The Nucleotide-Binding Sites of SUR1: A Mechanistic Model.

ATP-sensitive potassium (KATP) channels comprise four pore-forming Kir6.2 subunits and four modulatory sulfonylurea receptor (SUR) subunits. The latter belong to the ATP-binding cassette family of transporters. KATP channels are inhibited by ATP (or ADP) binding to Kir6.2 and activated by Mg-nucleotide interactions with SUR. This dual regulation enables the KATP channel to couple the metabolic state of a cell to its electrical excitability and is crucial for the KATP channel's role in regulating insulin secretion, cardiac and neuronal excitability, and vascular tone. Here, we review the regulation of the KATP channel by adenine nucleotides and present an equilibrium allosteric model for nucleotide activation and inhibition. The model can account for many experimental observations in the literature and provides testable predictions for future experiments.

Main Text

ATP-sensitive potassium (KATP) channels play important roles in a wide variety of physiological processes, including hormone secretion, neuronal function, cardiac excitability, and vascular tone (1). They comprise four Kir6.x (Kir6.1 or Kir6.2) subunits, which form a tetrameric pore, and four modulatory sulfonylurea receptor (SUR) subunits (Fig. 1 A). There are two SUR genes, SUR1 and SUR2, but the latter is differentially spliced to give SUR2A and SUR2B isoforms. These different SUR isoforms confer different properties upon Kir6.x.

Figure 1

Domain organization of the KATP complex. (A) Hetero-octameric complex of the KATP channel, showing the Kir6.x tetrameric pore (Kir6.1 or Kir6.2) surrounded by four SUR subunits. (B) Schematic representation of Kir6.x and SUR protein topologies, indicating the three hydrophobic TMDs (TMD0, TMD1, and TMD2) and the two NBDs (NBD1 and NBD2) of SUR.

KATP channels are modulated by numerous ligands, lipids, and drugs, but their most characteristic property is that they are regulated by cellular metabolism via changes in adenine nucleotide concentrations. Intracellular ATP and ADP cause channel inhibition (by binding to Kir6.x), whereas MgATP and MgADP stimulate channel activity by interacting with the nucleotide-binding sites (NBSs) of SUR (2, 3, 4, 5). In addition to conferring sensitivity to Mg-nucleotide activation, SUR enhances the unliganded channel open probability (Po) and increases the affinity for ATP block (5). It also endows the channel with sensitivity to therapeutic drugs such as sulfonylureas and K-channel openers, which inhibit and stimulate channel activity, respectively (5, 6). Sulfonylureas are used to treat type 2 diabetes and neonatal diabetes (7, 8), whereas the K-channel opener nicorandil is an antianginal agent (9).

Here, we briefly review our current understanding of the structure and function of the NBSs of SUR. We also provide a simple equilibrium model of nucleotide handling by the KATP complex that can account for most current data.

Structure of the nucleotide-binding domains

SUR belongs to the ATP-binding cassette (ABC) family of transporters (10, 11, 12, 13, 14). Like other ABC proteins, it consists of two sets of transmembrane domains (TMD1 and TMD2), each of which contains six helices, and two cytosolic nucleotide-binding domains (NBDs) (Fig. 1 B). SUR also has an additional N-terminal five transmembrane helices (TMD0). High-resolution structures of SURs or their isolated NBDs are not yet available. However, the NBSs of most ABC proteins feature a common structural fold (14). These highly conserved motifs adopt a bilobed architecture (Fig. 2). The larger lobe is a RecA-like domain that is found in other P-loop ATPases and contains the Walker A (WA) and B (WB) motifs, and functionally important aspartate (D-loop) and histidine (H-loop) residues. The smaller lobe, known as the α-helical subdomain, contains the ABC signature sequence (typically LSGGQ) and the Q-loop. The two lobes of the NBDs associate in an antiparallel sandwich dimer to form the two NBSs, each of which contains the WA and WB motifs of one NBD and the signature sequence of the other NBD. Thus, NBS1 comprises the WA and WB motifs of NBD1 and the signature sequence of NBD2, whereas NBS2 contains the WA and WB motifs of NBD2 plus the signature sequence of NBD1.

