NS1643 Interacts around L529 of hERG to Alter Voltage Sensor Movement on the Path to Activation

Abstract

Activators of hERG1 such as NS1643 are being developed for congenital/acquired long QT syndrome. Previous studies identify the neighborhood of L529 around the voltage-sensor as a putative interacting site for NS1643. With NS1643, the V_1/2 of activation of L529I (−34 ± 4 mV) is similar to wild-type (WT) (−37 ± 3 mV; P > 0.05). WT and L529I showed no difference in the slope factor in the absence of NS1643 (8 ± 0 vs. 9 ± 0) but showed a difference in the presence of NS1643 (9 ± 0.3 vs. 22 ± 1; P < 0.01). Voltage-clamp-fluorimetry studies also indicated that in L529I, NS1643 reduces the voltage-sensitivity of S4 movement. To further assess mechanism of NS1643 action, mutations were made in this neighborhood. NS1643 shifts the V_1/2 of activation of both K525C and K525C/L529I to hyperpolarized potentials (−131 ± 4 mV for K525C and −120 ± 21 mV for K525C/L529I). Both K525C and K525C/K529I had similar slope factors in the absence of NS1643 (18 ± 2 vs. 34 ± 5, respectively) but with NS1643, the slope factor of K525C/L529I increased from 34 ± 5 to 71 ± 10 (P < 0.01) whereas for K525C the slope factor did not change (18 ± 2 at baseline and 16 ± 2 for NS1643). At baseline, K525R had a slope factor similar to WT (9 vs. 8) but in the presence of NS1643, the slope factor of K525R was increased to 24 ± 4 vs. 9 ± 0 mV for WT (P < 0.01). Molecular modeling indicates that L529I induces a kink in the S4 voltage-sensor helix, altering a salt-bridge involving K525. Moreover, docking studies indicate that NS1643 binds to the kinked structure induced by the mutation with a higher affinity. Combining biophysical, computational, and electrophysiological evidence, a mechanistic principle governing the action of some activators of hERG1 channels is proposed.

Introduction

Well-orchestrated opening and closing or gating of ion channels in cardiac myocytes controls cardiac electrical excitation and relaxation that is directly coupled to the normal functioning of the heart (1). A component of this mechanism is the presence of potassium channels that shape signaling by selective permeation of K⁺ and voltage-dependent gating (2). Previous studies of voltage-dependent gating in K⁺ channels have reported substitution of three nonpolar residues of the S4 of the Shaw channel into Shaker (V369I, I372L, S376T; ILT motif) reproduced the voltage-dependent properties of activation of the Shaw channel (3–6). These amino-acid substitutions were quite conservative, suggesting that physiologically very important changes can be achieved with rather subtle modifications of nonpolar amino acids within the S4. Moreover, the Aldrich laboratory (5,6) provided direct evidence that the slowed activation was achieved by changing cooperative transitions late in the pathway toward opening of the channel. Their studies suggested that noncharged residues in the S4 could play a pivotal role in the cooperative interactions between subunits that preceded a final opening step. The general involvement of uncharged residues in model systems can be extended to clinically relevant human proteins such as a family of hERG1 channels.

The hERG1 gene (also referred to as KCNH2) encodes the α-subunit of an ion channel (Kv11.1, sometimes simply denoted as hERG1) underlying the rapid component in the delayed rectified potassium current (I_Kr) in cardiac myocytes (7). In the human heart, modulation of hERG1 currents reportedly has both therapeutic and proarrhythmic consequences (8,9). Wang and Rasmusson (10) have modeled the state-dependent changes in activation of hERG1 channels. Compared to the activation of Shaker, the activation kinetics of the ionic hERG1 current are very slow. Gating current kinetics indicate that voltage-sensor movements in the pathway toward activation of the hERG channel are also slow, although components of charge that move quickly have also been reported (10–13). In general, voltage-sensor movements reported from fluorescence studies mirror the movement of the majority of charge measured using gating currents (14).

While the best-characterized feature of hERG1 is drug interaction with its promiscuous intracavity blocking site by a variety of drugs, a rapidly emerging strategy focuses on channel activation by small molecules—i.e., channel activators (openers) (15). An implemented screening of novel drug candidates for their ability to attenuate hERG1 function has led to an identification of compounds capable of hERG1 current enhancement. A thorough examination of the effects of hERG1 activators in vitro and in intact cardiac tissue has not yet been completed. A drug-induced increase in the time-dependent I_Kr rather than its tail current could truncate the action potential and cause the short QT syndrome similar to that of gain-of-function mutations (16). One of the best-characterized hERG activators is NS1643 (17,18). It shifts the voltage-dependence of activation to hyperpolarized potentials, shifts the voltage-dependence of C-type inactivation to depolarized potentials, and increases tail-current amplitude. In a 2012 study, we used a structure-guided mutagenesis strategy to identify a number of novel amino acids that have proximity to NS1643 docked to a model of the hERG1 channel (18). One amino acid that was identified as potentially being involved in NS1643 binding was L529. However, the electrophysiological responses to NS1643 for that mutation were not examined. The reported binding region in the S4 was relatively large, spanning from L523 to V535 and from L553 to W568.

In our original article (18), radial cavities (10 Å) were created around a reference residue and then the reference amino acid was iteratively shifted to examine the characteristics of multiple radial cavities. The docking scores were obtained for each radial cavity to identify the most likely targets. We recognized that this approach had limitations because of the flexibility of this domain and the homology model itself. The L529 residue was identified as a potentially important residue but it was not highlighted as a key residue. Even so, we identified other residues (18) that were important determinants of pharmacologic response to NS1643. This study is based on the observation that L529 is equivalent to the hydrophobic isoleucine amino acid of the ILT domain (Table S1 in the Supporting Material) described by the Aldrich laboratory (5,6). Thus, a different experimental approach is being used to refine our understanding of the importance of residues in the large and flexible NS1643 binding domain of hERG1. It is also tempting to propose a possible coupling between an activator binding to the channel and subsequent gating modification via perturbing cooperative movements of the voltage-sensor between subunits. Accordingly, this study focuses on the impact of substitutions of nonpolar amino acids at the L529 site in the S4 of hERG1 on the characteristic response to NS1643. We combined biochemical, spectroscopic, electrophysiological, and theoretical studies to gain deeper understanding of functional and structural factors governing activator action and its dependence on nonpolar moieties in the S4 helix.

Materials and Methods

All animals were housed in the Animal Resource Centre of the Faculty of Medicine, University of Calgary, or at the Alberta or Simon Fraser University, Burnaby, British Columbia, Canada, using protocols in accordance with animal care guidelines established by the Canadian Council on Animal Care.

S4 sequences alignment

Table S1 shows the alignment of S4 sequences highlighting the ILT sequence identified by Aldrich’s laboratory in Shaker and Shaw as it relates to the equivalent hERG1 sequence (3–6). L529 in hERG1 appears to be equivalent to I of the ILT motif. This is part of the rationale for creating L529I hERG1. A second part of the rationale is that we previously reported that L529 has proximity to NS1643 when docked to a homology model of hERG1 (18).

