Identification of a Functionally Important Negatively Charged Residue Within the Second Catalytic Site of the SUR1 -Binding Domains Jeff D. Campbell,1,2 Peter Proks,1 Jonathan D. Lippiat,1 Mark S.P. Sansom,2 and Frances M. Ashcroft1

؉ Two very different subunits make up the K The ATP-sensitive K channel (KATP channel) couples ATP glucose metabolism to secretion in pancreatic channel (3). The pore is composed of four inwardly recti- ϩ ␤-cells. It is comprised of receptor (SUR)-1 fying K channel (Kir channel) subunits (Kir6.2), and each and Kir6.2 . Binding of Mg to the of these is associated with a (SUR) nucleotide-binding domains (NBDs) of SUR1 stimulates subunit (SUR1), which modulates the opening and closing channel opening and leads to membrane hyperpolariza- (gating) of the channel. Metabolically generated changes tion and inhibition of insulin secretion. To elucidate the in intracellular nucleotides influence channel gat- structural basis of this regulation, we constructed a molecular model of the NBDs of SUR1, based on the ing via interaction with both types of subunit. Thus, ATP crystal structures of mammalian proteins that belong to (and ADP) closes the channel by binding to Kir6.2, where- the same family of ATP-binding cassette transporter as channel opening is mediated by interaction of Mg nu- proteins. This model is a dimer in which there are two cleotides (MgATP and MgADP) with SUR1 (4,5). nucleotide-binding sites, each of which contains resi- SUR belongs to the ATP-binding cassette (ABC) family dues from NBD1 as well as from NBD2. It makes the of transporter proteins (6). Like other ABC transporters, it novel prediction that residue D860 in NBD1 helps coor- has two sets of transmembrane domains (TMDs), each dinate Mg nucleotides at site 2. We tested this predic- containing six membrane spanning helices, and two nucle- tion experimentally and found that, unlike wild-type channels, channels containing the SUR1-D860A muta- otide binding domains (NBDs). It also possesses an addi- tion were not activated by MgADP in either the pres- tional NH2-terminal set of five transmembrane helices (Fig. ence or absence of MgATP. Our model should be useful 1A). The NBDs of all ABC proteins are highly conserved for designing experiments aimed at elucidating the and contain several motifs involved in ATP binding and relationship between the structure and function of the hydrolysis, including a Walker-A (WA) and Walker-B (WB) KATP channel. Diabetes 53 (Suppl. 3):S123–S127, 2004 motif linked by a “signature” sequence (LLSGGQ) and two shorter sequences containing conserved glutamine (Gln- loop) and histidine (His-loop) residues (Fig. 1A). The ϩ crystal structures of a number of isolated NBDs suggest he ATP-sensitive K channel (KATP channel) plays a crucial role in insulin secretion by cou- that their topology is also highly conserved across both pling ␤-cell metabolism to insulin secretion (1). prokaryotes and eukaryotes. Furthermore, the X-ray struc- tures show that the NBDs form a dimer containing two At low glucose levels, the KATP channel is open, T nucleotide-binding pockets. These do not correspond to keeping the resting membrane potential hyperpolarized and preventing electrical activity and insulin secretion (2). the NBDs encoded in the primary sequence, however, Increased glucose metabolism results in closure of the because the ATP molecules are sandwiched between the ␤ KATP channels, leading to a depolarization of the -cell WA of one NBD and the signature sequence of the opposite membrane, initiation of electrical activity, opening of NBD (Fig. 1B). For this reason, we henceforth refer to the voltage-gated channels, and an influx of Ca2ϩ ions two nucleotide-binding pockets as site 1 (Walker motifs of that triggers the release of insulin granules. NBD1, linker NBD2) and site 2 (Walker motifs of NBD2, linker NBD1). Because the sequences of NBD1 and NBD2 of SUR1 are From the 1University Laboratory of Physiology, University of Oxford, Oxford, U.K.; and 2Laboratory of Molecular Biophysics, University of Oxford, Oxford, not identical, the structure of these two sites will also not U.K. be identical. This may provide a structural explanation for Address correspondence and reprint requests to Prof. Frances M. Ashcroft, University Laboratory of Physiology, University of Oxford, Parks Rd., Oxford, the observed differences in function between the two sites OX1 3PT, U.K. E-mail: [email protected]. of SUR1 (8). As described previously (7), it is possible to Received for publication 12 March 2004 and accepted in revised form equate the properties described for NBD1 in functional 25 May 2004. This article is based on a presentation at a symposium. The symposium and studies with those of site 1 and to equate those described the publication of this article were made possible by an unrestricted educa- for NBD2 with site 2 (8). Therefore, site 1 binds nucleo- tional grant from Servier. ABC, ATP-binding cassette; CFTR, cystic fibrosis transmembrane regulator; tides with higher affinity than site 2 (half-maximal inhibi- ϩ ϩ ␮ KATP channel, ATP-sensitive K channel; Kir channel, inwardly rectifying K tory concentration [IC50] of 4 and 60 mol/l for ATP and 26 channel; NBD, nucleotide-binding domain; SUR, sulfonylurea receptor; TAP, and 100 ␮mol/l for ADP at sites 1 and 2, respectively). In transporter associated with antigen processing; TMD, transmembrane do- addition, ATP is preferentially hydrolyzed at site 2. main; WA, Walker-A; WB, Walker-B. © 2004 by the American Diabetes Association. To examine these differences in nucleotide handling