Figure 2

Structure of the NBSs. (A) Homology model of the NBSs of SUR1 based on the heteromeric structure of TM287/288 (18). (B) Schematic representation illustrating the functionally important regions of the NBSs.

In ABC proteins, the WA (or P-loop) motif is a phosphate-binding loop that contains a highly conserved lysine, which coordinates the β and γ phosphates of ATP. The WB motif contains a conserved aspartate residue that coordinates Mg2+ and is crucial for ATP hydrolysis (14). The D-loop (containing an SALD motif) is responsible for conformational changes at the catalytic site that favor ATP hydrolysis (15). The Q-loop is approximately eight residues long, with a conserved glutamine that is thought to move into and out of the active site during the hydrolytic cycle (14). The Q-loop may also be important for the transmission of binding/hydrolysis events from the NBDs to the TMDs (16). The ABC signature sequence is located at the N-terminal end of a long helix that directs the positive end of its dipole toward the γ-phosphate of ATP (14).

SUR contains consensus (NBS2) and degenerate (NBS1) binding sites, which are so named because of the conservation of their sequences (or lack of it) compared with other ABC family members. ATP hydrolysis occurs primarily at NBS2 (17). The crystal structure of a heterodimeric ABC transporter (TM287/288) with a consensus and degenerate site has been solved (18, 19). Unlike ABC proteins with two consensus NBSs (20, 21), its NBDs remain in contact even in the absence of nucleotide. This dimerization is supported by a hydrogen-bonding network that mainly involves the D-loop. The D-loops also facilitate cross-communication between the two NBSs of TM287/288 throughout the transport cycle. A similar interaction appears to occur in SUR1, where the D- and H-loops are predicted to participate in formation of the NBD heterodimer. Point mutations at the predicted interface have dramatic effects on KATP channel function (22).

Clearly, a high priority in the field is to obtain an x-ray structure of the NBSs of SUR. Given that each NBS contains elements of both NBDs, crystallization of single NBDs is not sufficient. Ideally, the structure of the entire KATP channel complex would be obtained. However, this poses a considerable technical challenge. To date, only a single particle electron microscope structure has been published (23); however, its resolution is too low to permit identification of helices, let alone amino acids.

Nucleotide interactions at the NBDs

Binding studies have suggested that in addition to being structurally distinct, the two NBSs of SUR are functionally different. Covalent 8-azido-[32P]ATP labeling was used to explore binding at NBS1 and NBS2 of various SUR subtypes (17, 24). These studies revealed that NBS1 bound ATP in an Mg2+-independent fashion, whereas nucleotide binding at NBS2 required Mg2+. Furthermore, NBS1 was radiolabeled regardless of whether the 32P label was on the α or γ phosphate of ATP. Only 8-azido-α-[32P]ATP labeled NBS2. This suggests that NBS2, but not NBS1, hydrolyzes ATP. Detailed discussions of experiments that addressed nucleotide interactions with the NBSs of SUR can be found elsewhere (25, 26).

Equilibrium model for KATP channel gating

In spite of the structural complexity of the KATP channel, with eight subunits and 12 nucleotide-binding sites (one inhibitory site on each Kir6.x subunit and two stimulatory sites on each SUR), the main features of nucleotide regulation of the KATP channel can be explained with a simplified equilibrium gating model (Fig. 3).