Structural model of hERG1

Our model of the S1–S6 domains of hERG1 channel has been previously reported in Subbotina et al. (19) and Durdagi et al. (20) and has been independently validated by a number of laboratories (21,22). This homology model was used to guide mutagenesis to define potential binding sites of NS1643 to hERG1 (18). A number of putative binding sites were identified. One of the key binding sites was the S4 domain of hERG1. In this study, this homology model was used to help understand the potential molecular interactions of NS1643 in the neighborhood of L529.

Molecular dynamics on wild-type and L529I hERG1

A series of molecular-dynamics (MD) simulations were performed to test possible structural effects of L529I mutation and to refine binding pocket for future docking studies. All MD simulations were carried out using our previous homology models of hERG1 open and closed states with the NAMD program (23) and the CHARMM-27/CHARMM-36 force fields for proteins, ions, and phospholipids, and the TIP3P water model (24). All simulations were carried out at 323 K and 1 atm using periodic boundary conditions and the NPT ensemble. Similar to previous MD simulations of K⁺ channels, the particle-mesh Ewald algorithm was used for electrostatic interactions. K⁺ ions at the selectivity filter were used in the S0:S2:S4 positions according to previous studies. Each model was embedded into the DPPC membrane bilayer using the CHARMM-GUI membrane builder protocol. The simulation box contained one protein, DPPC molecules, 3 K⁺ ions, and pore water molecules in the intracellular cavity, solvated by 0.15 M KCl aqueous salt solution. Structures were minimized and equilibrated with gradually decreasing harmonic constraints (see CHARMM-GUI equilibration protocol for details) for 2 ns and then subjected to a 50-ns production run. Langevin dynamics with very weak friction coefficient was used to keep the temperature constant. The Langevin Nosé-Hoover method as implemented in CHARMM 36b1 (24) was used to maintain the pressure at 1 atm. The HELANAL module of the MDAnalysis software (25) was used to analyze the geometry of the helices on the basis of their Cα carbons alone. The geometry of the α-helix is characterized by computing the local helix axes and local helix origins for four contiguous Cα atoms (26). The softwares VMD (27) and PYMOL (28) were used to visualize trajectories.

Molecular docking

The methodology used to dock the NS1643 activator has been previously reported in Durdagi et al. (18). Briefly, the derived chemical coordinates of the drug were docked to the channel using GLIDE/INDUCED FIT DOCKING programs.

Molecular biology

The methods for site-directed mutagenesis have been previously reported in Durdagi et al. (18). The hERG1 constructs were transfected into mammalian hEK cells because their background potassium currents are small and no dofetilide-sensitive tail current is observed in untransfected hEK cells. For the parallel Xenopus oocyte studies, hERG1a channels were expressed in Xenopus laevis oocytes using a pBluescript SKII expression vector. The L520C mutation in the S3-S4 linker was used as a site for fluorophore labeling of the voltage sensor, while two endogenous extracellular cysteine residues (C445 and C449) in the S1-S2 linker were mutated to valine to prevent background labeling (29). Mutant L529I/L520C/C445V/C449V (henceforth called L529I L520C) constructs were generated with conventional overlap PCR using primers synthesized by Sigma Genosys (Oakville, Ontario, Canada) and sequenced using Eurofins MWG Operon (Huntsville, AL). Constructs were linearized with XbaI restriction endonuclease and cRNA was transcribed in vitro using the mMessage mMachine T7 Ultra cRNA transcription kit (Ambion, Austin, TX). Oocytes were prepared and injected with 50 nL cRNA (at concentrations of 1–30 ng/μL) as described in Cheng et al. (30).

Electrophysiology

Transfected hEK cells were grown on glass coverslips. The coverslips were placed in a 2-cc chamber, which was superfused at a rate of 2 mL/min at room temperature. The whole-cell voltage-clamp methods have been previously reported in Durdagi et al. (18). The pipette solution contained KCl 10 mM, K-aspartate 110 mM, MgCl₂ 5 mM, Na₂ATP 5 mM, EGTA 10 mM, HEPES 5 mM, and CaCl₂ 1 mM, corrected to pH 7.2 with KOH. The extracellular solution contained NaCl 140 mM, KCl 5.4 mM, CaCl₂ 1 mM, MgCl₂ 1 mM, HEPES 5 mM, and glucose 5.5 mM, corrected to pH 7.4 with NaOH. For hERG1 measurements the whole cell configuration of the patch-clamp method was used. The series resistance measured in the solution was <7 MΩ. Data were sampled at 1 kHz. The patch-clamp protocol is shown in Fig. 1 A. The holding potential was −80 mV. All tail currents reported from patch-clamp experiments represent dofetilide-sensitive currents.

Figure 1.

Slope decrease of the activation-voltage relationship for L529I after NS1643 application. Data are recorded in transfected hEK cells. (A and B, inset) Command protocol used to elicit the family of currents. Family of elicited currents for WT hERG1 (top row) versus L529I (bottom row). Mean normalized current-voltage relationships (far-right columns). Data under control conditions (●) and in the presence of 10 μM NS1643 (○) are compared. The n value for WT is 20 and for L529I is 25. Data are presented as mean ± SE. Comparing V_1/2 in control to that in NS1643; the P values were 0.00056 in WT and 0.07 in L529I. For k values, the P values were 0.08 and 0.00000009. The paired Student’s t-test was used. For statistical analysis, see Tables 1 and 3.

Voltage-dependence of activation

From a holding potential of −80 mV, cells were depolarized for 2 s to a range of voltages from –80 to +50 mV followed by a step to −120 mV to record the tail currents. The isochronal tail current-voltage plots were fit to a single Boltzmann function (18):

$$\frac{\mathit{Y}}{{\mathit{Y}}_{\text{max}}}=\frac{1}{\left(1+\text{exp}\left[\left({\mathit{V}}_{1/2}-V_m\right)/k\right]\right)}.$$ (1)

In this formulation, y (the y axis of the graph) is the normalized current for patch-clamp recordings and conductance (two-electrode voltage-clamp) or ΔF (voltage-clamp fluorimetry). V_1/2 is voltage of the half-maximal activation, and k is the slope factor. Using the Boltzmann function for steady-state activation may lead to an underestimation of the gating charge movement, z_g, during activation. Accordingly, another estimate was then used: the limiting slope of the logarithm of the open probability when channel opening is assessed near the threshold potential for channel activation. From a holding potential of −80 mV, voltage steps were applied at 2–4 mV increments at voltages negative to the threshold potential and from threshold to values +16 mV above the threshold values. Using this approach, the Z_a limit term can be calculated using the following formula (32):

$$Z_a=\underset{V_m}{\text{lim}}{k}_{\text{B}}T\frac{d\text{ln}{P}_O}{d{V}_m}.$$ (2)

Envelope of tails

The activation of hERG1 channels was examined at the V_1/2 potential in hEK cells. The protocol is shown later in Fig. 4. The cells are held at −80 mV and the membrane was stepped to the V_1/2 potential for various durations. The envelope of the tail currents was fitted with a single exponential function. The activation time delay is defined as the time where the exponential fit intersects the time axis. The data for the envelope of tails was also fit to a power function, as follows:

$$I=I_{K_{\text{inf}}}\left[{\left(1-\text{exp}\left(-t/\tau \right)\right)}^n\right].$$ (3)

The raw current and time data were processed with MATLAB (The MathWorks, Natick, MA). The MATLAB NLINFIT function was used to perform the nonlinear least-squares regression. The outcome parameters from this function were best-fit measures of τ; I_Kinf, the maximum current asymptote value; and n, the exponent. MATLAB uses the NLPARCI function to determine the 95% confidence intervals for the model parameters. The nonlinear regression was performed using data from t = 0 to the end time, which yielded the lowest mean residual error for the model fit (typically 0.3–0.5 s).

