Toward Linking Structure With Function in ATP-Sensitive K+ Channels

Beta-Cell K_ATP channels are hetero-octamers composed of two disparate subunits (Fig. 2): K_IR6.2, a two transmembrane helix pore forming subunit, and its regulatory subunit, the ATP-binding cassette (ABC) protein SUR1 (rev. in 40). K_IR6.2 is a member of the potassium inward rectifier family, channels that in physiological solutions conduct K⁺ better in the inward than outward direction (rev. in 41). Work on the structure(s) of K_IR6.0 pores necessary to understand their weak inward rectification and ATP-inhibitory gating has been limited. Eukaryotic ion channels are difficult to express and purify in the quantities needed for structural work. Thus, insight has come mainly from X-ray crystallographic studies on the cytoplasmic pore of mouse Kir3.1 (42), on the truncated bacterial channels KcsA (43) and KirBac1.1 (44), and on the use of these templates to generate homology models for the K_ATP channel (13) or various domains (45,46). KcsA provided the archetypal tetrameric K⁺ pore assembled from subunits with three helical elements: the outer M1, pore, and inner M2 helices. The selectivity filter, at the external face, is formed primarily from the main chain carbonyls of the GYG (GlyTyrGly) sequence between the M1 and pore helixes (47), whereas the bundle of four M2 helices at the internal face is thought to form the ligand-sensitive gate. An additional amphipathic helix, termed the slide helix, precedes M1 and appears to be lipid-anchored in KcsA (48). The slide helix is positioned perpendicular to the pore axis in the KirBac1.1 structure (44). The K_IR3.1 cytoplasmic pore and KirBac1.1 pore have provided insight into the organization of the long Mg²⁺ and polyamine binding permeation path that confers inward rectification.

The sulfonylurea receptors, SURs, are in the ABC transporter family (49). They have a modular structure (Fig. 2) with an MDR-like core consisting of two bundles of transmembrane domains (TMDs) linked to cytoplasmic nucleotide-binding folds (NBDs) by helical extensions of the TMDs termed intracellular coupling domains (ICDs) (50). Similar to eukaryotic channels, structural studies on eukaryotic transporters are limited. Insight has come from the structures of several bacterial ABC transporters including EC- and VC-MsbA, which are lipid-A exporters from E. coli (50) and Vibrio cholera (50), respectively, and a vitamin B12 importer, BtuCD (51). The VC-MsbA dimer (52) is currently the best structural template available for homology modeling of the multidrug-resistance protein (MDR)-like core of SUR1 (13), based on sequence homology, overall topology, and interacting NBDs. SURs, as well as several other ABCC family members (49), differ from the canonical MDR structure in having an additional NH₂-terminal helical bundle, TMD0, and a long cytoplasmic L0 linker connecting TMD0 with the core (Fig. 2). Although TMD0 and L0 of MRPs are critical for apical-basolateral routing and substrate transport (53–57), a functional role for the TMD0-L0 module in SURs remained unknown until recently (13,58).

While the basic octameric stoichiometry of K_ATP channels is well established (59–61), a basic structural problem has been understanding what parts of the SUR and K_IR subunits interact to activate a fully functional channel. In the absence of inhibitory ligands, K_ATP channels exhibit spontaneous activity characterized by bursts of openings and brief closings. Bursts are separated by interburst intervals, so that channels conduct ∼65–70% of the time (P_Omax up to ∼0.7). Expressed in the absence of SUR, homomeric (K_IR6.2ΔC)₄ pores, lacking their ER retention signals, display profoundly destabilized burst states—an ∼10-fold reduction in their P_Omax and a >30-fold reduction in their sensitivity to ATP. Therefore, we asked which domain(s) of SUR1 could convert the spiky activity of the pore to normal bursting inhibited with an IC_50(ATP) <10 μmol/l. Both the structure of the conserved K⁺ pore (43,62) and work using chimeras of K_IR6.2 and K_IR2.1, a subunit that does not associate with SUR1, implied the outer M1 helix was needed for coassembly (28). While there have been conflicting models of SUR/K_IR6.0 coupling (28,63,64), two recent studies have shown that TMD0 is the principal SUR coassembly domain (13,58). SUR1 TMD0 assembled stable complexes, or “mini-K_ATP channels,” with the (K_IR6.2ΔC)₄ pores resulting in restoration of full K_ATP-like bursting activity with about a fivefold increase in the P_Omax in ligand-free solutions (13,58) (Fig. 2).

