PT Huang
1, PW Lo
5, HH Liou
2, TB Lin
4, HC Spatz
6, YY Shiau
1,3*, KL Lou
1*
1Graduate Institute of Oral Biology, and 2Departments of Neurology and of Pharmacology, and 3Dentistry Department, NTU Hospital, Medical College, National Taiwan University, Taipei 10042, Taiwan; 4Department
of Physiology, Chung Shan Medical University, Taichung, Taiwan; 5Institut of Information Management National Chi Nan University, Taichung, Taiwan 5Institut für Biologie III, Albert-Ludwig-Universität Freiburg,
79104 Freiburg/Br., Germany.
* The corresponding author [email protected]
Abstract
The voltage-sensing domains of voltage-gated potassium channels Kv2.1 (drk1) contain four transmembrane segments in each subunit, termed S1 to S4. While S4 is known as the voltage sensor, the carboxyl terminus of S3 (S3C) bears a gradually broader interest concerning the site for gating modifier toxins like Hanatoxin and thus the secondary structure arrangement as well as its surrounding environment. To further examine the putative three-dimensional structure of S3C and to illustrate the residues required for Hanatoxin binding, which may then show influence on S4 segment to achieve the binding-resulted change in channel gating, molecular simulation and docking were performed, based on the solution structure of Hanatoxin and the structural information from lysine-scanning results for S3C fragment. Our data suggest that several basic and acidic residues of Hanatoxin are electrostatically and stereochemically mapped onto their partner residues on S3C helix, whereas some aromatic or hydrophobic residues located on the same helical fragment interact with the hydrophobic patch of the toxin upon binding. Therefore, a slight distortion of the S3C helix in direction possibly towards the N-terminus of S4 may exist. Such conformational change of S3C upon toxin binding was comprehended as a reasonable explanation for the observed shift in Hanatoxin binding-induced gating.
1. Introduction
The voltage-gated K+-channels comprise a large family of tetrameric membrane proteins that open and close in response to changes in membrane voltage.
Six putative transmembrane segments termed S1 through S6 are included in each subunit of the
tetramer. Among them, S5 through S6 assemble the central pore domain forming the K+-selective ion conduction pathway (MacKinnon and Miller, 1989;
MacKinnon and Yellen, 1990; Hartmann et al., 1991;
MacKinnon, 1991; Yellen et al., 1991; Yool and Schwarz, 1991; Liman et al., 1992; Heginbotham et al., 1994; Ranganathan et al., 1996; Armstrong and Hille, 1998). The crystal structure of a relatively simple prokaryotic K+ channel, KcsA, with two transmembrane segments in each subunit that are homologous to S5-S6 in voltage-gated K+ channels suggests that S5 and S6 are undoubtedly membrane spanning a-helices with the S5-S6 linker containing the most conserved region of all K+ channels which forms a short pore helix and the selectivity filter (Schrempf et al., 1995; Doyle et al., 1998). The first four transmembrane segments (S1-S4) of voltage-gated K+ channels do not contribute to the simple pore as in KcsA and in the inward rectifier K+ channels, and appear to underlie their unique voltage-sensing capabilities (Armstrong and Hille, 1998). However, the high-resolution structure of S1-S4 and its thus derived functional interpretation to illustrate the voltage-sensing mechanism are still not clear.
S4 is an unusual transmembrane segment that contains a large number of basic residues, which has been suggested by considerable studies be strongly involved in sensing changes in membrane voltage (Liman et al., 1991; Papazian et al., 1991; Perozo et al., 1994; Aggarwal and MacKinnon, 1996; Larsson et al., 1996; Mannuzzu et al., 1996; Seoh et al., 1996;
Yang et al., 1996; Smith-Maxwell et al., 1998;
Ledwell and Aldrich, 1999). In addition, a growing body of evidence suggests that S2 and S3 may also
52 participate in voltage sensing, especially S3 (Frech et al., 1989; Papazian et al., 1995; Planells-Cases et al., 1995; Seoh et al., 1996; Cha and Bezanilla, 1997;
Tiwari-Woodruff et al., 1997; Cha et al., 1999;
Monks et al., 1999; Li-Smerin et al., 2000; Takahashi et al., 2000; Winterfield and Swartz, 2000; Li-Smerin and Swartz, 2001). The C-termial part of S3 segment (S3C) is of particular interest because it has been identified as an important region for interaction with various gating modifier toxins (Rogers et al., 1996;
Swartz and MacKinnon, 1997b; Li-Smerin and Swartz, 1998, 2000; Winterfield and Swartz, 2000).