Figure 4.

Time-delay to activation in envelope of tails. Data were recorded in transfected hEK cells. The activation of the hERG1 channels at the V_1/2 potential was tested by depolarization of various durations. (A) Envelope of the tail currents was fitted to a power function with the exponent, n. N values for WT in control and NS1643 are 8, and for L529I n = 8. Baseline values for n, the exponent in the power model (see Materials and Methods), is significantly different comparing WT to L529I; however, NS1643 does not significantly affect the exponent value in either WT or L529I. (B) Fit to a power function (see Materials and Methods). The fit is established using nonlinear regression within MATLAB. Mean power function fits are n = 8 for both WT and L529I.

C-type inactivation

At potentials negative to the potassium equilibrium potential, the current deactivates very rapidly. Even at voltages of −70 to −60 mV the kinetics of the current decay is determined by both deactivation and inactivation. To minimize the contribution of contamination by deactivation, we assessed the V_0.3, a voltage where inactivation dominates. In our experiments, the relationship between the inactivation and the voltage varied substantially for the various mutations. Some mutations manifested positive shifts in the voltage dependence of inactivation, whereas others showed substantial negative shifts. Indeed, in some mutations, the maximum value of the ratio (150 ms/I_peak) was <0.5. To extract a parameter to represent the voltage dependence of inactivation (with little contamination of deactivation) in a large number of mutant channels, we used the V_0.3 measurement.

Two-electrode voltage-clamp in Xenopus oocytes

Macroscopic currents were recorded at room temperature (20–22°C) from Xenopus oocytes bathed in ND96 solution containing NaCl 96 mM, KCl 3 mM, CaCl₂ 0.5 mM, MgCl₂ 1 mM, and HEPES 5 mM, titrated to pH 7.4 with NaOH. In experiments using NS1643 (Tocris Bioscience, Minneapolis, MN), a 25 mM stock solution of the drug in DMSO (stored at −20°C) was diluted in ND96 to a final concentration of 30 μM. This concentration of NS1643 has been shown to cause near-maximal effects on wild-type (WT) hERG1 activation in oocytes (14,30,31). Microelectrodes were filled with 3 M KCl and had resistances of 0.2–2.0 MΩ. Voltage control and data acquisition were achieved using an OC-725C amplifier (Warner Instruments, Hamden, CT) connected to a Digidata 1440 interface and using pCLAMP10 software (Molecular Devices, Eugene, OR) run from a personal computer. Voltage protocols are described in the relevant text and figure legends. Current signals were acquired at a sampling rate of 10 kHz (16-bit) and low-pass-filtered at 4 kHz.

Voltage-clamp fluorimetry

The methods for fluorimetry have been previously reported in Thouta et al. (14). L520C was labeled with TMRM (tetramethylrhodamine-5-maleimide; Invitrogen, Carlsbad, CA) by bathing oocytes in a depolarizing solution: KCl 98 mM, MgCl₂ 1 mM, CaCl₂ 2 mM, and HEPES 5 mM, titrated to pH 7.4 with KOH, supplemented with 5 μM TMRM for 30 min at 10°C in the dark. Labeled oocytes were washed and stored in standard ND96 solution in the dark until voltage-clamped using an OC-725C amplifier, as described above. Fluorimetry was performed simultaneously with voltage-clamp on a TE2000S inverted microscope (Nikon, Mississauga, Ontario, Canada) outfitted with an epifluorescence attachment and photomultiplier-tube module (Cairn Research, Kent, UK) as described previously in Thouta et al. (14).

In brief, fluorescence emissions at >565 nm were collected by the photomultiplier-tube module via a 0.75 NA 20× objective lens focused on a TMRM-labeled oocyte excited by light at 525 ± 45 nm. Fluorescence signals were sampled at 10 kHz; traces represent an average of 3–15 sweeps and were filtered offline at 400 Hz. Photobleaching of the fluorophore during the activation protocol was corrected for by subtracting the fluorescence signal recorded at a constant holding potential (−130 mV) from the voltage-dependent test signals. Due to the length of the VCF protocols, experiments were not paired and recordings were made either in control ND96 solution or 30 μM NS1643. Conductance-voltage (G-V) curves were derived by plotting peak tail currents, normalized to the maximum peak tail current, as a function of the preceding voltage step. Fluorescence-voltage (F-V) curves were derived by plotting the amplitude of the fluorescence deflection (ΔF), normalized to the maximum ΔF, as a function of the voltage-step. Both G-V and F-V curves were fitted with a single Boltzmann function, as described above. Time-constants for the fluorescence deflection (F_on) during depolarization and for the return of fluorescence (F_off) at −130 mV after depolarization were obtained by fitting the relevant fluorescence reports with a single exponential function.

Statistical analysis

STATSVIEW (Abacus Concepts, Berkeley, CA) or SIGMAPLOT, Ver. 11.0 (Systat Software, San Jose, CA) were used to analyze the data. Data are presented as mean ± SE. The paired or unpaired Student’s t-test was used to compare data, with a two-tailed value of P < 0.05 designated as being significant. Nonlinear regression was used to fit the envelope of tails to a power function.

Results

L529I decreases the slope of the activation-voltage relationship in response to NS1643

Fig. 1 shows examples of a family of hERG1 currents elicited by command potentials (illustrated in the inset) before and after application of NS1643. Data are obtained from transfected hEK cells. The mean current-voltage relationships of the tail currents are shown in the right-hand panel, before (solid symbols) and after application of NS1643 (open symbols). NS1643 shifted the V_1/2 of activation by −20 ± 2 mV in WT hERG1 without alteration in the slope of the current-voltage relationship. In the drug-free state, the V_1/2 of activation of L529I hERG1 was −31 ± 3. After the application of NS1643, the V_1/2 values are very similar in WT (−37 ± 3 mV) whereas in L529I it was −34 ± 3 mV (not significant), but the slope of the current-voltage relationship was significantly and substantially flattened. Mean V_1/2 and slope factor data are shown in the right panels of Fig. 1.

Fig. 2 evaluates the impact of substitutions of various amino acids at the L529 site on the response to NS1643. Data are obtained from transfected hEK cells. Importantly, the mutations at the L529 site, at baseline, did not alter the slope factor of the current-voltage relationship. Fig. 2 D compares the mean Δ-change in response to NS1643 of the slope factors for the various mutations. L529A and L529M produced less flattening of the slope factors in response to NS1643 compared to that seen with L529I (see numerical values in Table 1). For activation, some substitutions shifted baseline V_1/2 to the depolarized potentials compared to WT (L529A, Fig. 2 A), whereas others shifted V_1/2 to the hyperpolarized potentials (L529M and L529I).

