Probing binding sites and mechanisms of action of an I(Ks) activator by computations and experiments

Abstract

The slow delayed rectifier (IKs) channel is composed of the KCNQ1 channel and KCNE1 auxiliary subunit, and functions to repolarize action potentials in the human heart. IKs activators may provide therapeutic efficacy for treating long QT syndromes. Here, we show that a new KCNQ1 activator, ML277, can enhance IKs amplitude in adult guinea pig and canine ventricular myocytes. We probe its binding site and mechanism of action by computational analysis based on our recently reported KCNQ1 and KCNQ1/KCNE1 3D models, followed by experimental validation. Results from a pocket analysis and docking exercise suggest that ML277 binds to a side pocket in KCNQ1 and the KCNE1-free side pocket of KCNQ1/KCNE1. Molecular-dynamics (MD) simulations based on the most favorable channel/ML277 docking configurations reveal a well-defined ML277 binding space surrounded by the S2-S3 loop and S4-S5 helix on the intracellular side, and by S4-S6 transmembrane helices on the lateral sides. A detailed analysis of MD trajectories suggests two mechanisms of ML277 action. First, ML277 restricts the conformational dynamics of the KCNQ1 pore, optimizing K(+) ion coordination in the selectivity filter and increasing current amplitudes. Second, ML277 binding induces global motions in the channel, including regions critical for KCNQ1 gating transitions. We conclude that ML277 activates IKs by binding to an intersubunit space and allosterically influencing pore conductance and gating transitions. KCNE1 association protects KCNQ1 from an arrhythmogenic (constitutive current-inducing) effect of ML277, but does not preclude its current-enhancing effect.

Figure 1

ML277 (ML) increases native I_Ks amplitude in adult ventricular myocytes and works from inside the cell membrane. (A) Tail I-V relationships of I_Ks in guinea pig ventricular myocyte, KCNQ1/KCNE1 (Q1/E1), or KCNQ1 (Q1 alone) expressed in COS-7 cells before and after application of ML (1 μM). Inset: voltage-clamp protocol. The open arrow in the right panel of A points to the apparent constitutive current component. (B) Normalized tail I-V relationships from the same cells shown in A. For each cell, the relationship between tail current (I_tail) and test pulse voltage (V_t) was fit with a simple Boltzmann function, I_tail = I_max/(1 + exp((V_0.5 − V_t)/k)), to estimate the maximal tail current (I_max), half-maximum activation voltage (V_0.5), and slope factor (k). The fraction activated (= I_tail/I_max) is plotted against V_t, superimposed on a curve calculated from the Boltzmann function. Insets: current traces before and after application of ML with V_h, V_t, and V_r marked. (C) Left: 1 μM ML shortened the APD of a guinea pig ventricular myocyte. Inset: ML structure. Middle: time course of changes in I_Ks amplitude in a guinea pig ventricular myocyte before, during, and after application of ML, followed by HMR1556 (HMR). Right: time course of changes in Q1 current amplitude expressed in an oocyte before and after intra-oocyte injection of ML, followed by TEA injection (estimated cytoplasmic concentrations based on assumed oocyte volume of 0.5 μl). Insets: current traces from time points marked by asterisks in the time courses. The following color scheme is used for all panels: gray, control; black, ML; dark gray, HMR or TEA. Light-gray shading in this and subsequent figures marks the time point of current measurement.

Figure 2

Docking ML to the Q1 and Q1/E1 models, and MD simulations. (A) Top 100 ML binding poses in Q1 (left), the E1-free cleft space of Q1/E1 (middle), and the E1-occupied cleft space of Q1/E1 (right). Q1 and E1 are shown as gray and purple ribbons, respectively. ML molecules are shown in rainbow colors. R1 and R2 demarcate binding regions 1 and 2. (B and C) 3D plots of the center of mass of ML molecules, with x-y plane projections, during 100 ns MD simulations. Four ML molecules were docked to Q1 in R1 or R2 (Q1_ML_R1 and Q1_ML_R2), and two ML molecules were docked to the equivalent positions in the E1-free cleft space of Q1/E1 (Q1/E1_ML_R1 and Q1/E1_ML_R2). The x and y axes are parallel and the z axis is normal to the plane of lipid bilayer. (D) RMSD of Q1 C_α atoms (top) and all-atoms of ML during MD trajectories. The same color scheme for ML molecules is used in B–D.