Among them, Hanatoxin (HaTx1), a 35-amino acid protein isolated from tarantula venom (Swartz and MacKinnon, 1995), shows an inhibition on Kv2.1 (drk1), which belongs to the shab K+ channel family, by shifting activation to more depolarized vo ltages (Swartz and MacKinnon, 1997a).
Recently, the solution structure of HaTx1 (Takahashi et al., 2000) has been determined and the hydrophobic patch of which the residues may interact with Kv2.1 upon binding was described (Fig. 1). The mechanism for the inhibition by this toxin is quite unique and distinct from other previously described K+ channel inhibitors. HaTx1 binds to Kv2.1 in each of the four voltage-sensing domains, not by bloking the pore, to achieve the inhibition (Swartz and MacKinnon, 1997b) . As previously described, the S3C has been proposed for the exact binding site (Swartz and MacKinnon, 1997b; Li-Smerin and Swartz, 1998; 2000; 2001). In the tryptophan-, alanine-, and lysine-scanning mutagenesis studies, a short non-helical stretch or kink of a conserved proline residue in S3 transmembrane segment was observed and its possible role in structural arrangement was discussed (O’Neil and DeGrado, 1990; MacArthur and Thornton, 1991; Blaber et al., 1993; Mathur et al., 1997; Monne et al., 1999; Hong and Miller, 2000; Li-Smerin et al., 2000; Li-Smerin and Swartz, 2001). Meanwhile, the existence of two helical fragments (S3N and S3C) for this segment, therefore, has been proposed after the helical periodicity was analyzed (Li-Smerin and Swartz, 2000; 2001).
However, the structural information illustrating the precise residues required for HaTx1-Kv2.1 binding and the thus derived molecular mechanism for the binding-induced shift in gating voltage are still absent due to the lack of complete structure of voltage-gated potassium channels in high resolution. Therefore, in present study, we have systematically docked the two molecules by presenting HaTx1 with its solution structure and Kv2.1 by modeling the C-terminus of S3 with restraints based on the possible structural arrangement of -helix deduced from previous lysine-scanning data. Molecular simulation has been performed and the detailed structural relationship depicting binding orientation around this region can
thus be discussed in this paper.
2. Method
Model building for Kv2.1 S3C fragment.
The Kv2.1 (gene accession number: 226432) S3C molecule (Val-271 to Val-282) was constructed via modification from fragment dictionary with geometry optimized using the consistent valence force field (CVFF) with Biopolymer module of Insight II software package (Molecular Simulation Inc., USA).
Atomic charges were computed using the semi-empirical MOPAC /AM1 method. The residues based on prediction of a-helix were individually regularized by energy minimization to give reasonable geometries.
Search for HaTx1 structure.
The coordinates for HaTx1 were obtained from Brookhaven Protein Databank in pdb file (PDB ID number 1D1H). Distribution of the surface charge via electrostatic potentials was performed by exhibiting the Connolly surface for HaTx1 residues.
Determination of starting orientations.
In principle, three criteria were used to determine the starting positions: stereochemistry, side -chain charge distribution, and previous structural information.
Inappropriate possibilities have been immediately excluded when definitely unreasonable combinations of alignment for docking were observed. Uncertain orientations were reserved and submitted for docking calculation to allow the computational results to perform the screening.
Docking simulation and energy comparison.
Upon docking, the total energies of electrostatic interactions and van der Waals contacts between the complexes of HaTx1 and S3C-binding model were compared. All docked complexes were subjected to 20 runs. Each run was composed of 500 cycles of simulated annealing and 200,000 steps of accepted/rejected configurations.
The values of all other default parameters were used. The alignment between docked and undocked molecules was performed by manually fitting the atomic coordinates of groups of residues that may be involved in the conserved interaction. Briefly, three-dimensional (3D) surfaces of the binding site enclose the most active members (after appropriate alignment) of the starting set of molecules. Note that errors in alignment can lead to incorrect, poorly predictive receptor surface models. This problem was overcome by using information obtained from previously related functional data. The surface was generated from a "shape field", in which the atomic coordinates of the contributing models were used to compute field values on each point of a 3D grid using a van der Waals function. A solvation energy correction term and the electrostatic charge complementarity’s method were used for energy evaluation. The solvation energy correction term is a penalty function that attempts to account for the loss
of solvation energy when polar atoms are forced into hydrophobic regions of the receptor surface. All the calculations and structure manipulations described above were performed with the Discover &
Docking/Insight II (2000) molecular simulation and modeling program (Molecular Simulation, San Diego, CA, USA; 950 release) on a Silicon Graphics Octane/SSE workstation..
