Introduction

Voltage gated potassium channels (Kv) are a family of structurally similar pore forming membrane proteins which are involved in a fine control of many physiological processes including neuronal and muscle excitability, immune cell signaling, secretion of hormones and neurotransmitters (Hayes 2004; Park et al. 2010; Shah and Aizenman 2014). Abnormal functioning of Kv channels leads to physiological disorders and requires correction that can be often achieved with channel blockers. Consequently, Kv channels are considered as potential drug targets for the treatment of a variety of diseases such as cardiovascular, neurological, and autoimmune disorders (inherited or acquired) and cancer. For example, blocking of Kv1.1 channel provides an antitumor effect (Jang et al. 2011) and alleviates symptoms of multiple sclerosis and spinal cord injury (Hayes 2004). Kv1.3 channel blockers are considered as perspective drugs for a number of autoimmune diseases (Pérez-Verdaguer et al. 2016).

Less is known about another member of Kv1 channel family, Kv1.6, which is involved in repolarization of neuronal action potential (Glazebrook et al. 2002) in electrically-excitable cells such as neurons of the central and peripheral nervous systems (Matus-Leibovitch et al. 1996; Schmidt et al. 1999; Chung et al. 2001; Alessandri-Haber et al. 2002; Glazebrook et al. 2002), muscle cells (Davies et al. 2002) and cardiomyocytes (Brahmajothi et al. 1997). Elevated expression of Kv1.6 channels in piriform cortex interneurons largely contributes to the development of pathological neuronal conduction in the case of epilepsy (Gavrilovici et al. 2012). Inhibition of Kv1.6 that results in increased duration of the action potential (Glazebrook et al. 2002) is an attractive approach to control epileptic activity of neurons.

Peptide toxins are high affinity pore blockers of Kv channels acting at pico- and nano molar concentrations (Dutertre and Lewis 2010; Bergeron and Bingham 2012). They are widely used to probe Kv channel structure, function and tissue distribution. Potential pharmacological applications attach considerable importance to peptide pore blockers guiding a search for advanced blockers of the target channels. However, cross activity to allied Kv-channels, which is a general property of many peptide toxins, significantly complicates their applications. In this way, rational design and development of potent and highly specific pore blockers is an evident way to overcome insufficient level of their selectivity. For this, a robust method for identification of molecular determinants of toxin-channel interactions and a convenient experimental approach to study peptide-channel complexes should be integrated.

As for the method, molecular modeling can help to reveal and compare interaction interfaces of peptides with Kv1.x channels (Gordon et al. 2013; Novoseletsky et al. 2016). Currently, there is no alternative to this approach, because of the difficulties in crystallizing Kv-channels and peptide-channel complexes. Structures of KcsA, KvAP, Kv1.2, chimeric Kv1.2–2.1 channels as well as two complexes of charybdotoxin with Kv1.2–2.1 and mutated KcsA were only resolved with a high resolution (Novoseletsky et al. 2016). High structural homology of both Kv1-channels and their ligands, peptide blockers, allows one to combine a homology modeling technique with a molecular dynamics simulation to describe the features of other complexes of interest.

In turn, bioengineering systems based on KcsA-Kv1.x hybrid channels can be used to search for new blockers in animal venoms and to study blocker affinity and selectivity to a variety of related Kv1.x channels (Anh et al. 2013; Legros et al. 2000, 2002; Kudryashova et al. 2013; Hoang et al. 2014; Kuzmenkov et al. 2015). KcsA-Kv1.x (x = 1–6) channels were developed by transferring the blocker-binding site located in the S5-S6 linker region of Kv1.x into the corresponding M1-M2 linker region of the bacterial KscA channel (Legros et al. 2000, 2002). A pore-forming region (especially outer vestibule of a pore domain) in KcsA-Kv1.x channels resembles structurally and pharmacologically the corresponding region in parental Kv1.x channels. The fact that KcsA-Kv1.x (x = 1, 3) are expressed in the plasma membrane of E.coli in the functional tetrameric form and able to bind specifically Kv1-blockers at the surface of spheroplasts (Nekrasova et al. 2009a, 2009b) facilitated their applications. Combination of a fluorescently labeled peptide blocker and spheroplasts with membrane-embedded KcsA-Kv1.x (х = 1, 3) channels was demonstrated to produce a convenient and easy-to-use analytical system (Kudryashova et al. 2013; Kuzmenkov et al. 2015), which complemented conventional approaches to the study of channel blockers based on radioligand and patch-clamp techniques.

Here we report on the development of analogous cellular bioengineering system based on spheroplasts bearing KcsA-Kv1.6 channel and demonstrate its applicability for characterization of Kv1.6 blockers. Data on the dependence of dissociation constants of blocker-channel complexes on ionic strength are presented. Molecular modeling of the complexes of KcsA-Kv1.6 and Kv1.6 with peptide pore blockers (agitoxin 2, kaliotoxin 1, OSK1) is performed for the first time. Structural features of complexes are compared, and interaction interfaces are described. Measurements of complexations between KcsA-Kv1.6 and several single point mutants of agitoxin 2 (AgTx2) are performed to verify some of the predicted interactions in a toxin-channel complex. The obtained data can be used to rationally enhance the selectivity of constructed peptide blockers to Kv1.6 or, in opposite, to decrease involvement of this channel in interactions with peptides targeted to other Kv1 channels.

Materials and Methods

Reagents

Tetraethylammonium chloride (TEA), kaliotoxin 1 (KTx) and 4-aminopyridine (4AP) were obtained from Sigma-Aldrich (USA). Recombinant OSK1 and AgTx2 were produced and purified as described earlier (Nekrasova et al. 2009a; Kudryashova et al. 2013). Hetlaxin isolated from the vietnamese Heterometrus laoticus scorpion venom (Anh et al. 2013; Hoang et al. 2014) was presented for investigation by Dr. Yu.N. Utkin. Scyllatoxin and margatoxin (MgTx) were supplied by Alomone Labs (Jerusalem, Israel). Production of recombinant single point mutants of AgTx2, namely, AgTx2(S7K), AgTx2(S11G), AgTx2(K19S), and AgTx2(R31G) is described in Supplementary materials. AgTx2 was labeled with 5(6)-carboxytetramethylrhodamine N-succinimidyl ester (R-AgTx2) as described earlier (Kudryashova et al. 2013).

Cloning and Expression of KcsA-Kv1.6

The gene coding for KcsA-Kv1.6, in which the 52–64 aa fragment of KcsA was changed for the 399–411 aa fragment of Kv1.6 was constructed by overlap extension PCR as described (Nekrasova et al. 2009a) using previously obtained pET28a-KcsA-Kv1.3 plasmid as a template. Kv1.6-specific primers (1.6-f and 1.6-r, Table S1), and pET-specific T7-promoter and T7-terminator universal primers were used for amplification of overlapping DNA fragments. These DNA fragments were further fused in the second round of PCR and ligated into NcoI/HindIII sites of pET28a to give pET28a-KcsA-Kv1.6 plasmid.

