Characterization of binding site of closed-state KCNQ1 potassium ...

3 downloads 23952 Views 1MB Size Report
May 10, 2005 - Molecular docking. InsightII/Builder was used to construct the ... best significant pharmacophore hypothesis in this study, characterized by the ...
BBRC Biochemical and Biophysical Research Communications 332 (2005) 677–687 www.elsevier.com/locate/ybbrc

Characterization of binding site of closed-state KCNQ1 potassium channel by homology modeling, molecular docking, and pharmacophore identification Lu¨-Pei Du a, Min-Yong Li a, Keng-Chang Tsai b, Qi-Dong You a,*, Lin Xia a b

a Department of Medicinal Chemistry, China Pharmaceutical University, Nanjing 210009, China Institute of Molecular Medicine, Department of Life Science, National Tsinghua Unviersity, Hsinchu 30013, Taiwan

Received 25 April 2005 Available online 10 May 2005

Abstract This investigation was performed to assess the importance of interaction in the binding of blockers to KCNQ1 potassium using molecular modeling. This work could be considered made up by three main steps: (1) the construction of closed-state structure of KCNQ1 through homology modeling; (2) the automated docking of three blockers: IKS-142, L-735821, and BMS-IKS, using DOCK program; (3) the generation and validation of pharmacophore for KCNQ1 ligands using Catalyst/HypoGen. The obtained results highlight the hydrophobic or aromatic residues involved in S6 transmembrane domain and the base of the pore helix of KCNQ1, confirming the mutagenesis data and pharmacophore model, and giving new suggestions for the rational design of novel KCNQ1 ligands.  2005 Elsevier Inc. All rights reserved. Keywords: Homology modeling; Molecular docking; Pharmacophore; KCNQ1 potassium channel; Blockers

KCNQ1, or KVLQT1 potassium channel, was discovered by positional cloning and is the founding member of the KCNQ, where it co-assembles with KCNE1 (minK), a 129-amino acid residue accessory subunit to form a channel complex that generates the slowly activating cardiac potassium current IKs, an important determinant of myocardial repolarization [1]. Selective IKs blockers were initially developed as class III antiarythmic agents due to their ability to prolong ventricular refractoriness [2]. Another potential therapeutic application of IKs blockers is peptic ulcer disease or diarrhea [3]. But, mutation in KCNQ1 is the most frequent cause of congenital long QT syndrome, an inherent disorder characterized by syncope, cardiac arrhythmias, and sudden death [4]. An understanding of the molecu*

Corresponding author. Fax: +86 25 8327 1351. E-mail address: [email protected] (Q.-D. You).

0006-291X/$ - see front matter  2005 Elsevier Inc. All rights reserved. doi:10.1016/j.bbrc.2005.04.165

lar features of the blocker-binding site of KCNQ1 potassium channel might facilitate the design of novel molecules that block this channel with less proarrhythmic risk. Currently, the knowledge about the binding mode of KCNQ1-blocker complexes is still much more limited. The results of mutation studies for putative interacting residues in S5 and S6 still require some comment. Seebohm et al. [5] found that mutations of Ile-337 and Phe340 could cause the greatest decrease in blocker potency. Recently, they also showed that mutant Leu-273 and Phe-340 channels can disrupt close states and modify inactivation [6]. There are three different chemical classes of selective blockers of the IKs channel. The first chemical class of IKs selective blockers is the chromans, such as IKS-142 (1), from Hoechst [7]. Another chemical class of IKs channel blockers reported is the benzodiazepines from

678

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

MSD, and a case in point is L-735821 (2) [8]. Just recently, a third highly potent benzamide IKs channel blocker family is described by BMS, for example, BMS-IKS (3), although little physiological or pharma-

cological data are available [9]. These compounds are listed in Fig. 1. To get a better understanding of ligand interactions, it is important to focus on the KCNQ1 3D structure

Fig. 1. Chemical structures of KCNQ1 potassium channel ligands, in which compounds 1–3 are molecules for molecular docking, compounds 4–23 are training set molecules for pharmacophore modeling, and compounds 24–36 are test set molecules for pharmacophore validation.