The sensitivity to inhibitory ATP was not increased by TMD0 or TMD0-L0, implying that interaction with another part of SUR will be required to account for the increased affinity observed in the full channel. As expected, in the absence of the drug-binding SUR core, mini-K_ATP channels are not stimulated by ADP and their activity is not modulated by sulfonylureas or K_ATP openers. The approach demonstrated that TMD0 is a major SUR domain that interacts with K_IR6.2. The interaction activates the pore and restores bursting, presumably via contacts of transmembrane helices. One expects that other parts of the receptor might interact with cytoplasmic domains of K_IR6.2, but we have been unable to identify stable complexes between K_IR6.2 and the SUR1 core or to demonstrate the core has a functional effect on the K_IR. The results imply that these additional interactions, presumably regulatory in nature, are transient and that L0 serves as a de facto “tether” linking the ABC core to the TMD0/K_IR complex. Additional work is necessary to define how the core is “docked” to the K_IR6.2/TMD0-L0 complex.

Analysis of channels assembled from K_IR6.2 and chimeric receptors SUR1∼SUR2A demonstrated that TMD0-L0 was a determinant of the isoform differences in slow gating (38). The importance of L0 in the control of the slow component of gating was strongly reinforced by the observation that including parts of L0 in mini-K_ATP channels dramatically altered their interburst kinetics. Inclusion of the initial third of L0 dramatically reduced the rate of the burst termination, resulting in “hyperactivation” seen in Nt232/K_IR6.2ΔC35 channels (Fig. 2). At a mechanistic level, this submembrane part of L0, predicted to contain a helix, appears to interact with a submembrane part of the pore to lock mini-K_ATP channels in the burst state. Inclusion of additional sequence attenuated this activation, implying L0 has a bidirectional effect and that the two actions, saturating versus attenuating the P_Omax, require structurally distinct segments of the linker. Existing data provide insight into which parts of K_IR6.2 are involved; deletion of the distal K_IR6.2 NH₂-terminus similarly locks ΔNK_IR6.2/SUR1 channels in the burst state (65–67). The results imply that an interaction of the peripheral K_IR NH₂-terminus with a cytoplasmic segment of L0 attenuates the P_Omax. This hypothesis is supported by the observation that a synthetic, hydrophilic peptide, Ntp, based on the first 33 K_IR6.2 NH₂-terminal residues, can compete with the endogenous NH₂-terminus for a putative inhibitory linker in SUR1 and increase the P_Omax by reducing transitions from the burst state (64). The proposed interactions between TMD0-L0 and the pore are color-coded in Fig. 2: inhibitory contacts (in blue) between the cytoplasmic domains of the two subunits counterbalance the stimulatory interactions (in red) between the submembrane L0 helix and the K_IR “slide” helix (44), which saturate the P_Omax.

Site-directed mutagenesis reduced the candidate residues involved in ATP binding, or its coupling to the ATP inhibition gate, to Lys185 and Gly334 in the COOH-domain and Arg50 in the NH₂-terminus of K_IR6.2. Substitutions at these positions decreased the IC_50(ATP) without affecting the P_Omax of homomeric K_IR pores and K_ATP channels in ligand-free solutions (65,68,69). Substitutions at other residues that affect ATP-binding, e.g., I182 (69,70), can alter the P_Omax. The first double mutant, K_IR6.2 Arg50Gln/Lys185Gln (65), demonstrated the additive contribution of NH₂- and COOH-terminal residues to ATP-induced stabilization of the closed state, consistent with the possibility that these two residues cooperate in the coordination of separate negatively charged phosphate groups of ATP in SUR1-containing channels. The GIRK1-based homology models of the K_IR6.0 C-domain (13,71) support the idea that Ile182, Lys185, and Gly334 are colocalized and that Lys185 and Gly334 probably contribute to the surface of the adenine- and β-phosphate-selective ATP-binding pocket that we have argued is conserved in both K_IR6.0 subunits (39,72). Our modeling of the complete pore indicated an additional key structural aspect of the ATP-inhibitory machinery. Consistent with the data on the R50Q/K185Q double-mutant channels (65), we proposed that the COOH-terminal binding pocket on one K_IR6.0 subunit is complemented by Arg50 from the neighboring K_IR6.0 subunit, producing four intersubunit ATP-binding pockets (inset, Fig. 2). The key feature of this structural model is that Arg50 immediately precedes (by three residues) the slide helix that we hypothesize contacts the submembrane helix of L0 and moves with it during interburst transitions. Repositioning of these elements thus has a dual effect: Movement of the activating (red) domains away from the pore axis opens the gate and will also move Arg50 away from the nucleotide-binding cavities, effectively opening the intersubunit pockets and thus increasing the rate of dissociation (K_d) for inhibitory ATP. This mechanism can explain the marked increase in the IC_50(ATP) by Nt232, which effectively mimics the action of full-length SUR saturated with stimulatory ligands. Conversely, movement of the inhibitory region (blue) of L0 is hypothesized to push the K_IR NH₂-terminus, thus repositioning Arg50 and increasing the affinity for nucleotides, while stabilizing a straight, closed KcsA- and KirBac1.1-like conformation of the M2 helical bundle, thereby reducing the mean P_O. This attenuation of the P_Omax by the inhibitory segments of L0 (see Nt288 in Fig. 2) mimics the effect of full-length SUR1 saturated with a hypoglycemic compound when tested in nucleotide-free solutions (73).