E.coli BL21(DE3) transformed with pET28a-KcsA-Kv1.6 were grown overnight at 30 °C in minimal M9 medium supplemented with 0.5 % glucose and 50 μg/ml of kanamycin and then inoculated in 50 ml of this medium in 0.75-l shake flasks to initial optical density of 0.05 (at 560 nm, OD560). In experiments related to the optimization of KcsA-Kv1.6 membrane presentation, the cells were grown at 37 °C with aeration (250 rpm) to OD560 of 0.6, 0.8, 1.2, or 1.9, then induced with 15, 50 or 100 μM isopropyl-β-D-1-thiogalactoside (IPTG) and further grown for 16, 24 or 40 h at different temperatures (20, 25, 30 or 37 °C).

The level of KcsA-Kv1.6 expression was analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). Densitometric analysis of SDS-PAGE gels was performed with Image J software (National Institute of Health, USA).

KcsA-Kv1.6 Binding Studies

Spheroplasts were prepared as described earlier (Kudryashova et al. 2013). Briefly, E.coli cells (3 ml, OD560 = 1.0) were treated with lysozyme (10 mg/l, 7 min) in buffer A (0.5 M sucrose, 10 mM Tris-HCl, pH 8.0), mixed with 3 ml of buffer B (0.6 mM EDTA, 10 mM Tris-HCl, pH 8.0) and incubated for 15 min at 4 °C. This procedure resulted in formation of spheroplasts ((6 ± 2) × 108 cells/ml) that were stabilized with 10 mM MgCl2 and stored at 4 °C.

Analysis of blocker binding to KcsA-Kv1.6 was performed in a buffer containing 0.25 M sucrose, 0.1 % bovine serum albumin, 10 mM MgCl2, 4 mM KCl, 0.3 mM EDTA and supplemented with either 50 mM Tris-HCl (pH 7.5), 50 mM NaCl (high salt buffer, HS-buffer) or 10 mM Tris-HCl (pH 7.5, low salt buffer, LS-buffer). KcsA-Kv1.6-spheroplasts were diluted with HS- or LS-buffer to a concentration of 1000 cells/μl.

For a saturation binding analysis, KcsA-Kv1.6-spheroplasts were incubated with increasing concentrations of R-AgTx2 in either HS- or LS-buffer. For competitive binding, spheroplasts were incubated with R-AgTx2 (1 nM in LS-buffer or 8 nM in HS-buffer) and increasing concentrations of pore blockers. Reactions mixtures were continuously stirred at room temperature during 1.5 h, transferred into wells of a flexiPERM silicon chamber (Perbio) attached to a cover slip, centrifuged for 6 min at 200×g and placed under a microscope.

Confocal fluorescent images of spheroplasts were recorded at ca. 0.25 μm lateral and 1.5 μm axial resolution with an inverted LSM 710 microscope (Carl Zeiss, Germany) using an αPlan-Apochromat oil immersion objective (×63, NA 1.46). Fluorescence of R-AgTx2 was excited with a He-Ne laser (543.5 nm, 12 μW on the sample) and registered in the 550–685 nm range. For quantitative analysis of R-AgTx2 binding to KcsA-Kv1.6, all parameters of the microscope that could affect recorded signal intensity were fixed. Spheroplast recognition in fluorescent images and estimation of an average fluorescence signal intensity per cell (I f ) was carried out with the ImageJ software (National Institutes of Health, USA). For each data point, I f values of 200–250 cells were averaged to get the I av value, and standard deviation was calculated. All measurements were repeated three times. Calculated dissociation constants were averaged over repeated experiments and presented as the mean ± standard error.

Computer Modeling of Channels and Complexes

Initial configurations of Kv1.6 and KcsA-Kv1.6 and their complexes with AgTx2 (KTx, OSK1) were prepared as follows. At first, models of channels and complexes were obtained by homology modeling with Modeller software (Sali and Blundell 1993) using eukaryotic channel Kv1.2 (PDB code 2A79) and the complex of charybdotoxin with eukaryotic chimeric channel Kv1.2–2.1 (PDB code 4JTA) as a templates, respectively. Next, toxins in the complex models were replaced by their experimentally obtained structures (PDB code 1AGT for AgTx2, 3ODV for KTx, 1SCO for OSK1, Fig. 1a). A voltage-sensor domain of Kv1.6 was excluded from the modeling since the analogous domain in Kv1.2 (PDB codes 2A79 and 3LUT), Kv1.2–2.1 (PDB code 2R9R) and charybdotoxin-Kv1.2–2.1 complex (PDB code 4JTA) is structurally independent from a pore domain and is not involved in toxin binding. Homologies between pore domains of Kv1.2–2.1 and Kv1.6, Kv1.2–2.1 (Kv1.2) and KcsA-Kv1.6 are 92 % and 42 %, respectively. Homologies between charybdotoxin and AgTx2, OSK1, KTx are 45 %, 39 %, 39 %, respectively (Fig. 1b). The high homology degree meets a main requirement of homology modeling technique: protein sequence should have over 30 % identity (Xiang 2006; Webb and Sali 2016) .

Fig. 1
figure 1

Structures of AgTx2, KTx, OSK1 and P-loops of KcsA, KcsA-Kv1.6 and Kv1.6 channels. (a) Three-dimensional structures of the toxins in a ribbon representation with Lys27 shown by rods. Colored AgTx2 structure visualizes localization of variable (red) and conservative (blue) residues in the toxins. (b) Alignment of sequences and secondary structure elements of the toxins according to PDB files 1AGT, 3ODV and 1SCO. Differing amino acids are marked with magenta. Pore plugging Lys27 is marked with green. Red and light red wavy lines indicate an α-helixes and 3/10-helixes, respectively, purple bridges – turns, blue lines - bending, yellow arrows – β-structures, black lines – no assigned secondary structure. (c) Alignment of amino acid sequences of P-loops of KcsA, KcsA-Kv1.6 and Kv1.6 channels. Yellow color marks a sequence transferred from Kv1.6 to KcsA to form the KcsA-Kv1.6 channel. Residues that are identical in Kv1.6 and KcsA are marked with blue

Full size image

Free channels and their complexes were put into the calculation cells (7 × 7 × 9 nm size) filled with approximately 11,000 water molecules (simple point charge model) and about 100 sodium and chloride ions. A minimal distance from a protein to a cell boundary was 1 nm. A potassium ion was placed in site S2 of a channel selectivity filter. Alternative K+ ion binding sites S1 and S3 were occupied by water molecules, because a presence of K+ ions in these sites was shown to prevent positioning of a pore blocking Lys residue of toxins in the selectivity filter (Chen and Kuyucak 2009). A set of five systems was prepared for each complex differed in the angle of toxin orientation (−10, −5, 0, 5 and 10°, where a positive direction is counterclockwise) relative to the charybdotoxin orientation in the 4JTA structure. The axis of a toxin rotation was a channel pore axis. The rotation range (±10°) was chosen on a basis of preliminary calculations that revealed a low probability of larger angles of a toxin orientation in a channel binding site (data not shown).

Molecular Dynamics

The Gromacs software package (www.gromacs.org) with the force field opls-AA (Jorgensen et al. 1996) was used for the molecular dynamics simulations. To optimize a system geometry, a potential energy minimization was performed by steepest descent and conjugate gradient methods followed by a system equilibration in the canonical (NPT: constant number of particles, volume, and temperature) and isothermal–isobaric (NVT: constant number of particles, pressure and temperature) ensemble. A temperature of 300 K and a pressure of 1 bar were maintained during the calculations. MD calculations were performed for 5 ns, whereas system equilibration was achieved during ca. 1 ns as followed from the analysis of the root mean-square deviation (RMSD) values of protein Cα-atoms. The residues of transmembrane helices of channels were frozen in the MD process to mimic a movement restricting action of membrane environment. We did not simulate the membrane around the channel because multiple MD studies revealed a good agreement of experimental data with simulated complexes between scorpion toxins and potassium channels obtained with such simplified approximation (Cui et al. 2002; Eriksson and Roux 2002; Wu et al. 2004; Ling and Yingliang 2007).