Figure 3

Up to 4 ML binding sites per Q1 channel, but fewer ML binding sites in Q1/E1. (A) Top: superimposed current traces from Q1 and Q1/E1 expressed in COS-7 cells, and Q1 expressed in oocytes, elicited by the voltage-clamp protocols diagrammed in the insets, under control conditions, and the steady-state effects of increasing concentrations of ML. Current amplitudes were measured at the end of the tail currents (open arrows). Bottom: ratio of current amplitude in the presence of ML to control (I_ML/I_C) plotted against ML concentrations ([ML], logarithmic scale) from the same cells as shown on top. Current traces and data points are coded according to the color scheme shown on the right. (B) Left: data summary. For each cell, response to ML was normalized to between 0 (control) and 1 (maximal ML effect, 10 μM in COS-7 expression and 50 μM in oocyte expression), and averaged over cells (n = 5, 3, 4, respectively). The values of fraction of maximum effect are plotted against [ML] superimposed on curves calculated from the modified Hill equation (top right). Bottom right: parameter values for the Hill equation.

Figure 4

Model prediction of ML binding space in the Q1 and Q1/E1 models. (A) Degree of ML contacts with Q1 during MD trajectories, with ML docked to Q1 (top) or E1-free space of Q1/E1 (bottom). Calculation of the degree of contacts (expressed in arbitrary units (A.U.)) is described in the Supporting Material. Transmembrane segments (S1–S6) and the pore (P) loop are highlighted by gray shading. Q1 residues tested in subsequent mutagenesis experiments are marked along the abscissa. (B) Three views of an ML binding space in the Q1 model. For clarity, only helices and loops that are directly involved in ML binding are shown (S2–S5 of one subunit (light gray ribbons) and S5-S6 of the adjacent subunit (dark gray ribbons)). The semitransparent surface represents Q1 residues predicted to interact with bound ML molecules. Eleven ML binding poses are included for illustration. Movie S1 provides a 3D view of the ML binding space in the context of the complete Q1 homology model.

Figure 5

Impact of mutating Q1 side chains predicted to interact with bound ML molecule on ML’s current-enhancing effect. (A) Representative current traces of Q1-WT and mutants expressed alone (top) or coexpressed with E1 (bottom) in COS-7 cells, under control conditions and the steady-state effects of 1 μM ML (gray and black traces, respectively). Left inset: voltage-clamp protocol. V_t is marked in each panel. (B) Data summary. Mutant data that differ from Q1-WT are marked by gray-black histogram bars.

Figure 6

ML and PIP₂ binding sites overlap and the cell membrane PIP₂ level affects ML potency. (A) Top: side view of Q1 homology model with 100 PIP₂ binding poses (eight in stick format with rainbow colors, and 92 as semitransparent structures). Bottom: the same view of Q1 with eight representative ML binding poses in R1, and the 100 PIP₂ poses as semitransparent structures (enlarged view in the yellow box). (B and C) Top: superimposed current traces recorded from Q1 alone or coexpressed with ci-VSP in oocytes, elicited by the diagrammed voltage-clamp protocol under the control conditions and in the presence of 50 μM ML (black and red traces, respectively). Bottom: test pulse I-Vs from the same oocytes under control conditions and in 50 μM ML. (D) Comparison of I_ML/I_C without versus with ci-VSP coexpression. Q1 current amplitudes were measured as peak tail currents at −40 mV after 2 s pulses to +60 mV.