DIABETES, VOL. 53, SUPPLEMENT 3, DECEMBER 2004 S123 KATP CHANNEL REGULATION BY SUR1

FIG. 1. A: Left, the transmembrane topology of SUR1. A schematic diagram of SUR1 showing the TMDs and the NBDs. Right, schematic showing the structure of the NBD of an ABC transporter. B: Model of the NBDs of SUR1 as viewed from the extracellular surface. The model shows the nucleotide sandwich dimer and the

conserved motifs. The WA motifs, WB motifs, and signa- ture sequence are yellow, orange, and green, respec- tively. The ATP molecule is shown in red, and Mg2؉ is shown in gray. NBD1 is blue and NBD1 is purple. Sites 1 and 2 are indicated. C: Expanded view of site 2 of the SUR1 model showing the proximity of D860 to the ATP ␥-phosphate in Site 2. further, we built a molecular model of the SUR1-NBD likely to be slight structural differences between the two heterodimer using a combination of mammalian and bac- NBDs. To take account of these differences, we used two terial crystal structures as templates. We then mutated different templates to construct a molecular model of the residues putatively identified in the model as being in- SUR1 heterodimer: NBD1 was modeled on the crystal volved in nucleotide handling or catalysis and tested their structure of NBD1 of mouse CFTR (15) and NBD2 was functional effects on KATP channel activity. modeled on the X-ray structure of the human TAP1 NBD (16). The two NBD models were then assembled in the RESEARCH DESIGN AND METHODS nucleotide sandwich dimer configuration by least-squares Modeling. The sequences of NBD1 and NBD2 of SUR1 were aligned with fitting to the bacterial NBD homodimer MJ0796 (11). The those of CFTR (cystic fibrosis transmembrane regulator) (protein database CFTR-NBD1 and TAP1-NBD sequences share 33 and 26% [PDB] code 1ROZ) and TAP1 (transporter associated with antigen processing) (PDB code 1JJ7), respectively, using ClustalW (9) and individual monomers sequence identity to the NBD1 and NBD2 of SUR1, respec- generated using MODELLER 6 (version 2) (10). The individual NBD models tively, which is the highest homology currently available were then least-squares fitted to the MJ0796 dimer crystal structure (PDB for NBDs of known structure. Additionally, the CFTR- code 1L2T) (11). Molecular graphics diagrams were generated using VMD NBD1 and TAP1-NBD2 structures were cocrystallized with (visual molecular dynamics) (12) and PovRay (http://www.povray.org). Molecu- lar dynamic simulations were performed as previously described (13). MgATP and MgADP, respectively. Thus, our model will Electrophysiology. Xenopus oocytes were defolliculated and coinjected represent a structure with ATP docked at NBD1 and ADP with ϳ5 ng each of mRNAs encoding Kir6.2 (GenBank accession no. D50581) docked at NBD2. It is thought that this corresponds to the and either wild-type or mutant SUR1 (GenBank accession no. L40624). The “active” conformation of SUR. final injection volume was ϳ50 nl per oocyte. Isolated oocytes were main- tained as previously described (14), and currents were studied 1–4 days after Predictions of the molecular model. Figure 1B shows injection. Macroscopic currents were recorded from giant excised inside-out the model of the NBD heterodimer of SUR1, with ATP patches at 20–24°C (14). The pipette solution contained (in mmol/l) 140 KCl, docked at site 1 and ADP docked at site 2. Given that site 1.2 MgCl , 2.6 CaCl , and 10 HEPES (pH 7.4 with KOH), and the internal (bath) 2 2 2 shows an enhanced ability to hydrolyze ATP (17), we solution contained 110 KCl, 1.44 MgCl2, 30 KOH, 10 EGTA, 10 HEPES (pH 7.2 with KOH), and nucleotides as indicated. Currents were recorded with an searched for negatively charged residues proximal to the Axopatch 200B amplifier (Axon Instruments), filtered at 0.2 kHz and sampled phosphate tail of ATP that could participate in coordina- at 0.5 kHz. tion and hydrolysis of the nucleotide. Two negatively charged residues, D1505 and E1506 in the WB motif of RESULTS AND DISCUSSION NBD2, were observed to line site 2. Mutation of either of Modeling the SUR1 NBD heterodimer. Because the these residues abolishes the ability of MgADP to stimulate sequence of NBD1 differs from that of NBD2, there are channel activity (5,18), and E1506 is of particular interest

