Islets and Type 2 Diabetes

Structure–function relationships in the β-cell KATP channel M. V. Mikhailov, E. A. Mikhailova and S. J. H. Ashcroft1 Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, Headington, Oxford OX3 9DU, U.K.

potential (for reviews see [1–4]). In response to an increase in blood glucose concentration, the enhanced rate of metabolism of the sugar within the β-cell causes changes in intracellular ATP and ADP concentrations, which lead to channel closure. The resulting membrane depolarization opens voltage-dependent Ca#+-channels and Ca#+ entry ensues, which triggers insulin release. Sulphonylureas cause similar closure of KATP channels by binding directly to the channel [5]. Diazoxide also binds to the channel to cause channel opening, hyperpolarization and insulin release [6]. The β-cell KATP channel contains two subunits, Kir6.2, an inwardly rectifying K+-channel that forms the pore, and SUR1, which contains the binding sites for sulphonylureas and diazoxide, and functions as a channel regulator. From hydropathy plots and by comparison with other members of the ATP-binding cassette protein family, SUR1 is predicted to have 17 transmembrane sequences (TMs) that are arranged in three groups of five, six and six transmembrane domains (TMDs), which are referred to as TMD0, TMD1 and TMD2 respectively [7]. Two large intracellular loops contain nucleotide-binding domains (NBDs) – NBD1 and NBD2. It is suggested that Kir6.2 possesses two TM regions together with a membrane-inserted region (or P-region), which is believed to constitute the actual pore [8]. There is good evidence that the functional channel contains four molecules each of SUR1 and Kir6.2 [9–11]. Our present studies are aimed at delineating the molecular interactions that are involved in assembly and ligand binding by KATP channel proteins. Our studies utilized a large number of recombinant SUR1 molecules. For identification, localization and purification purposes we placed tags at the C-terminus of SUR1 and at the N- or C-terminus of Kir6.2. These included green fluorescent protein (GFP) for localization within cells, and His (six histidine residues) for detection ' on gels and purification. We have also constructed a chimaeric molecule in which the C-terminus of SUR1 is linked to the N-terminus of Kir6.2 via a short linker peptide. This chimaera is active as a KATP channel [12].

Abstract The ATP-sensitive potassium (KATP) channel plays a key role in controlling β-cell membrane potential and insulin secretion. The channels are composed of two subunits, Kir6.2, which forms the channel pore, and SUR1, which contains binding sites for nucleotides and sulphonylureas and acts as a channel regulator. Our current studies are aimed at delineating the molecular interactions involved in assembly and ligand binding by KATP channel proteins. We have employed a complementation approach in which SUR1 halfmolecules are co-expressed in insect cells using a baculovirus system. Together with data from truncated SUR1 molecules and a fusion protein in which SUR1 is linked to Kir6.2, we have interpreted our findings in terms of a model for the structure of the KATP channel. The main features of the model are : (i) the C-terminal end of SUR1 is close to the N-terminus of Kir6.2; (ii) the two nucleotide binding domains (NBDs) of SUR1 – NBD1 and NBD2 – are in proximity; (iii) transmembrane helix 12 of SUR1 is orientated in such a way that it can make contact with Kir6.2; (iv) formation of the glibenclamide binding site requires that the two cytosolic loops (CLs) CL3 and CL8 are located close to each other; (v) there are homomeric interactions between the NBD1 domains of neighbouring subunits. We suggest that binding of glibenclamide leads to conformational changes in CL3 and CL8 leading to rearrangement of transmembrane helices. These effects are transmitted to Kir6.2 to result in channel closure.

Introduction The ATP-sensitive potassium (KATP) channel plays a key role in controlling β-cell membrane Key words : baculovirus, glibenclamide, insulin secretion, nucleotide-binding domain. Abbreviations used : CL, cytosolic loop ; GFP, green fluorescent protein ; KATP channel, ATP-sensitive potassium channel ; NBD, nucleotide-binding domain ; TM, transmembrane sequence ; TMD, transmembrane domain ; TMD, transmembrane domain. 1 To whom correspondence should be addressed (e-mail stephen.ashcroft!ndcls.ox.ac.uk).