Figure 3

Equilibrium gating model of Kir6.2/SUR1. (A) Schematic representing the gating of KATP channels as three interacting domains: the pore, the inhibitory NBS on Kir6.2, and the NBSs of SUR1. (B) Gating scheme in (A) expanded into a cubic model. C and O designate the closed and open states of the pore domain, respectively. The subscripts designate the occupancy of the two NBSs as either unbound (0) or nucleotide bound (A). The first subscript refers to the inhibitory site on Kir6.2 and the second refers to the NBDs. When both inhibition and activation by nucleotides are present, the open probability (Po) is described by

where [ANP] is the nucleotide concentration and all other symbols are as described in the main text. (C) Cubic model (solid lines) fit to the nucleotide activation/inhibition data (open circles) from Proks et al. (33). Because the experimental data were normalized, the model was also normalized. For activation curves, this was achieved by subtracting the Po at 0 nucleotide (L/(L + 1)) and dividing the difference by the maximum Po (EL/(EL + 1)) minus the unliganded Po. For inhibition curves, the equation was divided by the Po in the absence of nucleotide (L/(L + 1)). When both activation and inhibition were present, the model was normalized to the Po in the absence of nucleotide. (D) Simplified scheme for KATP activation via Mg-nucleotide binding at SUR1. This schematic assumes that SUR1 can hydrolyze MgATP, but that hydrolysis is not necessary for channel activation by MgATP.

Our model, which is based on the gating scheme used by Horrigan and Aldrich (27) to describe activation of BK channels, considers the channel complex as three independent domains—the pore, the inhibitory site on Kir6.x, and a stimulatory site on SUR1—in coupled equilibrium (Fig. 3 A). The pore opens and closes in a concerted step described by the constant L (where L = [open]/[closed] = (Po/(1 − Po)) (28, 29). All four inhibitory binding sites on Kir6.x are treated as a single site, which is in either an inhibited nucleotide-bound state or a permissive unbound state, as described by the affinity constant K1 (K1 = [Inhibited]/([Permissive][ANP]), where ANP is either ATP or ADP) (30, 31). Likewise, all eight SUR NBSs are simplified as a single domain that exists in an unbound resting state or in an MgANP-bound active state, where K2 = [Activated]/([Resting][MgANP]). For the purposes of this model, we have ignored the effects of PIP2, which have been modeled elsewhere (32).

The interactions between these domains are described by three more constants: D, E, and F. D describes the allosteric interaction between the inhibitory binding site and the pore domain, and is <1. E describes the interaction between nucleotide binding at the NBDs and the pore, and is >1. F describes the direct interaction, if any, between the NBDs and the inhibitory binding site on Kir6.x. This scheme expands into the cubic model shown in Fig. 3 B.

To understand the contribution of the different domains to channel gating, it is necessary to isolate their individual interactions. The pore domain can be studied alone in the absence of nucleotide. The intrinsic open probability (i.e., the Po in nucleotide-free solution) is ∼40% for Kir6.2/SUR1 channels, but is subject to rundown following patch excision (33). One can isolate nucleotide inhibition at Kir6.2 by examining the inhibitory effect of nucleotide binding in the absence of Mg2+, because nucleotide activation requires Mg2+ (2, 34). When Mg-nucleotides are applied to wild-type KATP channels, they elicit a mixture of inhibition and activation. Thus, to study activation by itself, it is necessary to remove inhibition. This can be done by taking advantage of a Kir6.2 mutation (G334D) that renders Kir6.2 completely insensitive to inhibition by ATP concentrations up to 10 mM, but has a minimal effect on Po (33, 35, 36). When coexpressed with SUR1, Kir6.2-G334D enables Mg-nucleotide activation to be measured directly (33).