Data Analysis

Preferred orientation of a toxin in a complex was defined by analyzing the distribution of a toxin orientation angle over the MD trajectory. Analysis of hydrophobic interactions, hydrogen and ionic bonds of a toxin having preferred orientation was performed using extended version of the Platinum software package (www.model.nmr.ru/platinum (Pyrkov et al. 2009)) for 70 selected trajectory frames. Strongly interacting residues were recognized as having considerable hydrophobic contact area (>0.2 Å2) and (or) strong hydrogen bonds (score > 0.5) and (or) short-range ionic interactions (distance between ions <6 Å) after averaging of these parameters determined in the selected trajectory frames. Such residues were considered to form an interaction interface.

Results

Design of KcsA-Kv1.6 and Cellular System Based on KcsA-Kv1.6

Design of KcsA-Kv1.6 hybrid channel proposed by Legros et al. (2002) is based on the substitution of a part of the extracellular P-loop (52RGAPGAQLITYPR64) in a bacterial KcsA channel with the homologous 399ADDDDSLFPSIPD411 sequence from human Kv1.6 (Fig. 1c). After this substitution homology between P-loops of KcsA-Kv1.6 and Kv1.6 achieves 83 %, and KcsA-Kv1.6 gets a capability to bind peptide pore blockers of Kv1.6 (Legros et al. 2002). P-loop contains the main determinants for the binding of outer pore blockers. Together with S5 and S6 transmembrane helices (M1 and M2 helices in KcsA) P-loop forms the pore domain of K+-channels. Even in the transmembrane helices of a pore domain homology between Kv1.6 and KcsA achieves 23 %. Thus, KcsA-Kv1.6 could serve as a close analog of the pore domain of Kv1.6 channel. Other S1-S4 transmembrane helices of Kv1.6 constitute the voltage-sensor domain, which is structurally independent from the pore domain and does not affect the P-loop region. The voltage-sensor domain is absent in KcsA and KcsA-Kv1.6 channels.

Development of the cellular system for the study of Kv1.6 blockers was initiated assuming that KcsA-Kv1.6 expressed in E.coli cells incorporates into plasma membrane and forms tetramers, in which the outer pore is oriented extracellularly. Such property was previously found for KcsA-Kv1.1 and KcsA-Kv1.3 (Nekrasova et al. 2009a, 2009b) and was used to bind peptide blockers at the surface of spheroplasts. Analogously, binding of a fluorescently labeled pore blocker R-AgTx2 to E.coli spheroplasts, which contained KcsA-Kv1.6 in the membrane, resulted in the appearance of fluorescent cells (Fig. S1 in Supplementary materials). AgTx2 was selected for fluorescent labeling due to its high affinity to Kv1.6 (Kd = 36 pM (Garcia et al. 1994)). Optimal conditions of cell growth, which favoured membrane incorporation of KcsA-Kv1.6, were established as described in Supplementary materials (induction with 50 μM IPTG at OD560 = 0.6–0.8 and cultivation of cells at 25 °C for 24 h or longer) and were further used in our studies.

Bright staining of KcsA-Kv1.6-presenting spheroplasts with R-AgTx2 (Fig. S1d) created a basis for the reliable quantitative analysis of ligand binding. This binding can be quantified with a laser scanning confocal microscopy (LSCM). Any pore blocker of Kv1.6 will compete with R-AgTx2 for the binding to KcsA-Kv1.6 and reduce fluorescence of cells in accordance with its affinity. This competition will help to recognize pore blockers among tested compounds and characterize quantitatively their affinities to KcsA-Kv1.6 and, therefore, to Kv1.6.

Characterization of KcsA-Kv1.6 System

Staining of KcsA-Kv1.6-bearing spheroplasts with R-AgTx2 was confirmed to be related to interactions between the blocker and KcsA-Kv1.6. No binding of R-AgTx2 (50 nM) to membrane of native E.coli spheroplasts or to spheroplasts prepared from KcsA-expressing E.coli cells was observed (Fig. 2a). Following our approach approbated for KcsA-Kv1.1 and KcsA-Kv1.3 (Kudryashova et al. 2013; Kuzmenkov et al. 2015), R-AgTx2 binding to KcsA-Kv1.6 on a spheroplast membrane can be characterized with an average fluorescence intensity per cell (I f ) and parameter I av being equal to I f averaged over a measured spheroplast sampling. A dependence of I av on R-AgTx2 concentration is observed in a sub- and low nanomolar range and saturated at R-AgTx2 concentration higher than 1 and 8 nM in LS- and HS-buffer, respectively (Fig. 2b). Frequency distribution of I f at the saturation of R-AgTx2 binding is described by a normal (Gaussian) distribution (R2 = 0.99, Fig. 2c). Therefore population of spheroplasts is uniform in terms of the amount of membrane-embedded KcsA-Kv1.6, and the I av parameter is valid for evaluation of the average amount of KcsA-Kv1.6 complexes with R-AgTx2.

Fig. 2
figure 2

Characteristics of KcsA-Kv1.6 system after optimization of membrane presentation of KcsA-Kv1.6. (a) Comparison of R-AgTx2 (10 nM) binding to KcsA-Kv1.6- and KcsA-bearing spheroplasts and to spheroplasts without recombinant channels in HS buffer. (b) Saturation curves of R-AgTx2 binding to KcsA-Kv1.6 in HS and LS buffers. (c) Fluorescence intensity (I f ) distribution of KcsA-Kv1.6-presenting spheroplasts stained with R-AgTx2 (10 nM) in HS buffer. Data were fitted with a normal (Gaussian) distribution (R2 = 0.99). (d) Comparison of MgTx, scillatoxin, 4-AP and AgTx2 for their ability to displace R-AgTx2 from complexes with KcsA-Kv1.6 in HS and LS buffers (black and white bars, respectively)

Full size image

If KcsA-Kv1.6 receptors are titrated with R-AgTx2 ligand at the condition of [L] >> [R], where [L] and [R] are concentrations of a ligand and receptor, respectively, then the following equation can be applied for the analysis of ligand-receptor interactions (Kudryashova et al. 2013; Kuzmenkov et al. 2015):

$$ {I}_{av}\left(\left[L\right]\right)={I}_{sat}\times \left[L\right]/\left({K}_d+\left[L\right]\right) $$
(1)

where [L] is R-AgTx2 concentration, K d is a dissociation constant, and I sat is a fitting parameter, which is equal to the I av value, when binding is saturated.

The I av ([L]) measurements did not depend on the concentration of cells (i.e. KcsA-Kv1.6 channels) in the 1000–10,000 cells/μl range (data not shown) and, therefore, met to the condition of [L] >> [R]. Fitting of I av ([L]) dependences with Eq. 1 gave us K d of 1.0 ± 0.3 and 0.14 ± 0.04 nM in HS- and LS-buffer, respectively.

