Structural modeling of the hERG potassium channel and associated drug interactions

Abstract

The voltage-gated potassium channel, K_V11.1, encoded by the human Ether-à-go-go-Related Gene (hERG), is expressed in cardiac myocytes, where it is crucial for the membrane repolarization of the action potential. Gating of the hERG channel is characterized by rapid, voltage-dependent, C-type inactivation, which blocks ion conduction and is suggested to involve constriction of the selectivity filter. Mutations S620T and S641A/T within the selectivity filter region of hERG have been shown to alter the voltage dependence of channel inactivation. Because hERG channel blockade is implicated in drug-induced arrhythmias associated with both the open and inactivated states, we used Rosetta to simulate the effects of hERG S620T and S641A/T mutations to elucidate conformational changes associated with hERG channel inactivation and differences in drug binding between the two states. Rosetta modeling of the S641A fast-inactivating mutation revealed a lateral shift of the F627 side chain in the selectivity filter into the central channel axis along the ion conduction pathway and the formation of four lateral fenestrations in the pore. Rosetta modeling of the non-inactivating mutations S620T and S641T suggested a potential molecular mechanism preventing F627 side chain from shifting into the ion conduction pathway during the proposed inactivation process. Furthermore, we used Rosetta docking to explore the binding mechanism of highly selective and potent hERG blockers - dofetilide, terfenadine, and E4031. Our structural modeling correlates well with much, but not all, existing experimental evidence involving interactions of hERG blockers with key residues in hERG pore and reveals potential molecular mechanisms of ligand interactions with hERG in an inactivated state.

Keywords: Rosetta; drug block; hERG channel; inactivation; potassium channel.

FIGURE 1

hERG WT channel model 1 (m1) and model 2 (m2) comparison. (A) Domain structure of hERG WT channel. Chains are colored according to the legend as CI-CIV. PD, pore domain; VSD, voltage sensing domain. Chain I is colored in blue, chain II is colored in yellow, chain III is colored in green, and chain IV is colored in violet. (B) Overlay of hERG WT m1 and m2 models. M1 model is in translucent light-blue cartoon with purple residues, m2 model in noodle wireframe and color coded according to RMSD, with orange residues. (C) Van der Waals pore radius comparison for hERG WT m1 and m2 models. Volume spheres are in blue, with surrounding residues in light blue. (D) hERG WT m1 (purple) and m2 (blue) pore radius, R, profiles as a function of z generated by the HOLE program (Smart et al., 1996).

FIGURE 2

hERG WT drug docking results for (A) dofetilide, (B) terfenadine and (C) E4031. Boxes show results of ligand docking to hERG WT model 1 (“WT m1”) and model 2 (“WT m2”) for neutral (“(0)”) and charged (“(1)”) form of each ligand (dofetilide, terfenadine, and E4031). Boxes represent top 50 scoring structures, with top-scoring structure as orange sticks (for carbon atoms) and remaining structures as blue centroids representing central nitrogen/charge group of the respective compound. Oxygen atoms are colored red, nitrogen—blue, sulfur—yellow. Hydrogens are not shown for clarity. Residue contacts to top-scoring structures are shown as sticks and labeled in bold red; residue contacts for all other ligands are shown as gray wires. Chains are colored as in Figure 1. Rosetta ligand interface score for each top-scoring structure is shown in Rosetta energy units (REU) in the bottom left corner of each box and summarized in Supplementary Table S1. Protein-ligand interaction profiles (PLIP) for top-scoring poses are shown in lower panels (below boxes) and oriented in approximately the same way. Key contact residues are labeled in black and correspond to bold red labels above. Key residues on protein chains that have been removed for clarity may not appear in upper panels (boxes) but are shown in lower panels for PLIP analysis.

FIGURE 3

hERG S641A model 1 (m1) and model 2 (m2) comparison. (A) Domain structure of hERG S641A. Chains are colored according to legend. PD, pore domain; VSD, voltage sensing domain. (B) Overlay of hERG S641A m1 models using the radial (rad, pink) and regional (reg, green) movemaps. WT hERG model structure is in gray. Single domain overlay on the left shows the 8th top-scoring structure for S641A m1 (top-scoring for all other models), whereas inset details on the right show top 50 representative structures. Top structure residues in inset detail are shown as sticks, with remaining shown as gray wires. Mutant residues are labeled in bold red. Chain I is colored in blue, chain II is colored in yellow, chain III is colored in green, and chain IV is colored in violet. (C) Overlay of hERG S641A m2 models using the radial (rad, pink) and regional (reg, green) movemaps. Depictions are identical to (C) (D) hERG WT m1 and hERG S641A m1 model pore volume profile comparison. Volume spheres are in blue, with surrounding residues in light blue. (E) hERG S641A m1 fenestration model top-view, with detail inset rotated by 90°. Fenestration tunnel (purple spheres) are generated by MOLE (Pravda et al., 2018). Fenestration-lining residues are labeled and shown in stick, with chains colored as before. (F) Pore radii profiles are generated by the HOLE program for hERG S641A m1 radial (blue) and regional (orange), WT m1 (gray) models.

FIGURE 4

hERG S641A drug docking results for (A) dofetilide, (B) terfenadine and (C) E4031. Boxes represent top 50 scoring structures, with top-scoring structure as orange sticks (for carbon atoms) and remaining structures as blue centroids representing central nitrogen/charge group of the respective compound. Atoms, residues, and chains are shown and colored as in Figure 2. Rosetta ligand interface score for each top-scoring structure and PLIP for top-scoring poses are shown as in Figure 2 and summarized in Supplementary Table S1.