The main action of oral hypoglycemic agents, both sulfonylurea and nonsulfonylurea compounds, is to stimulate insulin secretion by inhibiting K_ATP channels via their binding to SUR1. Early studies on structure-activity relationships indicated, over 30 years ago, that an effective pharmacophore consisted of three lipophilic or hydrophobic centers separated by a -CONH- group and a negative charge, respectively (Fig. 3). Recognition was posited to occur through interaction with a multifaceted bipartite pocket with distinct “A” and “B” binding sites within a “sulfonylurea receptor” (75–77,80; rev. in 81). More recent work has made it clear that the sulfonylurea group is not essential for hypoglycemic activity, but rather provides a properly positioned anionic group (78,79). The importance of this work is being rediscovered, and structure-function studies on SUR1 and SUR2 have begun to provide structural information about the nature of the “A” and “B” sites (rev. in 81). Chimeric receptors SUR1∼SUR2A were used to exploit the pharmacological differences between β-cell and cardiomyocyte channels and identify the importance of the TMD2 domain for high-affinity recognition of tolbutamide (73,82). A detailed consideration of the available structure-activity data suggests that TMD2 forms part of the “A” site and that SUR1 is able to accommodate the lipophilic center adjacent to the anionic group on the sulfonylurea moiety (i.e., the butyl side chain in tolbutamide, cyclohexane substituted with a propyl group in nateglinide, or the cyclohexyl ring in glibenclamide) while SUR2A is not. The implication is that SUR1 has a hydrophobic cavity able to accommodate these side chains and that this cavity is missing or occluded in SUR2. Substitution of Ser1237 with Tyr, the analogous residue in SUR2, reduced the apparent affinity of SUR1 for tolbutamide and glibenclamide (82). One possibility is that Ser1237 is in close proximity to the lipophilic center and that Tyr1206 occludes binding in SUR2A; the reverse substitution, Tyr1206Ser in SUR2B, does increase the affinity for glibenclamide several fold but not to the level of SUR1 (83). The interpretational difficulty lies in not being able to distinguish experimentally between a direct interaction of Ser1237 with the drug versus an indirect effect of the substitution at a more distant drug binding site. In the homology model of the SUR1 core, Ser1237 is at the end of one of the α-helices, which form the ICD (see Fig. 2). The ICDs are proposed to play a critical role in coupling the binding of stimulatory nucleotides to the NBD dimer with conformational changes in the substrate binding cavity of MsbA (52) or the maltose transporter (84). It is tempting to speculate that the A site lies at the level of the ICDs and that tolbutamide, or other insulin secretagogues, can “insert” between the halves of the core and restrict movement. This tentative structural picture provides a framework with which to interpret the abolishment of the “basal,” Mg-nucleotide-dependent stimulation of SUR1 by oral hypoglycemic agents (73,85,86), which is their principal mechanism of action in vivo.