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

to investigate the binding modes of these three chemical classes of ligands. Therefore, the objectives of this study include the construction of KCNQ1 potassium channel homology model, the analysis of its binding sites with blockers, and the identification of blocker pharmacophore to explore the binding mode of blocker–receptor complex. The results from this study should be useful in understanding the blocking mode of KCNQ1 potassium channel and in designing novel lead compounds.

Materials and methods Construction of homology model. The amino acid sequence of the human KCNQ1 potassium channel was retrieved from Swiss-Prot database [10] (Accession No. P51787, entry name KCNQ1_HUMAN) and aligned to the sequence of KcsA potassium channel (Accession No., 1K4C from the Protein Database Bank [11]) using the ClustalX multiple alignment program. A three-dimensional structural model of the S5/H5/S6 domains of KCNQ1 was then constructed based on the closed-state crystal structure of the corresponding domains of KcsA, with which it shares 46% homology and 36% identity. The rough homology models were generated from the sequence alignment (as seen in Fig. 2) using InsightII/Homology with default parameters that proposed loop conformation. The rough models were then refined by minimization with InsightII/Discover using the CVFF force field, van ˚ , dielectric value: 1 · r, by 1000 steps of steepest der Waals, cutoff: 9.5 A gradient followed by 10,000 steps of conjugate gradient until the RMS ˚ , then the models were gradient was less than 0.001 kcal/mol A ˚ thick water layers and optimized using embedded separately in 10 A molecular dynamics calculation by 30,000 steps in 300 K temperature. One nanosecond molecular dynamics was run for the model. In these building phases, the model quality was assessed using PROCHECK [12] and Verify-3D [13]. Molecular docking. InsightII/Builder was used to construct the molecular files for these three ligands represented in Fig. 1. All hydrogen atoms were presented for all of the constructed ligands. The ligands were optimized using GAMESS [14] at Hartree–Fock level with the 6-31 G* basis set. The partial charges on the atoms of the ligands were assigned by the Gastiger–Huckel method. All partial charges on the atoms of the KCNQ1 model were taken from the AMBER force field parameters. Docking of the ligands into the KCNQ1 homology model was performed by DOCK5 program [32]. The structures of obtained complexes were then analyzed using the Ligplot program to identify some specific contacts between the atoms of the ligand and receptor [16]. Pharmacophore modeling. In order to place appropriate chemical features onto the ligand–receptor complex so as to explore the interaction between ligands and KCNQ1 channel, we planned to construct a ligand-based pharmacophore for KCNQ1 ligands. The training set consisting of 20 compounds was selected by considering the structural diversity and wide coverage of activity range presented in Fig. 1. Activities are reported as IC50 values spanning from 6 nM to 0.2 mM

679

with 6 orders listed in Table 3 [7,15,18–28]. Each compound of the training set should provide new structural information to achieve a good pharmacophoric model in terms of predictive power and statistical significance. In this case where the chirality of a stereogenic center was not specified, all structures were built and minimized to the closest local minimum using the CHARMM-like force field within the Catalyst 4.9 program, and conformational analysis for each molecule was performed using the Poling algorithm, in which the setting number of conformer was limited to a maximum value of 250 using ‘‘best conformers generation’’ method with 20 kcal/mol energy cutoff. In the hypothesis generation process, a default uncertainty factor of 3 for each compound was defined, and four features, including hydrogenbond acceptor, hydrogen-bond donor, aromatic ring, and hydrophobic group, were selected to form the pharmacophore hypothesis using HypoGen module in Catalyst [17]. The HypoGen algorithm was forced to find pharmacophore models that contain at the most four of every feature. Hardware and software. InsightII 2000.3L [18] and Catalyst 4.9 [19] were used for molecular modeling on a SGI Origin 3800 workstation equipped with 48 · 400 MHz MIPS R12000 processors. The ab initio calculation (PC-GAMESS 6.3 [14]), docking calculation (DOCK 5.1 [32]), and docking analysis (Ligplot 4.22 [15]) were carried out on an Intel P4-based Redhat 9 Linux system.

