Elucidation of molecular mechanisms of sex-based arrhythmias

Female sex has been shown to be an independent risk factor for both inherited and acquired heart rhythm abnormalities, such as long QT syndrome (LQTS) and associated arrhythmias. Notably, female sex is a key element in up to 70% prevalence of drug-induced acquired LQTS, However, fundamental molecular mechanisms that explain this phenomenon are not well understood. Previous experimental and clinical studies suggested that it is likely related to differential levels of sex hormones (estradiol, progesterone and testosterone) playing opposite roles in pro-arrhythmia proclivities, exacerbating or mitigating effects of mutations or drugs on cardiac ion channels.  In the American Heart Association (AHA) sponsored career development award 19CDA34770101 "Elucidation of molecular mechanisms of sex-based arrhythmias" we focused on hormone interactions with the human Kv11.1 potassium channel (encoded by the hERG - human Ether-à-go-go-Related Gene), a major contributor to cardiac action potential repolarization and an anti-target for diverse drug molecules. We performed a comprehensive set of in silico atomistic modeling and simulations on hERG structure and function modulation by sex hormones in combination with hERG channel blockers with different proclivities for arrhythmogenesis.  These studies were informed by and will also guide electrophysiological experiments on cardiomyocytes and hERG-expressing HEK cells by our collaborators.  Molecular dynamics (MD) simulation and molecular docking data, presented here and validated by electrophysiological recordings, will provide us with quantitative estimates of such hormone modulatory effects and will be used for elucidation of molecular mechanisms of sex-dependent heart rhythm abnormalities and thus the ways to combat them via rational design of sex-specific cardiac safe pharmaceuticals and/or their dose adjustments. The dataset contains MD simulations of hERG channel - hormone and drug interactions as well as their analyses. Multi-microsecond-long umbrella sampling MD simulations as well as fragment-based Site-Identification by Ligand Competitive Saturation (SILCS) docking results were performed and presented here. 

Methods
The dataset was collected using molecular dynamcis (MD) simulations and fragment-based Site-Identification by Ligand Competitive Saturation (SILCS) molecular docking calculations of the wild-type hERG potassium channel model based on the cryogenic electron microscopy (cryo-EM structure) with PDB ID: 5VA2. Pore and voltage sensing domain residues 405-668 were used in the model. Rosetta structural modeling was used for de novo missing loop building.

hERG channel models in a ~260 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid bilayer soaked by 0.15 M aqueous KCl solution and a  hormone molecule initially placed  in aqueous solution under .  All simulations were run using NAMD 2.13 or later in the NPT ensemble at 310 K and 1 atm pressure using Nose-Hoover thermostat and Langevin piston barostat. Standard cutoff scheme and particle mesh Ewald (PME) will be used for non-bonded interactions. Standard CHARMM biomolecular force fields (C36 for lipids, CHARMM36m for protein and standard CHARMM ion parameters) and TIP3P water model were used for compatibility with our previous studies. One equilibration and 5 steered MD (SMD) “pulling” 90-ns long simulations were performed for each hormone (estradiol and dihydrotestosterone) using our previously developed protocol . The final frames from these simulations was be used as starting points for umbrealla sampling MD (US-MD) runs The simulation length was chosen to reach a well equilibrated system based on our prior experience. We needed to perform at least 5 SMD simulations with different initial conditions to avoid a bias in initial drug orientations in subsequent US-MD runs.

For US-MD simularions, we used hERG channel model in a POPC membrane solvated by 0.15 M aqueous KCl + 1 drug placed at different z positions along the channel axis in 0.5 Å intervals.  The same force field and general MD simulation parameters as described above was used. US-MD simulations with harmonic restraints on drug center of mass (COM) with respect to Ca COM of hERG selectivity filter (SF) (624SVGFG) residues were performed to compute drug free energy and diffusion coefficient profiles across the channel pore. Weak 0.2 kcal/mol/Å2 positional restraints for the SF backbone and the whole pore domain Ca atoms were used as was done in our previous simulations to preserve a channel conformational state and also minimize their effect on drug binding.  We used 90 US-MD simulation windows covering a range -50 £ z £ -5.5 Å, from the bottom of the SF and down through the pore, extending far enough into the solvent to get bulk-like free energy and diffusion coefficient values. Starting structures were taken from the 5 SMD simulations described above, choosing one of them at random for each z position to avoid an initial bias. 40 ns per window with an initial ~10 ns as equilibration, equivalent to amount of sampling used in similar studies  and in our previous calculations, where 40 ns per US-MD window was barely sufficient to reach a desired convergence. 

For SILCS simulations using 2022.2 software we used hERG channel models in a ~220 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) & 13 cholesterol lipid bilayer soaked by 0.15 M aqueous KCl solution along with multiple molecules of 9 different co-solvents (methanol, benzene etc.).  All simulations were run using Gromacs 2020.2 or later in the NPT ensemble at 310 K and 1 atm pressure. Standard cutoff scheme and particle mesh Ewald (PME) were used for non-bonded interactions. Standard CHARMM biomolecular force fields (C36 for lipids, CHARMM36m for protein and standard CHARMM27 ion parameters) and TIP3P water model were used for compatibility with our previous studies . Ten 100-ns MD simulations will be performed for each channel model as dictated by established SILCS protocol. Each 1 ns run is intervened by 200,000 steps of Grand Canonical Monte Carlo (GCMC) for co-solvent molecule insertion/deletion and translations/rotations. This was followed by SILCS MC docking of drug and/or hormone molecules into the hERG channel pore. Top-scoring (most favorable binding free energy) hERG channel - drug structures are available for further analysis.