channels (19).

Cells were harvested and centrifuged at 48,000g for 10 min at 4°C. The pellet was homogenized in ice-cold buffer (30 mmol/l Tris-HCl, pH 7.4) using an Ultra Turrax for 20 s. Centrifugation and homogenization were repeated, and the pellet was then resuspended in buffer and sucrose was added to a final concentration of 250 mmol/l. Protein concentration was measured using the Bio-Rad protein assay. Membranes were stored at −80°C.

Binding experiments were performed in 96-well OptiPlates (Packard). Membranes (5 μg protein/well for wild-type channels and 12 μg protein/well for mutant channels) were thawed on ice and incubated for 60 min at 37°C with 0.8 nmol/l [³H]repaglinide in 30 mmol/l HEPES (pH 7.4), in a total volume of 250 μl. Bound [³H]repaglinide was separated from free [³H]repaglinide by rapid filtration on a Filtermate Harvester (Packard) through UniFilter GF/B filterplates (Packard). Filterplates were washed five times with 450 μl water (room temperature) and dried. Scintillation cocktail (30 μl) (Microscint, Packard) was added to each well, and radioactivity was determined by counting the plates in a Microplate Scintillation and Luminescence Counter (Topcount-NTX; Packard). Nonspecific binding was determined in the presence of 1 μmol/l unlabeled repaglinide and was <5% of the total binding. Binding experiments were performed in triplicate (Kir6.2/SUR1) or duplicate (Kir6.2/SUR1[S1237Y]).

Whole-cell currents were recorded at 20–22°C using an EPC9 patch-clamp amplifier and Pulse+PulseFit version 8.07 software. The extracellular bath solution contained 140 mmol/l NaCl, 5 mmol/l KCl, 10 mmol/l HEPES, 1.8 mmol/l CaCl₂, and 20 mmol/l mannitol (pH 7.4 with NaOH). Cells were dialysed with intracellular solution containing 120 mmol/l KCl, 1 mmol/l MgCl₂, 5 mmol/l EGTA, 2 mmol/l CaCl₂, 20 mmol/l HEPES (pH 7.3 with KOH), 0.3 mmol/l K₂ ATP, and 0.3 mmol/l K₂ ADP. Cells were clamped at −80 mV, and currents were evoked by repetitive 200 ms, 10 mV depolarizing voltage steps. Signals were sampled at 20 kHz and filtered at 5 kHz. In some experiments, cells were held at −70 mV for a period of 10–20 min to study the time course of drug inhibition. In this case, currents were sampled at 25 Hz and filtered at 6.25 Hz.

Conformational analyses were performed using MacroModel version 6.0 (20). The compounds were subjected to a Monte Carlo search using the MMFF force field (21), with a solvation model for water (22). For each molecule, 5,000 conformations were generated, with an energy cut-off of 25 kJ/mol. The resulting low-energy conformations were imported into Sybyl version 6.6 (Tripos, St. Louis, MO), and all further manipulations were performed using this program. Initially, a number of alternative superimpositions were generated using the GASP algorithm (23), as implemented in Sybyl. The final pharmacophore models were generated by manually adjusting the superimpositions generated using GASP. In all cases, the resulting conformations were compared with the minimum-energy conformations found in the conformational search.

Data are presented as mean ± 1 SD unless otherwise stated. Concentration-response curves for drug-induced K_ATP current inhibition were constructed by expressing the current in the presence of the drug as a fraction of the current before the drug was added. Data were analyzed in Prism (GraphPad) using the four-parameter logistic equation:

where y is the current expressed as a percentage of that recorded before drug was added, a is the percentage of current remaining after maximal inhibition by the drug, IC₅₀ is the drug concentration that results in half-maximal inhibition, [L] is the concentration of drug, and n_H is the Hill coefficient.