We used the equilibrium binding scheme to fit (solid lines, Fig. 3 C) macroscopic data from Proks et al. (33). By first considering data in which nucleotides caused only activation (Kir6.2-G334D/SUR1 channels) or only inhibition (wild-type channels in Mg-free solution), we were able to greatly simplify the model. We made the assumption that the Kir6.2-G334D mutation does not influence nucleotide handling by SUR1. Additionally, we assumed the same value for L in all fits. L was calculated from a Po of 0.15 in the absence of nucleotides. This is smaller than the Po of 0.37 ± 0.07 reported for wild-type channels (33). However, Po is extremely variable both between patches and between preparations, probably due to channel rundown and other variable biological processes. Furthermore, adjusting L is reasonable as this parameter primarily determines the magnitude of MgADP activation relative to the current in the absence of nucleotide, which also varies widely among different experiments (33, 37, 38, 39). In terms of energetics, the difference between the measured value (0.37) and our chosen value (0.15) for Po is <1 kcal/mol. Finally, we assumed that changes in macroscopic current were due to changes in Po and not the number of channels (N). At the single-channel level, both MgADP and MgATP increased the Po of Kir6.2-G334D/SUR1 by increasing the mean burst duration and reducing the frequency and number of interburst closed states (33).

Based on the increase in the single-channel Po of Kir6.2-G334D/SUR1 channels elicited by saturating concentrations of MgADP and MgATP (33), which was identical for both nucleotides, we fixed the value of E, which reflects coupling of the NBDs to the pore, at 9.5. To fit the activation-only data (i.e., for Kir6.2-G334D/SUR1), we fixed K1 at 0 to reflect a lack of binding to the inhibitory site. Therefore, the only free parameter was the affinity constant K2. To fit the inhibition-only data (i.e., for Kir6.2/SUR1 in Mg-free solution), we fixed D at 10−6 because this provided the best fit to the inhibition measured experimentally at 60 mM ATP (33). K2 was set to 0 because there is no binding to the NBDs in the absence of Mg2+. Thus, the only free parameter was K1.

We then used the values generated by fitting the activation-only or inhibition-only curves to model the response of wild-type channels to Mg-nucleotides. The only remaining free parameter was F, which describes the direct interaction between the SUR1-activation sites and the Kir6.x-inhibitory sites. We set F at 1 because varying it in either direction caused the model to deviate substantially from the data for MgADP activation.

What does the model predict?

Although the model makes several simplifying assumptions, it adequately describes the MgADP and MgATP sensitivities of Kir6.2/SUR1. We believe it may be useful as a heuristic tool to understand the gating mechanism of KATP channels. The model makes several predictions:

Based on both our model and the data reported in Proks et al. (33), we propose that MgADP and MgATP activate Kir6.2-G334D/SUR1 channels (and by implication Kir6.2/SUR1 channels) to the same extent. Therefore, the only difference in the activation-only curves is in the affinity constants for the two nucleotides. The predicted K2 values are 5.7 × 104 M−1 (KD = 18 μM) for MgADP and 3.93 × 103 M−1 (KD = 250 μM) for MgATP. Thus, MgADP binds more tightly than MgATP to one of the NBSs of SUR1 (most likely NBS2).

In wild-type channels, where both activation and inhibition are observed, MgADP produces a bell-shaped concentration-response curve, with distinct activation and inhibition phases (33, 37, 38, 39). In contrast, the MgATP concentration-response curve resembles that obtained for inhibition alone, with no apparent activation component but a slightly greater Ki. This implies a limited role, if any, for MgATP hydrolysis in direct channel gating. This behavior is reproduced in our model and is a consequence of the fact that MgATP has a lower affinity for the NBSs than does MgADP. At high nucleotide concentrations, where both the NBSs and the inhibitory binding sites are occupied, inhibition dominates because of differences in energetics. Based on the Proks et al. (33) data and the fits from our model, nucleotide binding to the inhibitory site stabilizes the closed state by ∼8 kcal/mol, whereas binding to the NBSs only contributes ∼1.3 kcal/mol to stabilize the open state.

The allosteric coupling between the NBSs and the inhibitory Kir6.2 binding site can be adequately explained by the coupling of both domains to the pore, without the need to invoke a direct, functional interaction between the two (i.e., F = 1). In other words, nucleotide binding to SUR does not alter nucleotide binding to Kir6.2 (and vice versa) except via changes in channel gating.

