Biochimica et Biophysica Acta 1828 (2013) 1131–1142

Contents lists available at SciVerse ScienceDirect

Biochimica et Biophysica Acta

j ourna l homepage: www.e lsev ie r .com/ locate /bbamem

Identification of blocker binding site in mouse TRESK by molecular modeling andmutational studies

Songmi Kim a,1, Yuno Lee a,1, Hyun-Min Tak b, Hye-Jin Park b, Young-sik Sohn a, Swan Hwang a, Jaehee Han b,Dawon Kang b,⁎, Keun Woo Lee a,⁎⁎a Division of Applied Life Science (BK21 Program), Systems and Synthetic Agrobiotech Center (SSAC), Plant Molecular Biology and Biotechnology Research Center (PMBBRC),Research Institute of Natural Science (RINS), Gyeongsang National University (GNU), 501 Jinju-daero, Gazha-dong, Jinju, 660-701 Republic of Koreab Department of Physiology and Institute of Health Sciences, Gyeongsang National University School of Medicine, Jinju 660-751, Republic of Korea

⁎ Correspondence to: D. Kang,Department of PhysiologyGyeongsang National University School of Medicine, Jinju⁎⁎ Correspondence to: K.W. Lee, Division of Applied Life Scand Synthetic Agrobiotech Center (SSAC), Plant Molecularsearch Center (PMBBRC), Research Institute of Natural ScienUniversity (GNU), 501 Jinju-daero, Gazha-dong, Jinju, 660-755 772 1360; fax: +82 55 772 1359.

E-mail addresses: dawon@gnu.ac.kr (D. Kang), kwle1 Both authors contributed equally to the work.

0005-2736/$ – see front matter © 2012 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.bbamem.2012.11.021

a b s t r a c t

a r t i c l e i n f o

Article history:Received 4 May 2012Received in revised form 14 November 2012Accepted 16 November 2012Available online 27 November 2012

Keywords:TRESK (TWIK-related spinal cord K+ channel)K2P channelHomology modelingBlocker binding siteMolecular docking simulationMolecular dynamics simulation

TWIK (tandem-pore domain weak inward rectifying K+)-related spinal cord K+ channel, TRESK, a member ofthe tandem-pore domain K+ channel family, is the most recently cloned K2P channel. TRESK is highlyexpressed in dorsal root ganglion neuron, a pain sensing neuron, which is a target for analgesics. In thisstudy, a reliable 3D structure for transmembrane (TM) region of mouse TRESK (mTRESK) was constructed,and then the reasonable blocker binding mode of the protein was investigated. The 3D structure of themTRESK built by homology modeling method was validated with recommend value of stereochemical qual-ity. Based on the validated structure, K+ channel blocker-bound conformation was obtained by moleculardocking and 5 ns MD simulation with DPPC lipid bilayer. Our docking study provides the plausible bindingmode of known blockers with key interacting residues, especially, F156 and F364. Finally, these modelingresults were verified by experimental study with mutation from phenylalanine to alanine (F156A, F364Aand F156A/F364A) at the TM2 and TM4. This is the first modeling study for TRESK that can provide structuralinformation of the protein including ligand binding information. These results can be useful in structurebased drug design for finding new blockers of the TRESK as potential therapeutic target of pain treatment.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

Potassium channels (K+ channels) play important roles in the regu-lation of cell growth, cell excitability, cell volume, and proliferation, aswell as cell death. All K+ channels share a common functional featurewith respect to selectivity for K+ ions. It is associated with a conservedTVGYG motif which is important for selectivity filter of the pore-forming region (P-domain or P-loop) of the channel proteins [1–4].The K+ channels are characterized by the number of transmembrane(TM) domains (two, four, or six) and P-loops (one or two) [5].

The two-pore-domain K+ (K2P) channels contribute to the back-ground (or leak) K+ currents, which are responsible for the restingmembrane potential and excitability of several excitable andnonexcitable cell types in mammalian [6,7]. In heterologous systems,K2P channels could act as putative targets for various factors, including

and Institute of Health Sciences,660-751, Republic of Korea.ience (BK21 Program), SystemsBiology and Biotechnology Re-ce (RINS), Gyeongsang National01 Republic of Korea. Tel.: +82

e@gnu.ac.kr (K.W. Lee).

rights reserved.

drugs. The mammalian K2P channel family consists of 15 members,and it can be divided into six subfamilies based on amino acid homol-ogy, and their structural and functional properties: TWIK (TWIK-1,TWIK-2, and KCNK7), TREK (TREK-1, TREK-2, and TRAAK), TASK(TASK-1, TASK-3, and TASK-5), TALK (TALK-1, TALK-2, and TASK-2),THIK (THIK-1 and THIK-2), and TRESK [1,8–21]. The channels play akey role in cellular mechanisms of neuroprotection, anesthesia, pain,and depression. In addition, they are modulated by cellular lipids, phar-macological agents, polyunsaturated fatty acids, and volatile general an-esthetics [16]. Each subunit of K2P channels consists of four TMdomains(TM1-TM4) and two P-loops, and two subunits assemble to form afunctional dimeric K+ channel. Moreover, the channels possess an evo-lutionarily conserved extended extracellular loop between TM1 and P1(TM1-P1 loop), and the P-loop of K2P channels contains a consensussequence motif of TxG(Y/F)G [7,11,16,18,22,23].

TWIK (tandem-pore domain weak inward rectifying K+)-relatedspinal cord K+ channel (TRESK), a member of the tandem-pore domainK+ channel family, is the most recently cloned K2P channel. Its mRNAwas expressed in a specific region of the human spinal cord andmouse cerebellum, cerebrum, spleen, thymus as well as testis, cerebralcortex, and especially dorsal root ganglion (DRG) [8,21,24,25]. Hy-drophobicity analysis of the amino acid sequence for mouse TRESK(mTRESK) channel revealed that the channel has two P-loops, fourTM domains, and one N-glycosylation sites in the TM1-P1 linker

http://dx.doi.org/10.1016/j.bbamem.2012.11.021mailto:dawon@gnu.ac.krmailto:kwlee@gnu.ac.krhttp://dx.doi.org/10.1016/j.bbamem.2012.11.021http://www.sciencedirect.com/science/journal/00052736

1132 S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

region which are similar to the other K2P channels. In addition, theTRESK consists of extended intracellular loop between TM2 andTM3 and a short C-terminal and N-terminal regions which locatedon the intracellular side [21].

The TRESK is modulated by various intracellular and extracellularfactors such as arachidonic acid, Zn2+, volatile anesthetics, and GPCRagonists [8,26–29]. This channel is uniquely regulated by the calciumsignal compared to the other K2P channels [24]. Because mTRESKinclude an NFAT (nuclear factor of activated T-cells)-like consensusmotif for calcineurin binding such as PQIVID in mouse (PQIIIS inhuman) at its intracellular loop between TM2 and TM3 (TM2-TM3loop), it is activated in response to the calcium signal through bindingcalcineurin, a calcium/calmodulin-dependent protein phosphatase [30].Therefore, Cyclosporine A and FK506, immunosuppressive compounds,which are calcineurin inhibitors widely used as immunosuppressors inthe clinic, prevent TRESK activation. The activation of TRESK could in-duce activation of the transcription of new genes as well as regulationof several aspects of T-cell function, therefore the channel is consideredas a potential target to treat T-cell mediated immune dysfunction[24,30,31].