Figure 7

ML and E1 restrict Q1 pore dimension and optimize K⁺ conductance. (A) Top: representative tail current traces of Q1, Q1 treated with 25 μM ML, and Q1/E1 expressed in oocytes. Currents were recorded at −80 mV after a 2 s pulse to +60 mV. The bath solution contained 98 mM K⁺ (black traces) or 98 mM Rb⁺ (magenta) as the main cations. Bottom: summary of the Rb⁺ to K⁺ conductance ratio (G_Rb/G_K), calculated as [I_Rb/(−80 − E_Rb)]/[I_K/(−80 − E_K)], where I_Rb and E_Rb are the peak tail current amplitude and reversal potential in 98 mM Rb⁺, and I_K and E_K are the corresponding values in 98 mM K⁺. ^∗∗∗p < 0.001. (B) Top: pore radii of KcsA crystal structure (PDB ID: 1K4C) and KcsA snapshot during MD simulations (modified from Shrivastava and Sansom (30)). Middle: average pore radii of Q1 alone or Q1 with four ML molecules in region 1 or region 2 (Q1_ML_R1 and Q1_ML_R2). Bottom: average pore radii of Q1 alone or Q1/E1 (reproduced from Xu et al. (9)). Zero on the pore axis represents the C_α position of Thr in the signature sequence (TVGYG of KcsA, TIGYG of Q1). S1–S4: four K⁺ binding sites in the selectivity filter of KcsA crystal structure. Arrows in the middle and bottom panels indicate narrow regions in the averaged Q1 pore dimensions. (C) Pore domain snapshots of Q1, Q1_ML_R1, Q1_ML_R2, and Q1/E1. For clarity, only two diagonal Q1 subunits are shown (gold ribbons). Purple spheres, K⁺ ions. (D) Cartoon of working hypothesis (details in text).

Figure 8

(A and B) Effects of ML on gating kinetics of Q1 expressed in oocytes (A) or COS-7 cells (B). In all panels, control data are shown in gray and data in ML are shown in black. Top: time constant (τ) of deactivation plotted on a logarithmic scale against repolarizing voltage. Tail currents after the initial hooked phase were fit with a single-exponential function. Inset: tail currents at −80 mV as open circles superimposed on single-exponential fit. Bottom: τ of activation plotted on a logarithmic scale against depolarizing voltage. The activation phase of Q1 was fit with a single-exponential (oocyte) or double-exponential (COS-7) function with sigmoidal delay. Inset: current at +20 mV as open circles superimposed on single- or double-exponential fit. Enlarged views of the gray-shaded areas highlight ML-enhanced sigmoidal delay in Q1 activation (arrows).

Figure 9

PCA of Q1 backbone movements associated with ML or E1 binding. (A) Top to bottom: first principal component of Q1 C_α displacements calculated from single MD trajectories of Q1 and Q1/E1, and from combined MD trajectories of Q1+Q1_ML_R1, Q1+Q1_ML_R2, and Q1+Q1/E1. The values of C_α displacements in the four Q1 subunits are averaged and plotted as mean with SE bar against the Q1 position number. The transmembrane helices (S1–S6) and P loop are highlighted by gray shading. Horizontal bars along the abscissa denote the three regions highlighted in B. (B) Intracellular side view of Q1 with the first principal component of C_α displacements in the S2-S3 loop, S4-S5 helix, and S6_CT marked by green-white-red lines. A green-to-red transition signifies the direction of C_α movement. The length of the line signifies the magnitude of movement. Yellow and orange arrows indicate the general directions of movements in the VSD and PD, respectively. Channel domains are marked in the Q1 panel. The approximate E1 positions are indicated by dotted circles in the Q1+Q1/E1 panel.

Figure 10

Comparison of the ML binding site with those of other KCNQ activators. Top: sequence alignment between (KCN)Q1 and (KCN)Q2 in S2–S6 helices and the P loop. Color shading highlights equivalent Q1 and Q2 residues that are important for the effects of NH29 (blue), TMS (cyan), ztz240 (green), R-L3 (pink), ZnPy (purple), and retigabine (red). Center: side view of a partial Q1 homology model, showing helices involved in drug binding (S2–S5 of one subunit and S5-S6 of an adjacent subunit). Golden mesh, ML binding space; colored spheres, residues highlighted in the sequence alignment. Left and right: structures of KCNQ activators. Atom colors: carbon, gray; nitrogen, blue; oxygen, red; hydrogen, white; sulfur, yellow; Hg, cyan; Zn, olive green; Cl, green; F, light blue.