The same equation was used to fit the dose-inhibition curves for displacement of [³H]repaglinide binding, but in this case, y was specific binding and a was nonspecific binding. The experimentally measured IC₅₀ values for the competitive ligands were converted into K_i values using the Cheng-Prusoff equation:

where [L] is the concentration of [³H]repaglinide, and K_D is the equilibrium dissociation constant for [³H]repaglinide.

Tolbutamide was purchased from Sigma and glibenclamide from Research Biochemicals International. Nateglinide was synthesized at Novo Nordisk A/S and repaglinide at Boehringer Ingelheim (Biberach, Riss, Germany). Concentrated stock solutions were prepared in DMSO for subsequent dilution in buffer. The concentration of DMSO in the experiments did not exceed 0.1% and had no effect in either binding or electrophysiological experiments.

Radiolabelled repaglinide was prepared at the Department of Isotope Chemistry, Novo Nordisk A/S, by catalytic tritiation of the repaglinide precursor S(+)-2-ethoxy-4-[2-[[3-methyl-1-[2-(piperidinyl)-phenyl]4-buten-yl]amino]2-oxoethyl]-benzoic acid (17), which was kindly provided by Dr. M. Mark, Boehringer Ingelheim. A specific activity of ∼12 MBq/μg (144 Ci/mmol) was estimated from mass spectroscopy of the final product.

Whole-cell currents were recorded from HEK293 cells coexpressing Kir6.2 and either SUR1 or SUR1[S1237Y]. After establishment of the whole-cell configuration and dialysis with intracellular solution, there was a gradual increase in both Kir6.2/SUR1 and Kir6.2/SUR1[S1237Y] currents due to opening of K_ATP channels. As shown in Fig. 1A, tolbutamide (300 μmol/l) completely inhibited this K_ATP current (by 93 ± 4%, n = 4), and the effect of the drug was reversible upon return to the control solution. In contrast, tolbutamide had very little effect on Kir6.2/SUR1[S1237Y] currents (Fig. 1B). Glibenclamide (1 μmol/l) blocked Kir6.2/SUR1 channels by 98 ± 1% (n = 3), and this inhibition could not be reversed by 20 min of washing (Fig. 1A). Although glibenclamide also inhibited Kir6.2/SUR1[S1237Y] channels (by 67 ± 6%; n = 3), in this case, inhibition was largely reversed on wash-out of the drug (Fig. 1B). These findings are in agreement with earlier studies of rodent Kir6.2/SUR1 channels expressed in Xenopus oocytes (9).

Figure 2 compares the effects of repaglinide and nateglinide on wild-type and mutant K_ATP channels. Repaglinide (1 μmol/l) completely inhibited human Kir6.2/SUR1 currents (by 98 ± 1%, n = 5) (Fig. 2A), as previously reported for rat Kir6.2/SUR1 expressed in Xenopus oocytes (19). A similar extent of block was observed for Kir6.2/SUR1[S1237Y] (94 ± 1%, n = 4) (Fig. 2B), suggesting that S1237 is not required for repaglinide inhibition. In both cases, no significant recovery of current was observed on removal of the drug. In contrast, nateglinide (100 μmol/l) produced reversible inhibition of Kir6.2/SUR1 currents (96 ± 2%, n = 4) (Fig. 2A) but was without significant effect on Kir6.2/SUR1[S1237Y] channels (Fig. 2B).

Concentration-response relationships for repaglinide and nateglinide block of wild-type and mutant K_ATP channels are shown in Fig. 3. Repaglinide blocked Kir6.2/SUR1 and Kir6.2/SUR1[S1237Y] currents with similar potency: IC₅₀ = 21 nmol/l (95% CI 17–26) and IC₅₀ = 23 nmol/l (18–28), respectively. These results suggest that S1237 is not critical for repaglinide-induced channel inhibition.

In contrast to repaglinide, nateglinide blocked Kir6.2/SUR1 currents with an IC₅₀ of 0.8 μmol/l (95% CI 0.3–2.3) but produced <10% block of the mutant channel, even at a concentration of 100 μmol/l. Similar IC₅₀ values were found for Kir6.2/SUR1 expressed in COS-1 cells (24) and for native rat β-cell K_ATP currents (25). This suggests that the presence of a tyrosine at residue 1237 in SUR1 either prevents nateglinide binding or the transduction of binding into channel closure.