Several studies reported K2P homology modeled structure such asTASK-1, TASK-2 and TALK-2, and TREK-1 and TREK-2 [17,32–35]. TheTRESK is highly expressed in dorsal root ganglion neuron, a painsensing neuron, which is a target for analgesics. However, little isknown about the 3D structure and drug binding site of TRESKcompared with other K2P channels. Therefore, the main objectives ofthis study are to construct 3D structure of mTRESK and to predictK+ channel blocker binding site. In this study, reasonable modeledstructures of mTRESK were generated using homology modelingmethod. A suitable binding mode of known K+ channel blockersinto mTRESK structure was predicted, and key interacting residuesin blocker binding site were suggested in our molecular dockingstudy. In order to verify our modeling study, three TRESK mutants(F156A, F364A, and F156A/F364A) were constructed. The TRESK cur-rents were measured using whole-cell recording, a technique thatmeasures the movement of ions across the cell membrane. Pharmaco-logical agents were tested at the cellular levels. Our modeling resultsprovide the structural insight into blocker binding site of TRESK, thiswill be help to design structure-based drugs for TRESK as a potentialtherapeutic target in pain treatment.

2. Materials and methods

2.1. Prediction of transmembrane helices

ThemTRESK sequence including 394 amino acidswas obtained fromUniProt (http://www.uniprot.org) (UniProt ID: Q6VV64). TMHMMv2.0which is available at CBS (Center for Biological Sequence analysis,http://www.cbs.dtu.dk/services/TMHMM) was used to predict trans-membrane helices (TMHs) of mTRESK. The program is a method forprediction of TMHs based on a hidden Markov model [36].

2.2. Generation of 3D structure

Sequence database search was carried out with blastp and PSI-BLASTtools in NCBI (www.ncbi.nlm.nih.gov) to identify the homologue ofknown structure from the PDB for template proteins of mTRESK. Thecoordinates of MthK channel (PDB ID: 3LDC, 1.45 Å resolution) wereselected as a template to build the initial mTRESK structure. Sequencealignment between mTRESK and template was performed by ClustalW2program (http://www.ebi.ac.uk/Tools/clustalw2/index.html) [37]. Basedon the alignment, the 3D structure of mTRESK was constructed usingBuild homology model protocol within Discovery studio 2.5 (DS2.5)(Accelrys Inc., San Diego, USA). Two models were separately generateddepending on the extended TM1-P1 loop: Model 1 without loop, Model2 with edited helices in loop region. The build and edit protein tool was

applied to predict secondary structure as right-hand alpha helix for theextended loop, the range of these helices was calculated by Predictcommand in Secondary Structure from Sequence menu in DS2.5. Theorientation coordinates for monomer and homodimer of mTRESK struc-ture were obtained by superimposition of the constructed homologymodels with KcsA K+ channel structure (PDB ID: 1J95).

2.3. Solvated energy minimization and validation of 3D structure

Solvated energy minimization (EM) was carried out using theGROMACS program (version 4.5.1) [38,39] with GROMOS96 53a6force field to refine the homology modeled structure. The initial struc-tures were inserted to an orthorhombic water box (1 nm thickness),and the systemwas neutralized with the addition of 14 K+ counterionsby replacing water molecules. Long range electrostatics were handledusing the PME (particle mesh Ewald) method [40]. In a system, proteinalone consists of 4892 atoms and the entire system is made up of101228 atoms in 32112 water molecules. To remove the possible badcontacts from the initial structures, the steepest descent EM algorithmwas applied until energy convergence at 2000 kJ mol−1 nm−1.

The energy minimized structures were validated by three programssuch as PROCHECK, ProSA-web (https://prosa.services.came.sbg.ac.at/prosa.php), and ERRAT version 2.0 (http://nihserver.mbi.ucla.edu/ERRATv2/). The PROCHECK program was implemented to check thestereochemical quality of protein structures with Ramachandran plot[41]. The ProSA and ERRAT were also executed to evaluate theconstructed structures. The ProSA is an established tool which isfrequently employed in the refinement and validation of experimentalprotein structures, structure prediction, and modeling [42]. ERRAT is aprotein structure verification algorithm that is especially well-suitedfor evaluating the progress of crystallographic model building andrefinement [43].

2.4. Pore radius analysis

The HOLE program (http://hole.biop.ox.ac.uk/hole) [44,45] wasutilized for analysis and visualization of the pore dimensions in homol-ogy modeled mTRESK structure. The VMD program [46] was used tovisualize and to compare the pore surfaces between the template andmTRESK.

2.5. Molecular dynamics simulation in explicit membrane environment

The minimized mTRESK model was oriented perpendicularly tothe membrane plane by Add Membrane and Orient Molecule Tool inDS2.5 and then it was subjected to the molecular dynamics (MD) sim-ulation with explicit membrane [47–50]. The coordinate and topologyof the DPPC containing 128 lipids for GROMACS were obtained fromthe Web site of D. Peter Tieleman (http://moose.bio.ucalgary.ca).The proper membrane system of 512 DPPCs was constructed basedon the extension of 128 DPPCs. The GROMOS 53a6 force field [51]was applied with added lipid parameters developed by Berger et al.[52]. Initial structure was embedded in DPPC bilayer and solvatedwith SPC (simple point charge) water model [53]. The coordinateand topology of GROMACS for propafenone which shown the highestfitness score in pre-docking study were derived from PRODRG server[54] The systems were neutralized by the addition of 14 K+ counter-ions. The protein alone consists of 4448 atoms and the entire systemis made up of 129,303 atoms containing 33,135 water molecules and508 DPPC molecules by removed 4 lipids from upper leaflet wheninserting protein into the lipid bilayer (252 lipids in the upper, 256lipids in the lower leaflet). The steepest descent energy minimizationwas calculated to remove possible bad contacts from the initial struc-ture until energy convergence reached 1000 kJ/(mol∙nm). Thesimulation was performed in the NVT ensemble during 100 ps tostabilize the temperature of the system at 323 K. Continuously,

http://www.uniprot.orguniprotkb:Q6VV64http://www.cbs.dtu.dk/services/TMHMMhttp://www.ncbi.nlm.nih.govhttp://www.ebi.ac.uk/Tools/clustalw2/index.htmlhttps://prosa.services.came.sbg.ac.at/prosa.phphttps://prosa.services.came.sbg.ac.at/prosa.phphttp://nihserver.mbi.ucla.edu/ERRATv2/http://nihserver.mbi.ucla.edu/ERRATv2/http://hole.biop.ox.ac.uk/holehttp://moose.bio.ucalgary.ca

1133S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

equilibration of pressure is conducted under an NPT ensemble. Thesystem was subjected to normal pressure constant (1 bar) for 1 nsunder the conditions of position restraints for heavy atoms andLINCS [55] constraints for all bonds. Production run was carried outuntil 5 ns under periodic boundary conditions. Long range electro-statics interactions were calculated by the PME method. Cutoffdistances for the calculation of the short-range electrostatic and vander Waals interaction were 1.2 nm. The time step of the simulationwas set to 2 fs, and the coordinates were saved for analysisevery 2 ps. All analyses of MD simulations were carried out usingGROMACS package.

2.6. Molecular docking study

Molecular docking simulation was carried out to predict the bind-ing modes of the blockers into the cavity of the mTRESK structureusing GOLD 4.1 program [56]. The program uses genetic algorithm(GA) for docking of flexible ligands into protein binding sites. In thisstudy, the GA parameters were set as default values. The reasonableinternal binding site of mTRESK structure for K+ channel blockerswas identified by Find Sites from Receptor Cavities method in Defineand Edit Binding site tools of DS2.5. The blockers such as propafenone,quinine, lidocaine, and triethanolamine obtained from ChEMBL(https://www.ebi.ac.uk/chembldb) were used for the docking simu-lation. A radius was taken as 10 Å around the center of binding site inmTRESK. The flip and rotate options were applied for protonatedcarboxylic acids, ring-NHR, and ring NR1R1 during the dockingsimulation. The best docked poses of compounds were selectedbased on the GoldScore ranking. The GOLD fitness score calculatedby factors including H-bonding energy, van der Waals energy, andligand torsion strain has been optimized for the prediction of ligandbinding positions.