Hydrolysis of ATP at the NBDs is not required to explain the differences in gating between ATP and ADP. Our equilibrium model adequately represents the activation of Kir6.2/SUR1 channels without including any irreversible hydrolysis steps. This suggests that binding of MgATP (like MgADP) at NBS2 may directly promote channel opening in Kir6.2/SUR1 channels.

In almost all studies published to date, the experimental data were normalized in various ways to correct for rundown, and we have taken this into account in our modeling. If the rundown could be stabilized, however, and the data expressed as actual Po rather than normalized Po, the concentration-response curves for nucleotide effects on macroscopic currents would contain information regarding not only the concentration range over which nucleotides have their effects but also the energetic contributions that nucleotide activation or inhibition make to gating. This information would provide greater insight into the mechanism(s) underlying nucleotide modulation and the effects of disease-causing mutations in KATP channel subunits.

How valid are the model assumptions/predictions?

The main predictions of our model as they relate to Mg-nucleotide activation (as in Kir6.2-G334D/SUR1 channels) are shown in Fig. 3 D. As noted above, we made several simplifying assumptions. Chief among these was to treat the eight NBSs as a single unit with a single binding affinity constant. However, the reported number of SUR subunits required to activate KATP varies in the literature (40, 41, 42). We also modeled the inhibitory ATP binding sites as a single site, but this too is controversial (28, 31). Therefore, we also considered four alternatives (Fig. S1 in the Supporting Material). We modeled the NBSs as four separate subunits with cooperative binding and a concerted conformational change (Fig. S1 A), four subunits with independent binding and a concerted conformational change when all subunits are occupied (Fig. S1 B), and a Monod-Wyman-Changeux-type model in which binding to each subunit is independent and each subunit makes an identical energetic contribution to the pore domain (Fig. S1 C). We modeled the inhibitory Kir6.2 binding sites in the same way in each of these three models. We also fit the data with a mixed model in which the inhibitory Kir6.2 binding site was a single site and the binding sites on SUR1 were treated as four individual subunits, each with independent binding and gating contributions (Fig. S1 D). None of the alternative models fit the data sets as well as the model presented in Fig. 3 (compare Fig. S2, A–C, and Table S1), with the exception of the mixed model, which did not improve upon the fit to our initial model (compare Figs. 3 C, S2 D, and Table S1). None of the fits were improved by introducing terms to describe a direct interaction between the NBDs and the inhibitory binding sites (F).

The largest deviation between our simple allosteric model (Fig. 3) and the data from Proks et al. (33) is in the fit to wild-type channels at lower MgADP concentrations (1–30 μM), where activation dominates. Importantly, none of the alternative models improved on the fit over this concentration range. Therefore, the relatively poor agreement with the data is not a consequence of the simplifying assumption that the NBDs are treated as a single binding site. This deviation may instead challenge our assumption that there is no difference in the ability of SUR1 to stabilize the open state of the channel in wild-type versus Kir6.2-G334D channels. Alternatively, it may simply result from biological variability across experiments (33, 37, 38, 39).

As noted above, none of our fits had more than one free parameter. However, the values generated are only as accurate as our initial assumptions regarding L, D, and E. The value for E calculated from the single-channel open probability of Kir6.2-G334D/SUR1 channels in the absence and presence of Mg-nucleotides was 9.5 ± 1 (± SEM). We found that varying E between 8.5 and 10.5 (equivalent to E ± SEM) only affected the value generated for K2 by 7%. Changing the value of D by 100-fold had only minor effects (<<1%) on the K1 values generated. Changing our estimate for Po (and therefore L) to that reported for wild-type Kir6.2/SUR1 channels in the absence of nucleotide (0.37) had more significant effects, increasing our estimates of K1 by 35% and decreasing the K2 values generated from our fits by a factor of 1.8. Therefore, although we believe the major predictions of our model to be robust, some caution should be used in trusting the absolute value of the binding affinities it produces, unless Po can be closely monitored or controlled throughout an experiment.

What is the role of NBS1?