Figure 2.

Impact of substitutions of various amino acids at the L529 site. Data are recorded in transfected hEK cells. (A–C) Current-voltage relationships for various amino-acid substitutions at the L529 site are compared to WT. Data under control conditions (●) and in the presence of 10 μM NS1643 (○). (D) Δ-slope factors in response to NS1643. Data are presented as mean ± SE. The V_1/2 values and k values for each substitution are shown. (^∗) P < 0.05; (^∗∗) P < 0.01 by Student’s t-test. N values for L529A, L529M, and L529I are 5, 5, and 15, respectively. See also Tables 1 and 3.

Table 1.

Mean electrophysiological measurements recorded for hEK cells at room temperature, at baseline, and after NS1643 application (10 μM)

	k Activation					V1/2 Activation (mV)
	Baseline	NS1643	Δ	P	n	Baseline	NS1643	Δ	P	n
WT	8 ± 0	9 ± 0	1 ± 0	a	20	−17 ± 2	−37 ± 3	−20 ± 2	a	20
L529I	9 ± 0	21 ± 1	13 ± 1	a	25	−31 ± 3	−34 ± 4	−4 ± 4	NSb	25
L529A	10 ± 1	13 ± 1	3 ± 1	a	5	−7 ± 5	−20 ± 3	−13 ± 2	a,c	5
L529M	10 ± 1	12 ± 1	3 ± 1	a	5	−20 ± 2	−42 ± 3.5	−22 ± 3	a	5
K525C	18 ± 2	16 ± 2	−1 ± 2	NSb,c	5	−64 ± 7	−127 ± 6	−63 ± 10	a,b,c	5
L529I/K525C	34 ± 5	71 ± 10	38 ± 9	a,b,c,d	5	−63 ± 8	−120 ± 21	−57 ± 20	a,b,c	5
K525R	9 ± 0	24 ± 4	15 ± 4	a	5	−63 ± 4	−103 ± 3	−40 ± 3	a,b,c	5
L529I/ K525R	15 ± 1	26 ± 1	11 ± 1	a,b,c	5	−36 ± 4	−79 ± 4	−42 ± 5	a,b,d	5
R528C	11 ± 1	14 ± 1	3 ± 1	a	7	3 ± 1	−33 ± 4	−35 ± 4	a,c	7
L529I/R528C	12 ± 1	22 ± 2	10 ± 2	a	9	−28 ± 3	−59 ± 2	−31 ± 4	a,b,d	9
I567L	8 ± 0	9 ± 0	2 ± 0	a	5	−3 ± 3	−29 ± 4	−26 ± 6	a	5
I567L/L529I	9 ± 1	14 ± 1	6 ± 1	a,c	7	−23 ± 2	−41 ± 4	−18 ± 5	a,d	7

The paired Student’s t-test was used for comparison between baseline and NS1643. Otherwise, ANOVA analysis was used. Electrophysiological responses to NS1643 are shown as mean ± SE; n, number of experiments; NS, not significant compared to baseline.

^aP < 0.05 change compared to baseline.

^bP < 0.01 compared to WT.

^cP < 0.05 compared with L529I.

^dP < 0.05 comparing a double mutation to its individual component mutation other than L529I.

Fig. 3 shows the raw tail current data and mean ΔZ_a (Eq. 2) calculations for current-voltage relationships recorded near the threshold potential, comparing WT hERG1 to L529I, before and after NS1643 treatment. In the drug-free state, the Z_a values were similar in WT and L529I hERG1. For WT, the Z_a values were 4.2 ± 0.1 before and 3.9 ± 0.2 after NS1643. For L529I, the Z_a values were 4.4 ± 0.2 before and 3.3 ± 0.3 after NS1643. The ΔZ_a values due to NS1643 were −0.3 ± 0.1 for WT and −1.1 ± 0.2 for L529I (P < 0.05). Compared to L529A and L529M, only the L529I mutation produced a large change in ΔZ_a after treatment with NS1643. These data suggest that NS1643 significantly alters a voltage-dependent transition in the pathway to activation in L529I hERG1. The n values are 5 for WT and 5 for L529I.

Figure 3.

Activation near threshold. Representative data are recorded in transfected hEK cells. The family of currents (A and B) and the current-voltage relationships of tail currents (C and D) elicited by the command potential (inset, A) are shown for WT in the upper row (A and C) and L529I (B and D) in the bottom row. Baseline and data after NS1643 are shown. From a holding potential of −80 mV, voltage steps were applied at 2–4 mV increments at voltages negative to the threshold potential and from threshold to values +16 mV above the threshold values. Data under control conditions (●) and in the presence of 10 μM NS1643 (○) are compared. (E) The Z_a values are calculated and the ΔZ_a in response to NS1643 are compared. n = 5 for WT and n = 5 for L529I. (^∗) P < 0.05 by Student’s t-test.

To further evaluate the pathway to opening of the activation gate, the envelope of tails was elicited by the command potentials shown in Fig. 4. A representative example of the envelope of tails is shown in Fig. 4 B. The character of activation gating was analyzed by fitting the activation process to a power function. In the power function, the exponent, n, was the modeled value. Metrics were compared in WT and L529I hERG1 records before and after the application of NS1643. No significant change was noted in the NS1643-induced change in the kinetics of activation in either WT or L529I currents (P values given in the Fig. 4).

NS1643 binding alters voltage-sensor movement in L529I mutant channels

To support the notion that NS1643 alters voltage-sensor transitions before opening of the activation gate, voltage-clamp fluorimetry experiments were performed to track the voltage-sensor movement comparing L529I before and after treatment with NS1643 (Fig. 5). These experiments were performed in Xenopus oocytes where the effect of NS1643 on L529I mutant channels was qualitatively similar to that observed in hEK cells (compare Figs. 1 and 5 C). In oocytes, 30 μM NS1643 increased the slope factor from 8.9 ± 0.4 to 12.9 ± 1.2 (n = 5; paired data; P < 0.01) and left-shifted the G-V curve (the V_1/2 shifted from −46.4 ± 0.7 to −52.0 ± 2.0; P < 0.05). To perform fluorescence measurements of S4 movement, TMRM is covalently attached to L520C in the S3-S4 linker as previously shown in Thouta et al. (14), Cheng et al. (30). Fig. 5, A and B, shows typical ionic current and fluorescence data recorded at a range of potentials from TMRM-labeled L529I L520C channels in the absence and presence of 30 μM NS1643. Fig. 5 D shows that NS1643 shifted the V_1/2 of ionic current activation from −42.3 ± 2.0 mV to −50.5 ± 1.7 mV (n = 10; paired data; P < 0.01) and increased the slope factor, k (i.e., reducing the slope), from 13.5 ± 0.6 mV to 15.3 ± 0.5 mV (P < 0.01).

Figure 5.