2.7. Transfection

Mouse TRESK (mTRESK, GenBank accession number NM_207261)was a kind gift from Dr. Donghee Kim. The coding region of TRESK wassubcloned into pcDNA3.1/V5-His-TOPO vector (Invitrogen, Carlsbad,CA, USA). HEK-293A cells were seeded at a density of 2×105 cells per35-mm dish 24 h prior to transfection in Dulbecco's modified Eagle'smedium (DMEM) which included 10% fetal bovine serum (FBS;Invitrogen, Grand Island, NY, USA). HEK-293A cells were co-transfectedwith an mTRESK DNA in pcDNA3.1 and green fluorescent protein (GFP)in pcDNA3.1 using LipofectAMINE and Opti-MEM I Reduced SerumMedium (Invitrogen). The cells expressing GFP were detected with amicroscope (Axiovert 135; Carl Zeiss Jena GmbH, Jena, Germany)equipped with a mercury lamp light source. Cells were used 1–3 daysafter transfection.

2.8. Site-directed mutagenesis

Single and double point mutations in mTRESK cDNAwere done usinga QuikChange site-directed mutagenesis kit (Stratagene, Santa Clara, CA,USA) to generate penylalanin-to-alanine mutations according to theinstructions of the manufacturer (F156A, F364A, and F156A/F364A).Both strands of mutated TRESK DNA fragments were sequenced forconfirmation.

2.9. Electrophysiological studies

Electrophysiological recording was performed using a patch clampamplifier (Axopatch 200, Axon Instruments, Union City, CA). Whole-cellcurrentswere analyzedwith the pCLAMPprogram(Version 8). Bath solu-tion for whole-cell recording contained (mM): 135 NaCl, 5 KCl, 1 CaCl2, 1MgCl2, 5 glucose and 10 HEPES, and pipette solutions contained (mM):

150 KCl, 1 MgCl2, 5 EGTA and 10 HEPES (pH 7.3). All other chemicalswere purchased from Sigma Chemical Co. (St Louis, MO, USA).

2.10. Statistical analysis

Data were analyzed using one-way ANOVA with Tukey's test usedfor post hoc comparisons. All analyses were performed using SPSS(version18, Chicago, IL, USA) and represented as the mean±SD.Differences were considered significant at pb0.05.

3. Results and discussion

3.1. Prediction of number and topology for transmembrane helices

The number of transmembrane helices (TMHs) and transmembranetopology for mTRESK were predicted by TMHMM v2.0 server using ahidden Markov model. The results showed that TRESK has six TMHsconsisting of each three TMH at N-terminal and C-terminal, respective-ly. The second and fifth TMHs were identified as P-loop domains whichare including conserved P-loop sequences (TxGxG). The long extendedloops such as TM1-P1 and TM2-TM3 loops were observed in the plotshowing posterior probabilities of inside/outside/TM helix (Fig. S1).According to KangD et al., mTRESK has four TMHs, two P-loop domains,extracellular TM1-P1 loop, and the extended intracellular TM2-TM3loop [21]. The mTRESK topology obtained from TMHMM was used togenerate 3D structure of mTRESK.

3.2. Sequence alignment for the homology modeling

Sequence alignments of several reported K+ channel structures(PDB ID: 2AHZ for NaK channel, 1XL4 for KirBac3.1 channel, 3LDCfor MthK, and 1J95 for KcsA) were performed for selecting the propertemplate structure. Since MthK channel (PDB ID: 3LDC), which is inopen state rather than the closed state, was well aligned withmTRESK, the structure was used as template. All K+ channels have ahighly conserved TVGYG motif (P-loop sequence) which is respon-sible for selectivity filter. Since one subunit of mTRESK containstwo P-loops (P1-loop and P2-1oop) unlike other K+ channels,TM1-P1-TM2 and TM3-P2-TM4 regions of mTRESK were separatelyaligned with MthK channel (Fig. 1). The aligned sequence in P1helix region was edited by removing gap region (ESWTVS—LYW→ESWTVSLYW) which makes collapse in the secondary structure ofthe P1 helix. The sequence alignment result for TM1-P1-TM2 regionshowed 13.7% identity and 29.4% similarity with the template(Fig. 1A). When considering only TM region of TM1-P1-TM2 withoutTM1-P1 linker, sequence identity and similarity were slightly in-creased to 29.4% and 53.6%, respectively. This is due to structuralproperty of K2P family; they possess extended extracellular TM1-P1loop region. Although relatively low identity of TM1-P1-TM2 regionwas observed, P1-loop of the region was perfectly aligned. In case ofTM3-P2-TM4, sequence identity and similarity was 27.0% and 52.8%,respectively (Fig. 1B). From these sequence alignment results, thesequences of the both TM1-P1-TM2 and TM3-P2-TM4 regions inmTRESK were found to be similar with that of the template. TheP-loops were also found to be conserved well in mTRESK with thetemplate. Thus, the alignment was used to generate 3D structure ofmTRESK.

3.3. Generation of 3D structures using MthK as a template

Based on the TMHMM and sequence alignment results, the 3Dstructures of TM regions for mTRESK were generated using homologymodeling method (Fig. 2). The structure consists of six TMHscontaining TM1-TM4, P1, and P2 helices (Fig. 1A). Two differentmodels were separately constructed and named as Model1 andModel2 depending on type of the TM1-P1 loop. The Model1 suggests

https://www.ebi.ac.uk/chembldb

Fig. 1. Sequence alignment of mTRESK with MthK (PDB ID: 3LDC). Sequence alignment results for the region of TM1-P1-TM2 (A) and TM3-P2-TM4 (B). Sequence identities areindicated by asterisks (*), conservative substitutions by colons (:), and semi-conservative substitutions by dots (.). Extended extracellular loop region is denoted by green boxand predicted helix of Model2 is highlighted by orange color in the box. Each double-arrowhead lines represent TMH and P-loop (red) regions. TM topology sequences of mTRESKreported by Kang D et al. are shown in bold and underlined. Key residues highlighted in blue box are considered for mutation study. The gray arrow indicates the glycine residuewhich conserved in other K+ channels as a hinge residue.

1134 S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

the 3D structure of TM regions without the loop (Fig. 2A). The Model2represents not only TM regions but also the extended loop includingtwo helices which are predicted by Secondary Structure Predictionand constructed by Build and Edit Protein tools (Fig. 2B).

To obtain monomer and homodimer (dimer and tetramer, in caseof MthK) formation of mTRESK, crystal structure of KcsA (PDB ID:1J95) which consists of tetramer was used to form functional channelfor two models (Fig. 2C, D, and E). In order to refine the homologymodeled structures, calculation of solvated energy minimizationwas carried out and then the final structures were achieved.

After our modeling study was completed, the X-ray crystal struc-tures of the K2P channels, TWIK-1 and TRAAK, were currently solved[60,61]. The structures reveal the presence of a novel extracellularhelical cap domain that creates a bifurcated pore entryway. Althoughwe could not predict disulfide bond that covalently links the twosubunits, prediction of a part of two helices of the Model2 is corre-sponding with the crystal structure (Fig. S4A). Since the main aim ofthis study was to find out the blocker binding site rather than to con-struct the extracellular region, the results of this modeling study willbe continuously used for further study.