S124 DIABETES, VOL. 53, SUPPLEMENT 3, DECEMBER 2004 J.D. CAMPBELL AND ASSOCIATES

FIG. 2. A: Macroscopic currents recorded from oo- cytes coexpressing Kir6.2 and either SUR1, or SUR2B- D860A in response to a series of voltage ramps from ؊110 to ؉100 mV. 100 ␮mol/l ADP and 100 ␮mol/l ATP were added to the intracellular solution as indi- :cated by the bars. All solutions contained Mg2؉. B

Mean KATP conductance recorded in response to 100 ␮mol/l ADP, 100 ␮mol/l ATP, or 100 ␮mol/l ADP ؉ 100 ␮mol/l ATP in patches excised from oocytes coex- pressing Kir6.2 and either wild-type SUR1 (Ⅺ)or SUR1-D860A (f). The slope conductance (G) is ex- pressed as a fraction of the mean of that obtained in control solution before exposure to MgADP (Gc). The number of patches is three in each case. because its mutation to (E1506K) leads to congen- mutation abolishes the stimulatory effect of MgADP on the ital hyperinsulinism in humans (18). Our model also iden- channel and unmasks the inhibitory effect of ADP that is tified a third negatively charged residue, D860, which lies mediated via the inhibitory site on Kir6.2 (19). ϳ5 Å from the ␥-phosphate and could potentially interact Addition of 100 ␮mol/l MgATP produced a greater block with ATP (Fig. 1C). It is noteworthy that, in contrast to the of Kir6.2/SUR1-D860A channels than of wild-type chan- other negatively charged residues, which line site 2, D860 nels: 95 Ϯ 1% (n ϭ 3) compared with 85 Ϯ 5% (n ϭ 3) (Fig. is contributed by the opposite NBD. The equivalent resi- 2). This has also been observed for mutation of the WA due in NBD2 is D1512, which lines site 1. To examine lysine (20) or WB aspartate (5) in either NBD1 or NBD2. It whether D860 and D1512 are of functional importance, we can be attributed to loss of MgATP hydrolysis at the NBDs mutated them, individually, to alanine. of SUR1, which, in the case of the wild-type channel, Effects of the D860A mutation on Mg nucleotide enhances channel activity and shifts the relationship be- activation. Figure 2 shows that MgADP stimulates KATP tween channel inhibition and ATP concentration to higher channel activity approximately twofold in the absence of ATP levels. other nucleotides, in accordance with previous results MgADP was unable to stimulate the activity of channels (14). In contrast, when D860 in NBD1 of SUR1 was carrying the D860A mutation in the presence of ATP, in mutated to alanine, MgADP blocked rather than stimu- contrast to wild-type channels (Fig. 2). The equivalent lated channel activity. The extent of block (62 Ϯ 5%, n ϭ mutation in NBD2 of SUR1 (D1512A) did not result in 5) is similar to that seen in the absence of Mg2ϩ for functional channels. wild-type channels (ϳ60%) (14). This suggests that the Role of D860. Taken together, our results demonstrate a

