NMR investigation of the isolated second voltage-sensing domain of human Nav1.4 channel

Biochimica et Biophysica Acta 1859 (2017) 493–506

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta journal homepage: www.elsevier.com/locate/bbamem

NMR investigation of the isolated second voltage-sensing domain of human Nav1.4 channel A.S. Paramonov a,b, E.N. Lyukmanova a,b, M.Yu. Myshkin a,c, M.A. Shulepko a,b, D.S. Kulbatskii a,b, N.S. Petrosian a,c, A.O. Chugunov a, D.A. Dolgikh a,b, M.P. Kirpichnikov a,b, A.S. Arseniev a,c, Z.O. Shenkarev a,c,⁎ a b c

Shemyakin and Ovchinnikov Institute of Bioorganic Chemistry, Russian Academy of Sciences, Miklukho-Maklaya str., 16/10, Moscow 117997, Russia Lomonosov Moscow State University, Moscow 119991, Russia Moscow Institute of Physics and Technology (State University), Institutskiy Pereulok 9, Dolgoprudny, Moscow Region 141700, Russia

a r t i c l e

i n f o

Article history: Received 30 September 2016 Received in revised form 6 December 2016 Accepted 4 January 2017 Available online 6 January 2017 Keywords: Na+ channels Cell-free expression NMR spectroscopy Voltage-sensing domain

a b s t r a c t Voltage-gated Na+ channels are essential for the functioning of cardiovascular, muscular, and nervous systems. The α-subunit of eukaryotic Na+ channel consists of ~2000 amino acid residues and encloses 24 transmembrane (TM) helices, which form five membrane domains: four voltage-sensing (VSD) and one pore domain. The structural complexity significantly impedes recombinant production and structural studies of full-sized Na+ channels. Modular organization of voltage-gated channels gives an idea for studying of the isolated second VSD of human skeletal muscle Nav1.4 channel (VSD-II). Several variants of VSD-II (~150 a.a., four TM helices) with different Nand C-termini were produced by cell-free expression. Screening of membrane mimetics revealed low stability of VSD-II samples in media containing phospholipids (bicelles, nanodiscs) associated with the aggregation of electrically neutral domain molecules. The almost complete resonance assignment of 13C,15N-labeled VSD-II was obtained in LPPG micelles. The secondary structure of VSD-II showed similarity with the structures of bacterial Na+ channels. The fragment of S4 TM helix between the first and second conserved Arg residues probably adopts 310helical conformation. Water accessibility of S3 helix, observed by the Mn2+ titration, pointed to the formation of water-filled crevices in the micelle embedded VSD-II. 15N relaxation data revealed characteristic pattern of μs–ms time scale motions in the VSD-II regions sharing expected interhelical contacts. VSD-II demonstrated enhanced mobility at ps–ns time scale as compared to isolated VSDs of K+ channels. These results validate structural studies of isolated VSDs of Na+ channels and show possible pitfalls in application of this ‘divide and conquer’ approach. © 2017 Elsevier B.V. All rights reserved.

1. Introduction

Abbreviations: CF, cell-free; DC7PC, 1,2-diheptanoyl-sn-glycero-3-phosphocholine; DDM, n-dodecyl β-D-maltoside; DHPC, 1,2-dihexanoyl-sn-glycero-3-phosphocholine; DMPC, 1,2-dimiristoyl-sn-glycero-3-phosphocholine; DMPG, 1,2-dimiristoyl-sn-glycero3-phosphoglycerol; DOPG, 1,2-dioleoyl-sn-glycero-3-phosphoglycerol; FM, feeding mixture; FOS-12 (DPC), n-dodecylphosphocholine; FOS-14, n-tetradecylphosphocholine; Kv channel, K+ voltage-gated channel; LDAO, n-dodecyl-N,N-dimethylamine-N-oxide; LMPG, 1-myristoyl-2-hydroxy-sn-glycero-3-phosphoglycerol; LPN, lipid-protein nanodisc; LPPG, 1-palmitoyl-2-hydroxy-sn-glycero-3-phosphoglycerol; MD, molecular dynamics; MP, membrane protein; MSP, membrane scaffold protein; Nav channel, Na+ voltage-gated channel; POPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; POPE, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine; RM, reaction mixture; RMSD, root-mean-square deviation; RMSF, root-mean-square fluctuation; s.d., standard deviation; TM, transmembrane; τR, rotational correlation time; TROSY, transverse relaxation optimized spectroscopy; VGIC, voltage-gated ion channel; VSD, voltagesensing domain; VSD-II, isolated second VSD of human Nav1.4 channel; VSD-KvAP, isolated VSD of KvAP channel. ⁎ Corresponding author at: Str. Mikluho-Maklaya, 16/10, Moscow 117997, Russia. E-mail address: [email protected] (Z.O. Shenkarev).

http://dx.doi.org/10.1016/j.bbamem.2017.01.004 0005-2736/© 2017 Elsevier B.V. All rights reserved.

K+, Na+, and Ca2+ voltage-gated channels (Kv, Nav and Cav channels) are the structurally related integral membrane proteins (MPs), which belong to the superfamily of voltage-gated ion channels (VGICs or P-loop superfamily) [1,2]. These channels are involved in a wide range of physiological phenomena, including the excitability of cardiac, muscle and neuronal cells, propagation of nerve signals and secretion of hormones and neurotransmitters [1]. The VGICs have modular architecture and represent either non-covalent homotetramers of four individual subunits (Kv channels, bacterial Nav channels) or tetramers of four pseudosubunits linked within one polypeptide chain (eukaryotic Nav and Cav channels) (Fig. 1A) [1–3]. Each of the (pseudo-)subunits contains six transmembrane (TM) helices, four of which (helices S1–S4) form a voltage-sensing domain (VSD), and the remaining helices (S5 and S6) together with a ‘pore(P)-loop’ participate in formation of the pore wall and selective filter (pore domain, PD, Fig. 1B) [1–3]. The S4 helix of VSD, sometimes called ‘voltage sensor’, accommodates several positively charged

494

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

Fig. 1. Schematic representation of the spatial organization of the α-subunit of the human skeletal muscle Nav1.4 channel (SCN4A) in the cellular membrane and amino acid sequence of the Voltage-Sensing Domain (VSD) of the second pseudosubunit (VSD-II). (A). The polypeptide chain of Nav channel involves four (I, II, III, IV) pseudosubunits. The helices of VSDs (S1–S4) and of the Pore Domain (PD) (S5, S6) are shown in blue and gray, respectively. The positions of conserved Arg/Lys residues, responsible for the voltage gating, are marked by crosses. “P” and “G” denote possible sites of phosphorylation and glycosylation, respectively. IFM residues (1310−1312) are critical for fast inactivation. The fragment corresponding to VSD-II is encircled with the red frame. (B). The pseudo-tetrameric α-subunit of the Nav1.4 channel with one PD and four VSDs. (C). The three variants of VSD-II produced by cell-free expression in the present work. The additional residues introduced into VSD-II sequence are highlighted by gray background. The Cys residues mutated to Ser are underlined. The fragments corresponding to the S1–S4 TM helices as annotated in the UNIPROT database are shown by blue background. The sites of mutations associated with the periodic paralysis, paramyotonia congenita and myotonia are in red. The sites of conserved charged/polar residues in the S4 helix are numbered.

residues (usually four Arg), which are directly involved in the channel gating [4,5]. For the past years, large progress was achieved in the structural studies of homotetrameric VGICs. The crystal structures of eukaryotic and prokaryotic Kv channels [6,7], and several Nav channels from proteobacteria (e.g. NavAB and NavRH) [8,9] were determined. Unfortunately, this rapid progress does not significantly facilitate the studies of eukaryotic Nav and Cav channels. The α-subunit of these proteins contains four VSDs (S1–S4) and four pore-forming regions (S5–S6) interleaved in a single polypeptide chain (~ 2000 residues, 24 TM segments, Fig. 1A). Moreover, the α-subunit undergoes extensive posttranslational modifications and associates with auxiliary subunits within the membrane. Due to this complexity, recombinant production, handling and structural studies of the full-sized eukaryotic Nav and Cav channels are significantly complicated [3]. Presently only two structures of such channels were determined by cryoelectron microscopy: the low resolution structure of Nav channel from the eel Electrophorus electricus [10] and high resolution structure of Cav1.1 channel from rabbit skeletal muscle [11]. There is a variety of neurological and cardiovascular disorders caused by mutations in the VSDs of human Kv, Nav and Cav channels [2]. The VSDs of the different channels demonstrate marked variability and frequently possess unique binding sites [12]. This makes them attractive targets for development of new selective drugs against various channelopathies [12–14]. Notably, the four VSDs within one Nav channel (VSD-I–IV) have different properties and display different pharmacology [15]. The structural and biochemical data indicate that VSDs could behave as structurally independent units within the lipid membrane. The highresolution structures of homotetrameric Kv and Nav channels revealed that VSDs loosely associate with PD [16] and the domains are separated by a thin layer of lipid molecules [7,8,12]. The VSDs and their fragments demonstrate marked ‘portability’ [17], they can be grafted to voltage insensitive Kv channels in order to confer voltage-gating [18,19] or transplanted from mammalian Nav channel into homotetrameric Kv and Nav channels to transfer pharmacological properties [12,15]. In addition, there are two classes of membrane proteins, which membrane

domains are consist of VSDs without associated PD: voltage-gated proton channel Hv1 [20] and voltage-sensitive phosphatase VSP [21]. The several examples indicate that structure, dynamics and pharmacology of VSDs from the homotetrameric Kv channels could be studied using isolated variants of the domains [6,22–26]. However, the applicability of this ‘divide and conquer’ approach to structural studies of VSDs from eukaryotic Nav channels was not assayed previously. In this report we describe the results of the NMR study of the isolated VSD from the second pseudosubunit (VSD-II) of the human skeletal muscle Nav1.4 (SCN4A) channel. This domain represents important pharmacological target. It hosts the mutations linked to different onsets of periodic paralysis (R669H, R672H/G/S/C, R675G/Q/W) [27,28], paramyotonia congenita (I693T) [2] and myotonia (F671S) [29], and accommodates binding sites for spider ‘voltage sensor’ and β-scorpion toxins which modulate voltage-gating [15,30]. Three variants of isolated VSD-II with different N- and C-termini were produced by cell-free (CF) expression. The different membrane mimicking media [detergent micelles, lipid/detergent bicelles and lipid-protein nanodiscs (LPNs)] were assayed to find optimal conditions for the NMR study of VSD-II. In spite of the low stability of the isolated domain in the majority of tested media, we were able to characterize in detail the secondary structure and backbone dynamics of VSD-II in the micelles of anionic lysolipid LPPG. The determined secondary structure and observed pattern of μs–ms time scale motions within the VSD-II molecule confirmed its similarity to the VSDs of homotetrameric Kv and Nav channels. These results prove the validity of structural studies of isolated VSDs from eukaryotic Nav channels. The major obstacles which could arise in such studies are discussed. 2. Materials and methods 2.1. Materials All used phospholipids (DHPC, DC7PC, DMPC, DMPG, POPC and DOPG) and lysophospholipids (LMPG and LPPG) were products of Avanti Polar Lipids (Alabaster, AL). All undeuterated detergents (CHAPS, DDM, LDAO, FOS-12, FOS-14) were products of Anatrace Inc.