3.4. Validation for the structural quality

The PROCHECK, ProSA-Web, and ERRAT programs were performedto validate the refined model structures, and these results are detailedin Table 1. The PROCHECK, a well-known protein structure checkingprogram was carried out to check stereochemical quality of homologymodeled structures and MthK structure (Fig. S2). Ramachandran plotobtained from the PROCHECK showed that the 98.6% residues of MthKstructure and more than 90% residues of modeled 3D structures werelied in the most favored regions: 97.2% for Model1 and 93.2% forModel2 (Fig. S2A and B). Especially, the higher stereochemical qualityscore was observed in the Model1 compared to that of the Model2.There was no residue in disallowed region of Model1 structure andonly two residues in Model2 (Table 1). These results indicated that

our models are reasonable because the spot distributions were similarto the template structure. Moreover, overall Goodness factor (G-factor)indicating the quality of covalent and overall bond-angle distances was0.12 for Model1, 0.03 for Model2. The G-factor is a measure of the over-all ‘normality’ of the structurewhich provided from an average of all thedifferent G-factors of each residue in the structure [57]. Generally, theG-factor should be above -0.5 for a dependable structure. It means ourstructures are acceptable.

The z-score of twomodels was calculated using ProSA-web to checkoverall model quality. The proper z-score values (within −2.6) wereobserved in the twomodeled structures comparedwith that of the tem-plate (−1.65). As the result, our structures are within the range ofscores typically found for native proteins of similar size which areexperimentally determined protein chains in current PDB [42].

ERRAT is protein structure verification algorithm for non-bondedatomic interactions between different atom types. The recommendedvalue of ERRAT is greater than 50 for higher quality model, and thehigher scores indicate the better quality [43]. In the case of ourmodels, each ERRAT value was 66.88, and 76.78 for Model1 andModel2, respectively. Although low values of two models wereobserved when compared with that of the template which is 100,the models were found within the range of a high quality model. Allthe validation results revealed that our models were successfullyconstructed.

3.5. Structural comparison of pore region with template

AModel1was used to compare pore regionwith template (Fig. 3). Inorder to compare pore region of Model1 with that of template, the twostructures were superimposed as shown in the Fig. 3A. Each RMSDvalue for TM1-P1-TM2 region (Fig. 3A) and TM3-P2-TM4 region(Fig. 3C) was calculated as 0.26 and 0.28 Å, respectively. Moreover,the P-loops of the homology modeled structure were completely iden-tical, and it showed that our model can form properly pore which isresponsible for K+ ion permeation (Fig. 3B and D).

Fig. 2. Homology modeling structures of mTRESK with and without loop region. Structure of mTRESK monomer without TM1-P1 loop region (Model1) (A) and with extended loopincluding two predicted helices with extended flexible loop (Model2) (B). Potassium ions are represented as magenta sphere. The Model1 is superimposed with MthK (PDB ID:3LDC, gray). Each part of TM1-P1-TM2 and TM3-P2-TM4 is colored by orange and yellow, respectively. N-Glycosylation site (Asn83) is represented by stick model. Homodimericstructure of Model1 followed by orientation of KcsA (PDB ID: 1J95) structure (C). Top view (D) and bottom view (E) of panel (C) clearly show the orientation.

1135S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

The pore radius analysis of mTRESK and MthK channels wascarried out using HOLE program. The different gate region betweenmTRESK and MthK channels was observed through the analysis. Thenarrow pore radius (within 0.5–1.0 Å) of selectivity filter wasdetected in the both structures and the values are almost similar(Fig. 4). However, in the case of the gate region, the narrower pore ra-dius was investigated in mTRESK structure compared to the template(Fig. 4B and C). The radius of gate region in the MthK was approxi-mately 4Å, but that of the mTRESK structure was displayed as 2 Å(Fig. 4C). Although the both structures are in open state, aroundtwo fold differences were observed in the size of pore radius of gateregions. This reason is considered that the mTRESK structure hasfour phenylalanine residues, F156 of TM2 and F364 of TM4 in eachmonomer, around gate region (Fig. 4B). This result suggests thatthese residues can be form a narrow pore surrounding with volumeof the residues through the assembly of dimer.

Table 1Results of structure validation generated by PROCHECK, ProSA-web, and ERRAT tools.

Core (%) Allowed General Disallowed

MthK 98.6 1.4 0.0 0.0Model 1 97.2 2.8 0.0 0.0Model 2 93.2 5.8 0.0 1.1

a Overall G-factor score should be above −0.5 for a reliable model.b ERRAT score should be above 50 for a reliable model.

In order to compare pore regions between TRAAK and the mTRESK,we performed additional pore radius analysis for the TRAAK, it showedthat the regionwasmorewider thanmTRSEK. The gate region in TRAAKhas different amino acids such as glycine (TM2) and alanine (TM4)(Fig. S4).

3.6. Selection of representative structure for molecular docking

Generally, blockers of K+ channel can be bound on either side of themembrane at each binding site (external or internal) [58]. In this study,in order to find out blocker binding site from the overall structure ofmTRESK, we selected internal site which was identified by Define andEdit Binding site Tools in DS2.5. Sano Y et al. revealed that humanTRESK (hTRESK) was inhibited by K+ channel blockers such aspropafenone, glyburide (ATP-sensitive K+ channel blocker), lidocaine(Na+ channel blocker), quinine, quinindine, and triethanolamine

Overall G-factora Z-score (P1/P2) ERRATb score

0.41 −0.65 – 1000.12 −2.55 −2.48 66.880.03 −2.59 −2.55 76.78

image of Fig.�2

Fig. 3. Comparison of P-loops between mTRESK and MthK structures. The selectivity filter region showing with TVGYG motif of P1-loops (A) and TIGFG motif of P2-loops (C).Superimposition of P1-loops (B) and P2-loops (D) between mTRESK and MthK. The motif residue and K+ ion are represented as a stick model and a magenta sphere, respectively.

1136 S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

(non-selective K+ channel blocker) [8]. Since there is no reportedspecific blocker for the mTRESK, these reported blocker compounds in-cluding propafenone, quinine, lidocaine, and triethanolamine were se-lected as blockers of mTRESK (Fig. 5).

Fig. 4. Pore-lining surface and pore radius profiles. (A) Comparative illustrations of pore sizegray box) presented in mTRESK is shown in right panel. (B) Graph of pore radius profiles to cred (pore radiusb1.15 Å), green (1.15 Åbpore radiusb2.3 Å), and blue (pore radius>2.3 Å

To investigate the bindingmode of the blockers into the binding site,molecular dockingwas used and thenMD simulationwas carried out torefine blocker bounded structure in a dipalmitoylphosphatidylcholine(DPPC) bilayer (Fig. 6). Firstly, four blockers were docked into the

of MthK (gray) and mTRESK (orange). Detailed view of cavity and gate (highlighted byompare between MthK and mTRESK. The pore radii are represented as different colors:).

image of Fig.�4image of Fig.�3

Fig. 5. Chemical structures of K+ channel blockers for molecular docking simulation.

1137S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

internal binding site of the protein structure to obtain the initial blockerbound structure. All of the blockers were well docked into internalbinding site, and the conformations were placed in same position andbindingmode. Since the propafenone of blockers has shown the highestfitness score compared to the others, propafenone bound complex wasused to MD simulation.