Effects of NS1643 on the fluorescence report of voltage-sensor movement in L529I L520C and L520C channels. (A) Ionic current families recorded from one Xenopus oocyte expressing hERG L529I L520C mutant in control ND96 solution (●) and with 30 μM NS1643 (○). The cell was held at −100 mV and 2-s voltage steps were applied in 10-mV steps between −130 and +50 mV; tail currents were measured at −130 mV. (Arrows) Zero current level; current traces were not leak-subtracted. (B) Fluorescence signals recorded at different potentials from a TMRM-labeled cell in control solution (●) and from another cell in the presence of NS1643 (○). (C) G-V curves derived using standard tail current analysis (see Materials and Methods) from nine oocytes expressing L529I channels under control conditions and with NS1643. (D) G-V curves obtained from 10 TMRM-labeled oocytes expressing L529I L520C channels under control conditions and with NS1643. (E) The effects of NS1643 on the F-V curve of L529I L520C channels. For each cell, the amplitude of the downward fluorescence deflection (ΔF) at the end of each voltage pulse was normalized to the peak ΔF. The data represent 12 (control) and 13 (NS1643) experiments. (F) The G-V relationships obtained from eight TMRM-labeled oocytes expressing L520C channels under control conditions and with NS1643. (G) F-V relationships of L520C channels under control conditions (n = 12) and with NS1643 (n = 11). Voltage protocols used to collect data for (B)–(G) were the same as described for A. All data in (C)–(G) are presented as mean ≥ SE. (Solid and dashed lines) Fits to the Boltzmann function (see Materials and Methods). (^∗) Statistical significance compared to control data. See text and Table 2 for fit parameters.

Compared to L529I, the reduced effect of NS1643 in TMRM-labeled L529I L520C channels appears to be due to the L520C mutation, because the magnitude of the effect of NS1643 was similar in L529I L520C channels with and without TMRM (Table 2). Fig. 5 shows that the V_1/2 of the F-V curve was not significantly changed (P > 0.05) by drug binding; however, the slope factor of the F-V relation was significantly increased by NS1643 (from 20.2 ± 0.2 mV to 22.9 ± 0.5 mV; n = 12 and 13; unpaired data; P < 0.01). In control experiments, using TMRM-labeled L520C channels (Fig. 5 F), the V_1/2 of ionic current activation was left-shifted (P < 0.01) from −26.4 ± 1.2 mV in ND96 to −33.0 ± 1.9 mV in the presence of NS1643 (n = 8; paired data), similar to the effect observed in WT channels expressed in mammalian cells (Fig. 1 A). However, NS1643 had no effect (P > 0.05) on the slope factor of the L520C G-V curve (11.8 ± 0.2 mV to 12.6 ± 0.3 mV). In control fluorescence experiments, NS1643 had no significant effect on either V_1/2 or k values in L520C channels (Fig. 5 G and Table 2). These data suggest that NS1643 reduces the voltage-sensitivity of activation (increases the slope factor) in L529I channels via an effect on voltage-sensor movement, whereas the shift in the steady-state voltage dependence appears to result from an effect later in the activation pathway.

Table 2.

Summary of mean fit parameters for data collected from Xenopus oocytes, at baseline and after NS1643 application (30 μM)

	k (mV)			V1/2 (mV)			n
	Control	NS1643	P	Control	NS1643	P	Control	NS1643
GV curves
L529I	8.9 ± 0.4	12.9 ± 1.2	a	−46.4 ± 0.7	−52.0 ± 2.0	b	5 (paired)
L529I L520C	8.2 ± 0.3	10.1 ± 0.4	a	−37.3 ± 2.1	−45.7 ± 2.1	a	9 (paired)
L529I L520C with TMRM	13.5 ± 0.6	15.3 ± 0.5	a	−42.3 ±2.0	−50.5 ± 1.7	a	10 (paired)
L520C with TMRM	11.8 ± 0.2	12.6 ± 0.3	NS	−26.4 ± 1.2	−33.0 ± 1.9	a	8 (paired)
FV curves
L529I L520C	20.2 ± 0.5	22.9 ± 0.5	a	−44.5 ± 2.7	−48.9 ± 2.3	NS	12	13
L520C	17.8 ± 0.5	16.1 ± 0.9	NS	−24.8 ± 2.1	−28.2 ± 2.8	NS	12	11

Data are shown as mean ± SE; NS, not significant compared to control.

^aP < 0.01 compared to control.

^bP < 0.05 compared to control.

The time-course for the onset of the fluorescence deflection (F_on) during the pulse was not altered by the presence of NS1643 in L529I L520C. At +60 mV, the τ-value of F_on was 54 ± 7 ms in control (n = 12) and 52 ± 4 ms with NS1643 (n = 13); and at 0 mV, the corresponding values were 81 ± 7 and 60 ± 4 ms (P > 0.05). This suggests that the effects of the drug on the slope of the G-V and F-V relationships of the L529I L520C construct are unlikely related to an effect on activation kinetics. Interestingly, the τ-value for the return of fluorescence (F_off) at −130 mV in L529I L520C channels increased significantly from 118 ± 12 ms (n = 12) under control conditions to 217 ± 24 ms (n = 13, P < 0.05) in 30 μM NS1643. The influence of NS1643 on F_off suggests that the drug slows the rate of voltage-sensor return upon repolarization. These fluorescence studies suggest that the presence of NS1643 within its binding pocket alters activation and deactivation gating of hERG1 L529I channels by directly modifying voltage-sensor transitions.

Mutations in the S4 helix that alter the pharmacologic response to NS1643

Tables 3 and 4 show the mean raw data at baseline, after treatment with NS1643 and the Δ NS1643-induced changes in deactivation τ values and in the V0.3 inactivation for inactivation for all of the mutations in the neighborhood of the S4 helix. The conductance values for each of these mutations are presented in Table 4. To further explore whether an interface between NS1643 and charged and uncharged elements within the S4 could alter the pharmacologic response in the L529I channel, a series of mutations at sites in the neighborhood of L529 in the S4 helix were made. The mutations evaluated are shown in Fig. 6 E. We found that K525C has no impact on the slope factor of activation. However, the double mutation, K525C/L529I, even in the absence of NS1643, produced a very substantial reduction in the slope factor of activation. When NS1643 was applied to K525C/L529I channels, the slope factor was dramatically flattened with a slope factor of >70. The V_1/2 was also dramatically shifted to hyperpolarized potentials. In additional experiments, flattening of the slope factor in response to NS1643 was also more modestly flattened in the K525R mutation. These data indicate that not only the charge of the amino acid at K525 but also its size is a determinant of the response to NS1643. The slope factor of the double mutation K525R/L529I in the drug-free state mimics the effect of NS1643 in the L529I single mutant channel. In contrast, mutations at R528 had no substantial effects. These data point to an important interface of NS1643 with L529 and K525 that could partially shield this important charged amino acid in the voltage-sensor. In the reported hERG1 homology models, L529 and I567 are physically close together (18–20), thus we explored other potential hydrophobic interactions. However, the double mutation L529I/I567L did not substantially alter the slope factor of activation seen during NS1643 treatment.

Table 3.