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

(Maumee, OH). The deuterated D2O and d38-FOS-12 were products of CIL (Andover, MA). The membrane binding domain (44–243 fragment) of human apolipoprotein A-I without His6-tag (membrane scaffold protein, MSP, 23.3 kDa) was produced and purified as described in [31]. 2.2. Cell-free production of VSD-II samples Gene encoding a variant of isolated VSD-II (fragment 554–696 of human Nav1.4 channel) and C-terminal His6-tag (17.5 kDa, Fig. 1C) was amplified by PCR from overlapping synthetic oligonucleotides (Evrogen, Moscow, Russia) with optimized for E. coli codons. The gene was cloned into pET20b(+) (Novagen, Madison, WI) vector on the NdeI and HindIII restriction sites, and into pIVEX2.3d vector (Roche, Mannheim, Germany) on the NcoI and SmaI restriction sites. The VSDIINC gene (fragment 565–702 of Nav1.4, 16.5 kDa) and VSD-IIK (gene encoding VSD-II with six additional Lys residues on the N- and C-termini, 18.2 kDa) were obtained by PCR. The resulting plasmids pET20b(+)/ VSD-II, pIVEX2.3d/VSD-II, pIVEX2.3d/VSD-IINC and pIVEX2.3d/VSD-IIK were used as a template in continuous exchange CF expression system based on the E. coli S30 extract as described in [32]. The volume ratio of the reaction mixture (RM) to feeding mixture (FM) was 1:15. The dialysis membrane tubing with the cut-off of 12 kDa (Sigma Aldrich) was used to separate RM and FM. CF reactions were performed without addition of any membrane-mimicking components into RM. The synthesis was carried out at 30 °C with a gentle mixing for 20 h. Soluble and insoluble fractions of the RM after synthesis were separated by centrifugation for 15 min at 14000 rpm. The obtained fractions were analyzed by 12% Tris/Tricine SDS-PAGE. To remove RNA and DNA traces the obtained RM pellet was resuspended in 20 mM Tris/HCl, pH 8.0, 100 mM NaCl containing 5 μg/mL of DNAseI (Sigma) and 1.5 μg/mL of RNAse-A (Fermentas, Vilnius, Lithuania) and incubated for 30 min. 1 mL of this solution was used per 1 mL of RM. After that the sample was washed two times using the same buffer without enzymes and two times by deionized water (Milli-Q™, Millipore, Billerica, MA). Between washes the sample was centrifuged for 5 min at 14000 rpm. 15 N-labeled proteins were synthesized using 15N algal amino acid mixture (Isotec™, Sigma-Aldrich, Miamisburg, OH) at concentration of 3.7 mg/mL. 15N-Trp, 15N-Gln and 15N-Asp (CIL) were added to the RM and FM at concentrations of 2.3 mM, 1.3 mM and 1.3 mM, respectively. In some cases unlabeled Gln was used. For production of 13C-15N-labeled proteins 13C,15N algal amino acid mixture (Isotec™), 13C-15NTrp, 15N-Gln and 13C-15N-Asn (CIL) were used at the same concentrations. For the production of selectively labeled VSD-II the individual unlabeled (Sigma) and 15N- or 15N-13C-labeled amino acids (CIL) were used. 2.3. Reconstitution of VSD-II in to micelles, bicelles, and LPNs RM precipitate was solubilized with 300 μL 20 mM Tris/Ac, pH 5.0– 7.0 containing 3–5% (w/v) of appropriate detergent or detergent/lipid mixture. The several cycles of sonication using Transsonic T490DH ultrasonic bath (Elma, Singen, Germany) were used to speed-up the solubilization process. The solubility of VSD was controlled by SDS-PAGE and the concentration of detergents was monitored by 1D 1H NMR spectroscopy. Assembly of LPN particles containing VSD-II or VSD-IIK was done using protocols previously developed for VSD of KvAP channel (VSDKvAP) [31]. The proteins were solubilized from the RM precipitate in the buffer A (20 mM Tris/HCl, 250 mM NaCl, 1 mM NaN3, pH 8.0) containing 3% LPPG or LDAO (VSD-II or VSD-IIK, respectively). The proteins were mixed with MSP, lipids (DMPC, DMPG, POPC or DOPG) and sodium cholate at a 1:20:800:1600 M ratio. The final VSDs concentration was about 0.5 mg/mL. The mixtures were incubated overnight at 25 °C with gentle shaking. Reconstruction of the VSD/LPN complexes was initiated by the incubation with 0.5 g of Biobeads™ (BioRad, Hercules, CA)

495

per 1 mL of the reaction mixture during 2 h at 25 °C and gentle shaking. The VSD/LPN complexes were purified on Ni2+-column equilibrated in the buffer A. The column was washed by four volumes of the buffer A and six volumes of the buffer A with 50 mM imidazole. Fractions with VSD were eluted by the buffer A with 100 mM imidazole, dialyzed against 20 mM Tris/Ac, 10 mM EDTA, 1 mM NaN3, pH 7.0 and concentrated using Stirred Cells with Ultrafiltration Membranes, NMWL 10,000 (Millipore). For the structural study VSD-II samples in LPPG micelles were purified by Ni2+-chromatography. The appropriate amounts of LPPG were added into all the buffers used for chromatography. The detergent concentration at the column and in the protein samples was kept ≥0.5%. Instead of dialysis, the resulting samples were desalted by repeated cycles of dilution/concentration using Stirred Cells with Ultrafiltration Membranes, NMWL 10,000. 2.4. Measurement and analysis of NMR data The NMR spectra were recorded for 0.1–0.6 mM VSD-II, VSD-IINC or VSD-IIK samples on Bruker Avance-III 600 and Avance-III 800 spectrometers equipped with cryoprobes at 45 °С. For resonance assignment and relaxation measurements a 0.6 mM samples of uniformly 15N- and 13 15 C, N-labeled VSD-II (5% LPPG, 20 mM Tris/Ac pH 6.0, 1 mM NaN3, 5% D2O) were used. A series of TROSY and HSQC based triple-resonance (1H,13C,15N) NMR experiments was used for resonance assignment of backbone and side chains. The 3D HNCO, HNCACO, HNCA, HNCOCA, HNCACB and CBCACONH spectra were collected using non-uniform sampling method with 50% of sparse sampling. The 3D 15N-NOESYHSQC, 13C-NOESY-HSQC (τm = 80 ms) and 13C-HCCH-TOCSY were measured using conventional sampling scheme. To complete assignment, three selectively labeled samples were used: (#1) 13C,15N-Ile, (#2) 15N–Leu, 13C(1)-Val, (#3) 15N-Thr, 13C(1)-Ile, 13C(1)-Phe, 13 15 C, N-Trp. For selectively labeled samples 2D 15N-HSQC spectra and 2D 1H,15N planes from HNCO and HNCA experiments were measured. For sample (#1) 3D 15N-NOESY-HSQC spectrum was also acquired. Proton chemical shifts were referenced directly using trimethylsilylpropanoic acid (TSP). For 13C and 15N nuclei the indirect reference was used. Spectra were processed with MDDNMR [33] and TOPSPIN (Bruker) and analyzed in CARA. Secondary chemical shifts, content of α-helical secondary structure and random coil index order parameters (RCI-S2) [34] were calculated by the TALOS+ software [35]. The paramagnetic Mn2+ cations (in the form of MnCl2) were added to the 15N-VSD-II sample (70 μM, 2.5% LPPG) to the final Mn2+ concentration of ~1 mM (molar ratio of VSD-II/LPPG/Mn2+ ~ 1:700:14). 15Nrelaxation parameters (R1 and R2 relaxation rates and heteronuclear 15 N–{1H} NOEs) were measured for 81 non overlapped and non broadened 15NH groups at 80 MHz using the standard set of 15N-HSQC based pseudo 3D experiments. Chemical shift anisotropy/dipolar cross-correlation (ηXY) rates were measured for 118 15NH groups using constanttime amplitude modulated 3D TROSY-HNCO experiment [36]. Modelfree analysis of relaxation data was performed in FastModelFree software [37] using isotropic rotational model. The detailed description of statistical analysis of measured relaxation data is provided in the Supplementary materials and methods. Chemical shifts of N-formylated and N-deformylated forms of VSD-II have been deposited as BMRB entries 26992 and 26993, respectively. 2.5. Computer modeling Homology model of human Nav1.4 VSD-II was generated with MODELLER 8.2 [38] using the NavAb/hNav1.7-IV chimeric channel (PDB ID: 5EK0 [12]) as a template (see Fig. 7 for sequence alignment). A fragment of neuronal membrane-mimicking bilayer (POPC/POPE/ cholesterol, 196:98:98 molecules, respectively) was assembled and solvated inside the rectangular box with dimensions 9.3 × 9.3 × 16.2 nm3. System was energy minimized and heated to 37 °С during 480 ps. The

496

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

model of VSD-II was placed inside the equilibrated bilayer; some phospholipid and cholesterol molecules were removed for this purpose. The system was re-solvated using water molecules and one Na+ ion required to maintain electroneutrality. After that, the system was energy minimized, heated to 37 °С during 480 ps, and 270 ns molecular dynamics (MD) trajectory was calculated. All simulations were performed with GROMACS 5.1.4 package [39] using Gromos96 43a2x parameters set. Energy minimization was done using steepest descent algorithm, short-range electrostatic cut-off was 1.2 nm, short-range Van der Waals cut-off was 1.2 nm, SPC water model was used. MD calculations were carried out with a time step of 2 fs, imposing 3D periodic boundary conditions, in the Berendsen ensemble with semi-isotropic pressure of 1 bar and a constant temperature of 37 °С. All components of the system (water + ions, membrane lipids and protein) were coupled separately to a temperature bath using V-rescale thermostat. Electrostatic interactions were treated with the Particle Mesh Ewald (PME) algorithm (0.12 nm Fourier spacing). 3. Results

Fig. 2. SDS-PAGE analysis of CF synthesized VSD-II samples. 1 — Molecular mass markers; 2 — the RM precipitate after synthesis using pET20b(+)/VSD-II vector; 3 — the RM precipitate after synthesis using pIVEX2.3d/VSD-II vector; 4 — the VSD-II solubilized from the RM precipitate with 0.5% LMPG and diluted twice to 0.25% LMPG (k); 5–7 — “k” with 0.75% DMPC and 0.25%, 1.5%, 2.25% CHAPS, respectively; 8–10 — “k” with 0.75% DMPC and 0.25%, 1.5%, 2.25% DHPC, respectively; 11 — lane “10” after 2 h incubation at 45 °C; 12 — the VSD-IIK/LPN/DMPG sample. The bands corresponding to monomeric VSD-II or VSD-II-K and MSP are marked by the arrows. The bands corresponding to the aggregates of VSD-II are marked by asterisks.