DIABETES, VOL. 53, SUPPLEMENT 3, DECEMBER 2004 S125 KATP CHANNEL REGULATION BY SUR1

because an equivalent residue (D1512) lines site 1. How- A ever, there are likely to be differences in the way that D860 and D1512 function at their respective binding sites. This is because a nonconserved residue, R1379, which is NH2-terminal to the WA motif of site 2, lies directly adjacent to D860 (Fig. 3A), whereas no such positively charged residue exists in site 1. Mutation of the equivalent arginine residue in SUR2B (R1344) impaired ADP-induced channel activation as effectively as mutation of the WA lysine (21). This suggests both R1379 and D860 play an important functional role: potentially, R1379 has the ability to undergo alternating charge-charge interactions with ATP and D860. We investigated this hypothesis by molecular dynamics simulations of our SUR1 NBD model. We ran four simula- tions: the wild-type model with 1) ATP docked at both NBDs, or 2) ATP docked at NBD1 and ADP docked at NBD2, and 3, 4) the equivalent simulations of the D860A mutant. The wild-type simulations demonstrate that R1379 preferentially interacts with ADP rather than ATP (Fig. 3B). Furthermore, the D860A simulations demonstrate that removal of the negatively charged aspartate disrupts the interaction of R1379 with nucleotide; in contrast to the B wild-type simulations, R1379 more strongly interacts with the ATP than with ADP. The reduction in the predicted interaction of R1379 with ADP following mutation of D860 is consistent with the impaired ability of MgADP to stimulate channel activity (Fig. 2).

CONCLUSION We have demonstrated that an approach that combines molecular modeling predictions, simulations, and func- tional analysis is useful for understanding the structural biology of the NBDs of SUR1. Our results define a new residue, D860, which contributes to the complex mecha- nism by which nucleotide binding to these NBDs results in opening of the KATP channel.

ACKNOWLEDGMENTS We thank the Wellcome Trust and the Royal Society for support. J.D.C. is a Wellcome Trust Structural Biology FIG. 3. A: Detailed view of the interaction between R1379 and D860 in D.Phil student and a Linacre College Canadian National site 1. B: Molecular dynamics simulations of mutant and wild-type SUR scholar, and F.M.A. is the GlaxoSmithKline Research NBD models. The simulation system contained water molecules and Professor of the Royal Society. counter ions (not shown) and was simulated for a 1-ns equilibration run followed by a 1-ns production run from which data were analyzed. The distance between the R1379 side-chain nitrogens and the ATP ␤-phosphate is plotted for the wild-type model with ATP docked at both REFERENCES NBDs (blue), the wild-type model with ATP docked at NBD1 and ADP 1. Ashcroft FM, Rorsman P: Type-2 diabetes mellitus: not quite exciting docked at NBD2 (green), the D860A mutant model with ATP docked at enough? Hum Mol Genet 13:R21–R31, 2004 both NBDs (red), and the D860A mutant model with ATP docked at 2. Ashcroft FM, Harrison DE, Ashcroft SJ: Glucose induces closure of single NBD1 and ADP docked at NBD2 (black). potassium channels in isolated rat pancreatic beta-cells. Nature 312:446– 448, 1984 critical role for D860 in nucleotide activation of the KATP 3. Seino S, Miki T: Physiological and pathophysiological roles of ATP- channel. Because MgADP activation was abolished, the sensitive Kϩ channels. Prog Biophys Mol Biol 81:133–176, 2003 effect of this mutation cannot be attributed solely to loss of 4. Tucker SJ, Gribble FM, Zhao C, Trapp S, Ashcroft FM: Truncation of Kir6.2 MgATP hydrolysis. Instead, it must prevent either binding produces ATP-sensitive Kϩ channels in the absence of the sulphonylurea of Mg nucleotides to site 2 of SUR1 or transduction to receptor. Nature 387:179–183, 1997 5. Gribble FM, Tucker SJ, Haug T, Ashcroft FM: MgATP activates the beta Kir6.2 of the conformational change consequent on nucle- cell K-ATP channel by interaction with its SUR1 subunit. Proc Natl Acad otide binding. Additional experiments are required to SciUSA95:7185–7190, 1998 determine which of these possibilities is correct. 6. Higgins CF: Abc transporters: from microorganisms to man. Annu Rev Cell Although we cannot say whether D860 simply partici- Biol 8:67–113, 1992 7. Campbell JD, Sansom MSP, Ashcroft FM: Structural insights into the pates in nucleotide binding or also acts as a hydrolytic function of the nucleotide-binding domain of the human sulphonylurea residue, the presence of D860 alone is not responsible for receptor. EMBO Rep 4:1038–1042, 2003 the functional differences between site 1 and site 2. This is 8. Matsuo M, Tanabe K, Kioka N, Amachi T, Ueda K: Different binding