Mean electrophysiological measurements recorded for hEK cells at room temperature, at baseline, and after NS1643 application (10 μM)

	Deactivation (τ, ms)					V0.3 inactivation
	Baseline	NS1643	Δ/baseline	P	n	Baseline	NS1643	Δ/baseline	P	n
WT	129 ± 13	264 ± 31	1.0 ± 0.2	a	9	−25 ± 4	−19 ± 4	6 ± 1	a	6
L529I	104 ± 10	189 ± 10	0.8 ± 0.2	a	18	−23 ± 3	−12 ± 4	11 ± 3	a	11
L529A	56 ± 11	80 ± 9	0.4 ± 0.1	a	5	−18 ± 4	−12 ± 3	9 ± 5	NS	4
L529M	329 ± 62	403 ± 27	0.2 ± 0.3	NSb	5	−28 ± 14	−35 ± 3	−7 ± 12	NS	5
K525C	213 ± 18	319 ± 25	0.5 ± 0.2	a,b	5	−33 ± 2	−24 ± 3	9 ± 3	a	4
L529I/K525C	80 ± 14	255 ± 19	2.2 ± 0.5	a,c	8	−11 ± 3	−4 ± 4	7 ± 4	NS	10
K525R	120 ± 12	223 ± 19	0.9 ± 0.3	a	4	−26 ± 2	−19 ± 7	8 ± 5	NS	4
L529I/ K525R	56 ± 6	173 ± 27	2.1 ± 0.2	a	5	−23 ± 4	−4 ± 6	19 ± 3	a	4
R528C	289 ± 39	388 ± 56	0.3 ± 0.1	a,b,c	8	−26 ± 1	−14 ± 2	11 ± 2	a	8
L529I/R528C	89 ± 6	183 ± 14	1.1 ± 0.2	a,d	7	−11 ± 4	−9 ± 5	3 ± 2	NS	6
I567L	74 ± 16	160 ± 40	1.1 ± 0.1	a	5	−26 ± 4	−20 ± 4	7 ± 4	NS	9
I567L/L529I	106 ± 9	231 ± 25	1.1 ± 0.3	a	5	−17 ± 2	−4 ± 2	13 ± 2	a	18

The paired Student’s t-test was used in comparison between baseline and NS1643. Otherwise, ANOVA analysis was used. The V_0.3 inactivation is described in Materials and Methods. Electrophysiological responses to NS1643 are shown as mean ± SE. Δ for deactivation designates Δτ (ms)/baseline τ (ms); n, number of experiments; NS, not significant compared to baseline.

^aP < 0.05 change compared to baseline.

^bP < 0.01 compared to WT.

^cP < 0.05 compared with L529I.

^dP < 0.05 comparing a double mutation to its individual component mutation other than L529I.

Table 4.

Mean maximum channel conductances recorded for hEK cells at room temperature, at baseline, and after NS1643 application (10 μM)

	Conductance (nS)
	Baseline	NS1643	NS1643/baseline	P	n
WT	285 ± 51	324 ± 59	1.1 ± 0.1	NS	20
L529I	134 ± 14	157 ± 16	1.2 ± 0.1	NSa	25
L529A	182 ± 27	230 ± 36	1.3 ± 0.1	a	5
L529M	140 ± 17	145 ± 15	1.1 ± 0.1	NS	5
K525C	99 ± 20	115 ± 27	1.2 ± 0.1	NS	5
L529I/K525C	52 ± 12	81 ± 16	1.6 ± 0.4	NS a	5
K525R	179 ± 33	195 ± 31	1.1 ± 0.1	NS	5
L529I/ K525R	292 ± 39	326 ± 45	1.1 ± 0.1	NS	5
R528C	167 ± 19	215 ± 38	1.3 ± 0.1	NS	7
L529I/R528C	50 ± 9	97 ± 19	1.9 ± 0.3	a,b	9
I567L	162 ± 48	215 ± 54	1.3 ± 0.1	b	5
I567L/L529I	149 ± 31	157 ± 31	1.1 ± 0.1	NS	7

The paired Student’s t-test was used in comparison between baseline and NS1643. Otherwise, ANOVA analysis was used. The maximum channel conductances was calculated as maximum tail currents divided by the driving forces assuming the reversal potential of −80 mV, shown as mean ± SE. Statistics P were calculated as NS1643 compared to baseline, with the paired Student’s t test used; n, number of experiments; NS, not significant compared to baseline.

^aP < 0.01 compared to WT.

^bP < 0.05 change compared to baseline.

Figure 6.

Molecular site(s) of NS1643 mediated change in voltage-sensor movement. Data are recorded in transfected hEK cells. Representative examples of current-voltage relationships (A, K525C, ●; (B), L529I/K525C, ○; and (C), L529I/K525C after addition of 10 μM NS1643, □). (D) Average Boltzmann fitting results of K525C, L529I/K525C, and L529I/K525C in 10 μM NS1643 yielded V_1/2 values (mV) of −66 ± 3.8, −56.8 ± 5, and −120 ± 21; respectively, and k values (mV) of 16.4 ± 1.5, 37.4 ± 3.2, and 71.4 ± 10.2, respectively (n = 10, 10, and 5, respectively). (E) The slope factors of mutations at baseline and with 10 μM NS1643. N values for each group are shown above respective bars. (α) P < 0.01 compared to WT. (β) P < 0.01 versus L529I. (γ) P < 0.01 versus its single partner mutation. (^∗) P < 0.01 versus the same mutation at baseline. (δ) P < 0.01 in the presence of NS1643 compared to L529I. The paired Student’s t-test was used in comparison between baseline and NS1643. Otherwise, analysis of variance (ANOVA) was used. See also Tables 1 and 3. For paired mutation groups, n = 20, 25, 5, 5, 5, 5, 7, 9, 5, and 7.

MD and structural analysis for L529I mutant

To better understand the effect of the L529I mutation on the structure of hERG1 channel, we performed all-atom MD simulations with previously developed models of hERG1 channel in different conformational states. The MD data shows that the conservative mutation L529I resulted in a minimal perturbation to the hERG1 homology model global structure, but caused local changes in the voltage-sensing domain (VSD). The RMSD analysis from MD simulations for both model states (i.e., open and closed) shows that the mutation decreases flexibility of the VSD structure (Fig. 7 A). For closed and open states, the L529I mutation also had an effect on the helicity by introducing a kinked S4 helix (Table S1 and Fig. 7 B). That change affects the packing of the VSD, and also the structure of helix S3, especially in the closed state. The persistent contacts of L529 for WT and I529 in the mutant were analyzed for open and closed state models. The several salt-bridge interactions that form spontaneously within the VSD during simulations were also monitored for both states for WT and mutant. As expected, the conservative L529I mutation introduced only minimal perturbations. A complete analysis of the amino-acid packing is presented in Figs. S3–S9. One interesting observation is that the L529I substitution tends to affect stability of the key salt-bridges involving K525 and in particular, D466-K525. For WT, the distribution of distances spans to larger values indicating significant flexibility around this site. L529I not only shifts the average distance for this salt-bridge to shorter values, it also eliminates water-mediated salt-bridging.

Figure 7.