3. Results
Determination of the starting orientations.
Three criteria were used to determine the starting orientation for docking alignments: (i) the stereochemistry, (ii) the surface charge distribution, and (iii) the structural information from previous studies. The hydrophobic patch on HaTx1 surface (Takahashi et al., 2000) was very useful in providing directions for choices. Most of the combinations were commenced based on the search for required residues of S3C helix to interact with the hydrophobic patch.
The leading concern at this stage was the residue type that satisfied both in stereochemistry and hydrophobic/aromatic interactions, for which we took the suggested residues by Li-Smerin and Swartz (Li-Smerin and Swartz, 2000; 2001) into consideration and in the same time compared with the structural effects they might bring in three-dimension.
After such comparison, the residues surrounding the hydrophobic patch were observed, in order for the orientation of molecules to be manipulated for appropriate alignments. The electrostatic interaction was therefore the major premise for this step.
Docking simulation and energy calculation.
Eight possible conformations were chosen and submitted for docking calculation. In Fig. 2, the starting orientations for these eight possibilities are illustrated with the energy calculation results listed and compared. The last (8) orientation showed the best energy results after the simulation performance, with respect to both the electrostatic energy (-35.71 Kcal/mol) and the van der Waals contacts (268.85 Kcal/mol). It is interesting to note that (2), (3), (4), (7) and (8) have better van der Waals energies than others, whereas (2), (6) and (8) have better electrostatic energies. We have further examined the difference between (2) and (8) and found that in (2), Leu-5 and Ala-29 of HaTx1 (see next paragraph for more details) are not perfectly included in the hydrophobic patch area as in (8) to interact with Kv2.1, which may account for the minor loss of VDW energy in forming the best hydrophobic contacts. In addition, the disadvantage in electrostatic energy in conformation (2) is supposed to be compensated by forming hydrophilic interactions between the right-hand side residues (referred to the helix location in conformation (8) and see below for more discussions) of HaTx1 and Asn-279 of S3C, which are certainly less stable than in conformation (8).
Structural description for the binding site.
With such orientation, hydrophobic contacts should occur between Kv2.1 residue side chains of Val-271, Tyr-274, Leu-275, Val-282 and the residue side chains of Leu-5, Phe-6, Tyr-27, Ala-29, Trp-30 from the hydrophobic patch in HaTx1 (Figs. 3a and 3b). In addition, there are several charged or polar residues in Kv2.1 S3C surrounding this area to form electrostatic interactions with residues from HaTx1, which stabilize the binding between HaTx1 and Kv2.1 in a more efficient way (Fig. 3c). For example, salt-bridges were found between side-chains of Arg-24 from HaTx1 and those of Ser-281, Glu-277 from Kv2.1, whereas hydrogen-bonding networks were observed between Tyr-27 (HaTx1), Asp-25 (HaTx1) and Glu-277 (Kv2.1). In addition, intra-molecular interactions appeared in HaTx1 after docking simulation: salt-bridge between Arg-24 and Asp-25 and H-bonds for Tyr-27 and Asp-25. These are crucial interactions on the right-hand side of S3C helix (as seen in Fig. 3c). On the left-hand side, another salt-bridge was found between Ser-25 from HaTx1 and Asn-279 from Kv2.1. Such observation explains and verifies the reasonable requirement of those charged residues neighboring the hydrophobic patch in Hanatoxin for the toxin-channel binding function (Takahashi et al., 2000).
Conformational changes of S3C and HaTx1 upon binding.
In order to compare the difference in conformation between before and after the docking performance, we superimposed the two complexes of toxin and S3C helix structure together and the results are shown in Fig. 4. Both HaTx1 and Kv2.1 revealed significant conformational change due to binding. It is very fascinating and very important to note that the two molecules do not only present a common binding behavior, that is, to draw each other nearer to form tighter binding. Instead, or in addition to a tighter binding, they both move towards a similar direction, and consequently the S3c helix shows a slight distortion (refer to the moving direction in Fig.4).