High levels of specific [³H]repaglinide binding were observed for membranes prepared from HEK293 cells expressing Kir6.2/SUR1 but not for membranes isolated from HEK293 cells expressing the vector only. Saturation binding experiments showed that [³H]repaglinide binding was saturable, and a Scatchard transformation of the data indicated that binding to Kir6.2/SUR1 occurred in an apparent monophasic manner with a high capacity (B_max = 5.1 ± 1.8 pmol/mg protein, n = 5) (Fig. 4A). The K_D, estimated from nonlinear regression, was 0.40 ± 0.09 nmol/l (n = 5). Similarly, binding of [³H]repaglinide to Kir6.2/SUR1[S1237Y] revealed a single binding site (Fig. 4B) with a K_D of 0.31 ± 0.02 nmol/l and a B_max of 1.6 ± 0.2 pmol/mg protein (n = 3). Thus, repaglinide affinity for Kir6.2/SUR1 is unaffected by mutation of S1237 to tyrosine. The difference observed in B_max for wild-type and mutant channels is likely to be the result of the different expression systems (transient versus stable transfection).

Attempts to measure [³H]nateglinide binding to wild-type or mutant Kir6.2/SUR1 were not successful, possibly because of the low affinity and rapid unbinding of the drug (18). We therefore examined the ability of unlabelled nateglinide to displace [³H]repaglinide binding to membranes isolated from HEK293 cells expressing Kir6.2/SUR1 or Kir6.2/SUR1[S1237Y]. For comparative purposes, we also examined the effects of cold repaglinide, glibenclamide, and tolbutamide on [³H]repaglinide binding.

All drugs showed a monophasic displacement of [³H]repaglinide binding (Fig. 5). The K_i for the wild-type channel, estimated from the IC₅₀ values (see research design and methods), were 0.6 ± 0.3 nmol/l (n = 3) for repaglinide, 0.2 ± 0.1 nmol/l (n = 3) for glibenclamide, 0.24 ± 0.04 μmol/l (n = 3) for nateglinide, and 9.0 ± 3.2 μmol/l (n = 3) for tolbutamide. The Hill coefficients were close to unity in all cases (from 0.90 to 1.19). The K_i for repaglinide (0.6 ± 0.3 nmol/l) is not significantly different from the K_D estimated from the saturation experiments (0.40 ± 0.09 nmol/l). Furthermore, the relative order of affinity is glibenclamide ∼ repaglinide > nateglinide > tolbutamide, and is in close agreement with the relative order of potency for these drugs on inhibition of Kir6.2/SUR1 currents (26).

In the case of Kir6.2/SUR1[S1237Y], repaglinide displaced [³H]repaglinide binding with a K_i of 0.4 ± 0.2 nmol/l (n = 3), which is similar to that found for the wild-type channel. In contrast, the affinity for glibenclamide (K_i = 36 ± 6 nmol/l, n = 3) was 170-fold lower than that of the wild-type channel. Furthermore, nateglinide did not displace [³H]repaglinide binding to the mutant channel, even at the highest concentration tested (30 μmol/l), and tolbutamide (300 μmol/l) only marginally displaced [³H]repaglinide binding. The mean results are summarized in Table 1. The data are consistent with the idea that the nateglinide binding, as with that of tolbutamide, is abolished by the S1237Y mutation in SUR1. Furthermore, this mutation reduces the affinity for glibenclamide binding while leaving that of repaglinide unaffected.

The fact that S1237 is critical for nateglinide as well as tolbutamide inhibition raises the question of whether these drugs show structural similarities. We therefore carried out structural comparisons of repaglinide, glibenclamide, nateglinide, and tolbutamide (Fig. 6A).