3.1. Gene design and cell-free production of VSD-II In the human Nav1.4 channel the polypeptide fragments which flank different pseudo-subunits form relatively large cytoplasmic domains (Fig. 1), but within a pseudo-subunit the region between VSD and PD is short and has quite hydrophobic sequence. Therefore to isolate VSD-II without structural distortions we expressed the 554–696 fragment of the channel which covers not only the four predicted TM helices (S1–S4), but also involves additional N- and C-terminal regions. These regions should adopt the conformation of amphipathic α-helices (S0 and S45, respectively) which is probably situated on the membrane/ cytoplasm interface (see below). The sequence of the domain contains three Cys residues: two vicinal at the N-terminal region and one in the helix S1. Due to its cytoplasmic or TM localization, these cysteines probably do not participate in the disulfide bond formation. To reduce the aggregation tendency of the produced domain and to simplify handling of the protein samples these residues were mutated to Ser (Fig. 1C, underlined). For purification of the domain by Ni2+-chromatography 6 His residues (His6-tag) were introduced on the C-terminus. Usage of pET20b(+) based vectors as a template for continuous exchange CF synthesis of VSD-II resulted in the yield ~1 mg of the target protein from 1 mL of RM. To increase the production level, we cloned the VSD-II gene into pIVEX2.3d vector. Early it was reported that pIVEX vectors provide high-level CF production of MPs, e.g. the G-protein coupled receptors [40]. Indeed, usage of the pIVEX2.3d/VSD-II vector as a template for CF synthesis increased the yield of the protein up to 3.6 mg/mL (Fig. 2, compare lanes 2,3). 3.2. Screening of membrane mimetics for NMR study of VSD-II SDS-PAGE analysis revealed that CF synthesized VSD-II is presented exclusively as a precipitate in the insoluble fraction of RM. To transfer the protein into micelles of different detergents, the RM precipitate containing synthesized 15N-labeled VSD-II was directly solubilized by the target detergent. The relatively high purity of the solubilized domain (Fig. 2, lane 4) indicated that initial NMR screening of the membrane mimetics could be done without additional purification. The screening was guided by following criteria: (#1) sample stability at 45 °C; (#2) the quality of 2D 1H,15N-correlation (TROSY) spectra of VSD-II (number of signals, dispersion, line broadening). We expected to observe at the most 130–140 individual resonances of backbone HN groups (153 residues minus His-tag, five Pro, N-terminal residues and overlapped signals) and nine HN signals of Gly residues resonating in the specific region of the 15N-TROSY spectrum (Fig. 3, red rectangle). The results

of the screening are summarized in the Table 1, and representative spectra are shown in Fig. 3 and in Figs. S1–S3. Solubilization of VSD-II in the micelles of zwitterionic phosphocholine detergent FOS-12 (DPC) at pH ~5.0 resulted in the samples with the excellent stability. The dispersion of 1HN signals (from 7.5 to 8.7 ppm) observed in the TROSY spectrum was quite typical for helical MPs (Fig. 3A). N110 backbone HN resonances were observed, but some of them were significantly broadened. As a result only six Gly resonances with sufficient intensity were observed. The increase of the pH from 5.0 to 7.0 increased the signal broadening, probably due to exchange with solvent, thus making many signals unobservable (Fig. S1, Table 1). Use of zwitterionic phosphocholine FOS-14 with longer hydrophobic chain further decreased the spectra quality (Fig. 3E). So the observed broadening was not connected with the mismatch in the thickness of hydrophobic regions between the micelles and VSD-II. In spite of the good stability of VSD-II samples in FOS detergents, the quality of obtained spectra was not sufficient for structural investigations. The high quality TROSY spectrum was observed for VSD-II in the micelles of weakly cationic detergent LDAO (pKa of monomer ~5.0 [41]) at pH ~5.0 (Fig. 3B). The number of observed HN signals was close to expected and their line widths were significantly smaller than in FOS-12 micelles. At the same time, the stability of the VSD-II sample in LDAO micelles was low. The significant drop in the intensity of NMR spectra was observed during several hours of measurements at 45 °C (Fig. 3I). This pointed to the formation of large VSD-II aggregates, which are not observable by NMR. The aggregation of VSD-II was also observed at pH ~ 7.0, but lowering of the pH value to 4.0 significantly increased the sample stability (Fig. S2). Addition of FOS-12 to LDAO (1:3–3:1 M ratios) led to stabilization of the VSD-II sample at pH 5.0 and above. Nevertheless with the increase of FOS-12 fraction the resonances became more and more broadened and the spectra quality became worse (Figs. 3D, S1 and S2). Addition of mild non-ionic detergent DDM to VSD-II sample in LDAO micelles also increased the signal broadening; and the protein stability in this case was not significantly improved (Table 1). It should be noted that DDM itself was not able to solubilize VSD-II from the RM precipitate. The anionic lysophospholipids LMPG and LPPG efficiently solubilized VSD-II from the RM precipitate providing samples with long-term stability. The 2D 1H,15N-correlation spectra of VSD-II in these media were very similar to each other and showed relatively high quality (Figs. 3C and S3). The overall number of observed HN cross-peaks slightly exceeded the expected value, but was smaller (closer to the expectations) in the LPPG micelles. pH titration of the VSD-II/LPPG samples in the range from 5.0 to 6.0 revealed significant changes in the signal

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

497

Fig. 3. NMR screening of membrane mimetics for the study of VSD-II, VSD-IINC and VSD-IIK. (A-F) 2D 15N-TROSY spectra of VSD-II or VSD-IINC (0.1 mM, 45 °C) in different membrane mimetics. The used variant of the domain, detergent (concentration 3%) and pH value are shown on the legend in the each panel. The inserts show the downfield spectral regions with the 1H15Nε1 signals of Trp side-chain. The “Gly regions” of the spectra are marked by red dashed rectangle. (G) 2D 15N–TROSY spectrum of VSD-IIK (0.1 mM, 45 °C) in the LPN particles containing DMPG. (H) The 2D 1H,15N–plane (‘NOE on’) from the 15N-{1H}-NOE experiment (0.3 mM VSD-IIK, 3% LDAO, 45 °C). The peaks with negative intensities are shown in green. (I) NMR stability test for VSD-II/LDAO sample.

positions, probably associated with the changes in the ionization state of VSD-II side-chains. At the same time, the further increase of the pH to 6.5 led to negligible spectral changes, indicating that pH 6.0 is the optimal conditions for measurements of the NMR spectra. The short-chain zwitterionic lipids DHPC and DC7PC are frequently used for NMR studies of MPs due to their detergent-like properties [23,42]. Neither DHPC nor DC7PC were able to effectively solubilize VSD-II from the RM precipitate at pH in 5.0–7.0 range. Incorporation of VSD into DMPC/DHPC bicelles by direct solubilization of the RM precipitate was also not successful. The low efficiency of the VSD-II solubilization by these membrane mimetics could be connected with either with low ‘solubilization strength’ of DHPC and DC7PC per se or with instability of the VSD-II molecule in these media. To discriminate between these two possibilities, we made additional experiments. The domain was solubilized from RM precipitate by 0.5% (w/v) LMPG (pH 6.0) and after that the sample was diluted twice by concentrated DMPC/DHPC or DMPC/CHAPS solution (Fig. 2, lanes 5–10). Final VSD-II, LMPG and DMPC concentrations were 0.05 mM, 0.25% (5.2 mM) and 0.75% (11.1 mM), respectively. The final DHPC and CHAPS concentrations were varied from 0.25 to 2.25% (5.5–50 mM and 4.1–37 mM, respectively). It was found that increase in the CHAPS concentration above the 0.25% led to dramatic reduction in the amount of soluble VSD-II (Fig. 2, lanes 5–7). In contrast to that, increase in the DHPC concentration up to the 2.25% (LMPG/DMPC/DHPC molar ratio ~ 1:2.1:9.5) did not induce the immediate precipitation of the domain (Fig. 2, lane 10). At the

same time the 2-h incubation of this sample at 45 °C resulted in complete protein aggregation (Fig. 2, lane 11). Quite similarly, the stepwise addition of DC7PC to VSD-II/FOS-14 sample led to drastic decrease in the spectra quality and sample stability (Fig. S1, Table 1). Moreover the addition of DC7PC excess to VSD solubilized in LPPG (pH 6.0) induced the fast (20 min) irreversible precipitation of the protein. The obtained results indicated that VSD-II molecule is unstable in the solutions containing short-chain lipids and bicelles. The attempts to incorporate VSD-II into nanodiscs containing DMPC, DMPG, POPC or DOPG lipids failed due to extremely low yield of the protein incorporation. Notably, we employed the protocol which previously had been successfully used for the reconstruction of the LPNs containing isolated VSD-KvAP [31]. Most probably, the isolated VSD-II of Nav1.4 channel has low stability in the environment containing phospholipids and precipitates or forms aggregates during LPN formation. 3.3. Optimization of VSD-II sequence for stability enhancement The increase in stability of the VSD-II/LDAO sample observed upon lowering of pH value to 4.0 indicated that the aggregation tendency of the domain could be controlled by variation of its electric charge and the electric charge of the surrounding detergent molecules. The used variant of VSD has the theoretical pI value about 7.1 and it should be uncharged at neutral pH. To increase the electrostatic repulsion between the domain molecules we engineered two additional VSD-II variants

498

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

Table 1 Screening of membrane mimicking media for NMR study of VSD-II. pH

Quality of 2D 15N-TROSY spectrum

# of HN signals (Gly signals)

Stability (incubation at 45 °C, 10 d)a

5.0 7.0 5.0

+b −c −c

116 (6) 95 (4) 107 (5)

Stable Stable Stable

LDAO/FOS-12(3:1–1:3) LDAO/DDM (3:1)

5.0–7.0 4.0 5.0 5.0

+++ ++d +d +d

137 (9) 123 (8) 115 (8) 117 (8)

Unstable, ~60% aggregation in 2d Unstable, ~40% aggregation in 10d Stable Unstable, ~50% aggregation in 3 d

Anionic micelles LPPG LMPG

5.0–6.5 6.0

+++ ++e

147 (8) 153 (9)

Stable Stable

Membrane mimetic Zwitterionic micelles FOS-12 (DPC) FOS-14 Weakly cationic micelles LDAO

Mild non-ionic detergents, short-chain lipids, bicelles and lipid-protein nanodiscs DDM 5.0–7.0 n.o.f DHPC 5.0–7.0 n.o.f DC7PC 5.0–7.0 n.o.f DC7PC/FOS-14 (1:3–1:2) 5.0 −c DC7PC/LPPG (2:1) 7.0 n.o.f DMPC/DHPC (1:4) 5.0–7.0 n.o.f LMPG/DMPC/DHPC (~1:2:10) 6.0 – –c LMPG/DMPC/CHAPS (~ 1:2:5) 6.0 n.o.f LPNs with DMPC, DMPG, POPC or DOPG lipids 7.0 n.o.f

83 (4)

b10

No solubilization No solubilization No solubilization Unstable, ~60% aggregation in 4 h Unstable, precipitation in 20 m No solubilization Unstable, ~50% aggregation in 1 h Unstable, precipitation in 20 m Low yield of VSD-II incorporation

a Stability of the samples was estimated using 1D 15N-TROSY spectra measured at specified time. Degree of the aggregation was estimated from the intensities of the spectra, assuming that aggregate signals are not observable by NMR. b Some fraction of the signals were broadened. c Increased broadening as compared with FOS-12 at pH 5.0. d Increased broadening as compared with LDAO at pH 5.0. e Number of observed signals exceeds the expected value. f n.o. – not observed. NMR spectra were not measured due to low solubilization efficiency or low stability of the sample.