Results and discussion Evaluation of pharmacophore model Catalyst produces 10 hypotheses, and Hypo1 is the best significant pharmacophore hypothesis in this study, characterized by the highest cost difference, lowest error cost, closest weight cost to 2, and lowest root-meansquare divergence and has the best correlation coefficient. All 10 hypotheses have the same features: one aromatic ring and three hydrophobic groups. The null cost value of the top 10 hypotheses is 143.168, and the fixed cost value is 81.655. Configuration cost value is 13.258. All 10 hypotheses have a total cost close to the cost of the fixed hypothesis. The difference between the fixed cost and the null cost is 61.513 bits. Therefore, the probability that the cost difference of any hypothesis with the null hypothesis to be higher than 60 is small. The cost range (Dcost) between these hypotheses and the null hypothesis varies between 54.5 and 46.434 bits with a low cost range, 8.066 bits. Therefore, we can expect that for all these hypotheses, there is at least a 75–90% chance of representing a true correlation in the data. Table 1 shows these parameters of statistical significance and predicted power.

Fig. 2. Sequence alignment of KCNQ1 (KVLQT1) vs. KcsA. Only the sequences around the modeled region are given.

680

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

Table 1 Information of statistical significance and predictive power for top 10 hypotheses in pharmacophore modelinga

Table 2 ˚ ) between pharmacophore features for KCNQ1 The distances (A ligandsa

Hypothesis No.

Features

Feature

R

H1

H2

H3

1 2 3 4 5 6 7 8 9 10

RHHH RHHH RHHH RHHH RHHH RHHH RHHH RHHH RHHH RHHH

R H1 H2 H3

— 4.91 11.93 5.55

4.91 — 11.54 9.08

11.93 11.54 — 12.59

5.55 12.59 9.08 —

Training set Total cost

Dcost

rms deviation

Correlation (r)

88.668 88.757 90.221 92.403 93.649 93.827 95.309 95.891 96.055 96.734

54.5 54.411 52.947 50.765 49.519 49.341 47.859 47.277 47.113 46.434

0.834 0.843 0.925 1.037 1.095 1.103 1.167 1.193 1.195 1.223

0.953 0.952 0.942 0.926 0.917 0.916 0.906 0.902 0.901 0.896

a Null cost of top 10 score hypotheses is 143.168 bits. Fixed cost is 81.655 bits. Configuration cost is 13.258 bits. Abbreviation used for features: H, hydrophobic; R, aromatic ring.

In this case, the rms deviation value of the best hypothesis Hypo1, 0.834, represents a good quality for Hypo1. The correlation coefficient for Hypo1, 0.953, shows a good linear regression of the geometric fit index. Another validation method to characterize the quality of hypothesis is represented by its capacity for correct activity prediction. The difference between estimated activity values and experimental activity values is represented as error (ratio between the estimated and experimental activity), with a negative sign if the actual activity is higher than the estimated. In this study, the error values of all compounds were found to be less than

a

Abbreviation used for features: H, hydrophobic; R, aromatic ring.

10, that means a not more than one order difference between estimated and actual activity. The top scoring hypothesis mapped with molecule 4 (L-768673; IC50 = 8 nM) is depicted in Fig. 3. As the best active compound in the training set, molecule 4 shows a good fit with all features of Hypo1. In this case, the aromatic feature seems to be mapped by the condensed benzodiazepine plane, while these three hydrophobic features are fitted by trifluoromethyl group, phenyl ring, and trifluorophenyl ring, respectively. The distances between these pharmacophore features from this work are shown in Table 2. Test set validation of pharmacophore model The main purpose of a quantitative model is to identify active structures and to forecast their actual activity accurately. To verify if the hypothesis can also predict the activity of compounds that are structurally distinct from those included in the training set, we applied a test set of 13 molecules by different activity classes and of

Fig. 3. The Hypo1 hypothesis is mapped onto the structures of the most active blocker 4 (L-768673) in the training set and most active blocker 24 (L761710) in the test set. The blockers are shown in stick, which L-768673 is in green and L-761710 is in yellow. The pharmacophore features are color coded as follows: blue spheres, three hydrophobics (H), and yellow spheres, aromatic ring (R). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

681

Table 3 Actual and estimated IC50 of training set based on pharmacophore model Hypo1a No.