323

# 2002 Biochemical Society

Biochemical Society Transactions (2002) Volume 30, part 2

We have also constructed a large number of part molecules and truncated versions of SUR and have engineered a set of deletion mutants in which pairs of TM regions are removed from TMD1. The structure of SUR1 and the sites at which we have made deletions and cuts are illustrated in Figure 1. Most of the studies described here have utilized a baculovirus expression system to express recombinant proteins in cells of the insect Spodoptera frugiperda [12]. This has three major advantages. First, the very active baculovirus late promoter drives the synthesis of large amounts of protein so that putting our constructs under control of such a promoter gives us very high levels of expression. Secondly, it is easy to infect insect cells with several viruses simultaneously. Thirdly, and most importantly, the individual components of the channel, Kir6.2 and SUR1, can be expressed and studied individually [13]. This is in contrast with the situation in mammalian cells where targeting signals prevent expression of SUR1 or Kir6.2 alone at the plasma membrane [14]. Using the baculovirus system, we have shown that SUR1 half-molecules retain the ability to locate each other and re-assemble into active SUR1 [13,15,16]. It was, of course, essential for interpretation of the experiments to be sure that half-molecules do indeed reach the plasma membrane. This was demonstrated using GFP-tagged molecules [13]. Whereas GFP alone is randomly distributed within insect cells after infection, both N- and C-terminal SUR1 half-molecules were shown to be clearly targeted to the plasma membrane.

Importance of NBD1 for SUR1 assembly Substantial glibenclamide-binding activity was formed when two SUR1 half-molecules, divided after NBD1, were co-expressed. However, if two half-molecules, both of which lacked NBD1, were co-expressed no activity was observed [15]. Thus, NBD1 is important for SUR1 assembly and for the generation of glibenclamide-binding activity. This does not mean that NBD1 is involved in forming the actual binding site. Some clues as to the structural role of NBD1 were obtained from studies on purified NBD1 [13]. We found that purified NBD1 self-aggregates and a tetrameric form predominates, as determined by both gel filtration and sucrose-gradient ultracentrifugation. Since the intact KATP channel is a bi-tetrameric structure this suggests that selfinteractions of NBD1 play a role in stabilizing the structure of the KATP channel.

Importance of NBD2 for SUR1 assembly By contrast, similar experiments with NBD2 indicated that NBD2 does not have a critical role in the self-assembly of SUR1. There was little or no reduction in glibenclamide binding after coinfection of a SUR1 N-terminal half-molecule with a C-terminal moiety lacking NBD2 [16]. Furthermore, the glibenclamide-binding activity of whole SUR1 did not decrease when NBD2 was removed [13]. However, we have been able to show that significant interactions exist between NBD1 and

Figure 1 Topology of SUR1 and sites used for modification The diagram illustrates the three groups of TM regions of SUR1. The arrows indicate the sites at which SUR1 was divided or modified in the present study.

# 2002 Biochemical Society

324

Islets and Type 2 Diabetes

NBD2, in experiments using GFP-tagged SUR1 domains [13]. It was shown that NBD1–GFP is distributed throughout the insect cell. NBD2– GFP could be localized to the plasma membrane if the preceding TM regions of TMD2 are attached. When this TMD2–NBD2 construct (without GFP) was co-expressed with NBD1–GFP, which is distributed throughout the cell, we found that there was clear membrane localization of the NBD1–GFP to the plasma membrane. This indicates that there is interaction between NBD1 and NBD2. This was confirmed by a control experiment in which the same TM regions without NBD2 did not cause membrane localization of NBD1–GFP.

co-expressed with the corresponding C-terminal half-molecule [16]. Binding activity was retained when we removed TM12 and TM13. When we took the same construct and removed NBD2 we also retained activity. We made further cuts, first to remove cytosolic loop (CL) CL7, and then to remove TM17, and in each case no loss of binding activity was observed. Finally, we removed TM16 and activity was abolished. Thus the presence of TM14–TM15–CL8–TM16 in the C-terminal SUR1 half-molecule is sufficient to allow SUR1 assembly and formation of the glibenclamide-binding site.

Role of TM12 in SUR1 assembly When an N-terminal SUR1 half-molecule that contains TM12 and TM13 was co-expressed with the corresponding C-terminal half-molecule, glibenclamide-binding activity developed [15]. This was also observed when the N-terminal half contained TM12 and the C-terminal half contained TM13. If both TM12 and TM13 were deleted, binding activity was retained. However, if TM12 was retained but TM13 was deleted, binding activity was removed totally. We interpret these data as follows. TM12 and TM13 are not critical for assembly of SUR1 ; however, when TM12 is present by itself, it is unstable and disrupts the structure. It needs to interact with TM13 to stabilize the structure. We also found evidence for interaction of Kir6.2 with SUR1 and for the involvement of TM12 in this interaction [15]. Glibenclamidebinding activity generated by co-expression of an N-terminal SUR1 half-molecule containing TM12 and a C-terminal half-molecule containing TM13 was enhanced by Kir6.2. No such enhancement was seen with Kir4.1 ; hence this effect of Kir6.2 is specific. However, when TM12 was omitted from the N-terminal half-molecule we still obtained binding activity, but now Kir6.2 had no enhancing effect. We conclude that Kir6.2 specifically influences SUR1 assembly and that TM12 is critical for Kir6.2–SUR1 interaction.