(Fig. 1C). The first variant VSD-IINC contained shortened N-terminal sequence, that led to deletion of one positively charged Lys residue (pI ~6.7, charge −1 at neutral pH). The NMR spectra of VSD-IINC in FOS12 and LDAO micelles (Fig. S4) have overall appearance similar to one of the VSD-II. Moreover, the stability of this domain variant in LDAO micelles was not significantly improved as compared with the initial VSD variant (Fig. S4). The second variant VSD-IIK contained six additional Lys residues, three at the N-terminus and three at the C-terminus (pI ~ 9.7, charge +6). VSD-IIK variant demonstrated excellent stability in LDAO micelles; very small drop in intensity of NMR signals was observed after 17 days of incubation at 45 °C (Fig. S5). The partial backbone assignment of VSDIIK including complete S4 helix, and N- and C-terminal parts of the domain was obtained in this environment (Figs. S6 and S7). Surprisingly, the 15N-relaxation data revealed significant intramolecular mobility of the domain. The measured 15N-{1H}-NOE values were b 0.6 (Fig. S7), with the values for S4 helix of only 0.36 ± 0.08 (mean ± s.d.). Moreover, the ‘NOE-on’ 2D plan from the 15N-{1H}-NOE spectrum of VSD-IIK revealed large number of negative cross-peaks belonging to other, unassigned, parts of the molecule (Fig. 3H). These observations indicate that the overall dynamics in the large parts of the VSD-IIK molecule, including S4 helix, is dominated by ‘fast’ motions at sub-nanosecond time scale. Please note, that 15N-{1H}-NOE values ≤0.4 (and of course negative values) are usually observed only in the unstructured regions (extended loops, prolonged N- and C-terminal tails) of ‘well-folded’ proteins. Most probably the introduction of N- and C-terminal Lys residues destabilized the spatial structure of the domain leading to dissociation of S4 helix out of the VSD TM helical bundle and to exclusion of this rather polar fragment from the hydrophobic micelle interior on its surface or into the surrounding solvent. Interestingly, the relatively fast aggregation of the VSD-IIK variant was observed in the FOS-12 environment (Fig. S5). Using VSD-IIK variant we were able to incorporate the domain into LPN particles containing DMPG lipid (Fig. 2, lane 12), but the quality of obtained 15N-TROSY spectrum was very poor (Fig. 3G). This spectrum

was characterized by the relatively low dispersion of chemical shifts and small number of the individually resolved resonances. The NMR spectrum of VSD-IIK in LPN did not superimpose well with spectra of VSDIIK in the LDAO micelles and with spectra of VSD-II in various membrane mimetics. We should mention, that 2H-labeled (deuterated) samples of membrane proteins are usually required for NMR studies in nanodiscs environment. However, the previous study of isolated VSDKvAP in the complex with LPNs [31] revealed much better NMR spectra, even in the cases where undeuterated protein was used. Relying on the previous results, we could propose that VSD-IIK incorporated into the LPN particles is incorrectly folded or aggregated. 3.4. Secondary structure of VSD-II in LPPG micelles The environment of LPPG micelles at pH 6.0 was chosen for further structural study of VSD-II. The analysis of VSD-II NMR spectra measured at these conditions revealed the presence of two sets of the backbone signals for the N-terminal residues (Met1-Leu14), which correspond to the two structural forms of the domain with the relative population of 3:2. The observed heterogeneity was connected with incomplete processing of N-terminal formyl group, which is typical for proteins synthesized in CF systems based on bacterial extracts. The 1H–15N cross-peak of Met1 residue observed for the major form of the protein (Fig. 4) confirmed the presence of unprocessed N-terminal formyl. This modification induced the formation of short helix at residues Lys3-Trp7, but did not influence the overall conformation of the protein molecule (Fig. 5). The almost complete backbone and side-chain resonance assignments (≥90% of 1HN, 15NH, 13Cα, 1Hα and 13C′, and ~85% of 13Cβ of residues belonging to the VSD sequence) were obtained for the 13C,15Nlabeled VSD-II (Figs. 4, 5). The 15N-labeled Gln was used for the CF synthesis; therefore 13C chemical shift data were unavailable for the Gln88Gln89 dipeptide segment (assignment of these residues was achieved using 15N-NOESY-HSQC spectrum). At the same time the weak 13Cα and 13C′ resonances were identified for Gln112 in HNCO and HNCA

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

499

Fig. 4. 2D 15N–HSQC spectrum of VSD-II in LPPG micelles. The spectrum was measured at 800 MHz for a 0.6 mM sample of 13C,15N-labeled VSD in 5% (100 mM) LPPG, pH 6.0, 45 °C. The obtained resonance assignments are shown. The resonances of side-chain groups are marked with superscript “s”. The resonances of Asn and Gln NH2 groups are connected by dotted lines. The encircled cross-peaks are under the drawing threshold.

spectra. The intensities of these signals were equal to the 2–3% of the corresponding intensities for neighboring 13C-labeled residues. The observed enrichment over the natural 13C abundance (~1%) could be explained by a transaminase activity of the CF extract, which convert 13 C-labeled Glu into Gln. Secondary structure of VSD-II calculated from the backbone and side-chain chemical shifts revealed four major helical segments S1–S4 (Fig. 5). The low values of α-helical probability observed in the middle

of S1 and S2 helices are probably the artifacts of calculation due to the incomplete resonance assignment in these regions. Although the location of S1–S4 helices approximately coincided with the predictions from the sequence analysis (e.g. listed in the UNIPROT database), the several discrepancies and additional secondary structure elements were observed by NMR (Fig. 5). (#1) Additional helix S0 (Pro12Val24) was observed at the N-terminus of the domain. (#2) Helix S2 started from the short ‘pseudohelical’ segment S2a (Met51-Asp56)

Fig. 5. NMR data that define the secondary structure of VSD-II in LPPG micelles. The unassigned residues are underlined and the cloning artifact residues are shown in gray on the protein sequence. The secondary structures as annotated in the UNIPROT database and determined from the present NMR data are shown on the top. The α-helices are shown by cylinders, and the elements with distorted α-helical conformation (probably 310-helices) are hatched. Vertical dashed lines highlight boundaries of the secondary structure elements. The conserved Arg/Lys residues that are responsible for voltage gating are marked. Data obtained for the major and minor (N-formylated and N-deformylated) variants of VSD-II are shown by bars and open diamonds, respectively. (α-Helix) – Probability of α-helical conformation calculated in the TALOS+ program [35]. Cut-off value (0.8) used for identification of helical segments is shown by dashed line. Please note that TALOS+ could not correctly calculate the α-helical probability for unassigned residues. (Δδ) – Secondary chemical shifts, the deviation of the 13 α 13 C , C′, and 13Cβ chemical shifts from random coil values (ppm). Down-field values (positive deviation) of the 13Cα and 13C′ resonances and up-field value of the 13Cβ resonance indicate a backbone helical conformation. (Δδ 1HN/ΔT) – Temperature coefficients of chemical shifts of amide protons (ppb/K) measured in range 25–45 °C using 3D HNCO spectra. The protons with the absolute value of coefficient smaller than 4.5 ppb/K (black bars, dashed line) are probably participate in the hydrogen bond formation [43]. (IMn2+/I0) – The relative changes of 15N-HSQC cross-peaks intensity induced by addition of Mn2+ to the VSD-II/LPPG sample. The shielding from the attenuation by paramagnetic cations (IMn2+/ I0 N 0.2, dashed line) indicates that the corresponding residue is probably situated in the hydrophobic region of the micelle. The unobservable and overlapped signals (where data could not be obtained) are shown by negative bars. Error bars in the figure correspond to expected standard deviation of experimental/calculated data points estimated from the measured noise level in the NMR spectra and error propagation analysis.

500

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

which is characterized by relatively low values of α-helical probability (~0.5). (#3) Short α-helix S23 (Pro83-Phe87) was observed between the TM segments S2 and S3. (#4) The α-helical structure of the S4 segment was distorted between the first and second conserved Arg residues (Arg118–Arg121). (#5) The domain structure ended by the S45 helix (Pro133-Gly143), which is separated from the S4 segment by three residue turn (Lys130-Trp132). The small absolute values of temperature coefficients of chemical shifts of amide protons (|Δδ1HN/ΔT| b 4.5 ppb/K, Fig. 5) indicated that the majority of HN groups within the S1, S2, S4 and S45 helical segments participate in the formation of intramolecular hydrogen bonds. The obtained data also revealed the formation of stable hydrogen bonds within the pseudohelical segment S2a and within the Arg118–Arg121 region of S4 segment, where destabilization of α-helical conformation was observed. Probably, these parts of the domain adopt the conformation of 310-helix, which could not be adequately predicted from the chemical shifts. The large temperature coefficient (~ −7 ppb/K) of Leu123 residue revealed the free HN group at the site of junction between 310and α-helix in the S4 segment. The large Δδ1HN/ΔT values observed in the middle of S0, in the first half of S2, and at N- and C-termini of S3 could indicate the relatively high structural plasticity of the corresponding helices.