Before the MD simulation, the implicit membrane was added tomTRESK structure using Add Membrane and Orient Molecule protocol

Fig. 6. Final conformations of Model1 (A) and Model2 (B) with the implicit membrane. Replayer and docked propafenone (represented by magenta stick model). The RMSD of the Cα

to better understand the topology of the membrane as well as togain orientation of the structure in membrane environment (Fig. 6Aand B). The 5 ns MD simulation of the propafenone bounded mTRESKmodel embedded in a lipid bilayer (Fig. 6C) was performed to acquireadjusted protein conformation by propafenone as well as to refinehomology model in explicit membrane environment. The backboneroot mean square deviation (RMSD) achieves stability after approxi-mately 2500 ps and its average value was 0.42 nm during simulationtime. In addition stability of propafenone also adjusted after about1500 ps and maintained 0.2 nm until the end of the simulation(Fig. 6D). Based on these RMSD analyses, representative structures(4570 ps frame) as ligand bound form was selected from the last2 ns snapshots. Refined representative structure was utilized tomolecular docking study.

3.7. Prediction of blocker binding site

Molecular docking studywas carried out to find out suitable blockerbinding site using refined representative structure (Fig. S3). As previousdocking simulation, four blockers such as propafenone, quinine, lido-caine, and triethanolamine were docked to representative structure ofmTRESK (Fig. S3A) and all of blockers showed similar binding modeto internal binding site of the mTRESK (Fig. S3B). The highest fitnessscore (64.77) was obtained in the case of propafenone, and therespective score of the other compounds, quinine, lidocaine, andtriethanolamine was 53.90, 46.78, and 30.56, respectively. Among theblockers, especially the triethanolamine showed the lowest fitnessscore (30.56). The reported experimental study from Sano Y et al. alsoshowed that triethanolamine (2 mM) blocked TRESK by 34%, whereaspropafenone (0.05 mM), quinine (0.1 mM), and lidocaine (1 mM)

resentative structure of Model2 obtained from 5 ns MD simulation with DPPC lipid bi-atom for mTRESK structure (D). The inset shows the RMSD calculated for propafenone.

image of Fig.�6image of Fig.�5

Fig. 7. Prediction of blocker binding site of mTRESK. Binding modes of propafenone (A), quinine (B), lidocaine (C), and Triethanolamine (D). The π–π interactions are represented asred lines and interacting key residues as stick models.

1138 S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

inhibited TRESK by more than 70% [1,8]. The reported experimentalresults were found to be in accordance with our docking results show-ing the effect of pi (π) interactions in ligand binding. In the dockingresults, three blockers except for triethanolamine were participated inthe π interactions with hydrophobic residues of mTRESK (Fig. 7).Several key interacting residues for the docked blockers were detected,and its residues are listed in Table 2. Almost interacting residues in eachcomplex were located in TM2 (I152, F156, L157, L159, T160, and D161)and TM4 (F364, I365, F367, K368, and L369). In addition, T335 belongsto P2-loop was detected as hydrophobic interacting residue with lido-caine and triethanolamine.

Especially, all blockers but triethanolamine contain aromatic ringgroups (Fig. 5), the blockers (propafenone, quinine, and lidocaine)have formed π–π stacking interactionwith two phenylalanine residues,F364 and F367 which are located in TM4 of mTRESK (Fig. 7A, B and C).According to hERG K+ channel study by Masetti M et al. [59], aromaticresidues including tyrosine and phenylalanine were formed cation–πand π–π stacking interaction with K+ channel blockers. Our dockingresults also showed similar the π–π stacking interactions which wereobserved between the aromatic ring groups of the blockers and the

Table 2Key interacting residues in blocker binding obtained from final docked conformations.

Blockers Hydrophobic contacts Hydrogenbond

π–πinteractions

Propafenone Ile152(B), Phe156(A, B), Leu157(B),Leu159(A), Thr160(B), Asp161(B),Ile293(A), Phe364(B), Ile365(A),Lys368 (B)

_ Phe364 (A),Phe367 (A)

Quinine Ile152(B), Phe156(A, B), Leu157(B),Leu159(A), Thr160(A), Phe364(B),Lys368(A), Leu369(B)

_ Phe364 (A),Phe367 (A)

Lidocaine Ile152(B), Phe156(A,B), Leu159(A),Thr160(A), Thr335(A), Phe364(B),Leu369(B)

_ Phe364 (A),Phe367 (A)

Triethanolamine Ile152(B), Phe156(A,B), Leu157(B),Thr334(A), Thr335(A), Lys368(A)

Phe364 (A) _

two phenylalanine residues. Particularly, the quinine has more π–πstacking interaction compared with others due to the presence ofquinoline group (Fig. 7B). Whereas, the triethanolamine docking resultrevealed that the F364 residue of mTRESK structure was formed onlyone hydrogen bond with H atom of hydroxyl group in triethanolamine(Fig. 7D). Therefore, π interaction seems to be strong effect on blockerbinding in mTRESK.

Interestingly, when we compared this docking result with previ-ous pore radius analysis, we found that the key interacting residuesF156 and F364 were also involved into the formation of narrow cavityin gate region. It can be concluded that the phenylalanine residues arecritical key residues for both binding of the blockers and forming thenarrow gate region. From our molecular docking simulation, we canunderstand not only plausible binding mode of known K+ channelblockers into the mTRESK but also significant key residues of mTRESK.

3.8. Modulation of mTRESK by pharmacological agents

In order to validate our modeling study, we tested the affinity ofF156 and F364 for mTRESK channels expressed in HEK-293A cells inrespond to TRESK blockers. Phenylalanine residues of mTRESK weremutated to alanine (F156A, F364A, and F156A/F364A) at the TM2 andTM4, and each mutant was expressed in HEK-293A cells. In physiologi-cal bath solution containing 5 mMKCl, the membrane potential of cellsunder the whole-cell configuration was held at −80 mV, and then aramp pulse (−120 to +60 mV; 1-s duration) was applied. In controlcells transfected with plasmid containing GFP alone, the ramp pulseelicited very small currents (b0.2 nA at +60 mV; n=6). However, incells transfected with TRESK/GFP DNA, the same ramp pulse producedlarge currents (b10 nA at +60 mV; n=6; Fig. 8) with reversal poten-tials near -80 mV, indicating that wild-type (WT) and mutant mTRESKformed a functional K+ channel in the plasma membrane of HEK-293cells. All of the mutants expressed functional currents in transfectedcells, as judged by measurement of single-channel and whole-cellcurrent (Fig. 8A). The single-channel was observed before recordingwhole-cell currents (upper left hand side in Fig. 8A). WT and mutantmTRESK were studied under same conditions, but cells showed

image of Fig.�7

2

4

3

6

2

4

3

6

1

2

5

10

2

4

2

4

3

6

4

8

4

8

2

4

ConWash

Quinine

ConWash

Quinine

ConWash

Quinine

ConWash

Lido

ConWash

Bupi

Wild-type

ConWash

ConWash

Bupi

ConWash

Bupi

ConWashBupi

ConWash

Lido

ConWash

Quinine

-120 -60 60 -120 -60 60 -120 -60 60 mV

-120 -60 60 -120 -60 60 -120 -60 60 mV

-120 -60 60 -120 -60 60 -120 -60 60 mV

-120 -60 60 -120 -60 60 -120 -60 60 mV

F156A/F364A

F364A

F156A

Lido

ConWash

Lido

nAnA nA

nAnA nA

nAnA nA

nAnA nA

0

50

100

#*#* #*

***

Wild type F364A F156A/F364A

Rel

ativ

e cu

rren

t(%

of i

nhib

ition

)

Bupi

Lido

Quinine

A

B

F156A

Fig. 8. Modulation of mTRESK mutants by TRESK blockers. (A) Potential binding sites were mutated to prevent binding of blockers to TRESK transmembrane domains, and the mu-tants were expressed in HEK-293A cells. Whole cell currents were recorded using the ramp pulse protocol before and after application of bupivacaine, lidocaine, and quinine. Thecell membrane potential was held at −80 mV, and then a voltage ramp (−120 to +60 mV; 1-s duration) was implemented out at room temperature. Con, Bupi, and Lido indicatecontrol, bupivacaine, and lidocaine, respectively. The single-channels shown in upper left hand side were recorded before forming whole-cell configuration in physiological solu-tion. (B) Summary of the effects of bupivacaine, lidocaine, and quinine on TRESK channel activity in whole-cell mode. Bar graphs show % inhibition of TRESK currents produced byblockers for mutants at +60 mV. Each bar represents means±S.D. of six independent experiments. *pb0.05 and #pb0.05 indicate significant difference from the value observedwith wild-type and F364A mutant, respectively.