Summary of MD simulations for closed and open state homology models of WT and L529I hERG1. (A) RMSD for all subunits, showing the backbone fluctuations for open and closed states of WT hERG1 (top) and L529I (bottom). WT is depicted in open state (black) and closed state (blue), L529I is depicted in open state (gray), and closed state (light blue). (B) Structure comparison between the S4 helix (in red) of WT (top) and L529I (bottom) for closed and open states of hERG1. Only one subunit is shown, and the S1 segment is omitted for clarity. To see this figure in color, go online.

NS1643 binding sites from docking studies

To evaluate the binding mode of NS1643 to WT and L529I variants of hERG1, we performed high-precision docking to MD-refined structures of the WT and mutant forms of the channel. NS1643 was docked to closed and open state WT and L529I mutant structures extracted from clustering of MD trajectories. We considered kinked and straight helical structures for our docking studies of open and closed states in the case of mutant. Importantly, MD simulations of WT do not display any kinked structures for simulations of open and closed states, while L529I simulations display a predominantly kinked structure of S4. Three potential binding sites in the vicinity of L529 (interface between helices S1, S2, S3, S4, and S5 of the adjacent subunit) were explored in the BS1, BS2, and BS3 sites (Figs. S10, S11, and S12). All of them have similar binding affinities (see Table S2), rendering identification of a unique binding site difficult.

It is quite possible that the NS1643 drug binds to all these binding sites simultaneously with similar affinity, and that the L529I mutation is affecting the binding not through a direct interaction with the drug but by inducing a structural change that affects the different binding sites to differing extents. It is also worth noting that additional work extending simulations other than docking studies will help to resolve specificity for each of the sites better. The BS2 site was previously described as IC2 (18), and was successfully used recently for the rational design, chemical synthesis, and experimental evaluation of new hERG1 activators. For the purpose of this study, we focused on the BS1 site (see Fig. 8), where the drug is in proximity with L529 and K525, the two key residues identified experimentally. The drug interacts directly with K525 in the open state. This interaction is completely obliterated in the closed state, where K525 is salt-bridging to D466. Although it is possible that this salt-bridge could be affected by the drug binding to the closed state, further studies such as MD including the drug are necessary to prove that. For both the open and closed cases, the docking shows that the mutant binding site is rearranged when the S4 helix is kinked and the binding affinity for NS1643 is a little higher compared to the straight-less-curved S4 helix in WT (see Table S2).

Figure 8.

NS1643 binding site in the vicinity of L529 and K525 (BS1) from docking to closed and open state hERG1 homology models. (A) Docking of NS1643 to open (right, gray) and closed (left, blue) homology models states of WT and L529I hERG1. The best docked structures to the BS1 binding sites are shown. (B) Diagram for NS1643 interactions in the BS1 binding sites for open (right) and closed (left) states. Residues are represented as colored spheres with the residue name and number; the color indicates the residue type. Interactions with the protein are marked with lines between the ligand atoms and residues. The protein pocket is displayed with a line, colored with the color of the nearest protein residue. Gaps in the line show the accessibility of the binding pocket. The cutoff used to display the interactions was 4.5 Å. To see this figure in color, go online.

Discussion

The main innovative findings of this study are that various substitutions at the L529 residue can shift the baseline V_1/2 of activation to the left (L529I; −31 ± 3 mV) or to the right (L529A; −7 ± 5 mV) compared to WT (−17 ± 2 mV), but after treatment with NS1643, all substitutions at the L529 site flatten the IV curve (Tables 1 and 3). Moreover, the extent of flattening of the slope factor by NS1643 for L529I was significantly greater than seen with the other substitutions (Table 1). In WT channels, we propose that NS1643 is coordinated in a putative binding site in proximity to L529 and K525 (BS1, Fig. 8). We propose that binding of NS1643 to this putative binding site changes its topology destabilizing a salt-bridge between K525 and D466, which consequently destabilizes the resting state of the voltage sensor. In the baseline drug-free state, the L529I mutation was not associated with a flatten-the-slope factor (9 ± 0.4 mV with L529I compared to 8 ± 0.2 mV with WT), but in the presence of NS1643, the slope factor for activation of L529I is significantly flattened (22 ± 1.5 mV for L529I versus 9 ± 0.3 mV for WT). Moreover, when NS1643 was applied to K525C/L529I channels, the slope factor was very dramatically flattened and shifted to hyperpolarized potentials. NS1643 shifts the V_1/2 of activation of both K525C and K525C/L529I to hyperpolarized potentials (−131 ± 4.4 mV for K525C and −120.4 ± 21.3 mV for K525C/L529I), but the response (Δ) of the slope factor to NS1643 for K525C/L529I was +38 ± 9.1 vs. −1 ± 2.3 mV for K525C, P < 0.001). For the double mutation I567L/L529I, the baseline value (9 ± 1 was similar to L529I (9 ± 0), but after NS1643 the slope factor of the double mutation (14 ± 1) was less than that seen with L529I (21 ± 1; P < 0.05). These data suggest that the I567L mutation may modify the spatial interaction of NS1643 with L529. The shallow slope of the G-V relationship is indicative of reduced voltage sensitivity, and this is supported by our limiting slope analysis, which shows that ΔZ_a was reduced by NS1643 in L529I, but not WT channels. Modeling studies of ionic current data suggest that the activation pathway in hERG1 channels involves progression through two voltage-dependent steps separated by a voltage-independent step that becomes rate-limiting at strongly depolarized voltages (10,33). This scheme was subsequently modified, based on gating current data, to detail two consecutive and independent voltage-sensitive transitions within each α-subunit of the channel that occur before the voltage-independent step (12). Despite these studies, understanding of hERG1 activation gating lacks the depth of that for Shaker channels and this can challenge mechanistic interpretation of data.

A power analysis of the envelope of tails data indicates that the exponent of the power function, n, was significantly changed in the baseline condition in L529I compared to WT. These data indicate that the L529I mutation at baseline appears to alter the extent of the cooperativity between subunits in the pathway toward activation. However, NS1643 did not significantly alter the exponent values in either WT or L529I. So, while there may be evidence for a change in cooperativity, it does not appear to explain the difference in the slope factor observed during NS1643 when comparing WT to L529I. These data are consistent with the idea that NS1643 alters movements of the voltage-sensor before opening of the activation gate. The envelope of tails does not interrogate the movements of the voltage-sensor; instead, it assesses kinetics of opening of the activation gate. To assess movements of the voltage-sensor, a fluorimetric reporter of movements of the S4 was used. In an accompanying article, a kinetic model is used to elucidate state-dependent transitions on the path toward opening of the activation gate (34).