To probe the topology of VSD-II in the LPPG micelles we used paramagnetic Mn2+ ions. These cations have affinity for negatively charged phosphate groups of the detergent molecules (as well as for carboxyl groups of Asp and Glu residues) [44]; therefore we expected to observe the strong attenuation of NMR signals in the VSD-II regions which are not shielded in the hydrophobic micelle interior. The shielding from the paramagnetic relaxation enhancement was observed in all helical regions of the domain except the N-terminal parts of S1 and S2 (S2a) helices, and entire S23–S3 segments (Fig. 5). Probably these unshielded parts of the domain together with the interhelical loops have contact with the solvent. 3.5. Backbone dynamics of VSD-II in the LPPG micelles The observation of a distribution in intensities of cross-peaks in the 3D HNCO spectrum (Fig. 6) indicated the presence of significant intramolecular mobility within the VSD-II molecule. Strong or weak (relative to average) intensities of HNCO peaks are most probably the consequences of motions at ps–ns or μs–ms time scales, respectively. According to these data, the extensive ps–ns motions were presented in N- and C-terminal regions of the VSD-II, S2a helix, and S3–S4 loop including Nterminal part of the S4 helix. The most pronounced patches of μs–ms

Fig. 6. NMR data reveal dynamical properties of VSD-II in LPPG micelles. The full set of the measured 15N-relaxation data (800 MHz, 45 °C) together with the results of ‘model-free’ analysis are shown in the Fig. S6. The protein sequence and secondary structure are shown using same codes as in the Fig. 5. Data obtained for the major and minor (N-formylated and Ndeformylated) forms of VSD-II are shown by closed diamonds and open squares, respectively. (Ln[IHNCO]) – Intensity (log values) of peaks in 3D HNCO spectrum. The level corresponding to the average intensity is shown by a dashed line. The intensities larger than twice average are shown by light gray bars. The intensities smaller than half average are shown by dark gray bars. The asterisks and negative bars denote unobservable (Pro residues and residues after 15N-Gln) and unassigned HNCO peaks, respectively. The data for Nterminal residues are the sums of the values for N-formylated and N-deformylated VSD-II. (τR) – The effective rotational correlation time was calculated from the R2/R1 ratio. The overall τR value derived by the ‘model-free’ analysis (15.4 ns) is shown by broken line. (R1 × R2) The residues displaying R1 × R2 product N18 s−2 could be subjected to exchange fluctuations in the μs–ms time scale [45]. Residues demonstrating nonzero contributions of μs–ms exchange (REX) to R2 rates are marked by stars. (S2) – The generalized order parameters calculated by ‘model-free’ analysis of 15N relaxation data (symbols) and random coil index order parameters (RCI-S2, lines) [34]. RCI-S2 values for N-deformylated VSD-II are shown by gray line. The residues displaying 15N-{1H}-NOEb 0.6 or S2 b 0.75 are subjected to high-amplitude motions on the ps–ns time scale. Error bars in the figure correspond to expected standard deviation of experimental/calculated data points estimated from the measured noise level in the NMR spectra and error propagation analysis.

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

exchange fluctuations were located in the S23 helix and in the middle of the S1 and S2 helices (Fig. 6). The weak HNCO intensities were also observed for the Val99 residue in the middle of the S3 helix and for Leu135 in the N-terminal part of S45. Obtained data explain the inability to assign the signals of the residues in the middle of the S1 and S2 helices; the corresponding resonances are probably broadened beyond the detection limit due to the μs–ms time scale motions. A detailed investigation of the VSD-II backbone dynamics was done using 15N-relaxation measurements and ‘model-free’ calculations (Figs. 6 and S8). The presence of μs–ms fluctuations in the S3, S4 and S45 helices was confirmed by observation of the large values of R1 × R2 products, which exceeded the 18 s− 2 threshold level (at 800 MHz NMR spectrometer [45], Fig. 6), and by comparison of the R2 and ηXY rates (Fig. S9). In addition, the μs–ms exchange contributions (REX) to the transverse relaxation (R2) rates were observed for the Val24 (S0), Phe43 (S1), Asp95 and Val99 (S3), Thr134 and Lys140 (S45) residues (Fig. 6, marked by stars and Fig. S8). Please note, that the relaxation data for the large fragments of the S1 and S2 helices were not available due to the signal broadening, and observed REX contributions could be significantly underestimated due to the presence of extensive dynamics in the ps–ns time scale (see below). The relatively low values of steady-state 15N-{1H}-NOE (from 0.6 to 0.8, Fig. 6) and the results of the ‘model-free’ calculations revealed the presence of nanosecond time scale motions in all regions of the VSD-II backbone (characteristic time of 0.9 ± 0.3 ns [mean ± s.d.], Fig. S8). Interestingly, the values of generalized order parameters (S2 N 0.75, Fig. 6) indicated that these motions in the helical regions of VSD-II have relatively small ‘amplitude’. At the same time, the N- and C-terminal regions of the domain, interhelical loops, S2a helix, N-terminal half of S4 helix, and turn connecting S4 and S45 helices demonstrate high-amplitude mobility at ps–ns time scale (S2 b 0.75). The good correspondence was observed between S2 values derived from 15N-relaxation and the random coil index order parameters (RCI-S2 [34]) calculated from the chemical shifts (Fig. 6). 15 N-relaxation data also provide information about rotational diffusion of macromolecules in solution. The overall rotational correlation

501

time of VSD-II (τR = 15.4 ns at 45 °C), determined by ‘model free’ analysis, was consistent with reorientation of the large protein/detergent complex having hydrodynamic Stokes radius (RH) ~30.0 Å and apparent MW about 42 kDa. The estimated radius fits the sum of the expected cross-sectional radius of the four-helical VSD-II bundle (~ 10 Å) and the length of the detergent molecule (~ 20 Å). The comparison of the overall τR value with the effective τR values calculated for each residue from the R2/R1 ratios (Fig. 6) and from ηXY rates (Fig. S8) revealed anticipated difference. The effective τR values were lower because they contain contributions from the motions at nanosecond time scale. The NMR data measured at different temperatures (in range 25– 45 °C, Figs. S10 and S11) and at different VSD-II concentrations (from 0.75 mM to 0.06 mM, Fig. S12) were compared. Results indicated that the observed patterns of ps–ns and μs–ms timescale motions represent inherent property of VSD-II solubilized in LPPG and are not the consequence of the relatively high temperature and protein concentration used for NMR study (45 °C and 0.6 mM, respectively). 3.6. Computer modeling of VSD-II in the lipid bilayer Comparison of VSD-II secondary structure determined in the LPPG environment with the structures of known Kv and Nav channels revealed close correspondence with the structure of NavAb chimera with VSD-IV of the human Nav1.7 channel [12] (see Fig. 7 and Section 4.3 below). Therefore the VSD of this chimeric channel was used as a template for modeling of VSD-II structure. The obtained homology model (Fig. 8A) nicely corresponds to the secondary structure of VSDII in LPPG environment. To estimate stability of the proposed VSD-II model, it was placed inside the lipid bilayer and 270 ns full-atom MD trajectory was calculated (Fig. 8C). Obtained data revealed that the structural model is conformationally stable; the trajectory became equilibrated during first 70 ns of simulation with the RMSD value from the starting structure of 0.38 ± 0.02 nm (mean ± s.d., Fig. S13). Comparison of the starting VSD-II model with the structures obtained by clustering of the last 200 ns of MD trajectory revealed well preserved positions of TM helices (Fig. 8AB). In accordance with experimentally

Fig. 7. Sequence alignment and secondary structure of second VSD of human Nav1.4 channel and VSDs of homotetrameric Kv and Nav channels which structures have been experimentally determined. The topology of VSD-II in the lipid membrane which agrees with the NMR data is shown in the insert. The sequences of VSDs were manually aligned using conserved aromatic/ hydrophobic (green) and charged (red and blue) residues. The α- and 310-helices are highlighted by gray and green background. The distorted α-helices (probably 310-helices) observed in VSD-II are shown by yellow background. The sites of conserved charged/polar residues in the S4 helix are numbered. Protein Data Bank entries and references for the secondary structures presented in this figure are as follows: NavAB (PDB ID: 3RVY, [8]), chimera of NavAb with VSD-IV of human Nav1.7 (NavAB/1.7-IV, PDB ID: 5EK0 [12]), Kv1.2/2.1 chimera (PDB ID: 2R9R, [7]), KvAP (PDB ID: 2KYH [22,23]), KCNQ1 [25]. The residues of KCNQ1 which were not included in the sequence of isolated VSD used for NMR study are in gray.

502

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

observed distribution of ps–ns timescale motions (Fig. 6), RMSF calculations (Fig. 8D) pointed to the increased mobility of the interhelical loops and N- and C-terminal parts of the domain. However some subtle structural features of the starting VSD-II model were not preserved during MD simulation. For example, the fragment of 310-helix in the middle of S4 segment adopted α-helical conformation and membrane associated S0, S23 and S45 helices became significantly disordered (Fig. 8AB, yellow and cyan). Several phenomena could be responsible for the observed differences ranging from nonideality of used force-field and imperfect membrane parameters (composition, charge, hydrophobic thickness, etc) to the absence of neighboring channel domains. Taking into account that the secondary structure of starting VSD-II model better corresponds to the measured NMR parameters (Fig. 8A), we used this model to visualize experimental data about dynamics and topology of VSD-II in LPPG environment (see Fig. 8EFG and Section 4.4 below). 4. Discussion A typical VSD, involving four TM helices, has molecular mass 15– 20 kDa. Nowadays the high-resolution NMR spectroscopy in the special membrane mimicking environment (detergent micelles, lipid-detergent bicelles, lipid-protein nanodiscs, etc.) is a method of choice for structural studies of membrane proteins of this size [46]. However, for successful NMR study of isolated VSD the several additional requirements should be met. (#1) The VSD sample should be sufficiently stable for prolonged measurements, which is frequently conducted at elevated temperature (at least 4–7 days at 40–50 °C). (#2) NMR spectra of the protein studied should be suitable for extraction of structural and dynamic information; at least the backbone resonance assignment should be possible. (#3) Most importantly, to provide valuable structural information, isolated VSD should at least partially preserve structural and dynamics properties of the native domain within the full-length channel complex. The previous NMR study of VSD-KvAP revealed the dependence of the stability, structure and dynamics of the domain from the properties of the used membrane mimetic [31]. Thus the rational selection of the membrane mimicking environment is crucial for success in NMR studies of isolated VSDs. 4.1. Aggregation of isolated VSD-II in media containing phospholipids The effective solubilization of VSD-II from RM precipitate after CF synthesis was achieved using a number of moderately harsh onechain zwitterionic (FOS-12, FOS-14), anionic (LMPG, LPPG), and partially cationic (LDAO) detergents. At the same time, the subsequent attempts to transfer the solubilized protein into media containing phospholipids or lipid-like two-chain detergents (DC7PC micelles, DMPC/DHPC or DMPC/CHAPS bicelles, and DMPC, POPC, DMPG or DOPG nanodiscs) were failed due to aggregation of VSD-II (see Fig. 2, lane 11). In addition relatively fast (half-time ~ 2 days) aggregation of VSD-II was observed in pure LDAO micelles (Fig. 3I). The aggregation of helical membrane proteins in a lipid bilayer and in various membrane mimetics is not uncommon problem. For example, extensive channel-to-channel lateral aggregation was previously observed for the full-length KvAP channel in asolectin vesicles [47]. Moreover, isolated VSD-KvAP demonstrated irreversible aggregation in LDAO micelles, although at much slower rate (~7 days) than VSD-II [31]. The fact that samples of isolated VSD-KvAP were stable in the environment of DC7PC micelles, DMPC/DHPC bicelles and nanodiscs containing various lipids [23,31] emphasized the high oligomerization tendency of isolated VSD-II. The transfer of a protein from harsh detergents into lipid-containing environment, like vesicles, bicelles or nanodiscs [48,49], is the most common approach for the in vitro folding of helical MPs. Therefore, the observed aggregation of VSD-II in lipid-containing media is possibly not the consequence of the incorrect folding of the domain after initial