1139S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

different amount of mTRESK currents. These results suggest that eachcell may control transcription and translation levels differently. Underwhole-cell conditions, extracellular application of 100 μM bupivacaine,1 mM lidocaine, and 100 μM quinine decreased WT currents by 82±

4%, 83±5%, and 82±8%, respectively. F156A and F364A mutants areless sensitive to quinine, lidocaine, and bupivacaine compared withWT. In F156A and F364A mutants, bupivacaine, lidocaine, and quinineinhibited the currents by 70±10% and 58±10%, 78±12% and 40±

image of Fig.�8

1140 S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

12%, and 76±8% and 38±9%, respectively. The mTRESK currents gen-erated by thesemutants were similar to those generated byWT (Fig. 8).

To test the importance of phenylalanine at the two amino acid res-idues (F156 and F364), a double mutant was constructed and testedfor sensitivity to quinine, lidocaine, and bupivacaine. F156A/F364A dou-ble mutant was insensitive to the blockers. Bupivacaine, lidocaine, andquinine slightly inhibited the currents of F156A/F364A mutant by15±14%, 20±15%, and 18±11%, respectively. The % of inhibition byblockers was significantly small compared with that in WT currents(pb0.05, Fig. 8A and B). Also the amounts reduced by blockers inF156A/F364A current were also significantly lower than those inF364A current (Fig. 8B).

Additional computational study was carried out to investigate thebinding mode difference between the mutants and WT. Three com-pounds such as bupivacaine, lidocaine, and quinine which were usedin experimental study were docked into each mutant structure. Incomparison of the mutants with WT, each fitness score for thecompounds with mutants was slightly lower than for those with WT.As shown in Table S1, the fitness scores are not showing definitecorrelation with measured current inhibition for the blockers. In thisstudy, the fitness scoreswere used to find out the proper ligand bindingpositionwith key interacting residues rather than to findout quite exactcorrelation with binding affinities because the fitness score has beenoptimized for the prediction of ligand binding positions [56]. In bindingmode analysis, we found that while the single mutants (F156A orF364A) showed π interactions with aromatic ring group of the com-pounds, the doublemutant (F156A/F364A) had no corresponding inter-actions (Fig. S5). Thus, these results are consistent with experimentalstudy, showing low current in double mutants.

To identifywhether non-blockers of TRESK also affect F156 and F364sites of TRESK, 9-anthracene carboxylic acid (9-AC), a chloride channelblocker, and tetraethylammonium (TEA), a K+ channel blocker but notTRESK, were applied to WT and mutant TRESK currents overexpressedin HEK-293A cells. As shown in Fig. 9, the 9-AC did not inhibit TRESKcurrents in WT and all mutants (F156A, F364A, and F156A/F364A).

3

6

3

6

3

6

3

6

Con TEA

nA

Wild-type

Con TEA

9-AC Con

9-AC Con

nA

-120 -60 60 mV -120 -60 60 mV -

-120 -60 60 mV -120 -60 60 mV -

F156AnA

-20-10

01020

% o

f cha

nges

(tre

atm

ent/c

ontr

ol)

nA

A

B

Wild type FF156A

Fig. 9. Effect of 9-AC and TEA on TRESK mutants. (A) No significant effect of 9-AC and TEchemicals treated (9-AC and TEA). Black and gray lines indicate control and treatment, rewhole-cell mode. Bar graphs show % changes of TRESK currents produced by 9-AC and TEA aative indicates inhibition.

TEA slightly inhibited TRESK currents. The inhibition amounts by TEAwere 4±8%, 4±7%, 2±8%, and 9±4% in WT, F156A, F364A, F156A/F364A, respectively. The effects of 9-AC and TEA on TRESK mutantswere similar to those on TRESK WT. In docking result of the TEA,completely different binding mode with other blockers showing thelowest fitness scores of 24.90, 29.23, 22.33, and 27.88 for WT, F156A,F364A, and F156A/F364A, respectively (Table S1). This computationalresult indicates that non-blocker of TRESK cannot bind properly withcavity region of the protein. These results suggest that each non-blockermight have a different binding site for each protein channel.

4. Conclusions

The aims of this study were to obtain reasonable 3D structureand to find out the proper binding site of K+ channel blockers. The3D structure was built by homology modeling with consideringthe TM topology of mTRESK obtained from TMHMM prediction.The minimized final structures were achieved for subsequent molec-ular docking and MD simulations. In pore radius analysis, similarpore radius of selectivity filter was observed in mTRESK comparedwith template structure, whereas pore dimension of gate regionwas narrower than that of the template. This narrow dimensionwas caused by four phenylalanine residues consisting of F156 andF364 located in each monomer. The representative structure of theligand-adjusted protein conformation was obtained from moleculardocking and 5 ns MD simulation with DPPC lipid bilayer. The confor-mation was used for final molecular docking for the blockers. Fromour docking study, the plausible binding mode of known K+ channelblockers was investigated into the mTRESK internal binding site, andkey interacting residues were identified. Especially, two phenylala-nine residues, F156 and F364, were considered as critical key resi-dues for blocker binding. This prediction was tested by mutationsof F156A, F364A, and F156A/F364A, and the results support the im-portance of these residues. Based on these results we can concludethat our modeled structure and docking results are reliable as a model

3

6

3

6

3

6

3

6

Con TEA

nA

9-AC Con

120 -60 60 mV -120 -60 60 mV

120 -60 60 mV -120 -60 60 mV

F156A/F364A

nA

F364A

9-ACCon

nA

Con TEA

nA

9-AC

TEA

364A F156A/F364A

A on currents produced from WT and mutants. Same experiment as in Fig. 8 exceptspectively. (B) Summary of the effects of 9-AC and TEA on TRESK channel activity int +60 mV. Each bar represents means±S.D. of six independent experiments. The neg-

1141S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

for finding the bindingmode properly. Our study is first modeling studyof TRESK to provide structural knowledge including ligand bindinginformation. Furthermore, this would be helpful to structure-baseddrugs design for TRESK as a potential therapeutic target in the reduc-ing of pain.

Acknowledgements

This work was supported by the National Research Foundation ofKorea (NRF) Grant funded by the Korean Government (NRF-2012-Fostering Core Leaders of the Future Basic Science Program).This research was supported by Pioneer Research Center Program(2009-0081539) and Management of Climate Change Program (2010-0029084) through the NRF funded by the Ministry of Education, Scienceand Technology (MEST) of Republic of Korea. And this work was alsosupported by the Next-Generation BioGreen 21 Program (PJ008038)from Rural Development Administration (RDA) of Republic of Korea.This work was supported by a grant from the NRF of Korea, which isfunded by the Korean Government (KRF-2008-313-E00026, NRF-2010-0017234, and NRF-2012-0000300).

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.bbamem.2012.11.021.