Our model does not discriminate between Mg-nucleotide binding to NBS1 and NBS2, but it has been proposed that nucleotide interactions with NBS2 primarily drive channel opening (3, 43, 44). What, then, is the role of NBS1?

A solely structural role is suggested by CFTR, an ABC transporter that is also a Cl−-permeable ion channel. As is the case in SUR1, NBS1 of CFTR is degenerate. In CFTR, channel opening is initiated by binding of MgATP to NBS2 and terminated by ATP hydrolysis at NBS2 (45). In contrast, ATP stays bound to NBS1 for several gating cycles (46, 47). Thus, bound ATP at NBS1 of CFTR can be considered a structural element that allows for dimerization of the NBDs when NBS2 is occupied (48).

In contrast, the functional effect of nucleotide binding to NBS1 of SUR1 is not completely understood. In part, this is because no measurements of equilibrium nucleotide binding exist. It is uncertain whether mutations in NBS1 influence ATP binding (as they do 8-azido-[32P]ATP binding) (49), and SUR2A appears to handle nucleotides differently from SUR1 (or SUR2B) (50, 51, 52). Furthermore, ATP hydrolysis properties differ when measured for isolated NBDs, SUR, or the whole KATP channel complex (23, 53). Thus, extrapolation from data obtained in isolated domains or subunits to KATP channel gating has to be treated cautiously.

Results from analyses of NBS1 mutations are inconclusive. Mutation of the Walker A lysine in NBS1 (K719 in SUR1) reduced high-affinity labeling of 8-azido-[α-32P]ATP (49), but it is unclear whether the same is true of ATP binding. It is possible that this mutation does not alter ATP binding, as it decreased ATP hydrolysis by SUR1 without affecting the Km for ATP (53) and reduced MgATP activation without affecting the EC50 (39). Although the K719A mutation strikingly shifted the EC50 for MgADP activation (39), this may have been mediated via an allosteric effect on NBS2, perhaps by preventing NBD dimerization. Complete deletion of NBD1 produced channels that were not activated by MgADP and were either less stable at the plasma membrane or not correctly trafficked (54). On balance, the available data suggest that NBD1 of SUR1 has an important structural role, but its influence on channel activity remains obscure. If it is indeed merely a necessary structural element, then what is being measured in experiments, and in our model, may be occupancy of NBS2.

Interestingly, mutation of the Walker A lysine (K707A) of NBS1 of SUR2A had no effect on Mg-nucleotide activation (52), which may support the idea that, in this case at least, ATP handling at NBS1 does not influence channel activity. Why the same mutation in SUR2B, which differs from SUR2A only in its last 42 amino acids, prevents MgADP activation is a puzzle.

Is MgATP hydrolysis necessary for activation?

In our model, hydrolysis of MgATP is not required for channel activation, implying that both MgADP and MgATP can promote channel activation simply by binding more tightly to the active conformation of the NBDs. The act of hydrolysis per se is not necessary to promote channel opening, as MgADP is able to activate. However, it remains a possibility that MgATP must first be hydrolyzed to MgADP to activate channels.

Numerous studies have shown that isolated NBD2 (and to a lesser extent NBD1) of SUR1 and SUR2 can hydrolyze ATP (25, 43, 50, 53), as can SUR1 (53) and Kir6.2/SUR1 (23). In the case of Kir6.2/SUR2A, conditions that lock the channel in a posthydrolytic state favor channel opening (43). Orthovanadate, which occupies the same position as the γ-phosphate of ATP, inhibits ATPase activity and stimulates channel opening in the presence of ATP. This suggests that the MgADP·Pi state is able to stabilize channel opening. Likewise, MgADP is able to stimulate channel opening. It is worth noting, however, that the orthovanadate effect develops much more slowly than direct activation of Kir6.2-G334D/SUR1 by MgATP or MgADP. BeF, which arrests the ATPase cycle in a prehydrolytic configuration, impairs the ATPase activity of SUR1 and SUR2 (53, 55) and reduces Kir6.2/SUR2A channel opening produced by MgADP (43). This suggests that the MgADP·BeF state, analogously to a prehydrolytic ATP-bound state, does not induce channel activity, but it does not necessarily imply that MgATP is unable to directly stimulate channel opening, perhaps by stabilizing a state that resembles a posthydrolytic conformation of the NBDs.