detergent solubilization, but could represent the feature of the isolated VSD-II molecule. Most probably the other parts of the Nav1.4 channel are needed to protect VSD-II from aggregation within the lipid bilayer. The obtained results do not obligatory indicate that isolated VSD-II is unable to adopt its native 3D fold in the lipid membrane or membrane mimetic. The VSD-II aggregates could be formed by lateral aggregation of the natively folded domain molecules, or nucleated from the non-native structures temporary generated by dynamic excursions from the correct fold. 4.2. Increased positive charge protect VSD-II from the aggregation The results obtained for the engineered VSD-IIK variant of the domain, containing six additional Lys residues at N- and C-termini, indicated that the absence of electrostatic repulsion between the VSD-II molecules is probably the major factor which favors aggregation of the domain. Contrary to the isolated VSD-II (theoretical charge ±0 at neutral pH), the VSD-IIK variant (charge +6) demonstrated excellent stability in the LDAO environment (Fig. S5) and successfully got through the nanodisc incorporation procedure (Fig. 2 lane 12 and Fig. 3G). Thus, the increase in the overall charge could make an isolated VSD more amenable to the NMR analysis. The results of previous NMR studies support this idea. Indeed the almost complete backbone resonance assignments were obtained for the VSDs of KvAP and KCNQ1 channels, which are cationic (charge +5 and +10, respectively) [22,23,25], while the study of anionic hERG VSD (charge − 3) resulted only in a partial (~30%) assignment of backbone NMR resonances [50]. The analysis of human Nav channel sequences (Table S1) revealed that the isolated VSD-I, VSD-III and VSD-IV of all the nine channel isoforms (Nav1.1– 1.9) should be positively charged at neutral pH (charge + 2 to + 7) and only VSD-II of Nav1.1–1.7 channels are uncharged or have small negative charge (−1 to ±0). So, we could expect that the VSD-I, VSDIII and VSD-IV of human Nav channels will demonstrate higher stability in membrane mimicking media and represent more suitable targets for NMR studies as compared with VSD-II. However, some caution concerning the above idea should be expressed. The excessive charge density could not only protect an MP from aggregation, but also could disrupt it tertiary structure by intramolecular repulsion. The drastic increase of ps–ns dynamics observed for the S4 helix of VSD-IIK (Figs. 3H and S7) probably is a consequence of the VSD unfolding induced by electrostatic repulsion between N- and C-termini of the domain (between the S1 and S4 helices). The other way to influence the electric charge of MP/micelle complex is the increase of the charge of detergent head-groups. Indeed, the concentration dependence of 15N-TROSY spectra (Fig. S12) and 15 N relaxation data measured in the anionic lysolipid LPPG (Fig. 6) was consistent with monomeric VSD-II in complex with the micelle (τR ~ 15 ns at 45 °C, RH ~ 30 Å). The inability of VSD-II to integrate into nanodiscs with negatively charged lipids (DMPG or DOPG) indicated that anionic bilayer cannot protect the domain from aggregation. This provides additional argumentation in favor of the lateral aggregation of the VSD-II molecules in the media containing phospholipids (see above). In this case the anionic micelles having quasi-spherical shape effectively prevent contacts between TM regions of the different domain molecules. 4.3. Secondary structure of micelle solubilized VSD-II The results of detergent screening revealed that only anionic lysolipids are compatible with the structural NMR study of VSD-II. There is no straightforward way to test the folding or ‘functionality’ of the isolated domain in the membrane mimicking media. Therefore to analyze the folding of VSD-II we compared structural and dynamical data obtained in the LPPG micelles with the results of previous structural studies of homotetrameric Nav and Kv channels and their isolated

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

503

Fig. 8. (A) VSD-II structure (Ala11-Ser145 residues) modeled using template of NavAb/hNav1.7-IV chimeric channel (PDB ID: 5EK0 [12]). Secondary structure of the domain is color coded according to the NMR data obtained in LPPG environment. The region of distorted α-helix in the S4 helix (Arg118–Arg121) probably adopting 310-helical conformation is marked as S4/310. (B) Six structures of VSD-II obtained after clustering of the ‘stable’ part (last 200 ns) of MD trajectory using distance cut-off of 0.2 nm. Backbone of the domain is colored as in panel A. The side chains of positively and negatively charged residues are shown by blue and red, respectively. The conserved polar and aromatic/hydrophobic residues are in magenta and green. (C) The view of VSD-II in the POPC/POPE/cholesterol (2:1:1) bilayer after MD calculation. Hydrophobic atoms of lipids and cholesterol are shown in yellow and orange, respectively. Oxygen, phosphorus, and nitrogen atoms of phospholipids are shown by red, orange, and blue, respectively. The proximal parts of the membrane, water, and ions are removed for clarity. The five conserved positively charged residues from the S4 helix (Arg118, Arg121, Arg124, Lys127, Lys 130) and three conserved aromatic/hydrophobic residues from the central parts of S1, S2 and S3 helices (Ile37, Phe70, Val99) are shown in cyan and green, respectively. (D) Root-mean-square fluctuation (RMSF) of VSD-II calculated over ‘stable’ part (last 200 ns) of MD trajectory. Secondary structure of the domain determined by NMR is shown on a separate line. TM helices are shown in black and distorted/membrane associated helices in gray. (E,F) Residues of VSD-II affected by dynamic processes on the ps–ns and μs–ms timescales are shown on the homology model of the domain (see legend and Fig. 6). (G) Residues of VSD-II demonstrating attenuation of 15N-HSQC cross-peaks due to the contact with paramagnetic Mn2+ ions (IMn2+/I0 b 0.2) and residues which are protected from the contact with Mn2+ are shown on the homology model of the domain (see legend and Fig. 5). In addition to positively and negatively charged side chains shown in blue and red, respectively, the His residues are drawn in magenta. The putative mechanism of Mn2+ access to the central part of S3 helix is shown by gray arrows. The negatively charged residues located in the N- and C-terminal parts of S1, S2, and S3 helices are underlined. The S0 helical segment is omitted in the panels E, F and right G for clarity.

VSDs. (In the present study we did not directly assay the tertiary structure of the domain.) The secondary structure of micelle solubilized VSD-II (Fig. 7) closely corresponds to the structures of other VSDs, and especially to the structures of NavAb channel [8] and NavAb/hNav1.7-IV chimera [12]. The obtained homology-based model of VSD-II (Fig. 8A) and results of its MD simulation in lipid bilayer (Fig. 8BC) illustrate this good correspondence with the NMR data, which covers not only the four TM helices (S1–S4), but also ‘additional’ peripheral helical elements associated with the membrane surface. The starting S0 helix, associated with the cytoplasmic interface of the membrane, was found in all the VGICs studied to date [7–9,11,12] and other proteins which contain VSDs, namely voltage-gated proton channel Hv1 [20] and voltage-sensitive phosphatase Ci-VSP [21]. The short disordered helical element S2a observed at the N-terminus of the VSD-II S2 helix nicely corresponds to the isolated turn of the α-helix presented in the NavAb and NavAb/hNav1.7-IV structures [8,12] (Figs. 7 and 8A). Moreover the short S23 helix, located

in the cytoplasmic loop of VSD-II, was also presented in VSDs of KvAP and NavAb channels [6,8,22,23]. One of the most surprising observations of the present study is the destabilization of the α-helical conformation in the middle of S4, between the first and second conserved Arg residues. The position of this distorted region coincides with the position of the 310-helical insert in the S4 α-helix of the NavAb and NavAb/ hNav1.7-IV chimera (Figs. 7 and 8A) [8,12]. The VSD-II structure ends with the membrane associated amphipathic S45 helix which is supposed to be a part of the linker to the pore domain (Fig. 8). The corresponding helical element was also found in majority of VSDs studied to date (Fig. 7). The some details of VSD-II structure differentiate it from the VSDs of Kv channels. The S1-S2 and S3-S4 loops of VSD-II are relatively short and do not contain additional helices (S1b, S12, and S34) which were observed in Kv1.2/2.1 chimera [7] (Fig. 7). The S3 helix in VSD-II is continuous and is not divided into two parts (S3a and S3b) as observed in KvAP and Kv1.2/2.1 channels [6,7,22,23]. Moreover the region of 310-

504

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

helical conformation in the S4 helix of Kv1.2/2.1 VSD is located after the third conserved Arg residue and goes up to the N-terminus of S45 [7] (Fig. 7). Interestingly, the secondary structure of VSD-II shows significant differences from isolated VSD of the KCNQ1 channel recently studied by NMR. In this cardiac channel the S1, S2 and S4 helices of VSD have unusual length and their termini significantly protrude into aqueous environment [25]. 4.4. Topology and dynamics of micelle solubilized VSD-II It is generally assumed that VSD experiences significant conformational rearrangement during voltage-activation coupled with translocation of the gating charges from intracellular to extracellular side of the membrane. The detailed mechanism of this conformational transition is still under debate [51–53], but the NMR data obtained for VSD-KvAP indicated that these large-scale motions could be echoed in the activated domain state (observed at zero TM potential) [22]. This VSD state is characterized by the relatively weak helix-helix interactions, resulting in the unusually high ‘fluidity’ of the protein structure, and characteristic μs–ms time scale motions in the regions sharing inter-helical contacts [22]. The four-helical bundle of activated VSD has an overall hourglass shape; and the major interhelical contacts became localized in the central parts of the TM helices [6–8]. The conserved hydrophobic residues forming these contacts Ile37(S1), Phe70(S2) and Val99(S3) (green in Figs. 7 and 8, the numbers are given for VSD-II) separate extracellular and intracellular water-filled crevices, which protrude from both sides of the lipid membrane to the central part of the domain (central part of the membrane) [47,54,55]. The observed pattern of μs–ms mobility (Fig. 8F) and the data obtained using paramagnetic Mn2+ cations (Fig. 8G) are in general agreement with the similar topology of VSD-II in LPPG micelles. Indeed, the μs–ms motions were detected for conserved residues in the central parts of the S1 (Ile33–Ile37 residues), S2 (residues around Phe70), S3 (Asp95, Val99 residues), and S4 (Val125 residue) helices, and in the N-terminal part of the S45 helix (Thr134, Leu135 residues). The similar μs–ms motions at the Val residue in the middle of S3 (homologues to Val99 in VSD-II) were previously observed not only in VSD-KvAP, but also in VSD of the KCNQ1 channel [25]. The significant broadening (in some cases beyond the detection limit) of resonances belonging to S1 and S2 helices of VSD-II induced by μs–ms fluctuations (Fig. 8F) indicates that S1 helix does not stably pack against S2. At the same time this observation indirectly confirms that S1 and S2 helices of VSD-II could contact each other within the LPPG micelle, and at least temporary form two-helical bundle. The protection from the contact with Mn2+ revealed that significant parts of the S1, S2 and S4 helices of VSD-II are shielded from the aqueous environment and, therefore, located in the hydrophobic region of the micelle (Fig. 8G). To explain Mn2 + protection data obtained for the other parts of the domain, we should additionally consider the presence of electrostatic interactions between Mn2+ cations and charged protein groups, and the fact that the water-filled crevices could provide access of the ions into VSD interior. Thus, large density of the positive charges (Lys15 and Lys17 residues) probably enhances Mn2+ protection of the membrane associated helix S0 (Fig. 8G). Similarly, two Lys residues (130 and 140) are responsible for the enhanced protection of S45 segment as compared to the neighboring S23 helix. Notably, the fully solvent exposed Lys3 residue does not provide the similar level of Mn2+ protection to the neighboring backbone fragment (see Fig. 5). On the other hand, the negatively charged Asp95 and Glu105 side-chains located in the N- and C-terminal parts of S3 helix could mediate access of Mn2 + ions to the hydrophobic residues in the central part of the S3 helix (Fig. 8G, gray arrows). Interestingly, the similar pairs of negatively charged residues (Asp30/Glu47 and Asp56/Glu73) found in the S1 and S2 segments, respectively, do not support Mn2+ access to the central parts of the corresponding helices (Fig. 8G). In this respect, it should be mentioned that the enhanced solvent accessibility of the S23-S3a