References

[1] D. Kim, Physiology and pharmacology of two-pore domain potassium channels,Curr. Pharm. Des. 11 (2005) 2717–2736.

[2] L. Heginbotham, Z. Lu, T. Abramson, R. MacKinnon, Mutations in the K+ channelsignature sequence, Biophys. J. 66 (1994) 1061–1067.

[3] L. Heginbotham, T. Abramson, R. MacKinnon, A functional connection betweenthe pores of distantly related ion channels as revealed by mutant K+ channels,Science 258 (1992) 1152.

[4] M.S.P. Sansom, I.H. Shrivastava, J.N. Bright, J. Tate, C.E. Capener, P.C. Biggin,Potassium channels: structures, models, simulations, Biochim. Biophys. Acta1565 (2002) 294–307.

[5] S. Choe, Potassium channel structures, Nat. Rev. Neurosci. 3 (2002) 115–121.[6] D.A. Bayliss, J.E. Sirois, E.M. Talley, The TASK family: two-pore domain background

K+ channels, Mol. Interv. 3 (2003) 205.[7] S.A.N. Goldstein, D. Bockenhauer, I. O'Kelly, N. Zilberberg, Potassium leak channels

and the KCNK family of two-P-domain subunits, Nat. Rev. Neurosci. 2 (2001)175–184.

[8] Y. Sano, K. Inamura, A. Miyake, S. Mochizuki, C. Kitada, H. Yokoi, K. Nozawa, H.Okada, H. Matsushime, K. Furuichi, A novel two-pore domain K+ channel,TRESK, is localized in the spinal cord, J. Biol. Chem. 278 (2003) 27406.

[9] M. Fink, F. Lesage, F. Duprat, C. Heurteaux, R. Reyes, M. Fosset, M. Lazdunski, A neu-ronal two P domain K+ channel stimulated by arachidonic acid and polyunsaturatedfatty acids, EMBO J. 17 (1998) 3297–3308.

[10] S.A.N. Goldstein, D.A. Bayliss, D. Kim, F. Lesage, L.D. Plant, S. Rajan, InternationalUnion of Pharmacology. LV. Nomenclature and molecular relationships of two-Ppotassium channels, Pharmacol. Rev. 57 (2005) 527.

[11] D.P. Lotshaw, Biophysical, pharmacological, and functional characteristics ofcloned and native mammalian two-pore domain K+ channels, Cell Biochem.Biophys. 47 (2007) 209–256.

[12] P. Enyedi, G. Czirjak, Molecular background of leak K+ currents: two-pore domainpotassium channels, Physiol. Rev. 90 (2010) 559.

[13] F. Lesage, E. Guillemare, M. Fink, F. Duprat, M. Lazdunski, G. Romey, J. Barhanin,TWIK-1, a ubiquitous human weakly inward rectifying K+ channel with a novelstructure, EMBO J. 15 (1996) 1004.

[14] D.J. Pountney, I. Gulkarov, E. Vega-Saenz de Miera, D. Holmes, M. Saganich, B.Rudy, M. Artman, W.A. Coetzee, Identification and cloning of TWIK-originatedsimilarity sequence (TOSS): a novel human 2-pore K+ channel principal subunit,FEBS Lett. 450 (1999) 191–196.

[15] A.J. Patel, F. Maingret, V. Magnone, M. Fosset, M. Lazdunski, E. Honore, TWIK-2, aninactivating 2P domain K+ channel, J. Biol. Chem. 275 (2000) 28722.

[16] E. Honore, The neuronal background K2P channels: focus on TREK1, Nat. Rev.Neurosci. 8 (2007) 251–261.

[17] G. Sandoz, D. Douguet, F. Chatelain, M. Lazdunski, F. Lesage, Extracellularacidification exerts opposite actions on TREK1 and TREK2 potassium channelsvia a single conserved histidine residue, Proc. Natl. Acad. Sci. 106 (2009)14628.

[18] F. Duprat, F. Lesage, M. Fink, R. Reyes, C. Heurteaux, M. Lazdunski, TASK, a humanbackground K+ channel to sense external pH variations near physiological pH,EMBO J. 16 (1997) 5464–5471.

[19] C. Karschin, E. Wischmeyer, R. Preisig-Muller, S. Rajan, C. Derst, K.H. Grzeschik,J. Daut, A. Karschin, Expression pattern in brain of TASK-1, TASK-3, and a tandempore domain K+ channel subunit, TASK-5, associated with the central auditorynervous system, Mol. Cell. Neurosci. 18 (2001) 632–648.

[20] F. Duprat, C. Girard, G. Jarretou,M. Lazdunski, Pancreatic two P domain K+ channelsTALK 1 and TALK 2 are activated by nitric oxide and reactive oxygen species,J. Physiol. 562 (2005) 235–244.

[21] D. Kang, E. Mariash, D. Kim, Functional expression of TRESK-2, a new memberof the tandem-pore K+ channel family, J. Biol. Chem. 279 (2004) 28063.

[22] A.D. O'Connell, M.J. Morton, M. Hunter, Two-pore domain K+ channels—molecularsensors, Biochim. Biophys. Acta 1566 (2002) 152–161.

[23] F. Lesage, M. Lazdunski, Molecular and functional properties of two-pore-domainpotassium channels, Am. J. Physiol. Renal Physiol. 279 (2000) F793.

[24] G. Czirjak, Z.E. Toth, P. Enyedi, The two-pore domain K+ channel, TRESK, isactivated by the cytoplasmic calcium signal through calcineurin, J. Biol. Chem.279 (2004) 18550.

[25] D. Kang, D. Kim, TREK-2 (K2P10. 1) and TRESK (K2P18. 1) are major backgroundK+ channels in dorsal root ganglion neurons, Am. J. Physiol. Cell Physiol. 291(2006) C138.

[26] C.H. Kindler, C.S. Yost, Two-pore domain potassium channels: new sites of localanesthetic action and toxicity, Reg. Anesth. Pain Med. 30 (2005) 260–274.

[27] C.H. Kindler, M. Paul, H. Zou, C. Liu, B.D. Winegar, A.T. Gray, C.S. Yost, Amide localanesthetics potently inhibit the human tandem pore domain background K+

channel TASK-2 (KCNK5), J. Pharmacol. Exp. Ther. 306 (2003) 84.[28] D. Kang, G.T. Kim, E.J. Kim, J.H. La, J.S. Lee, E.S. Lee, J.Y. Park, S.G. Hong, J. Han,

Lamotrigine inhibits TRESK regulated by G-protein coupled receptor agonists,Biochem. Biophys. Res. Commun. 367 (2008) 609–615.

[29] D.Y. Huang, B.W. Yu, Q.W. Fan, Roles of TRESK, a novel two-pore domain K+

channel, in pain pathway and general anesthesia, Neurosci. Bull. 24 (2008)166–172.

[30] G. Czirjak, P. Enyedi, Targeting of calcineurin to an NFAT-like docking site is re-quired for the calcium-dependent activation of the background K+ channel,TRESK, J. Biol. Chem. 281 (2006) 14677.

[31] J. Han, D. Kang, TRESK channel as a potential target to treat T-cell mediatedimmune dysfunction, Biochem. Biophys. Res. Commun. 390 (2009) 1102–1105.

[32] W. Treptow, M.L. Klein, The membrane-bound state of K2P potassium channels,J. Am. Chem. Soc. 132 (2010) 8145–8151.

[33] M.I. Niemeyer, F.D. Gonzalez-Nilo, L. Zuniga, W. Gonzalez, L.P. Cid, F.V. Sepulveda,Neutralization of a single arginine residue gates open a two-pore domain,alkali-activated K+ channel, Proc. Natl. Acad. Sci. 104 (2007) 666.