Importance of TMD0 for SUR1 assembly Co-expression of a C-terminal SUR1 halfmolecule with an N-terminal moiety, from which the whole of TMD0 had been omitted, resulted in the formation of substantial glibenclamide-binding activity [16]. Moreover, glibenclamide-binding activity was retained in a SUR1 construct that lacked NBD2 and TMD0. Hence TMD0 is not involved in SUR1 assembly or glibenclamide binding.

Importance of TMD1 for SUR1 assembly To explore the role of TMD1, we constructed Nterminal half-molecules in which successive pairs of TM regions were deleted [16]. These were coexpressed with a C-terminal SUR1 half-molecule. Deletion of TM6 and TM7 abolished the formation of glibenclamide-binding sites. This was also the case for deletion of TM7 and TM8, TM8 and TM9, or TM10 and TM11. However, the construct from which TM9 and TM10 were deleted retained substantial binding activity. Similar findings were obtained in whole SUR1 constructs : derivatives lacking TM6 and TM7, TM7 and TM8, TM8 and TM9, or TM10 and TM11 had no glibenclamide-binding activity, but activity was retained in a deletion lacking TM9 and TM10. Therefore, TMs of TMD1 are of critical importance in SUR1 assembly.

Glibenclamide-binding site Glibenclamide is a bifunctional reagent containing both a sulphonylurea moiety and a benzamide functionality. Since no binding activity can be detected by the sole expression of either N- or C-terminal half-molecules of SUR1, but substantial activity is formed when both half-molecules

Importance of TMD2 for SUR1 assembly In these experiments we used an N-terminal SUR1 half-molecule which leads to formation of substantial glibenclamide-binding activity when

325

# 2002 Biochemical Society

Biochemical Society Transactions (2002) Volume 30, part 2

are co-expressed [13], two regions of SUR1 are needed for glibenclamide binding We propose that CL3 constitutes the key element making up the N-terminal part of the glibenclamide-binding site. Glibenclamidebinding activity of SUR1 was barely affected by removal of NBD2 [13,16] and the further deletion of TMD0 [16]. However, the additional removal of CL3 abolished glibenclamide binding [16]. A similar requirement for CL3 was obtained from studies involving the expression of halfmolecules. We showed, using GFP-linked proteins, that the requirement for CL3 is not simply due to a failure of constructs lacking this loop to

reach the plasma membrane. We conclude that CL3 is a key component of the N-terminal part of the glibenclamide-binding site. The initial data described above implicated the region TM14–TM15–CL8–TM16 in the C-terminal element of the glibenclamide-binding site. Investigation of glibenclamide binding after co-expression of N-terminal SUR1 halfmolecules with different truncated variants of C-terminal half-molecules [16] showed that : (i) deletion of NBD2 from the C-terminus led to an increase in glibenclamide-binding activity, which may be due to facilitated folding; (ii) further deletion of TM17 did not prevent glibenclamide

Figure 2 Model for the β-cell KATP channel The upper panel depicts the suggested transmembrane topology of the β-cell KATP channel. The lower panel provides cross-sectional views within the membrane (A) and within the cytosol (B). The interactions and roles of individual regions deduced from the studies described here are indicated.

# 2002 Biochemical Society

326

Islets and Type 2 Diabetes

removal of most pairs of TMs causes a conformational change resulting in disruption of the CL3–CL8 interaction and hence glibenclamide binding. Only in the case of removal of TM9 and TM10 is conformation retained sufficiently to permit continued association of CL3 and CL8. We suggest that binding of glibenclamide leads to conformational changes in the CL3–CL8 region, leading to rearrangement of TM regions. These effects are transmitted to Kir6.2 to result in channel closure. One could speculate that binding of ATP might lead to conformational changes in NBD1, which could result in similar changes in the CL3–CL8 conformation and result in channel closure.