Mutation of the catalytic residues of NBS2 of SUR1 impairs ATPase activity (53). If these mutations only prevent hydrolysis, they would be expected to affect activation by MgATP, but not MgADP. However, mutation of the Walker A lysine or Walker B aspartate in NBS2 of SUR1 impairs activation by both MgADP and MgATP (2, 4, 34); indeed, it appears to have a larger effect on MgADP activation (39). This raises the possibility that these mutations affect either nucleotide binding directly or the conformational change that couples binding to activation. Thus, the catalytic residues in the NBDs of SUR may have functional effects beyond MgATP hydrolysis. Experiments in which SUR1 was expressed without Kir6.2 showed that binding of ATP (or ATP analogs) under nonhydrolytic conditions (e.g., no Mg2+, catalytically dead mutations) was still competent to cause a conformational change in SUR1 as assayed by a change in glibenclamide affinity (56, 57). Therefore, although it seems likely that KATP channels can hydrolyze MgATP, this may not be an absolute requirement for channel activation via SUR. In future studies, single-channel and kinetic analyses could be used to identify KATP gating transitions that are not at equilibrium (e.g., irreversible hydrolysis steps), as has been demonstrated for CFTR (45).

Interactions between nucleotide binding at SUR and Kir6.x

The final question is whether or not there is a direct, functional interaction between the NBDs of SUR1 and the inhibitory site of Kir6.2. Numerous studies (58, 59, 60, 61, 62, 63) have proposed such an interaction, but there is little direct evidence. Our model predicts that such an interaction is not necessary to explain the data.

How, then, might nucleotide occupancy at the NBSs be relayed to the pore? In the crystal structure of the ABC transporter McJD (64), the first cytosolic loop of TMD1 (and TMD2) contacts both NBDs, whereas the second interacts only with the NBD from the opposite monomer. It is possible the corresponding cytosolic loops (4, 7) in SUR may form coupling helices that transmit conformational changes in the NBSs (produced by nucleotide binding/activation) to the TMDs. Both the cytoplasmic N-terminal (58, 61, 63) and C-terminal (60) regions of Kir6.2 have been proposed to interact with SUR. Likewise, many regions of SUR, including TMD0 of SUR1 (59), the loop connecting TMD0 to TMD1 of SUR1 (61), and the linker between TMD2 and NBD2 of SUR2A (62), have been suggested to modify gating of Kir6.2. To date, however, none of these interacting regions have been shown to directly transduce binding of Mg-nucleotides at SUR to gating changes at Kir6.2.

Conclusions

Although there has been much work on the structure and function of the NBSs of SUR and how they regulate KATP channel activity, our understanding is still far from complete. This is in part because there are few experiments in which nucleotide binding and ATPase activity have been measured in the intact KATP channel complex. The available results suggest that lack of Kir6.2, or the TMDs of SUR, influences nucleotide handling by the NBSs (compare (23) and (53)). Thus, it is hard to compare electrophysiological and biochemical data directly. A further problem is that SUR1 and SUR2 show differences in nucleotide handling—again, why they do so remains unclear. Although a high-resolution structure of the KATP channel, or even SUR, will undoubtedly provide many novel insights, it is unlikely to resolve the issue of precisely how MgATP binding/hydrolysis at NBS2 opens the channel. Functional data on nucleotide handling will still be required. We look forward to seeing how the results of such diverse approaches will combine to elucidate the working of the KATP channel, and hope that our (admittedly simple) model will provide testable predictions for experiments to come.

Author Contributions

M.C.P. performed the modeling. N.V., F.M.A., and M.C.P. wrote the manuscript.