Applicability of KcsA-Kv1.6 system for recognition of pore blockers and estimation of their affinity was verified with known pore blockers of Kv1 channels, including peptides AgTx2, OSK1, KTx, MgTx and hetlaxin as well as low-molecular-weight organic molecules TEA and 4-AP. Peptide scyllatoxin, a pore blocker of a small conductance calcium activate potassium channel KCa2.3, was used as a negative control. AgTx2, OSK1, KTx and TEA were found to compete with R-AgTx2 for the binding to KcsA-Kv1.6 in HS- and LS-buffers (Fig. 3a, b). Hetlaxin displaced R-AgTx2 from the complexes with KcsA-Kv1.6 in LS-buffer (Fig. 3b), while 4-AP, MgTx and scyllatoxin were not active at the tested concentrations (60 mM, 0.53 μM and 1.6 μM, respectively, Fig. 2d).

Fig. 3
figure 3

Competition between R-AgTx2 and different ligands for binding to KcsA-Kv1.6-presenting spheroplasts in HS (a, c) and LS (b) buffers

Full size image

The displacement of R-AgTx2 by the blockers was measured when the concentration of free R-AgTx2 was much higher than that of the bound ligand, and analyzed with the equation:

$$ {I}_{av}={I}_m/\left(1+{10}^{\left(\mathit{\lg}\left[C\right]-\mathit{\lg}\right(IC}{{}^{50}}^{\left)\right)}\right) $$
(2)

where [C] is the concentration of the added competitive ligand, I m is I av at [C] = 0, and IC 50 is the ligand concentration that displaces 50 % of R-AgTx2 from the complex with KcsA-Kv1.6. The estimated IC 50 values were used to calculate the apparent dissociation constants K ap of competitive ligands:

$$ {K}_{ap}=I{C}_{50}/\left(1+\left[L\right]/{K}_d\right) $$
(3)

K ap value of 90 ± 20 pM obtained with KcsA-Kv1.6 for AgTx2 in HS buffer is similar to K d of 36 pM measured for this recognized blocker in patch-clamp experiments on Xenopus oocytes expressing Kv1.6 channel (Garcia et al. 1994). TEA, nonspecific blocker of K+ channels, binds to KcsA-Kv1.6 with K ap being equal to 8 ± 3 mM, which corresponds well to its K d of 1.7–7 mM determined in physiological measurements on the Kv1.6 channel (Coetzee et al. 1999). Blocker 4-AP does not compete with R-AgTx2 for the binding to KcsA-Kv1.6 (Fig. 2d) because its binding site is situated on the cytoplasmic side of a Kv1-pore domain (Kirsch et al. 1993). MgTx, which is a highly active blocker of Kv1.1-Kv1.3 channels (www.kaliumdb.org), does not bind to KcsA-Kv1.6 till submicromolar concentrations (Fig. 2d). As expected, scyllatoxin demonstrates no binding to KcsA-Kv1.6 (Fig. 2d). To our best knowledge there are no published data concerning ability of MgTx to block Kv1.6 and scyllatoxin to interact with Kv1 channels.

According to analysis performed with KcsA-Kv1.6 channel in HS buffer, OSK1 is a high affinity blocker of Kv1.6 (K ap  = 190 ± 60 pM). This result is in contradiction with the published statement about inactivity of OSK1 on Kv1.6 (Mouhat et al. 2005, 2006). It has to be noted that this statement was not supported by presentation of any experimental data.

In HS buffer, affinity of KTx to KcsA-Kv1.6 (K ap  = 110 ± 40 pM) is similar to that of AgTx2. To our best knowledge no K d value of KTx was determined for Kv1.6 channel in physiological experiments. At the same time Kv1.6 blocking activity of KTx was demonstrated in the measurements on mammalian cells (Hao et al. 2013).

Hetlaxin is a toxin recently found in the venom of Heterometrus laoticus scorpion (Hoang et al. 2014). It interacts with the extracellular vestibule of the K+-conducting pore of KcsA-Kv1.3 and KcsA-Kv1.1 channels in HS buffer with K ap values of 59 ± 6 nM and 0.8 ± 0.3 μM, respectively (Anh et al. 2013; Hoang et al. 2014). Hetlaxin has K ap of 60 ± 20 nM in LS buffer but does not disturb R-AgTx2 complexes with KcsA-Kv1.6 in HS-buffer even at 0.47 μM (Fig. 3a, b), thus demonstrating low affinity to Kv1.6.

Similar to hetlaxin, affinities of AgTx2, KTx and OSK1 to KcsA-Kv1.6 are several times lower in HS buffer than in LS-buffer. In LS buffer, K ap values of KcsA-Kv1.6 complexes with AgTx2, KTx and OSK1 are 24 ± 14, 22 ± 12 and 70 ± 20 pM, respectively. The most evident reason of ionic strength dependence of toxin binding to KcsA-Kv1.6 is essential contribution of ionic interactions between toxin and channel charged residues into the complex formation and interference of buffer ions with these interactions. A complex with TEA is less sensitive to ionic strength. Its K ap is 6 ± 2 mM in LS buffer.

The presented data show that the KcsA-Kv1.6 cellular system recognizes both small organic molecules and peptides that are able to bind to the extracellular vestibule of the Kv1.6 channel pore. This system discriminates the Kv1.6 pore blockers by their affinity and allows quantitative evaluation of blocker activity in a convenient mix-and-measure regime. The KcsA-Kv1.6 cellular system together with previously designed KcsA-Kv1.1 and KcsA-Kv1.3 systems (Kudryashova et al. 2013; Kuzmenkov et al. 2015) provide extended opportunities to search for and study of Kv1-channel blockers.

In order to extend prospective applications of the new cellular system to rational design of peptide blocker derivatives with advanced properties for functional and biomedical applications, it has to be combined (as discussed in Introduction section) with molecular modeling of blocker-channel complexes. Moreover, modeling of the complexes is aimed to improve our understanding of their structural determinants.

Modeling of Kv1.6 and Kcsa-Kv1.6

Kv1.6 channel belongs to the Shaker-related family of potassium channels, which have a similar structure (Tian et al. 2014). A pore domain of Kv1 channels assembled from transmembrane helices S5, S6 and the P-loop region in a homo-tetrameric structure is a target of peptide blockers. The pore domain of Kv1.6 is highly homologous to that of Kv1.1 (83 %), Kv1.2 (85 %) and Kv1.3 (87 %).

Since the structures of Kv1.6 and KcsA-Kv1.6 channels remain unresolved, the molecular modeling technique is intended to fill in this gap using the known structure of Kv1.2 channel (Banerjee et al. 2013) for homology modeling. We have created a model of a KcsA-Kv1.6 pore domain and repeated previously performed (Mondal et al. 2007) modeling of a Kv1.6 pore domain for direct comparison. In contrast to a previous work (Mondal et al. 2007) the structure of Kv1.6 (Kcsa-Kv1.6) pore domain obtained by homology modeling was further optimized by MD, and detailed description of structural features of Kv1.6 (Kcsa-Kv1.6), which can be important for peptide blocker binding, is presented.