helices, but not S1 and S2, was previously observed in the full-length KvAP channel and its isolated VSD by cysteine scanning mutagenesis and NMR spectroscopy, respectively [22,56]. The obtained data supports the formation of water-filled crevices in the spatial structure of micelle embedded VSD-II. Probably, negatively charged residues from the S3 helix protrude into these crevices (Fig. 8G). The one of the most surprising findings of the present study is the nanosecond time scale motions observed in all the regions of VSD-II (Fig. S8). The values of generalized order parameters revealed that these motions have relatively high amplitude only in two distinct regions of the VSD-II molecule (S2 b 0.75, Fig. 8E). One of these regions formed by extracellular loops of the domain (S1–S2 and S3–S4), while second includes three spatially neighboring fragments contacting cytoplasmic side of the membrane (N-terminal parts of S1 and S3 helices and C-terminal part of S4). Contrarily, the nanosecond timescale motions in the central parts of the TM (S1–S4) helices and membrane associated S0, S23 and S45 helical elements have relatively small amplitude (S2 N 0.75, Fig. 8E) and stability of the corresponding protein fragments is comparable to well-folded regions of MPs. Therefore the observed nanosecond timescale motions probably do not induce unfolding/dissociation of the VSD helical bundle. The significant nanosecond time scale mobility previously was not observed for VSD-KvAP solubilized in zwitterionic (FOS-12/LDAO or DC7PC) micelles [22,23]. Therefore, the enhanced mobility on nanosecond time scale could either be the consequence of weakening of interhelical contacts in the environment of anionic (LPPG) micelles, or reflect the structural feature of isolated VSD of Nav channel. We think that both reasons are operative. On the one hand, previously we observed enhancement of ps–ns mobility in VSD-KvAP upon the protein transfer from FOS-12/LDAO to LPPG micelles (Z.O.S. unpublished observation). On the other hand, the recent NMR study of KCNQ1 VSD in the LPPG environment did not revealed enhanced ps–ns time scale mobility [25]; the reported 15N-{1H}- NOE values were larger than presently observed for VSD-II. This indicates that VSD-II of Nav1.4 channel indeed is characterized by enhanced ps–ns mobility in comparison with VSDs of Kv channels. We could speculate that the larger ‘fluidity’ and inherently smaller stability of the spatial structure represent the features of Nav VSDs discriminating them from VSDs of the Kv channels. The conformational plasticity of Nav VSDs could be, at least in part, responsible for the faster activation of Nav channels (as compared with Kv) upon depolarization of excitable membranes [57]. 5. Conclusions According to our knowledge the present investigation for the first time assayed the validity of structural studies of isolated VSDs from the eukaryotic Nav channels and provided experimental insight into the human sodium channel structure. Using micelle solubilized variant of isolated VSD-II of skeletal muscle Nav1.4 channel we were able to characterize its secondary structure and obtain the detailed information about the domain dynamics. The several lines of evidence support the presence of the quasi-correct VSD-II fold in the micellar environment: (#1) conserved secondary structure and topology of protein/membrane interactions; (#2) characteristic pattern of μs–ms motions in the TM helices of VSD-II, which are probably connected with the fluctuations in the interhelical packing; (#3) the marked destabilization of the VSD-II structure observed upon introduction of excessive positive charges at N- and C-termini of the domain. We should note, however, that the tertiary structure of VSD-II was not directly detected in the present NMR study and, thus, the obtained conclusions should be considered with some level of caution. The present investigation revealed the major pitfalls which could arise upon application of the ‘divide and conquer’ approach to the studies of the eukaryotic Na+ channels. The covalent linkage of the four pseudo-subunits in one joint polypeptide chain of the Nav channel probably provides additional stabilization to the structure of voltage-

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

sensing domains as compared to the Kv channels. Therefore, the isolation of VSD from the sodium channel complex could lead to significantly higher aggregation tendency and increased intramolecular mobility in a membrane mimicking environment. These properties require more careful screening of membrane mimetics and critical assessment of the obtained structural and dynamic data. The application of proposed ‘divide and conquer’ approach could significantly facilitate structure-function investigations of Nav channels and stimulate further pharmacological developments. The samples of isolated VSDs in the membrane mimicking environment could be used for the structural studies of toxin-channel interactions [26] and for screening of various isoform-selective small molecule ligands, the prototypes of drugs for treatment of hyperexcitability disorders. Transparency document The Transparency document associated with this article can be found, in the online version.

Acknowledgements Access to the computational facilities of the Supercomputer Center of St. Petersburg Polytechnic University is greatly appreciated. The work was supported by the Russian Ministry of Science and Education (Project ID RFMEFI61615X0044), Russian Academy of Sciences (Program “Molecular and Cellular Biology”), and Russian Foundation for Basic Research (project № 12-04-01712). Computer modeling and the manuscript preparation was performed with the financial support of the Russian Science Foundation (project № 16-14-10338). Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.bbamem.2017.01.004.

References [1] B. Hille, Ion Channels of Excitable Membranes, third ed. Sinauer Associates Inc., Sunderland, MA, 2001. [2] F.M. Ashcroft, Ion Channels and Disease, first ed. Academic Press, San Diego, CA, 1999. [3] Y. Chen-Izu, R.M. Shaw, G.S. Pitt, V. Yarov-Yarovoy, J.T. Sack, H. Abriel, R.W. Aldrich, L. Belardinelli, M.B. Cannell, W.A. Catterall, W.J. Chazin, N. Chiamvimonvat, I. Deschenes, E. Grandi, T.J. Hund, L.T. Izu, L.S. Maier, V.A. Maltsev, C. Marionneau, P.J. Mohler, S. Rajamani, R.L. Rasmusson, E.A. Sobie, C.E. Clancy, D.M. Bers, Na+ channel function, regulation, structure, trafficking and sequestration, J. Physiol. 593 (2015) 1347–1360. [4] S.I. Börjesson, F. Elinder, Structure, function, and modification of the voltage sensor in voltage-gated ion channel, Cell Biochem. Biophys. 52 (2008) 149–174. [5] K.J. Swartz, Sensing voltage across lipid membranes, Nature 456 (2008) 891–897. [6] Y. Jiang, A. Lee, J. Chen, V. Ruta, M. Cadene, B.T. Chait, R. MacKinnon, X-ray structure of a voltage-dependent K+ channel, Nature 423 (2003) 33–41. [7] S.B. Long, X. Tao, E.B. Campbell, R. MacKinnon, Atomic structure of a voltage-dependent K+ channel in a lipid membrane-like environment, Nature 450 (2007) 376–382. [8] J. Payandeh, T. Scheuer, N. Zheng, W.A. Catterall, The crystal structure of a voltagegated sodium channel, Nature 475 (2011) 353–358. [9] X. Zhang, W. Ren, P. DeCaen, C. Yan, X. Tao, L. Tang, J. Wang, K. Hasegawa, T. Kumasaka, J. He, J. Wang, D.E. Clapham, N. Yan, Crystal structure of an orthologue of the NaChBac voltage-gated sodium channel, Nature 486 (2012) 130–134. [10] C. Sato, Y. Ueno, K. Asai, K. Takahashi, M. Sato, A. Engel, Y. Fujiyoshi, The voltage-sensitive sodium channel is a bell-shaped molecule with several cavities, Nature 409 (2001) 1047–1051. [11] J. Wu, Z. Yan, Z. Li, X. Qian, S. Lu, M. Dong, Q. Zhou, N. Yan, Structure of the voltagegated calcium channel Cav1.1 at 3.6 Å resolution, Nature 537 (2016) 191–196. [12] S. Ahuja, S. Mukund, L. Deng, K. Khakh, E. Chang, H. Ho, S. Shriver, C. Young, S. Lin, J.P. Johnson, P. Wu, J. Li, M. Coons, C. Tam, B. Brillantes, H. Sampang, K. Mortara, K.K. Bowman, K.R. Clark, A. Estevez, Z. Xie, H. Verschoof, M. Grimwood, C. Dehnhardt, J.C. Andrez, T. Focken, D.P. Sutherlin, B.S. Safina, M.A. Starovasnik, D.F. Ortwine, Y. Franke, C.J. Cohen, D.H. Hackos, C.M. Koth, J. Payandeh, Structural basis of Nav1.7 inhibition by an isoform-selective small-molecule antagonist, Science 350 (2015) aac5464.