In agreement with earlier studies (17), repaglinide and glibenclamide can be fitted to the same pharmacophore model, with the benzoic acid fragment of repaglinide superpositioned on the acidic arylsulphonamide part of glibenclamide (Fig. 6B). The distal aromatic rings are copositioned, whereas the carbonyl oxygens of the amide groups are capable of interacting with the same putative hydrogen bond donor in SUR1. As both molecules are very flexible, the model shows only one of the several alternative spatial arrangements of the three pharmacophore groups (benzoic acid/arylsulphonamide, distal aromatic ring, and carbonyl oxygen).

Figure 6C shows that although nateglinide does not contain a sulphonylurea moiety, its three-dimensional structure is similar to that of tolbutamide and to the sulphonylurea part of glibenclamide. In this model, the phenylpropionic acid part of nateglinide is superimposed on the arylsulphonamide part of the sulphonylureas. In addition, the carbonyl oxygens and the hydrophobic tails are superpositioned. As both tolbutamide and nateglinide are smaller than glibenclamide, the benzamide part of glibenclamide is depicted in a random extended conformation.

The ability of the new insulin secretagogues repaglinide and nateglinide to inhibit recombinant human Kir6.2/SUR1 channels was investigated and compared with that of tolbutamide and glibenclamide. Structurally, glibenclamide consists of a sulphonylurea group similar to tolbutamide and a nonsulphonylurea moiety that resembles the benzoic acid derivative meglitinide. Previous mutagenesis and affinity-labeling studies have provided evidence for a model of the glibenclamide binding site in which contributing amino acid residues are derived from two distinct regions of SUR1: the cytosolic loop linking TMs 15 and 16 (in which S1237 is critical) and the cytosolic loop linking TMs 5 and 6 (8,14,15). The TM15-16 loop is essential for high-affinity inhibition of K_ATP channel currents by drugs that resemble tolbutamide in structure. In the present study, three different lines of evidence suggest that nateglinide, but not repaglinide, interacts with this region of the channel: binding data, electrophysiological data, and structural considerations.

The K_D for [³H]repaglinide binding obtained in this study (0.4 nmol/l) is lower than that reported for [³H]repaglinide binding to intact βTC3 cells (6.4 nmol/l) (27) but similar to the K_i (1.8 nmol/l) found for repaglinide inhibition of [³H]glibenclamide binding to RIN-m5F cells (18). The K_i for inhibition of [³H]repaglinide binding to human Kir6.2/SUR1 by nateglinide presented here (235 nmol/l) is in accordance with that of human SUR1 expressed in COS-1 cells (IC₅₀ = 280 nmol/l) (24) and that of native K_ATP channels in HIT-T15 cells (K_i = 248 nmol/l [28] or 434 nmol/l [29]) or RIN-m5F cells (K_i = 170 nmol/l [18]). In the present study, tolbutamide and nateglinide were found to fully displace [³H]repaglinide binding to Kir6.2/SUR1 channels. This suggests that the binding sites for tolbutamide/nateglinide and repaglinide either overlap or are allosterically coupled.

Mutation of serine 1237 in SUR1 to tyrosine, which is known to influence channel inhibition by tolbutamide and glibenclamide (8), abolished nateglinide but not repaglinide inhibition. This finding is consistent with the idea that the binding site for nateglinide overlaps with that of tolbutamide and includes residue 1237. In contrast, the binding site for repaglinide does not include S1237 (it is unlikely that repaglinide could interact with S1237 and not be affected by a mutation at this position, because a serine to a tyrosine represents a rather large change in side-chain volume).