Insight into the location of the B site comes from affinity labeling studies and efforts to define the glibenclamide binding site. Analysis of peptides affinity labeled by ¹²⁵I-iodoglibenclamide, presumably by photoactivation of the carbonyl group in the benzamido moiety, placed the labeling site in the TMD0-L0 segment (87). When SUR1 and K_IR6.2 are in K_ATP channels, both subunits are labeled by ¹²⁵I-azidoglibenclamide, where the active group is a nitrene on the benzamido ring generated by photolysis and thus positioned to react with residues in the B site (59,88,89). Mikhailov et al. (90) showed that TMD0 is not required for high-affinity glibenclamide binding to SUR1. Similarly, TMD0 is not required, whereas L0 is necessary, for affinity labeling (W.H.V.-C., unpublished data). Alanine substitutions around Trp²³², a conserved residue in L0, identify a cluster of amino acids critical for affinity labeling (Fig. 3). Substitution of Trp²³² with alanine results in loss of affinity labeling, whereas replacement with cysteine retains labeling. The results are consistent with the nitrene on the benzamido group being in close proximity to Trp²³², but again conformational change with action at a more distant site cannot be ruled out completely. A conservative interpretation is that the benzamido group of glibenclamide lies near the Trp232 region of L0, which forms part of the B site.

Deletion of the NH₂-terminus of K_IR6.2 reduced its affinity labeling with ¹²⁵I-azidoglibenclamide (64). Data in Fig. 3 show that a fragment of K_IR6.2 containing M1 but lacking M2 and the COOH-terminal cytoplasmic domain makes a stable complex with SUR1 and is affinity labeled. Progressive deletion of residues from the NH₂-terminus of this M1 fragment abolished affinity labeling without affecting its association with SUR1. Short NH₂-terminal deletions of K_IR6.2 also impaired the coupling of tolbutamide binding to SUR1 and consequent attenuation of the P_O in the absence of nucleotides (65). The results support the idea that the outer helix of K_IR6.0 subunits specifically interact with SUR (28), as well as the conservative interpretation that the principal affinity-labeling site is either in the first half of the K_IR6.2 NH₂-terminus or that the peripheral NH₂-terminus is essential for positioning more proximal residues near the nitrene (64).

Chimeric receptors were also used to localize regions important for recognition of several channel openers, including diazoxide, cromakalim, and pinacidil. The action of diazoxide required TMD1 and part of the first nucleotide binding domain, NBD1 (91), a localization supported by a study on the novel opener NNC 55-9216 (92). A region of TMD2 was shown to be critical for the action of cromakalim and pinacidil (91,93,94), and Moreau et al. (94) identified two residues in the last transmembrane helix of SUR—Leu1249 and Thr1253 in SUR2 and Thr1286 and Met1290 in SUR1. Swapping these two residues conferred sensitivity to cromakalim and pinacidil on SUR1 and abolished sensitivity in SUR2A. In the homology model of the SUR1 core (13,81), Met1290 is near the top of TMD2 facing the putative drug-binding cavity (see Fig. 2). The results are consistent with the conclusion that both halves of the SUR core are required for specific interaction with K_ATP channel openers (91). The positioning of Met1290 in central cavity, as well as earlier data indicating that the action of the potassium channel openers depends on the presence of hydrolysable MgATP (95–97; however, see 98), suggests a parallel with transport mechanisms in which site accessibility is dependent on nucleotide hydrolysis and implies that binding of an opener in the central cavity can trap a specific conformation of the core and thus potentiate channel activity.

Current research is providing an increasingly detailed picture of β-cell K_ATP channel structure, including homology models of the ion-conducting pore and the nucleotide- and drug-binding SUR1 core. A role for TMD0-L0 in the coupling of SUR1 to K_IR6.2 has been demonstrated, and the importance of these interactions for control of gating has been emphasized. This structural picture provides an increasingly sophisticated framework with which to interpret mutations and polymorphisms that affect glucose homeostasis by altering the normal coupling mechanism to produce both inactive and hyperactive channels. The spectrum of diseases linked with K_ATP channels, from neonatal to adult diabetes through congenital hyperinsulinism, is broad and of considerable social, medical, and economic importance. Increased understanding of the structure and coupling mechanism(s) in K_ATP channels will lead to improved pharmaceuticals directed at these disorders.

This work was funded in part by grants from the National Institutes of Health-National Institute of Diabetes and Digestive and Kidney Diseases (J.B. and L.A.-B.) and the American Heart Association (A.P.B.).