Compound

Actual Ki (nM)

Estimated Ki (nM)

Error

Reference

4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

L-768673 L-761334 L-365260 S-5557 Ebastine Haloperidol Azimilide BTHP Mibefradil DDPH Terodiline Ambasilide Nisoldipine Clofilium Disopyramide Triamterene N-Acetylprocainamide Pentobarbital Propofol Nifedipine

0.008 0.011 0.21 0.56 0.8 2.6 3 9.3 12 13 30 32 40 50 82 100 100 180 250 360

0.011 0.016 0.59 1.5 1 8.6 2 5.6 5.1 2.2 120 80 32 16 34 84 80 540 280 80

+1.4 +1.5 +2.8 +2.8 +1.3 +3.3 1.5 1.7 2.4 6 +4.1 +2.5 1.2 3.2 2.4 1.2 1.3 +3 +1.1 4.5

[20] [20] [21] [7] [22] [22] [22] [18] [19] [15] [23] [22] [24] [22] [25] [22] [26] [27] [28] [28]

a The error column shows the ratio of estimated activity to actual activity (or the ratio of actual/estimated, if that gives a number greater than 1, in which case the number is negative).

different structural information [7,19,20,22,25,26,29–31]. All test set molecules were built and minimized as well as used in conformational analysis like the training set molecules. The structural data for the test set are shown in Fig. 1. The correlation coefficient generated using the test set, 0.856, shows a good correlation between actual and estimated activity. In this test set analysis, out of 13 compounds, the error values of 11 compounds were found to be less than 10, that means a not more than one order difference between estimated and actual activity. The most potent compound 24 (L-761710, IC50 = 10 nM) in the test set was then selected to show the mapping of this

molecule on Hypo1 represented in Fig. 3. Compound 24 also fits all features including in Hypo1. Cross-validation of pharmacophore model To further evaluate the statistical relevance of the model, the Fischer method was applied to check whether there is a strong correlation between the chemical structures and the biological activity. With the aid of the CatScramble program, the experimental activities in the training set were scrambled randomly, and the resulting training set was used for a HypoGen run. In this validation test, we select the 95% confidence level, and the 19

Table 4 Actual and estimated IC50 of test set based on pharmacophore model Hypo1a No.

Compound

Actual IC50 (lM)

Estimated IC50 (lM)

Error

Reference

24 25 26 27 28 29 30 31 32 33 34 35 36

L-761710 L-763540 Bepridil SQ-23791 Chromanol 293B Terfenadine NE-10133 CPU-86017 Cibenzoline Thiopentone Indapamide Diphenhydramine Thymol

0.01 0.03 0.82 5 6.2 10 12 14 25 56 100 130 200

0.074 0.006 3.4 23 82 1.1 96 3.7 10 630 30 85 290

+7.4 5 +4.1 +4.6 +13 9.5 +8 3.7 2.4 +11 3.4 1.5 +1.5

[20] [20] [19] [22] [7] [22] [29] [30] [25] [22] [22] [26] [31]

a The error column shows the ratio of estimated activity to actual activity (or the ratio of actual/estimated, if that gives a number greater than 1, in which case the number is negative).

682

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

Table 5 Results from cross-validation using CatScramblea Trial No.

Total cost

Fixed cost

rms deviation

Correlation (r)

Configuration cost

81.65

0.834

0.953

13.258

67.272 67.272 75.616 80.946 76.785 67.272 67.272 79.742 67.272 75.431 79.763 76.274 67.272 79.916 76.211 75.43 73.567 67.272 80.388

2.755 2.755 2.427 2.191 1.977 2.755 2.755 2.106 2.755 2.393 2.451 2.444 2.755 1.906 2.435 2.291 2.294 2.755 1.728

0 0 0.476 0.606 0.697 0 0 0.645 0 0.498 0.459 0.465 0 0.622 0.472 0.555 0.555 0 0.678

0 0 7.219 12.549 8.388 0 0 11.345 0 7.033 11.366 7.877 0 11.519 7.814 7.033 5.17 0 11.991

Results for unscrambled 88.668 Results for scrambled 1 143.168 2 143.168 3 135.781 4 128.99 5 115.999 6 143.168 7 143.168 8 124.092 9 143.168 10 134.111 11 141.399 12 138.024 13 143.168 14 116.876 15 137.634 16 127.929 17 126.521 18 143.168 19 110.323 a

Null cost = 143.168.