The combined approach of analyzing ionic currents, fluorimetric measurement of voltage-sensor movement, and molecular dynamic simulations, is the strength of this study. Our data indicates that NS1643 alters the voltage-sensitivity of movements of the voltage-sensor or alters the potential cooperative movements of the voltage-sensor that precede the activation gate opening. It appears that NS1643 may partially shield the K525 residue or alter the allosteric tertiary structure of the voltage sensor during its movements. Our examination of the response of the S4 to quite conservative substitutions, K525R and L529I, suggests that both size and tertiary structure as well as charge/or hydrophobicity influence the response to NS1643. It appears that it may not just be shielding of charged residues but that NS1643 may produce an allosteric tertiary structure change of the voltage-sensor during its transitions toward opening of the activation gate. Interestingly, a 2013 study showed functional interactions of K525 with a putative gating-charge transfer center formed by F463 and D466 that stabilize the closed state (30). In Shaker, the highly conserved phenylalanine (F290) forms a hydrophobic plug that acts to shuttle S4 positive charges across the membrane during gating. Therefore, one possibility is that NS1643 interferes with the interactions of K525 with the gating-charge transfer center to destabilize the resting state of the voltage sensor and reduce charge movement upon activation. Consistent with this, parallel recording of fluorescence signals, which reflect movement of the voltage sensor, confirm that NS1643 alters movement of the voltage sensor in L529I channels.

Fluorimetry reporter of movement of the S4 voltage sensor

Voltage-clamp fluorimetry has been used widely to report upon conformational dynamics of voltage sensors in numerous channels (31,35–43) including hERG1 (14,29,36–39,44). Fluorophores attached at L520C in the S3-S4 linker of hERG1 channels report slow depolarization-induced environmental changes that mirror the time course and voltage dependence of ionic current. Consistent with the notion that fluorophores labeling the hERG1S3-S4 linker report specifically upon voltage-sensor movement, it was reported in 2012 that perturbations altering pore opening did not affect the fluorescence signal (35). Thus, it appears that fluorophores attached in the S3-S4 linker of hERG1 track the bulk of the charge movement that is relatively slow, a conclusion that is supported by gating current measurements (12,35). Here, we report the effects of NS1643 binding on the dynamics of voltage-sensor movement in hERG1 L529I channels. These data suggest that NS1643 reduces the voltage sensitivity of voltage-sensor movement and slows the return of the voltage sensor to its resting configuration upon repolarization. These data suggest that the presence of NS1643 within its binding pocket alters activation and deactivation gating of hERG1 channels by directly modifying voltage-sensor transitions.

MD simulations and binding of NS1643: L529I induced structural rearrangements in the transmembrane S4 domain

The MD data shows that in both closed and open states of hERG1, the L529I mutation increases rigidity of the local elements and affects helicity by introducing a kinked S4 helix. That change affects the packing of the VSD and the structures of helix S2 and S3, especially in the closed state (Fig. 7). In addition to this, docking studies suggest a potential binding site in the vicinity of L529 and K525 (BS1, Fig. 8; and see Figs. S11 and S12). Taking these results together, we suggest that the L529 residue is a key element in determining the VSD local structure and packing, and it is involved in NS1643 binding not through direct interactions with the drug but by regulating the appropriate structure of the binding sites. Additionally, the K525 residue interacts directly with the drug in the open state, probably stabilizing it. This interaction is missing in a closed channel, illuminating significance of the K525-NS1643 interaction for stabilization of the open state. One hypothesis suggested here is that the drug could also affect this salt-bridge interaction, destabilizing the closed state and thus favoring the channel opening. This not only provides a tentative explanation for observed experimental results, but also testable hypotheses for a kinetic modeling. In the accompanying article (34), we performed kinetic modeling of currents using experimental data to test different gating schemes in the presence of NS1643 and to connect to atomistic simulations. Kinetic models allow isolation of the gating transition explicitly affected by the presence of NS1643.

Amino acids in the transmembrane S4 domain are candidates for an interaction site that mediates the hERG1 opener pharmacologic activity of NS1643

In our previous work, we suggested the presence of multiple potential binding sites for NS1643 in hERG1 (18). One potentially important binding domain of hERG1 may be within the S4 in the neighborhood of L529. Mutations producing some of the greatest impact on the pharmacologic response to NS1643 exist in the S4 domain. For example, the leftward shift in activation by NS1643 in the T526M channel was −59 ± 9 mV compared to −18 ± 3 mV in WT hERG1. The mutation of L564A in the S4 abrogates the ability of NS1643 to increase the magnitude of the tail current. NS1643 abrogates the inward conductance seen in a mutation located at the cytosolic junction of the S4 segment, D540K (18). These data suggested that NS1643 prevented the inward movement of the S4 segment of the D540K channel. The results of this study provide further evidence suggesting that NS1643 interacts to alter function of the S4 and its movements before opening of the activation gate. The fluorescence studies also suggest that presence of NS1643 within its binding pocket alters activation and deactivation gating of hERG1 channels by modifying voltage-sensor transitions. Finally, amino acids L529 and K525 of hERG1 are candidates for an interaction site that mediates the hERG1 opener pharmacologic activity of NS1643. The presence of NS1643 in this putative binding site appears to alter activation and deactivation gating of hERG1 channels by modifying voltage-sensor transitions.

Conclusions

A potential hypothesis that explains coupling between drug binding and channel gating that is consistent with our data from a multitude of methods is schematically summarized in Fig. 9. In this hypothetical model, the strength of the salt-bridge between D466 and K525 is diminished as the hydrophobic molecule, NS1643, interdigitates with this domain in the WT channel. The presence of NS1643 in this site would diminish the latching effect of the salt-bridge in the closed state and allow the electromotive force to more easily move the voltage sensor, while stabilizing the open state through a direct interaction with K525. We recognize that this is a simplistic representation of a complex conformational change occurring in the voltage-sensor movement before the opening of the activation gate. To better understand the effect of the mutation and the NS1643 drug in terms of a kinetic mechanism, Markov models were used to describe the gating kinetics in an accompanying study. Consistent with what is proposed here, the modeling of ionic current data shows that the drug is affecting the early transitions to the open state and captures the dramatic perturbation to these early transitions caused by the L529I mutation (Perissinotti et al. (34)).

Figure 9.

Cartoon representation of how the interaction of NS1643 and key residues can influence voltage-sensor movement in open and closed states in WT and L529I. The structure and packing of the voltage-sensing domain (VSD) is affected by the L529I mutation and it is shown in different colors for WT (green) and L529I (blue). The interaction with NS1643 (magenta star) is shown for only one VSD subunit in the respective darker color. In this model, the strength of the salt-bridge between D466 and K525 present in the closed state (charges highlighted in red) is diminished as the hydrophobic molecule, NS1643, interdigitates with this domain (charges highlighted in orange). In the open state, the drug is directly interacting with K525 (interaction highlighted in yellow) probably stabilizing the open state over the closed. To see this figure in color, go online.

Author Contributions

J.G. performed and analyzed ionic voltage-clamp studies and revised the manuscript. Y.M.C. performed and analyzed voltage-fluorimetry studies and revised the manuscript. T.C. directed voltage-clamp fluorimetry studies and wrote the manuscript. L.L.P. performed molecular dynamic simulations, analysis and structure clustering, molecular docking studies, and wrote and revised the manuscript. S.D. helped with the molecular docking studies and revised the manuscript. J.P.L.-M. designed and performed site-directed mutagenesis and helped revise the manuscript. S.Y.N. directed molecular modeling and wrote and revised the manuscript. H.J.D. conceived the idea and wrote the manuscript.