505

[13] P. Li, Z. Chen, H. Xu, H. Sun, H. Li, H. Liu, H. Yang, Z. Gao, H. Jiang, M. Li, The gating charge pathway of an epilepsy-associated potassium channel accommodates chemical ligands, Cell Res. 23 (2013) 1106–1118. [14] P. Kornilov, A. Peretz, B. Attali, Channel gating pore: a new therapeutic target, Cell Res. 23 (2013) 1067–1068. [15] F. Bosmans, M.-F. Martin-Eauclaire, K.J. Swartz, Deconstructing voltage sensor function and pharmacology in sodium channels, Nature 456 (2008) 202–208. [16] S.Y. Lee, A. Lee, J. Chen, R. MacKinnon, Structure of the KvAP voltage-dependent K+ channel and its dependence on the lipid membrane, Proc. Natl. Acad. Sci. 102 (2005) 15441–15446. [17] A.A. Alabi, M.I. Bahamonde, H.J. Jung, J.I. Kim, K.J. Swartz, Portability of paddle motif function and pharmacology in voltage sensors, Nature 450 (2007) 370–375. [18] Z. Lu, A.M. Klem, Y. Ramu, Ion conduction pore is conserved among potassium channels, Nature 413 (2001) 809–813. [19] C. Arrigoni, I. Schroeder, G. Romani, J.L. Van Etten, G. Thiel, A. Moroni, The voltagesensing domain of a phosphatase gates the pore of a potassium channel, J. Gen. Physiol. 141 (2013) 389–395. [20] K. Takeshita, S. Sakata, E. Yamashita, Y. Fujiwara, A. Kawanabe, T. Kurokawa, Y. Okochi, M. Matsuda, H. Narita, Y. Okamura, A. Nakagawa, X-ray crystal structure of voltage-gated proton channel, Nat. Struct. Mol. Biol. 21 (2014) 352–357. [21] Q. Li, S. Wanderling, M. Paduch, D. Medovoy, A. Singharoy, R. McGreevy, C.A. Villalba-Galea, R.E. Hulse, B. Roux, K. Schulten, A. Kossiakoff, E. Perozo, Structural mechanism of voltage-dependent gating in an isolated voltage-sensing domain, Nat. Struct. Mol. Biol. 21 (2014) 244–252. [22] Z.O. Shenkarev, A.S. Paramonov, E.N. Lyukmanova, L.N. Shingarova, S.A. Yakimov, M.A. Dubinnyi, V.V. Chupin, M.P. Kirpichnikov, M.J.J. Blommers, A.S. Arseniev, NMR structural and dynamical investigation of the isolated voltage-sensing domain of the potassium channel KvAP: implications for voltage gating, J. Am. Chem. Soc. 132 (2010) 5630–5637. [23] J.A. Butterwick, R. MacKinnon, Solution structure and phospholipid interactions of the isolated voltage-sensor domain from KvAP, J. Mol. Biol. 403 (2010) 591–606. [24] Q. Li, S. Wanderling, P. Sompornpisut, E. Perozo, Structural basis of lipid-driven conformational transitions in the KvAP voltage-sensing domain, Nat. Struct. Mol. Biol. 21 (2014) 160–166. [25] D. Peng, J.H. Kim, B.M. Kroncke, C.L. Law, Y. Xia, K.D. Droege, W.D. Van Horn, C.G. Vanoye, C.R. Sanders, Purification and structural study of the voltage-sensor domain of the human KCNQ1 potassium ion channel, Biochemistry 53 (2014) 2032–2042. [26] S. Ozawa, T. Kimura, T. Nozaki, H. Harada, I. Shimada, M. Osawa, Structural basis for the inhibition of voltage-dependent K+ channel by gating modifier toxin, Sci. Rep. 5 (2015) 14226. [27] K. Jurkat-Rott, B. Holzherr, M. Fauler, F. Lehmann-Horn, Sodium channelopathies of skeletal muscle result from gain or loss of function, Pflugers Arch. 460 (2010) 239–248. [28] J.B. Kim, M.H. Kim, S.J. Lee, D.J. Kim, B.S. Lee, The genotype and clinical phenotype of Korean patients with familial hypokalemic periodic paralysis, J. Korean Med. Sci. 22 (2007) 946–951. [29] N. Dupré, N. Chrestian, J.P. Bouchard, E. Rossignol, D. Brunet, D. Sternberg, B. Brais, J. Mathieu, J. Puymirat, Clinical, electrophysiologic, and genetic study of non-dystrophic myotonia in French-Canadians, Neuromuscul. Disord. 19 (2009) 330–334. [30] M. Gurevitz, Mapping of scorpion toxin receptor sites at voltage-gated sodium channels, Toxicon 60 (2012) 502–511. [31] Z.O. Shenkarev, E.N. Lyukmanova, A.S. Paramonov, L.N. Shingarova, V.V. Chupin, M.P. Kirpichnikov, M.J. Blommers, A.S. Arseniev, Lipid-protein nanodiscs as reference medium in detergent screening for high-resolution NMR studies of integral membrane proteins, J. Am. Chem. Soc. 132 (2010) 5628–5629. [32] E.N. Lyukmanova, Z.O. Shenkarev, N.F. Khabibullina, G.S. Kopeina, M.A. Shulepko, A.S. Paramonov, K.S. Mineev, R.V. Tikhonov, L.N. Shingarova, L.E. Petrovskaya, D.A. Dolgikh, A.S. Arseniev, M.P. Kirpichnikov, Lipid-protein nanodiscs for cell-free production of integral membrane proteins in a soluble and folded state: comparison with detergent micelles, bicelles and liposomes, Biochim. Biophys. Acta 1818 (2012) 349–358. [33] V.Y. Orekhov, I. Ibraghimov, M. Billeter, Optimizing resolution in multidimensional NMR by three-way decomposition, J. Biomol. NMR 27 (2003) 165–173. [34] M.V. Berjanskii, D.S. Wishart, A simple method to predict protein flexibility using secondary chemical shifts, J. Am. Chem. Soc. 127 (2005) 14970–14971. [35] Y. Shen, F. Delaglio, G. Cornilescu, A. Bax, TALOS+: a hybrid method for predicting protein backbone torsion angles from NMR chemical shifts, J. Biomol. NMR 44 (2009) 213–223. [36] J.H. Chill, J.M. Louis, J.L. Baber, A. Bax, Measurement of 15 N relaxation in the detergent-solubilized tetrameric KcsA potassium channel, J. Biomol. NMR 36 (2006) 123–136. [37] R. Cole, J.P. Loria, FAST-Modelfree: a program for rapid automated analysis of solution NMR spin-relaxation data, J. Biomol. NMR 26 (2003) 203–213. [38] M.A. Martí-Renom, A.C. Stuart, A. Fiser, R. Sánchez, F. Melo, A. Sali, Comparative protein structure modeling of genes and genomes, Annu. Rev. Biophys. Biomol. Struct. 29 (2000) 291–325. [39] M.J. Abraham, T. Murtola, R. Schulz, S. Páll, J.C. Smith, B. Hess, E. Lindahl, GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers, SoftwareX 1–2 (2015) 19–25. [40] E.N. Lyukmanova, Z.O. Shenkarev, N.F. Khabibullina, D.S. Kulbatskiy, M.A. Shulepko, L.E. Petrovskaya, A.S. Arseniev, D.A. Dolgikh, M.P. Kirpichnikov, N-terminal fusion tags for effective production of g-protein-coupled receptors in bacterial cell-free systems, Acta Nat. 4 (2012) 58–64. [41] Y. Imaishi, R. Kakehashi, T. Nezu, H. Maeda, Dodecyldimethylamine oxide micelles in solutions without added salt, J. Colloid Interface Sci. 197 (1998) 309–316.

506

A.S. Paramonov et al. / Biochimica et Biophysica Acta 1859 (2017) 493–506

[42] C. Fernández, C. Hilty, G. Wider, K. Wüthrich, Lipid-protein interactions in DHPC micelles containing the integral membrane protein OmpX investigated by NMR spectroscopy, Proc. Natl. Acad. Sci. U. S. A. 99 (2002) 13533–13537. [43] N.J. Baxter, M.P. Williamson, Temperature dependence of 1H chemical shifts in proteins, J. Biomol. NMR 9 (1997) 359–369. [44] Z.O. Shenkarev, S.V. Balandin, K.I. Trunov, A.S. Paramonov, S.V. Sukhanov, L.I. Barsukov, A.S. Arseniev, T.V. Ovchinnikova, Molecular mechanism of action of β-hairpin antimicrobial peptide arenicin: oligomeric structure in dodecylphosphocholine micelles and pore formation in planar lipid bilayers, Biochemistry 50 (2011) 6255–6265. [45] J.M. Kneller, M. Lu, C. Bracken, An effective method for the discrimination of motional anisotropy and chemical exchange, J. Am. Chem. Soc. 124 (2002) 1852–1853. [46] D. Nietlispach, A. Gautier, Solution NMR studies of polytopic α-helical membrane proteins, Curr. Opin. Struct. Biol. 21 (2011) 1–12. [47] L.G. Cuello, D.M. Cortes, E. Perozo, Molecular architecture of the KvAP voltage-dependent K+ channel in a lipid bilayer, Science 306 (2004) 491–495. [48] J.L. Popot, Folding membrane proteins in vitro: a table and some comments, Arch. Biochem. Biophys. 564 (2014) 314–326. [49] Z.O. Shenkarev, E.N. Lyukmanova, I.O. Butenko, L.E. Petrovskaya, A.S. Paramonov, M.A. Shulepko, O.V. Nekrasova, M.P. Kirpichnikov, A.S. Arseniev, Lipid-protein nanodiscs promote in vitro folding of transmembrane domains of multi-helical and multimeric membrane proteins, Biochim. Biophys. Acta 1828 (2013) 776–784.

[50] H.Q. Ng, Y.M. Kim, Q. Huang, S. Gayen, A.A. Yildiz, H.S. Yoon, E.K. Sinner, C. Kang, Purification and structural characterization of the voltage-sensor domain of the hERG potassium channel, Protein Expr. Purif. 86 (2012) 98–104. [51] M.Ø. Jensen, V. Jogini, D.W. Borhani, A.E. Leffler, R.O. Dror, D.E. Shaw, Mechanism of voltage gating in potassium channels, Science 336 (2012) 229–233. [52] U. Henrion, J. Renhorn, S.I. Börjesson, E.M. Nelson, C.S. Schwaiger, P. Bjelkmar, B. Wallner, E. Lindahl, F. Elinder, Tracking a complete voltage-sensor cycle with metal-ion bridges, Proc. Natl. Acad. Sci. U. S. A. 109 (2012) 8552–8557. [53] P. Gosselin-Badaroudine, L. Delemotte, A. Moreau, M.L. Klein, M. Chahine, Gating pore currents and the resting state of Nav1.4 voltage sensor domains, Proc. Natl. Acad. Sci. 109 (2012) 19250–19255. [54] X. Tao, A. Lee, W. Limapichat, D.A. Dougherty, R. MacKinnon, A gating charge transfer center in voltage sensors, Science 328 (2010) 67–73. [55] D. Krepkiy, M. Mihailescu, J.A. Freites, E.V. Schow, D.L. Worcester, K. Gawrisch, D.J. Tobias, S.H. White, K.J. Swartz, Structure and hydration of membranes embedded with voltage-sensing domains, Nature 462 (2009) 473–479. [56] E.J. Neale, H. Rong, C.J. Cockcroft, A.J. Sivaprasadarao, Mapping the membrane-aqueous border for the voltage-sensing domain of a potassium channel, J. Biol. Chem. 282 (2007) 37597–37604. [57] J.J. Lacroix, F.V. Campos, L. Frezza, F. Bezanilla, Molecular bases for the asynchronous activation of sodium and potassium channels required for nerve impulse generation, Neuron 79 (2013) 651–657.