Fig. 4. The structure of blocker IKS-142-KCNQ1 homology model complex was analyzed using the Ligplot 4.22 program to identify some specific contacts between atoms of ligand and receptor.

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

spreadsheets were generated by the CatScramble command. These random spreadsheets were used to generate hypothesis using exactly the same features as those used in generating the initial hypothesis (see Table 4). The results of the CatScramble runs are listed in Table 5. The data of cross-validation clearly indicate that all values generated after randomization produced hypotheses with no predictive value. Besides, out of 19 runs, only two, trials 5 and 19, had a correlation value near 0.7, but their rms deviations and configuration costs were very high, which is not desirable for a good hypothesis. This cross-validation also provided strong confidence on the initial pharmacophore Hypo1. IKS-142-KCNQ1 homology model complex analysis Fig. 4 reports the main interaction between IKS-142 and KCNQ1 model. More in detail the complex analysis highlights the following major residues involved: (1) Thr-312 plays a key role realizing a hydrogen bond with the oxygen atom of the chroman ring. (2) Phe-340 forms hydrophobic interactions with two methyl groups at methanesulfonamide moiety, as indicated by mutagenesis studies. (3) Thr-311, Thr-312, Ala-336, Ile-337, and Phe-340 form a hydrophobic network in which they interact with the trifluoropropyl chain. The mapping of IKS-142 onto pharmacophore Hypo1 and the docking result are represented in Fig. 5. In this case, we can find that the docking conformation is overlaid with the pharmacophoric conformation

683

very well. The aromatic feature is fitted with a phenyl plane at the chroman ring. One hydrophobic interaction is mapped by trifluoropropyl chain, while another is mapped by two methyl groups. According to the docking result, these two hydrophobic features are also indicated as two hydrophobic pockets formed by Thr-311, Thr-312, Ala-336, Ile-337, and Phe-340. The third hydrophobic feature failed to be mapped. L-735821-KCNQ1 homology model complex analysis The analysis of L-735821 complex shows the residues involved in blocker-binding site (as seen in Fig. 6). On the whole L-735821 is located in a site lined by three hydrophobic/aromatic regions: the first, formed by Ile-337, interacts with the N-methyl group and condensed benzodiazepine ring, the second, characterized by Thr-312, Ile-337, and Phe-340, interacts with benzyl ring, and the third, consisting of Ala344 and Gly-345, contacts the difluorophenyl substitute. L-735821 can also be fitted with all features of Hypo1: the condensed benzodiazepine ring has a good mapping with the aromatic feature (in interaction with Ile-337) and three hydrophobic features are fitted with N-methyl group (in interaction with Ile-337), benzyl ring (in interaction with Thr-312, Ile-337, and Phe-340), and dichlorophenyl group (in interaction with Ala-344 and Gly-345) reasonably, similar to that reported by docking analysis. The good overlay between conformation

Fig. 5. Pharmacophore Hypo1 mapped with IKS-142 and the key residues in the binding site of KCNQ1 homology ribbon model. The residues are represented as line, and IKS-142 is shown in stick, which docking conformation is in green and pharmacophoric conformation is in blue. H-bond is represented as yellow line. The pharmacophore features are color coded as follows: blue spheres, three hydrophobics (H); yellow sphere, aromatic ring (R). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

684

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

Fig. 6. The structure of blocker L-735821-KCNQ1 homology model complex was analyzed using the Ligplot 4.22 program to identify some specific contacts between atoms of ligand and receptor.