4. Discussion
We have systematically docked the two molecules by presenting HaTx1 with its solution structure (Takahashi et al., 2000) and Kv2.1 by modeling the C-terminus of S3 with restraints based on the possible structural arrangement of -helix deduced from previous lysine-scanning data (Li-Smerin and Swartz, 2001). From our results, the precise residues required for Hanatoxin binding onto voltage-gated potassium channel Kv2.1 are described (Fig.3). With respect to this point, we found that both hydrophobic and electrostatic interactions are required to stabilize the toxin-channel binding. This verifies the idea proposed by Takahashi and co-workers (Takahashi et al., 2000).
In addition to the essential structural
54 description, we are also quite interested in the structural-functional correlations that may explain the shift of activation to more depolarized voltages (Swartz and MacKinnon, 1997a) due to HaTx1 binding to Kv2.1. Chimeras constructed using the Kv2.1 and shaker K+ channels suggested the required sequence for HaTx1 binding in Kv2.1, but not in shaker (Swartz and MacKinnon, 1997b). Our preliminary data (not shown here) based upon the docking analysis for S3C region with shaker sequence suggested the existence of a much weaker but still possible binding between HaTx1 and the putative shaker S3C. This seemed to be in contradiction with the observation proposed by Swartz and MacKinnon (1997b). A possible explanation would be that the spatial freedom around S3C-S4N region in combination with required binding sequence might both play crucial roles in such binding-resulted influence of gating. However, this idea is just a speculation, more detailed structural analyses using shaker substitutions are still under investigation.
Regarding the conformational changes observed in our results, the moving direction for both HaTx1 and Kv2.1 molecules after simulations should be discussed in more details (Fig.4). If this direction is similar as that to push S3c from its original position towards N-terminus of S4 upon toxin binding, such motion may reduce the extent of rotation and movement of S4 in the membrane with depolarization and therefore channels would require higher voltage for activation. Very similar effect was observed in the periodic perturbation study by deletions in the S3 -S4 linker (Gonzalez et al., 2000;
2001). In other words, in Kv2.1, either the motion of S3c upon toxin binding or shortening of the S3-S4 linker might be able to result in the same effect:
spatial restriction for S4 move ment. Meanwhile, we should not exclude the possibility that extension of the S3-S4 linker may compensate such effect, as described in the previous paragraph for the case of shaker substitution.
The next question comes for the range of conformational change. From Fig. 4, the maximal displacement of the residues occurs in the HaTx1 structure (~7Å), which results in a distance of motion larger than 5Å in S3c. We would suggest that such range of motion should be sufficient to interfere the rotation and movement of S4 with depolarization. To sum up, the Hanatoxin binds to Kv2.1 via S3C helix and might push S3C towards the N-terminus of S4 segment (S4N) due to the binding-induced conformational change. The range of such movement is quite large and enables S3C to be at S4N’s close vicinity to interfere or reduce the freedom of space for S4 segment to translocate outwards membrane upon open gating (Cha et al., 1999). This therefore brings the required activation potentials to the more depolarized ones (Swartz and MacKinnon, 1997a).
Conclusion .
Our structural analysis with docking simulation has for the first time provided objective and more direct evidence for describing how Hanatoxin and the voltage-gated potassium channel Kv2.1 interact upon binding and which type of conformational changes can be brought through such binding. The functional implication in the binding-resulted gating shift from our structural observation has been also derived. Due to the absence of recombinant proteins, more advanced functional analysis by way of such as direct binding assays to prove our proposed hypothesis are still not so easily available. Crystallization trials for related channel molecules are of course under investigation but without definite promise.
Nevertheless, such structural information described in this paper should be sufficient to provide directions for further detailed electrophysiological examination.
We should be able to anticipate a Kv2.1 variant with and only with all the residues required for the Hanatoxin binding being mutated, which is derived from our simulation, in the presence of Hanatoxin, to behave like a wild-type channel in the absence of the same toxin.
5. Acknowlegement
The authors would like to thank Miss Yi -Chun Tsai at Hitron Technology for her technical advice. We are also very grateful to Drs. Francisco Bezanilla and Robert J. French for their advisory discussion and helpful suggestions. This work was supported in part by Grants from National Sciences Council of Taiwan:
NSC 90-2321-B-002-005 and NSC 89-2320-B-002-234 for LKL and NSC 90-2314-B-002-341 for SYY.