FIGURE 5

hERG S641T model 1 (m1) and model 2 (m2) comparison. (A) Domain structure of hERG S641T mutant models. Chains are colored according to the legend. PD, pore domain; VSD, voltage sensing domain. Chain I is colored in blue, chain II is colored in yellow, chain III is colored in green, and chain IV is colored in violet. (B) Overlay of hERG S641T m1 models using the radial (rad, pink) and regional (reg, green) movemaps. WT model structure is in gray. Single domain overlay shows top-scoring structures, whereas inset details show top 50 structures. Top structure in inset detail is shown as stick, with remaining structures shown as gray wires. Mutant residue is labeled in bold red. (C) Overlay of hERG S641T m2 models using the radial (rad, pink) and regional (reg, green) movemaps. Depictions are identical to (B) (D) Comparison of residue 641 (labeled bold red) and associated residue positions surrounding the selectivity filter (SF) for hERG WT and S641T m1/m2 models. Top row: top view of hERG channel pore, bottom row: side-view of S/T641 interaction details. Residues are shown as spheres and colored as in Figure 2. (E) Pore radii generated by the HOLE program for hERG S641T m1 radial (blue) and regional (orange) as well as WT m1 (gray) models.

FIGURE 6

hERG S641T drug docking results for (A) dofetilide, (B) terfenadine and (C) E4031. Boxes represent top 50 scoring structures, with top-scoring structure as orange sticks (for carbon atoms) and remaining structures as blue centroids representing central nitrogen/charge group of the respective compound. Atoms, residues, and chains are shown and colored as in Figure 2. Rosetta ligand interface score for each top-scoring structure and PLIP for top-scoring poses are shown as in Figure 2 and summarized in Supplementary Table S1.

FIGURE 7

hERG S620T model 1 (m1) and model 2 (m2) comparison. (A) Domain structure of hERG S620T. Chains are colored according to legend. PD, pore domain; VSD, voltage sensing domain. Chain I is colored in blue, chain II is colored in yellow, chain III is colored in green, and chain IV is colored in violet. (B) Overlay of hERG S620T m1 models using the radial (rad, pink) and regional (reg, green) movemaps. WT model structure is in gray. Single domain overlay shows top-scoring structures, whereas inset details show top 50 structures. Top structure in inset detail is shown as sticks, with remaining shown as gray wires. Mutant residue is labeled in bold red. (C) Overlay of hERG S620T m2 models using the radial (rad, pink) and regional (reg, green) movemaps. Depictions are identical to (B) (D) Comparison of residue 620 (labeled bold red) and associated residue positions surrounding the SF in hERG WT and S620T models. Top row: detailed side-view of residue 620 interactions, bottom row: top view of hERG channel pore. Residues are shown as spheres and colored according to their respective chains (see legend in (A) and Figure 2). (E) Pore radii, R, profiles along the z axis are generated by the HOLE program for hERG S620T m1 radial (blue) and regional (orange), WT m1 (gray) models.

FIGURE 8

hERG S620T drug docking results for (A) dofetilide, (B) terfenadine and (C) E4031. Boxes represent top 50 scoring structures, with top-scoring structure as orange sticks (for carbon atoms) and remaining structures as blue centroids representing central nitrogen/charge group of the respective compound. Atoms, residues, and chains are shown and colored as in Figure 2. Rosetta ligand interface score for each top-scoring structure and PLIP for top-scoring poses are shown as in Figure 2.

FIGURE 9

Relative stability of WT and mutant models using the (A) radial and (B) regional movemaps. Single domain representations are shown on left with regions specified as movable by movemap depicted in blue, and key residues labeled. RMSD: C_α root-mean-square deviation in Angstroms. RMSD for m1 and m2 mutant models were measured against m1 and m2 WT hERG channel models, respectively. Error bars represent Max and Min values; Mean (x); Median (horizontal bar).

FIGURE 10

Interface score distributions for hERG channel interactions with dofetilide. Left Panels: Box-and-whisker plots of interface scores in REU. Right Panels: Probability density charts plotting probability density against interface scores (REU) All plots utilize 2,000 generated structures. Plots are defined for (A) neutral (0) model 1, (B) charged (1) model 1, (C) neutral (0) model 2 and (D) charged (1) model 2 structures. Error bars represent Max and Min values; Mean (x); Median (horizontal bar).

FIGURE 11

Interface score distributions for hERG channel interactions with terfenadine. Left Panels: Box-and-whisker plots of interface scores in REU. Right Panels: Probability density charts plotting probability density against interface scores (REU). All plots utilize 2,000 generated structures. Plots are defined for (A) neutral (0) model 1, (B) charged (1) model 1, (C) neutral (0) model 2 and (D) charged (1) model 2 structures. Error bars represent Max and Min values; Mean (x); Median (horizontal bar).

FIGURE 12

Interface score distributions for hERG channel interactions with dofetilide. Left Panels: Box-and-whisker plots of interface scores in REU. Right Panels: Probability density charts plotting probability density against interface scores (REU). All plots utilize 2,000 generated structures. Plots are defined for (A) neutral (0) model 1, (B) charged (1) model 1, (C) neutral (0) model 2 and (D) charged (1) model 2 structures. Error bars represent Max and Min values; Mean (x); Median (horizontal bar).