in pharmacophore mapping and in molecular docking also supports this binding mode. Results of these two studies have been incorporated in Fig. 7. BMS-IKS-KCNQ1 homology model complex analysis The docking study of BMS-IKS into the KCNQ1 homology model also shows an almost similar binding site to the mutagenesis study. In this case, the trifluoroalkyl chain possibly has a hydrophobic interaction with Gly-348, while the phenyl ring is presumed to form a hydrophobic or aromatic interaction with Ala-341 and Ala-344. Moreover, the cyclopentyl ring is inserted into another hydrophobic pocket, lined by Thr-312 and Phe340. Fig. 8 schematically shows the binding site with BMS-IKS in KCNQ1 channel. Fig. 9 depicts the conformation of BMS-IKS mapped onto pharmacophore Hypo1 and the binding site of KCNQ1 channel. BMS-IKS maps closely with Hypo1, which is characterized by four features. The 1,2,4oxadiazole plane in BMS-IKS is properly mapped into

aromatic feature, while the phenyl bridge (indicated to Ala-341 and Ala-344 in KCNQ1) and trifluoropropyl chain (indicated to Gly-348 in KCNQ1) are accurately treated as hydrophobic groups, respectively. However, the third hydrophobic feature does not overlap with any substructure in this compound. In summary, we have determined the aromatic or hydrophobic interactions in ligand–channel complexes using homology modeling, molecular docking, and pharmacophore identification, and localized the putative KCNQ1 channel-binding site for ligands to specific residues located on S6 transmembrane domain and the base of the pore helix. The binding models of blockers show clearly the mechanism of how three chemical classes of ligands bind to the KCNQ1 channel. These models match both with the topology of binding region of KCNQ1 and with both mutagenesis data well. This characterization in the binding site could be used in the near future as the starting point for molecular dynamics simulations in order to analyze the dynamic behavior of ligand interaction, to stabilize of the inacti-

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

685

Fig. 7. Pharmacophore Hypo1 mapped with L-735821 and the key residues in the binding site of KCNQ1 homology ribbon model. The residues are represented as line, and L-735821 is shown in stick, which docking conformation is in green and pharmacophoric conformation is in blue. The pharmacophore features are color coded as follows: blue spheres, three hydrophobics (H); yellow sphere, aromatic ring (R). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

Fig. 8. The structure of blocker BMS-IKS-KCNQ1 homology model complex was analyzed using the Ligplot 4.22 program to identify some specific contacts between atoms of ligand and receptor.

686

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687

Fig. 9. Pharmacophore Hypo1 mapped with BMS-IKS and the key residues in the binding site of KCNQ1 homology ribbon model. The residues are represented as line, and BMS-IKS is shown in stick, which docking conformation is in green and pharmacophoric conformation is in blue. The pharmacophore features are color coded as follows: blue spheres, three hydrophobics (H); yellow sphere, aromatic ring (R). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

vated or closed-state of KCNQ1 channel, and to design novel blockers of KCNQ1 channel for the treatment of cardiac arrhythmia.

Acknowledgments This work is supported in part by a grant (NSC922313-B-007-002) from the National Science Council. The Catalyst and InsightII studies were conducted at the National Center for High Performance Computing, Taiwan.

References [1] M.C. Sanguinetti, A. Zou, Molecular physiology of cardiac delayed rectifier K+ channels, Heart Vessels Suppl. 12 (1997) 170–172. [2] M.W. Veldkamp, Is the slowly activating component of the delayed rectifier current, IKsÕ absent from undiseased human ventricular myocardium? Cardiovasc. Res. 40 (1998) 433–435. [3] F. Grahammer, A.W. Herling, H.J. Lang, A. Schmitt-Graff, O.H. Wittekindt, R. Nitschke, M. Bleich, J. Barhanin, R. Warth, The cardiac K+ channel KCNQ1 is essential for gastric acid secretion, Gastroenterology 120 (2001) 1363–1371. [4] J. Barhanin, B. Attali, M. Lazdunski, IKs, a slow and intriguing cardiac K+ channel and its associated long QT diseases, Trends Cardiovasc. Med. 8 (1998) 207–214. [5] G. Seebohm, J. Chen, N. Strutz, C. Culberson, C. Lerche, M.C. Sanguinetti, Molecular determinants of KCNQ1 channel block by a benzodiazepine, Mol. Pharmacol. 64 (2003) 70–77.