Results of the modeling show that the extracellular moiety of the Kcsa-Kv1.6 (Kv1.6) pore domain is characterized by a broad and rather shallow plateau of a square shape (ca. 10 nm2) surrounding a pore mouth. The plateau is bordered by unstructured P-loops rising up by ca. 1 nm (Fig. 4c, d). While main structural features of Kv1.6 (KcsA-Kv1.6) pore domain are similar to those of Kv1.x (x = 1–3) channels described previously (Chen and Kuyucak 2012), a net charge of Kv1.6 (KcsA-Kv1.6) exterior is much more negative (−24) as compared to the exterior of Kv1.1 (−20), Kv1.2 (−16) and Kv1.3 (−16). Here it is assumed that pKa of histidines is not significantly affected. High negative charge seems to facilitate recognition of Kv1.6 on a plasma membrane by cationic peptide blockers. Characteristic arrays of D400, D401, D402 and D403 (D53-D56) residues situated at the top of P-loops (Fig. 4e, f) are responsible for 16 of 24 negative charges of Kv1.6 (KcsA-Kv1.6) and can navigate a peptide blocker accelerating initial diffusion-limited stage of its binding due to multivalent ionic interactions. Here and below, to avoid repeated references to a channel type, residues of KcsA-Kv1.6 and Kv1.6 are typed by normal type and italics, respectively. Ligand and channel residues are denoted by 3-letter and 1-letter symbols, respectively.

Fig. 4
figure 4

Structural models of KcsA-Kv1.6 (a, c, e, g) and Kv1.6 (b, d, f, h) channels: top exterior (a, b, e-h) and side (c, d) views. Ribbons and solid lines of different color (a-d) show backbones of channels subunits. “Clouds” restrict regions of side chains. Blue circles (a-d) are potassium ions in a channel pore. A side view (c, d) is shown for clarity without a front subunit and a backbone of a back subunit. (e, f) A distribution of sidechains of exposed basic (blue), acidic (red) and aromatic (green) residues. A grey color palette encodes different layers of a channel from top (light grey) to deep (dark grey). Orange ovals show positions of characteristic arrays of D400, D401, D402 and D403 (D53-D56) residues situated at the top of P-loops. (g, h) A distribution of sidechains of exposed hydrophobic (brown) and hydrophilic (blue) residues. Yellow ovals indicate positions of characteristic L405 (L58) residues in P-loops. Color sphere presentation is used to show sidechains

Full size image

Side chains of basic amino acids of Kv1.6 (KcsA-Kv1.6) are practically hidden (Fig. 4e, f) similar to Kv1.1 and Kv1.3 (Chen and Kuyucak 2012), whereas Kv1.2 has exposed arginine at the top of a P-loop. Distributions of aromatic residues in the exterior of Kv1.6 (KcsA-Kv1.6, Fig. 4e, f) and Kv1.x (x = 1–3) (Chen and Kuyucak 2012) channels were found to be similar except for Y429 (Y82), which is present in Kv1.1 and Kv1.6 (KcsA-Kv1.6), but not in Kv1.3.

A presence of L405 (L58) in a P-loop creates four well defined hydrophobic sites within the hydrophilic exterior of Kv1.6 (KcsA-Kv1.6, Fig. 4g, h), which are absent in the exterior of Kv1.x (x = 1–3) channels (Chen and Kuyucak 2012). Evidently, these hydrophobic sites can noticeably affect peptide blocker binding to Kv1.6 (KcsA-Kv1.6).

During MD simulation the channel selective filter and pore region of Kv1.6 (KcsA-Kv1.6) did not undergo significant changes, and their final structure was very similar to that of Kv1.2 in a crystal (RMSD of Cα-atoms was 1 Å). This result is in accordance with a general opinion that a pore region is highly conserved in the potassium channels of eukaryotic and prokaryotic species and has essentially the same structure (Gonzalez et al. 2012).

Modeling of Toxin Complexes

General Description of Complexes

According to the MD studies AgTx2, KTx and OSK1 preserved their initial secondary structure (namely, three β-strands and an α-helix crosslinked with three disulfide bridges, Fig. 1a) in complexes with KcsA-Kv1.6 and Kv1.6 (Fig. 5, 6 and 7). Rigidity of the toxin secondary structure during binding corresponds to the data reported for crystallized complexes of charybdotoxin with Kv1.2–2.1 and modified KcsA (Yu et al. 2005; Banerjee et al. 2013) as well as to molecular modeling studies of complexes of some other channels with toxins (Rashid and Kuyucak 2014).

Fig. 5
figure 5

Structures of complexes of AgTx2 with KcsA-Kv1.6 (a, c, e) and Kv1.6 (b, d, f): top exterior (a, b) and side (c-f) views. Ribbons and solid lines show backbones of channel subunits (red, grey, orange and yellow) and toxin (green). Front and back subunits are removed on side views (c-f) for clarity. Interacting residues are shown using stick presentation

Full size image
Fig. 6
figure 6

Structures of complexes of KTx with KcsA-Kv1.6 (a, c, e) and Kv1.6 (b, d, f): top exterior (a, b) and side (c-f) views. Ribbons and solid lines show backbones of channel subunits (red, grey, orange and yellow) and toxin (green). Front and back subunits are removed on side views (c-f) for clarity. Interacting residues are shown using stick presentation

Full size image
Fig. 7
figure 7

Structures of complexes of OSK1 with KcsA-Kv1.6 (a, c, e) and Kv1.6 (b, d, f): top exterior (a, b) and side (c-f) views. Ribbons and solid lines show backbones of channel subunits (red, grey, orange and yellow) and toxin (green). Front and back subunits are removed on side views (c-f) for clarity. Interacting residues are shown using stick presentation

Full size image

Starting from different orientations AgTx2 and KTx relaxed in complexes with Kv1.6 (KcsA-Kv1.6) during MD experiments to the charybdotoxin orientation in the 4JTA structure, whereas OSK1 turned to −5 ° in the plane perpendicular to the pore axis. In dynamics, RMSD of Cα-atoms of toxins in preferable orientation were ca. 0.5 Å. In the equilibrated complexes (Figs. 5, 6 and 7, S2), a flat surface formed by the second and, partially, third β-strands of the toxins was in contact with the plateau of a pore domain, whereas the toxin α-helix and first β-strand faced away from a channel pore. The second β-strand of the toxins was roughly parallel to the plateau of a pore domain (Table S3) and even produced direct hydrophobic contacts (Phe25, Met29) with some amino acids around the pore (Tables 1, 2 and 3). Lys27 of the second β-strand of the toxins preserved intrusion into the pore (Figs. 5, 6 and 7, S2) that was an essential feature of the used template (charybdotoxin complex) and had a similar intrusion depth in the studied complexes (Table S2). The toxins were bound asymmetrically relative to a channel pore: an α-helix side of the toxins touched channel P-loops rising around the plateau, whereas an opposite side (a region of the first β-strand) was far from the plateau boundary (Figs. 5, 6 and 7).

Table 1 AgTx2 interactions with KcsA-Kv1.6 and Kv1.6 according to molecular modeling
Full size table

Complexes were stabilized with variegated interactions of 15–19 toxin residues with 10 channel residues (Figs. 5, 6 and 7, S2, Tables 1, 2 and 3). All channel subunits were involved in complex formation but in different extent. OSK1 formed considerable number of bonds with subunits 3 and 4, whereas for AgTx2 and KTx, interactions with subunit 3 dominated. Here number 3 was assigned to the subunit having major interactions with the α-helix of toxins, and numeration occurred clockwise if to see from an extracellular side (Figs. 5, 6 and 7).