binding; (iii) further deletion of TM16 abolished binding. Taking into account that deletion from the whole SUR1 molecule of either TM17 alone, or TM16 and TM17 together, had no significant influence on glibenclamide binding, we conclude that TM16 is essential for molecular assembly of the SUR1 half-molecules. Deletion of the CL between TM13 and TM14 (CL7) from the N-terminus of the C-terminal half-molecule had no significant effect on glibenclamide binding, therefore indicating a lack of involvement of CL7 in glibenclamide binding. By combining data from C- and N-termini truncation, we conclude that CL8 is the only cytosolic structure of SUR1 that is required for formation of the C-terminal component of the glibenclamide-binding site. Moreover, a point mutation in CL8 decreased glibenclamide binding affinity more than 100-fold [16]. This mutation has been previously shown to abolish the high affinity blocking of KATP currents by tolbutamide [17]. We therefore propose that CL8 forms the C-proximal part of the glibenclamidebinding site. The fact that removal of TM16 did not abolish glibenclamide binding by a truncated SUR1 but abolished binding when tested in a Cterminal half-molecule indicates that TM16 is not involved in the formation of the glibenclamidebinding site, but plays an essential role in molecular assembly of the SUR1 half-molecules.

These studies were supported by grants from Diabetes UK, the Wellcome Trust and Eli Lilly Co. Ltd.

References 1 2 3 4 5 6 7 8

Model for the KATP channel

9

A model for the structure of the KATP channel based on the above studies is shown in Figure 2. The main features are as follows : (i) the activity shown by the chimaeric SUR1–Kir6.2 protein suggests that the C-terminal end of SUR1 is close to the N-terminus of Kir6.2; (ii) the interaction we have observed between NBD1 and NBD2 places these domains in proximity; (iii) TM12 is orientated in such a way that it can make contact with Kir6.2; (iv) formation of the glibenclamidebinding site requires that CL3 and CL8 are situated close together; (v) self interaction of NBD1 from neighbouring subunits suggests that there are homomeric interactions between the NBD1s of neighbouring subunits; (vi) TM16 is involved in interaction with TMD1. The observed effects of removal of pairs of TM sequences from TMD1 can be accounted for as follows : the

10 11 12 13 14 15 16 17

Ashcroft, S. J. H. (2000) J. Membr. Biol. 176, 187–206 Aguilar-Bryan, L. and Bryan, J. (1999) Endocr. Rev. 20, 101–135 Aguilar-Bryan, L., Bryan, J. and Nakazaki, M. (2001) Recent Prog. Horm. Res. 56, 47–68 Ashcroft, F. M. and Gribble, F. M. (1999) Diabetologia 42, 903–919 Ashcroft, S. J. H. and Ashcroft, F. M. (1992) Biochim. Biophys. Acta 1175, 45–59 Ashcroft, F. M. and Gribble, F. M. (2000) Trends Pharmacol. Sci. 21, 439–445 Tusnady, G. E., Bakos, E., Varadi, A. and Sarkadi, B. (1997) FEBS Lett. 402, 1–3 Babenko, A. P., Aguilar-Bryan, L. and Bryan, J. (1998) Annu. Rev. Physiol. 60, 667–687 Clement, J. P., Kunjilwar, K., Gonzalez, G., Schwanstecher, M., Panten, U., Aguilar-Bryan, L. and Bryan, J. (1997) Neuron 18, 827–838 Inagaki, N., Gonoi, T. and Seino, S. (1997) FEBS Lett. 409, 232–236 Shyng, S. L. and Nichols, C. G. (1997) J. Gen. Physiol. 110, 655–664 Mikhailov, M. V., Proks, P., Ashcroft, F. M. and Ashcroft, S. J. H. (1998) FEBS Lett. 429, 390–394 Mikhailov, M. V. and Ashcroft, S. J. H. (2000) J. Biol. Chem. 275, 3360–3364 Zerangue, N., Schwappach, B., Jan, Y. N. and Jan, L. Y. (1999) Neuron 22, 537–548 Mikhailov, M. V., Mikhailova, E. A. and Ashcroft, S. J. H. (2000) FEBS Lett. 482, 59–64 Mikhailov, M. V., Mikhailova, E. A. and Ashcroft, S. J. H. (2001) FEBS Lett. 499, 154–160 Ashfield, R., Gribble, F. M., Ashcroft, S. J. H. and Ashcroft, F. M. (1999) Diabetes 48, 1341–1347

Received 1 November 2001

327

# 2002 Biochemical Society