[6] G. Seebohm, P. Westenskow, F. Lang, M.C. Sanguinetti, Mutation of colocalized residues of the pore helix and transmembrane segments S5 and S6 disrupt deactivation and modify inactivation of KCNQ1 K+ channels, J. Physiol. 563 (2005) 359–368. [7] U. Gerlach, Blockers of the slowly delayed rectifier potassium IKs channel: potential antiarrhythmic agents, Curr. Med. Chem. Cardiovasc. Hematol. Agents 1 (2003) 243–252. [8] H.G. Selnick, N.J. Liverton, J.J. Baldwin, J.W. Butcher, D.A. Claremon, J.M. Elliott, R.M. Freidinger, S.A. King, B.E. Libby, C.J. McIntyre, D.A. Pribush, D.C. Remy, G.R. Smith, A.J. Tebben, N.K. Jurkiewicz, J.J. Lynch, J.J. Salata, M.C. Sanguinetti, P.K. Siegl, D.E. Slaughter, K. Vyas, Class III antiarrhythmic activity in vivo by selective blockade of the slowly activating cardiac delayed rectifier potassium current IKs by (R)-2-(2,4trifluoromethyl)-N-[2-oxo-5-phenyl-1-(2,2,2-trifluoroethyl)-2,3dihydro-1H-benzo[e][1,4]diazepin-3-yl]acetamide, J. Med. Chem. 40 (1997) 3865–3868. [9] J. Lloyd, J.B. Schmidt, G. Rovnyak, S. Ahmad, K.S. Atwal, S.N. Bisaha, L.M. Doweyko, P.D. Stein, S.C. Traeger, A. Mathur, M.L. Conder, J. DiMarco, T.W. Harper, T. Jenkins-West, P.C. Levesque, D.E. Normandin, A.D. Russell, R.P. Serafino, M.A. Smith, N.J. Lodge, Design and synthesis of 4-substituted benzamides as potent, selective, and orally bioavailable I(Ks) blockers, J. Med. Chem. 44 (2001) 3764–3767. [10] A. Bairoch, R. Apweiler, The SWISS-PROT protein sequence database: its relevance to human molecular medical research, J. Mol. Med. 75 (1997) 312–316. [11] Y. Zhou, J.H. Morais-Cabral, A. Kaufman, R. MacKinnon, Chemistry of ion coordination and hydration revealed by a K+ ˚ resolution, Nature 414 (2001) 43–48. channel-Fab complex at 2.0 A [12] R.A. Laskowski, J.A. Rullmannn, M.W. MacArthur, R. Kaptein, J.M. Thornton, AQUA and PROCHECK-NMR: programs for checking the quality of protein structures solved by NMR, J. Biomol. NMR 8 (1996) 477–486. [13] D. Eisenberg, R. Luthy, J.U. Bowie, VERIFY3D: assessment of protein models with three-dimensional profiles, Methods Enzymol. 277 (1997) 396–404.