Note that the algorithm of Kv1.6 (KcsA-Kv1.6) complex modeling was significantly modified as compared to previously published approach (Mondal et al. 2007). We performed homology modeling of channel-toxin complexes instead of molecular docking to obtain a more reliable model, optimized structures using MD and for the first time described all contacts formed between Kv1.6 (KcsA-Kv1.6) and studied toxins using advantages offered by the program Platinum.

Channel Interaction Interfaces

Channel amino acids that interact with toxins form two main clusters (Tables 1, 2 and 3, Figs. 5, 6 and 7). The first cluster is situated in the P-loop and includes three residues: D402, D403, L405 (D55, D56, L58) in complexes with AgTx2 and OSK1; D403, L405, P407 (D56, L58, P60) in a complex with KTx. Second cluster of interacting residues includes Y425 and G426 (Y78, G79) of the selective filter and four next residues (D427, M428, Y429, M431; D80, L81, Y82, V84) forming a plateau around the channel pore. One more residue interacting with the toxins is found in a pore helix (D411; D64).

It is clearly seen that patterns of interacting amino acids are similar for KcsA-Kv1.6 and Kv1.6 (Tables 1, 2 and 3, Figs. 5, 6 and 7). KcsA-Kv1.6 reproduces bundles of interactions formed by L405, D427 and Y429 in the Kv1.6 complexes with studied toxins, a particular bundle of bonds generated by D403 in the complex with OSK1 and a rather moderate number of bonds produced with toxins by D402, D411, M428 and M431. At the same time invariant interactions of these residues are supplemented with additional bonds that vary for KcsA-Kv1.6 and Kv1.6 (Tables 1, 2 and 3). Totally, molecular modeling predicts that 81, 77 and 75 % of bonds in (KcsA-Kv1.6)-AgTx2, (KcsA-Kv1.6)-KTx and (KcsA-Kv1.6)-OSK1 complexes are the same as in the Kv1.6-AgTx2, Kv1.6-KTx and Kv1.6-OSK1 complexes, respectively. And vice versa, 72, 93 and 73 % of bonds in the Kv1.6-AgTx2, Kv1.6-KTx and Kv1.6-OSK1 complexes are the same as in (KcsA-Kv1.6)-AgTx2, (KcsA-Kv1.6)-KTx and (KcsA-Kv1.6)-OSK1 complexes.

Among the interacting residues of Kv1.6, residues D402, D403, L405 and P407 of a P-loop and Y429 can be responsible for a channel-specific modulation of peptide blocker binding, since they differ (totally or partially) from the residues situated in analogous positions of related Kv1.x (x = 1–3) channels. D402 and L405 are different from analogous residues in Kv1.x (x = 1–3). P407 of Kv1.6 is present in Kv1.2 but not in Kv1.1 and Kv1.3. D403 distinguishes Kv1.6 from Kv1.3 and partially from Kv1.1 and Kv1.2, where a homological glutamic acid residue is present. Y429 of Kv1.6 is present in Kv1.1 but not in Kv1.2 and Kv1.3.

As mentioned above, L405 (L58) and Y429 (Y82) produce two remarkable bundles of interactions. L405 (L58) forms multiple hydrophobic contacts with an every studied toxin (Tables 1, 2 and 3). From five to eight toxin residues interact with L405 (L58). Among them there are variable toxin residues as well as Leu(Ile)15 and Arg24 that are conservative in the studied toxins. In complexes with AgTx2 and KTx, residue L405 (L58) of subunit 3 has the highest number of hydrophobic contacts as compared to L405 (L58) of other subunits. Residues L405 (L58) of subunits 3 and 4 are highly involved in the OSK1 binding. A bundle of hydrophobic contacts in a P-loop is a characteristic feature of Kv1.6 (KcsA-Kv1.6) complexes.

Y429 (Y82) produces a bundle of hydrophobic contacts and hydrogen bonds with 6–8 toxin amino acids (Tables 1, 2 and 3). Residues Y429 (Y82) of all subunits are involved in the binding of the toxins, but participation of subunit 3 prevails in the complexes with AgTx2. Conservative Leu(Ile)15, Met29 and Thr36 as well as variable residues of the toxins interact with Y429 (Y82).

Ionic and hydrogen bonds of D402 (D55) were found in all complexes except for the KTx-Kv1.6 complex. Multiple salt bridges and hydrogen bonds produced by D403 (D56) are a feature of the complex with OSK1. In contrast, involvement of D403 (D56) in the stabilization of complexes with KTx and AgTx2 is noticeably less. The single bond of P407 (P60) was revealed in the complex with KTx only. Obviously, structural differences of peptides results in multiple features of their interactions with the channel-specific residues.

Residues D411, Y425, G426, D427 and M428 of Kv1.6 are conservative in Kv1.x (x = 1–3) and manifest noticeable conservatism in interactions with toxins. Though there are still many gaps in understanding the structure of Kv1.x (x = 1–3) complexes with AgTx2, KTx and OSK1, the published data support this statement. Residue D411 forms an ionic bond with invariant cationic Arg24 of AgTx2, KTx and OSK1 (Tables 1, 2 and 3) like its analog (D381) in Kv1.3-AgTx2 complex (Gao and Garcia 2003). A role of Y425 (Tables 1, 2 and 3, Fig. 5, 6 and 7) is the same as the role of an analogous tyrosine residue in Kv1.х (х = 1–3) (Eriksson and Roux 2002; Rashid and Kuyucak 2012; Wang et al. 2015), namely, to bind a toxin Lys residue (Lys27 for AgTx2, KTx and OSK1) that occludes a channel pore. Residue G426 is differently involved in interactions with AgTx2, KTx and OSK1 (Tables 1, 2 and 3), but its analog (G396) in Kv1.3 is also bound to Ser11 of AgTx2 (Gao and Garcia 2003). Residue D427 produces a hydrogen bond with conservative Asn30 of the studied toxins and some additional bonds that vary for the toxins (Tables 1, 2 and 3). Similar interaction (D375-Asn30) is found in the crystal of Kv1.2-charibdotoxin complex (Banerjee et al. 2013) and predicted for KTx-Kv1.3 (D402-Asn30) (Yu et al. 2004). Hydrophobic interaction of M428 with invariant Phe25 of the toxins (Tables 1, 2 and 3) is reproduced in the Kv1.3-AgTx2 complex (Gao and Garcia 2003).

For complexes of several toxins with channels Kv1.2–2.1 (Banerjee et al. 2013), Kv1.1 (Kohl et al. 2015), Kv1.2 (Jouirou et al. 2004) and Kv1.3 (Gao and Garcia 2003; Yu et al. 2004; Rashid and Kuyucak 2012) it was concluded that analogs of Y425, G426 and D427 act as an anchor on the channel surface for the bound toxins. Our data allows extension of this conclusion to Kv1.6 complexes. But in spite of the conservatism, residues Y425, G426, D427 seem to be not sufficient for toxin binding, since KcsA channel, in which they are present in analogous positions, does not bind peptide pore blockers of Kv1-channels. R64D mutation (analogous to D411 in Kv1.6) is required to impart toxin binding property to KcsA (MacKinnon et al. 1998). Therefore, D411 together with Y425, G426 and D427 should be considered as a basic docking assembly for toxins in Kv1.x (x = 1–3) and Kv1.6. In Kv1.6, in addition to bonds of Y425 and D411 described above, hydrogen and, surprisingly, hydrophobic interactions are realized between D427 and five (OSK1) or six (KTx, AgTx2) toxin residues. Next, G426 produces either hydrophobic contacts (OSK1) or both hydrophobic contacts and hydrogen bonds (AgTx2, KTx) with three toxin residues.