Glibenclamide produced a reversible block of Kir6.2/SUR1[S1237Y]. In previous studies, it was not possible to detect [³H]glibenclamide binding to this channel, a finding that was suggested to be because the drug rapidly dissociates from SUR during the washing procedure of the binding assay (8). Our binding data provide additional support for this hypothesis. Thus, glibenclamide displaced [³H]repaglinide binding to the mutant channel with a much lower potency than for the wild-type channel, consistent with a larger dissociation rate constant for glibenclamide binding to Kir6.2/SUR1[S1237Y]. The lower binding affinity of the mutant channel is also consistent with the fact that 1 μmol/l glibenclamide did not inhibit the channel completely in patch-clamp experiments. Nateglinide, like tolbutamide, inhibited wild-type K_ATP currents reversibly. In contrast, repaglinide inhibition of both wild-type and mutant channels was not reversible. This suggests that the binding affinity of repaglinide is enhanced by interaction with additional residues in SUR1 and that this interaction is not disrupted by the S1237Y mutation.

The effect of the S1237Y mutation on nateglinide-induced K_ATP channel inhibition could be due to either reduced drug binding or an impaired ability of SUR1 to transduce drug binding into channel closure. However, because Kir6.2/SUR1[S1237Y] retains the ability to be blocked fully by repaglinide and glibenclamide, the transduction mechanism does not appear to be compromised by the mutation. (We assume that the transduction mechanism is common for all drugs used in this study.) Furthermore, the mutation impairs the ability of nateglinide to displace [³H]repaglinide binding, consistent with a reduced affinity for the former drug.

The residue S1237 is positioned in the intracellular loop between TM segments 15 and 16. The importance of this residue for binding and channel inhibition by both tolbutamide and nateglinide strongly argues for an intracellular binding site for these drugs. Nateglinide has been proposed to act from the extracellular side and not to be internalized into the β-cells (30,31). The results presented here, however, suggest that the binding site for nateglinide resides on the intracellular side of SUR1 and shares at least one point of interaction with the binding site for tolbutamide.

The modeling studies demonstrate that glibenclamide and repaglinide can be fit to a common pharmacophore model, whereas nateglinide, tolbutamide, and glibenclamide can be fit to another model that is distinct from that of glibenclamide/repaglinide. The conformations of glibenclamide in the two models differ slightly, but we cannot rule out the possibility that the large number of conformational degrees of freedom enable glibenclamide to adopt a conformation that simultaneously fits both models. Most importantly, the structural investigations illustrate that the acidic moiety is the only pharmacophore group shared by all four drug molecules. They further reveal that the amide part of glibenclamide is shared with repaglinide, whereas the sulphonylurea part of glibenclamide is shared with both tolbutamide and nateglinide. This provides additional support for the idea that nateglinide shares (at least in part) a common site of interaction on SUR1 with sulphonylureas.

The impact of these observations on the efficacy of repaglinide and nateglinide in vivo, or in the clinic, is uncertain. However, our data suggest that the structural determinants involved in nateglinide binding overlap with those of sulphonylureas such as tolbutamide. In contrast, the properties of repaglinide binding and inhibition differ from those of the sulphonylureas. This is of relevance in terms of other known differences in the mechanism of action of repaglinide and that of other insulin secretagogues.

In addition to closing K_ATP channels, sulphonylureas (e.g., tolbutamide) have been shown to interact directly with the secretory machinery of β- and α-cells, thereby stimulating Ca²⁺-dependent exocytosis of insulin and glucacon, respectively (32,33). Likewise, tolbutamide stimulates direct exocytosis of somatostatin and growth hormone in vitro (34–36). This has led to the suggestion that a sulphonylurea binding protein exists on the secretory granules (37,38). It is of interest that nateglinide, but not repaglinide, shares the ability to stimulate exocytosis (34–36). It is possible, although unproven, that the shared characteristics of the binding sites for nateglinide and sulphonylureas, demonstrated in this study, help explain these differences.

In conclusion, our data indicate that S1237 in SUR1 is a prerequisite for the ability of tolbutamide and nateglinide to act as K_ATP channel inhibitors. In contrast, glibenclamide (partially) and repaglinide (wholly) interact with regions of SUR1 that are distinct from S1237.

We thank Tina Moeller Tagmose and Jesper Boeggild Kristensen for synthesis and radiolabelling of repaglinide. F.M.A. is the GlaxoSmithKline Royal Society Research Professor.