L.-P. Du et al. / Biochemical and Biophysical Research Communications 332 (2005) 677–687 [14] M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S. Gordon, J.H. Jensen, S. Koseki, N. Matsunaga, K.A. Nguyen, S.J. Su, T.L. Windus, M. Dupuis, J.A. Montgomery, General atomic and molecular electronic structure system, J. Comput. Chem. 14 (1993) 1347–1363. [15] N. Zhong, Q.Z. Fang, Y. Zhang, J.S. Xia, J.Q. Qian, Z.N. Zhou, Inhibitory effects of DDPH on two components of delayed rectifier potassium current in guinea pig ventricular cells, Acta Pharmacol. Sin. 23 (2002) 601–618. [16] A.C. Wallace, R.A. Laskowski, J.M. Thornton, LIGPLOT: a program to generate schematic diagrams of protein–ligand interactions, Protein Eng. 8 (1995) 127–134. [17] H. Li, J. Sutter, R. Hoffmann, Pharmacophore Preception, Development and Use in Drug Design, International University Line, La Jolla, California, 2000, pp. 171–189. [18] Y. Li, L. Fu, W. Yao, G. Xia, M. Jiang, Effects of benzyltetrahydropalmitine on two components of the delayed rectifier K+ current in Guinea pig ventricular myocytes, Cardiovasc. Drugs Ther. 16 (2002) 317–325. [19] C. Chouabe, M.D. Drici, G. Romey, J. Barhanin, Effects of calcium channel blockers on cloned cardiac K+ channels IKr and IKs, Therapie 55 (2000) 195–202. [20] G.L. Stump, G.R. Smith, A.J. Tebben, H. Jahansouz, J.J. Salata, H.G. Selnick, D.A. Claremon, J.J. Lynch Jr., In vivo canine cardiac electrophysiologic profile of 1,4-benzodiazepine IKs blockers, J. Cardiovasc. Pharmacol. 42 (2003) 105–112. [21] J.J. Salataa, H.G. Selnickb, J.J. Lynch Jr., Pharmacological modulation of I(Ks): potential for antiarrhythmic therapy, Curr. Med. Chem. 11 (2004) 29–44. [22] U. Gerlach, IKs channel blockers: potential antiarrhythmic agents, Drugs Future 26 (2001) 473–484. [23] S.E. Jones, T. Ogura, L.M. Shuba, T.F. McDonald, Inhibition of the rapid component of the delayed-rectifier K+ current by therapeutic concentrations of the antispasmodic agent terodiline, Br. J. Pharmacol. 125 (1998) 1138–1143.

687

[24] P. Daleau, M. Khalifa, J. Turgeon, Effects of cadmium and nisoldipine on the delayed rectifier potassium current in guinea pig ventricular myocytes, J. Pharmacol. Exp. Ther. 281 (1997) 826–833. [25] H. Satoh, Comparative actions of cibenzoline and disopyramide on I(Kr) and I(Ks) currents in rat sino-atrial nodal cells, Eur. J. Pharmacol. 407 (2000) 123–129. [26] P. Geelen, B. Drolet, E. Lessard, P. Gilbert, G.E. OÕHara, J. Turgeon, Concomitant block of the rapid (IKr) and slow (IKs) components of the delayed rectifier potassium current is associated with additional drug effects on lengthening of cardiac repolarization, J. Cardiovasc. Pharmacol. Ther. 4 (1999) 143–150. [27] K. Knobloch, J. Brendel, B. Rosenstein, M. Bleich, A.E. Busch, K.J. Wirth, Atrial-selective antiarrhythmic actions of novel Ikur vs. Ikr, Iks, and IKAch class Ic drugs and beta blockers in pigs, Med. Sci. Monit. 10 (2004) 221–228. [28] B.M. Heath, D.A. Terrar, Separation of the components of the delayed rectifier potassium current using selective blockers of IKr and IKs in guinea-pig isolated ventricular myocytes, Exp. Physiol. 81 (1996) 587–603. [29] A.E. Busch, K. Malloy, W.J. Groh, M.D. Varnum, J.P. Adelman, J. Maylie, The novel class III antiarrhythmics NE-10064 and NE10133 inhibit IsK channels expressed in Xenopus oocytes and IKs in guinea pig cardiac myocytes, Biochem. Biophys. Res. Commun. 202 (1994) 265–270. [30] Y.P. Ma, X.M. Hao, G.Q. Zhang, P.A. Zhou, C.H. Wu, D.Z. Dai, Blocking effect of CPU 86017 on slowly activated rectifier potassium current in single ventricular myocytes of guinea pigs, J. Chin. Pharm. Univ. 31 (2000) 121–125. [31] J. Magyar, N. Szentandrassy, T. Banyasz, L. Fulop, A. Varro, P.P. Nanasi, Effects of thymol on calcium and potassium currents in canine and human ventricular cardiomyocytes, Br. J. Pharmacol. 136 (2002) 330–338. [32] T.J. Ewing, S. Makino, A.G. Skillman, I.D. Kuntz, DOCK 4.0: search strategies for automated molecular docking of flexible molecule databases, J. Comput. Aided Mol. Des. 15 (2001) 411–428.