During optimization of complexes in MD experiments, the structure of a selective filter of KcsA-Kv1.6 (Kv1.6) channel did not undergo significant changes (RMSD of Cα atoms =1 Å), whereas the structure of P-loops varied greater (RMSD of Cα atoms =2.5 Å). Similar to Kv1.x (x = 1–3) channels, selective filter of Kv1.6 (KcsA-Kv1.6) is a very conservative and rigid region that is not adjusted to toxins, and formation of complexes within this region corresponds well to a “lock and key” mechanism. P-loops are much more tunable to different interactions with toxin residues. In the studied complexes, P-loop of subunit 1 has a greatest mobility that can be explained by the least number of bonds formed between subunit 1 and toxins (Tables 1, 2 and 3).

Features of Toxin Interactions with Kv1.6 and KcsA-Kv1.6

Interactions of Homologous Residues of Toxins

Representatives of αKTx3 subfamily of scorpion toxins, AgTx2, OSK1 and KTx, are highly homologous: 29 of 38 residues are invariant in all three peptides, 34 residues are invariant in AgTx2 and KTx (Fig. 1). At the same time, twelve invariant residues do not participate directly in the binding of the toxins to Kv1.6 (KcsA-Kv1.6). Their primary role (together with other homologous residues) seems to be maintaining a characteristic secondary structure as well as its proper spatial organization that are essential factors contributing to toxin binding to both Kv1.6 (KcsA-Kv1.6) and Kv1.x (x = 1–3). Among interacting amino acids seven homologous toxin residues (Leu (Ile)15, Arg24, Phe25, Lys27, Met29, Asn30, Pro37) form a network of bonds, which is reproduced in all studied complexes and includes one ionic, multiple hydrophobic and several hydrogen bonds (Tables 1, 2 and 3). This steady interaction network is complemented with additional bonds of the invariant residues, which vary from toxin to toxin. In particular, Phe25 and Met 29 are involved in several additional interactions that differ for AgTx2, KTx and OSK1 (Tables 1, 2 and 3).

Five residues (Leu (Ile)15, Arg24, Phe25, Lys27 and Met 29) are especially prominent in the network because they produce knots of interactions (three and more bonds) in all studied complexes (Tables 1, 2 and 3). His34 in OSK1 and Thr36 in KTx produce toxin-specific knots of interactions.

If we compare AgTx2 and KTx complexes only, Gly1, Val2, Ser11 and Thr36 are added to the list of invariant residues that produce similar bonds. This extended similarity of interactions is explained by highest homology and the same orientation of AgTx2 and KTx in the binding site of Kv1.6 (KcsA-Kv1.6).

Most of residues forming the steady interaction network are situated in the second β-sheet having 100 % homology in the studied toxins (Fig.1b) and producing extensive and tight contact with a plateau around the channel pore (Tables 1, 2 and 3, S2). It is reasonable to suppose that the steady interaction network formed by invariant residues is a basis of complexes of AgTx2, OSK1 and KTx with Kv1.6 (KcsA-Kv1.6) and probably with other Kv1-channels. In accordance with this supposition, involvement of Leu15, Arg24, Phe25, Lys27, Met29 and Asn30 in KTx binding to Kv1.3 was determined experimentally (Aiyar et al. 1995; Hidalgo and Mackinnon 1995; Naranjo and Miller 1996; Ranganathan et al. 1996) and demonstrated with molecular docking (Aiyar et al. 1996). Molecular modeling predicts that Arg24, Phe25, Lys27, Met29 and Asn30 of AgTx2 interact with Kv1.3 (Lipkind and Fozzard 1997; Gao and Garcia 2003). Mutations of Arg24, Phe25, Lys27, Asn30 and Pro37 were shown to decrease considerably the AgTx2 binding to a Shaker K+ channel from Drosophila (Gross and Mackinnon 1996; Ranganathan et al. 1996). Theoretical functional maps of OSK1 complexes with Kv1.x (x = 1–3) channels highlight importance of the following interactions: Arg24, Lys27 and Met29 with Kv1.1; Arg24, Phe25, Lys27 and Met29 with Kv1.2; Arg24, Phe25, Lys27, Met29 and Asn30 with Kv1.3 (Mouhat et al. 2005).

Interactions of Toxin Variable Residues

Variable amino acids in the studied toxins are: residues 3, 7, 9 and, partially, 15 for KTx versus AgTx2; residues 3, 7, 9, 10, 16, 20 and 31 for OSK1 versus AgTx2 and KTx (Fig. 1).

As mentioned above Pro3/Asp3 and Ser7/Lys7 of AgTx2/KTx do not interact with Kv1.6 (KcsA-Kv1.6) (Tables 1 and 2). Interactions of Ile15/Leu15 of AgTx2/KTx with Kv1.6 (KcsA-Kv1.6) are practically identical (Tables 1 and 2). Thr9 of AgTx2 forms a hydrogen bond with a channel, whereas Ser9 of KTx does not participate in the binding. Therefore, variable amino acids in AgTx2 and KTx do not introduce noticeable differences in the interactions of these toxins with Kv1.6 (KcsA-Kv1.6). This is a definite reason of similar affinities of AgTx2 and KTx to Kv1.6 pore revealed with the KcsA-Kv1.6 channel. Like 3D structure (Fig. 1a), which is preserved in complexes (Figs. 5 and 6), and similar repertoire of bonds formed by homologous residues (Tables 1 and 2) are two other essential factors that equalize interactions of AgTx2 and KTx with Kv1.6 pore.

Table 2 KTx interactions with KcsA-Kv1.6 and Kv1.6 according to molecular modeling
Full size table

As for OSK1 (Table 3), variable residues 16, 20 and 31 were not involved in the interactions with Kv1.6 (KcsA-Kv1.6). Variable residues 3, 7, 9 and 10 contributed to the interactions producing hydrophobic contacts (Ile3 and Ile10) or forming hydrogen and ionic bonds (Lys7 and Lys9) with channel residues. This should enhance OSK1 binding to the channel as compared to AgTx2 with non-interacting residues 3,7 and 10, but there are additional essential differences in the bonds produced by these toxins that result finally in slightly weaker affinity of OSK1 to Kv1.6 pore. In particular, Val2, Gly26 and Arg31 participate in the interactions of AgTx2, whereas Cys8, Lys32 and His34 are involved in the binding of OSK1.

Table 3 OSK1 interactions with KcsA-Kv1.6 and Kv1.6 according to molecular modeling
Full size table

Interactions of Toxins with L405 (L58) of a Channel

As pointed out above hydrophobic sites in the P-loops formed by L405 (L58) are characteristic feature of Kv1.6 (KcsA-Kv1.6). This feature is recognized by toxins. Five-six amino acids of toxins minimize unfavorable exposure of hydrophobic groups by forming hydrophobic contacts with L405 (L58) of a channel (Tables 1, 2 and 3). The contacts of Ile15(Leu15), Arg24 and Pro37 are reproduced for all toxins. AgTx2 and KTx have additional similar hydrophobic interactions of Val2 and Phe25 with L405 (L58). Hydrophobic interactions of OSK1 with L405 (L58) are strengthened by Thr36 as well as by Cys8 and Lys19 or Lys9 and Ile10 in the case of KcsA-Kv1.6 or Kv1.6, respectively.