6. Reference
Aggarwal SK and MacKinnon R (1996) Neuron 16 1169-1177
Armstrong CM and Hille B (1998) Neuron 20 371-380
Blaber M Zhang XJ and Matthews BW (1993) Science 260 1637-1640
Cha A and Bezanilla F (1997) Neuron 19 1127-1140 Cha A Snyder GE Selvin PR and Bezanilla F (1999) Nature 402 809-813
Doyle DA Morais Cabral J Pfuetzner RA Kuo A Morais Gulbis J Cohen SL Chait BT and MacKinnon R (1998) Science 280 69-77
Frech GC van Dongen AM Schuster G Brown AM and Joho RH (1989) Nature 340 642-645
Gonzalez C Rosenman E Bezanilla F Alvarez O and Latorre R (2000) J Gen Physiol 115 193-207
Gonzalez C Rosenman E Bezanilla F Alvarez O and Latorre R (2001) PNAS USA 98 9617-9623
Hartmann HA Kirsch GE Drewe JA Taglialatela M Joho RH and Brown AM (1991) Science 251 942-944
Heginbotham L Lu Z Abramson T and MacKinnon R (1994) Biophys J 66 1061-1067
Hong KH and Miller C (2000) J Gen Physiol 115 51-58
Larsson HP Baker OS Dhillon DS and Isacoff EY (1996) Neuron 16 387-397
Ledwell JL and Aldrich RW (1999) J Gen Physiol 113 389-414
Liman ER Hess P Weaver F and Koren G (1991) Nature 353 752-756
Liman ER Tytgat J and Hess P (1992) Neuron 9 861-871
Li-Smerin Y and Swartz KJ (1998) PNAS USA 95 8585-8589
Li-Smerin Y Hackos DH and Swartz KJ (2000) J Gen Physiol 115 33-49
Li-Smerin Y and Swartz KJ (2000) J Gen Physiol 115 673-684
Li-Smerin Y and Swartz KJ (2001) J Gen Physiol 117 205-218
MacArthur MW and Thornton JM (1991) J Mol Biol 218 397-412
MacKinnon R and Miller C (1989) Science 245 1382-1385
MacKinnon R and Yellen G (1990) Science 250 276-279
MacKinnon R (1991) Nature 350 232-235
Mannuzzu LM Moronne MM and Isakoff EY (1996) Science 271 213-216
Mathur R Zheng J Yan Y and Sigworth FJ (1997) J Gen Physiol 109 191-199
Monks SA Needleman DJ and Miller C (1999) J Gen Physiol 113 415-423
Monne M Hermansson M and von Heijne G (1999) J Mol Biol 288 141-145
O’Neil KT and DeGrado WF (1990) Science 250 646-651
Papazian DM Timpe LC Jan YN and Jan LY (1991) Nature 349 305-310
Papazian DM Shao XM Seoh SA Mock AF Huang Y and Wainstock DH (1995) Neuron 14 1293-1301 Perozo E Santacruz-Toloza L Stefani E Bezanilla F and Papazian DM (1994) Biophys J 66 345-354 Planells-Cases R Ferrer-Montiel AV Patten CD and Montal M (1995) PNAS USA 92 9422-9426
Ranganathan R Lewis JH and MacKinnon R (1996) Neuron 16 131-139
Rogers JC Qu Y Tanada TN scheuer T and Catterall WA (1996) J Biol Chem 271 15950-15962
Schrempf H Schmidt O Kummerlen R Hinnah S Müller D Betzler M Steinkamp T and Wagner R (1995) EMBO J 14 5170-5178
Seoh SA Sigg D Papazian DM and Bezanilla F (1996) Neuron 16 1159-1167
Smith-Maxwell CJ Ledwell J L and Aldrich RW (1998a) J Gen Physiol 111 399-420
Smith-Maxwell CJ Ledwell JL and Aldrich RW (1998b) J Gen Physiol 111 421-439
Swartz KJ and Mackinnon R (1995) Neuron 15 941-949
Swartz KJ and Mackinnon R (1997a) Neuron 18 665-673
Swartz KJ and Mackinnon R (1997b) Neuron 18 675-682
Takahashi H Kim JI Min HJ Sato K Swartz KJ and Shimada I (2000) J Mol Biol 297 771-780
Tiwari-Woodruff SK Schulteis CT Mock AF and Papazian DM (1997) Biophys J 72 1489-1500
Tiwari-Woodruff SK Schulteis CT Mock AF and Papazian DM (1997) Biophys J 72 1489-1500