S126 DIABETES, VOL. 53, SUPPLEMENT 3, DECEMBER 2004 J.D. CAMPBELL AND ASSOCIATES

properties and affinities for ATP and ADP among sulfonylurea receptor Shriver S, Thibodeau PH, Thomas PJ, Zhang M, Zhao X, Emtage S: subtypes, SUR1, SUR2A, and SUR2B. J Biol Chem 275:28757–28763, 2000 Structure of nucleotide-binding domain 1 of the cystic fibrosis transmem- 9. Thompson JD, Higgins DG, Gibson TJ: CLUSTAL W: improving the brane conductance regulator. EMBO J 23:282–293, 2004 sensitivity of progressive multiple sequence alignment through sequence 16. Gaudet R, Wiley DC: Structure of the ABC ATPase domain of human TAP1, weighting, position-specific gap penalties and weight matrix choice. Nu- the transporter associated with antigen processing. EMBO J 20:4964–4972, cleic Acid Res 22:4673–4680, 1994 2001 10. Sali A, Blundell TL: Comparative protein modelling by satisfaction of 17. Matsuo M, Kioka N, Amachi T, Ueda K: ATP binding properties of the spatial restraints. J Mol Biol 234:779–815, 1993 nucleotide-binding folds of SUR1. J Biol Chem 274:37479–37482, 1999 11. Smith PC, Karpowich N, Millen L, Moody JE, Rosen J, Thomas PJ, Hunt JF: 18. Huopio H, Reimann F, Ashfield R, Komulainen J, Lenko HL, Rahier J, ATP binding to the motor domain from an ABC transporter drives Vauhkonen I, Kere J, Laakso M, Ashcroft F, Otonkoski T: Dominantly formation of a nucleotide sandwich dimer. Mol Cell 10:139–149, 2002 inherited hyperinsulinism caused by a mutation in the sulfonylurea recep- 12. Humphrey W, Dalke A, Shulten K: VMD: visual molecular dynamics. J Mol tor type 1. J Clin Invest 106:897–906, 2000 Graph 14:33–38, 1996 19. Tucker SJ, Gribble FM, Proks P, Trapp S, Ryder TJ, Haug T, Reimann F, 13. Campbell JD, Biggin PC, Baaden M, Sansom MSP: Extending the structure Ashcroft FM: Molecular determinants of K-ATP channel inhibition by ATP. of an ABC transporter to atomic resolution: modelling and simulation EMBO J 17:3290–3296, 1998 studies of MsbA. Biochemistry 42:3666–3673, 2003 20. Gribble FM, Tucker SJ, Ashcroft FM: The essential role of the Walker A 14. Gribble FM, Ashfield R, Ammala C, Ashcroft FM: Properties of cloned motifs of SUR1 in K-ATP channel activation by Mg-ADP and diazoxide. ATP-sensitive Kϩ currents expressed in Xenopus oocytes. J Physiol EMBO J 16:1145–1152, 1997 (Lond) 498:87–98, 1997 21. Matsushita K, Kinoshiita K, Matsuoka T, Fujita A, Fujikado T, Tano Y, 15. Lewis HA, Buchanan SG, Burley SK, Conners K, Dickey M, Dorwart M, Nakamura H, Kurachi Y: Intramolecular interaction of SUR2 subtypes for Fowler R, Gao X, Guggino WB, Hendrickson WA, Hunt JF, Kearins MC, intracellular ADP-induced differential control of K(ATP) channels. Circ Lorimer D, Maloney PC, Post KW, Rajashankar KR, Rutter ME, Sauder JM, Res 90:554–561, 2002

DIABETES, VOL. 53, SUPPLEMENT 3, DECEMBER 2004 S127