Interactions of Toxin Charged Residues

Positively charged amino acids of toxins are supposed to play an essential role in channel recognition, navigation and binding. AgTx2, OSK1 and KTx have a high positive charge, but differ in amount and distribution of charged residues (Fig. 1b). As compared to least charged AgTx2, OSK1 has two additional cationic amino acids, whereas KTx has one cationic and one anionic additional residues. No direct relationships were observed between a number of cationic and anionic amino acids and toxin affinities. Some of charged amino acids (including variable) do not produce bonds with Kv1.6 (KcsA-Kv1.6): Lys32 and Lys38 in AgTx2; Glu3, Lys7 and Lys32 in KTx; Glu16 and Lys20 in OSK1 (Tables 1, 2 and 3). Among invariant cationic residues Lys19 and, especially, Arg24 and Lys27 are involved (see above) in the binding in all complexes (Tables 1, 2 and 3). Interactions of other charged amino acids vary for the studied toxins and sometimes differ for Kv1.6 and KcsA-Kv1.6 (Tables 1, 2 and 3). Ionic bonds of Arg31 contribute to particular interactions of AgTx2 and KTx with Kv1.6 and KcsA-Kv1.6 as compared to OSK1, which lacks this residue. Arg24 of KTx produces a specific salt bridge in addition to invariant one (Table 2). Lys7, Lys9, Arg12 and Lys32 of OSK1 form a set of distinctive ionic bonds with aspartates of the D-array of Kv1.6 (KcsA-Kv1.6).

As discussed above, a D-array is a characteristic feature of Kv1.6 (KcsA-Kv1.6), and cationic amino acids of the studied toxins recognize it producing ionic and hydrogen bonds with D402(D55) and D403(D56). This recognition is most pronounced for OSK1 (Table 3), in which five cationic residues form the extended set of interactions with D402(D55) and D403(D56). Two and three cationic residues interact with these aspartates in AgTx2 and KTx, respectively (Tables 1, 2).

Seven (eight) of ten charged residues of OSK1 participate in complexation with Kv1.6 (KcsA-Kv1.6), and a number of ionic bonds is highest in these complexes as compared to the complexes with AgTx2 and KTx. AgTx2 forms the least number of ionic bonds in the complexes, but its affinity is slightly higher than that of OSK1. For certain, affinities of the toxins to Kv1.6 (KcsA-Kv1.6) are defined not only by ionic interactions, but also by the combination of hydrogen bonding and hydrophobic interactions (Tables 1, 2 and 3).

Interactions of AgTx2 Mutants with KcsA-Kv1.6

As discussed above, residues of KcsA-Kv1.6 (Kv1.6) that interact with toxins form two clusters: the first one comprises the residues, which are specific for KcsA-Kv1.6 (Kv1.6), whereas the second cluster includes residues that are mainly invariant in relative Kv1-channels. It is reasonable to suppose that substitutions in toxin residues interacting with the first cluster can be used to modulate affinity and specificity of blockers to KcsA-Kv1.6 (Kv1.6). In addition, construction of such toxin mutants can help to verify the developed models in experiments and check reliability of molecular modeling predictions. We have produced four AgTx2 mutants having single point mutations (S7 K, S11G, K19S, R31G) and measured their binding to KcsA-Kv1.6. Mutant AgTx2(S7 K) was designed to verify a molecular modeling prediction that Ser7 is not involved in the interactions of AgTx2 with KcsA-Kv1.6 (or Kv1.6, Table 1). Similar Kap values obtained for AgTx2 (K ap  = 70 ± 10 pM) and AgTx2(S7 K) (K ap  = 55 ± 15 pM, Fig. 3c) supported this prediction. According to data presented in Table 1, substitution S11G should disrupt hydrogen bonds formed by Ser11 with channel residues and weaken toxin-channel interactions. Exactly, binding of KcsA-Kv1.6 to the toxin was considerably impaired by S11G mutation (K ap  = 350 ± 60 pM, Fig. 3c). To check the effect of ionic interaction of Lys19 with a channel (Table 1) and estimate its strength, this charged residue was replaced with serine. Competitive binding measurements revealed fivefold reduction of AgTx2(K19S) affinity to KcsA-Kv1.6 (K ap = 370 ± 50 pM) as compared to AgTx2 (Fig. 3c), which was evidently associated with disappearance of the Lys19-related ionic bond with D55 of KcsA-Kv1.6. Involvement of another charged residue, Arg31, in ionic interactions with KcsA-Kv1.6 (Table 1) was confirmed by an increase in K ap to 300 ± 80 pM after R31G substitution (Fig. 3c). Lys19 and Arg31 of AgTx2 interact with aspartates (D55 and D56) of KcsA-Kv1.6 (Table 1), which are a feature of Kv1.6 P-loops. Thus, disruption of these bonds accompanied with a decrease in stability of AgTx2-Kv1.6 complex could be a way to increase selectivity of toxin interactions with Kv1.3 and (or) Kv1.1.

Conclusion

As demonstrated, the developed cellular system, which is based on the membrane presentation of hybrid KcsA-Kv1.6 channel and fluorescent detection of pore blocker interactions, is a new convenient analytical tool for investigation of molecular determinants of affinity and specificity of peptide blockers to Kv1.6 pore. Fundamentally, this system utilizes evolution-related homology and structural similarity of pore domains of eukaryotic and bacterial potassium channels as well as accumulated knowledge on structural elements forming a toxin binding site in the outer vestibule of Kv1 channels. The presented results are a successful attempt to create artificial high-affinity membrane-embedded receptor and to control its expression, membrane transfer and presentation in E.coli cells by bioengineering means.

Comparative molecular modeling of toxin complexes with KcsA-Kv1.6 and Kv1.6 revealed reproduction of the binding mode and essential intermolecular bonds in both types of complexes that complementary validated an application of KcsA-Kv1.6-based system for investigation of toxin-Kv1.6 channel interactions. Moreover, KcsA-Kv1.6 system is suitable for seeking new peptide blockers in animal venoms, and these studies are in progress now.

Complexes of three recognized Kv1 channel blockers (AgTx2, OSK1 and KTx) with a pore domain of Kv1.6 channels, which were created for the first time using homology modeling approach, fill in the gap in our understanding of structural features of peptide blocker interactions with Kv1.6 channel. Ten channel residues participated in the toxin binding comprise two clusters: residues, which are situated in the P-loop and vary in relative Kv1 channels and conserved residues, which are involved in interactions with toxins in different Kv1 channels. The first cluster can be a target to modulate (enhance or reduce) binding of designed peptide blockers to Kv1.6 as confirmed with point mutations introduced in AgTx2. Taking into account the published data, the second cluster seems to be a general basis for toxin binding to relative Kv1 channels.

Definitely, the rational design of selective peptide channel blockers, which are required in physiology investigations of Kv1 channels and in prospective medical applications, can be assisted with molecular modeling techniques. These computational experiments are intended to take into account fine features of multipoint peptide blocker interactions with several relative Kv1 channels.