[34] A. Kollewe, A.Y. Lau, A. Sullivan, A structural model for K2P potassium channelsbased on 23 pairs of interacting sites and continuum electrostatics, J. Gen. Physiol.134 (2009) 53.

[35] A.K. Streit, M.F. Netter, F. Kempf, M. Walecki, S. Rinne, M.K. Bollepalli, R.Preisig-Muller, V. Renigunta, J. Daut, T. Baukrowitz, A specific two-pore domainpotassium channel blocker defines the structure of the TASK-1 open pore,J. Biol. Chem. 286 (2011) 13977.

[36] A. Krogh, B.E. Larsson, G. Von Heijne, E.L.L. Sonnhammer, Predicting transmem-brane protein topology with a hidden Markov model: application to completegenomes1, J. Mol. Biol. 305 (2001) 567–580.

[37] M. Larkin, G. Blackshields, N. Brown, R. Chenna, P. McGettigan, H. McWilliam, F.Valentin, I. Wallace, A. Wilm, R. Lopez, Clustal W and Clustal X version 2.0, Bioin-formatics 23 (2007) 2947.

[38] H. Berendsen, D. van der Spoel, R. Van Drunen, GROMACS: a message-passingparallel molecular dynamics implementation, Comput. Phys. Commun. 91(1995) 43–56.

[39] D. van der Spoel, E. Lindahl, B. Hess, A. van Buuren, E. Apol, P. Meulenhoff, D.Tieleman, A. Sijbers, K. Feenstra, R. van Drunen, GROMACS, User ManualVersion 4.0, The GROMACS development team, Groningen, The Netherlands,2006.

[40] T. Darden, D. York, L. Pedersen, Particle mesh Ewald: an N log (N) method forEwald sums in large systems, J. Chem. Phys. 98 (1993) 10089.

[41] R.A. Laskowski, M.W. MacArthur, D.S. Moss, J.M. Thornton, PROCHECK: a programto check the stereochemical quality of protein structures, J. Appl. Crystallogr. 26(1993) 283–291.

[42] M. Wiederstein, M.J. Sippl, ProSA-web: interactive web service for the recog-nition of errors in three-dimensional structures of proteins, Nucleic AcidsRes. 35 (2007) W407.

[43] C. Colovos, T.O. Yeates, Verification of protein structures: patterns of nonbondedatomic interactions, Protein Sci. 2 (1993) 1511.

[44] O.S. Smart, J.M. Goodfellow, B. Wallace, The pore dimensions of gramicidin A,Biophys. J. 65 (1993) 2455–2460.

[45] O.S. Smart, J.G. Neduvelil, X. Wang, B. Wallace, M.S.P. Sansom, HOLE: a programfor the analysis of the pore dimensions of ion channel structural models* 1, J.Mol. Graph. 14 (1996) 354–360.

[46] W. Humphrey, A. Dalke, K. Schulten, VMD: visual molecular dynamics, J. Mol.Graph. 14 (1996) 33–38.

[47] P.J. Bond, M.S.P. Sansom, Membrane protein dynamics versus environment:simulations of OmpA in a micelle and in a bilayer, J. Mol. Biol. 329 (2003)1035–1053.

[48] C. Domene, P.J. Bond, M.S.P. Sansom, Membrane protein simulations: ion chan-nels and bacterial outer membrane proteins, Adv. Protein Chem. 66 (2003)159–193.

[49] R.J. Law, C. Capener,M. Baaden, P.J. Bond, J. Campbell, G. Patargias, Y. Arinaminpathy,M.S.P. Sansom, Membrane protein structure quality in molecular dynamics simula-tion, J. Mol. Graph. Model. 24 (2005) 157–165.

http://dx.doi.org/10.1016/j.bbamem.2012.11.021http://dx.doi.org/10.1016/j.bbamem.2012.11.021

1142 S. Kim et al. / Biochimica et Biophysica Acta 1828 (2013) 1131–1142

[50] J.A. Lemkul, D.R. Bevan, A comparative molecular dynamics analysis of the amyloid[beta]-peptide in a lipid bilayer, Arch. Biochem. Biophys. 470 (2008) 54–63.

[51] C. Oostenbrink, A. Villa, A.E. Mark, W.F. Van Gunsteren, A biomolecular force fieldbased on the free enthalpy of hydration and solvation: the GROMOS force-fieldparameter sets 53A5 and 53A6, J. Comput. Chem. 25 (2004) 1656–1676.

[52] O. Berger, O. Edholm, F. Jähnig, Molecular dynamics simulations of a fluidbilayer of dipalmitoylphosphatidylcholine at full hydration, constant pressure,and constant temperature, Biophys. J. 72 (1997) 2002–2013.

[53] H. Berendsen, J. Postma, W. Van Gunsteren, J. Hermans, Interaction models forwater in relation to protein hydration, Intermol. Forces 331 (1981).

[54] A.W. Schuttelkopf, D.M.F. Van Aalten, PRODRG: a tool for high-throughput crys-tallography of protein–ligand complexes, Acta Crystallogr. D: Biol. Crystallogr.60 (2004) 1355–1363.

[55] B. Hess, H. Bekker, H.J.C. Berendsen, J.G.E.M. Fraaije, LINCS: a linear constraintsolver for molecular simulations, J. Comput. Chem. 18 (1997) 1463–1472.

[56] M.L. Verdonk, J.C. Cole, M.J. Hartshorn, C.W. Murray, R.D. Taylor, Improvedprotein–ligand docking using GOLD, Proteins: Struct. Funct. Bioinf. 52 (2003)609–623.

[57] R. Laskowski, M. MacArthur, D. Smith, D. Jones, E. Hutchinson, A. Morris, B.Naylor, D. Moss, J. Thornton, PROCHECK Manual, Version 3.0, Oxford MolecularLtd., Oxford, UK, 1994.

[58] M.J. Lenaeus, M. Vamvouka, P.J. Focia, A. Gross, Structural basis of TEA blockade ina model potassium channel, Nat. Struct. Mol. Biol. 12 (2005) 454–459.

[59] M. Masetti, A. Cavalli, M. Recanatini, Modeling the hERG potassium channel in aphospholipid bilayer: molecular dynamics and drug docking studies, J. Comput.Chem. 29 (2008) 795–808.

[60] S.G. Brohawn, J. delMármol, R.MacKinnon, Crystal Structure of theHumanK2P TRAAK,a Lipid-and Mechano-Sensitive K+ Ion Channel, Science's STKE 335 (2012) 436.

[61] A.N. Miller, S.B. Long, Crystal Structure of the Human Two-Pore Domain Potassi-um Channel K2P1, Science's STKE 335 (2012) 432.

Identification of blocker binding site in mouse TRESK by molecular modeling and mutational studies1. Introduction2. Materials and methods2.1. Prediction of transmembrane helices2.2. Generation of 3D structure2.3. Solvated energy minimization and validation of 3D structure2.4. Pore radius analysis2.5. Molecular dynamics simulation in explicit membrane environment2.6. Molecular docking study2.7. Transfection2.8. Site-directed mutagenesis2.9. Electrophysiological studies2.10. Statistical analysis

3. Results and discussion3.1. Prediction of number and topology for transmembrane helices3.2. Sequence alignment for the homology modeling3.3. Generation of 3D structures using MthK as a template3.4. Validation for the structural quality3.5. Structural comparison of pore region with template3.6. Selection of representative structure for molecular docking3.7. Prediction of blocker binding site3.8. Modulation of mTRESK by pharmacological agents

4. ConclusionsAcknowledgementsAppendix A. Supplementary dataReferences