Introduction and Motivation

This project week was carried out in the group of Dr. Bert de Groot in the Department of Theoretical and Computational Biophysics of Prof. Helmut Grubmüller at the Max Planck Institute for Biophysical Chemistry. It was supervised by Dr. Ulrich Zachariae. The goal was to set up a new molecular dynamics simulation box for the hERG potassium channel in a lipid membrane environment to be used for further studies on hERG structure and function.

Since my main thesis work deals with the investigation of structure and function of a chimeric potassium channel, KcsA-Kv1.3, by solid-state NMR methods, and since my group has close ties and collaborations with the Grubmüller department and the de Groot group, for me, the project week in this group was an obvious choice. I was able to learn more about and get hands-on experience in molecular dynamics simulations, which are of great importance for NMR and structural biology in general, and work on a system related to my own thesis work.

Background

Ion channels are membrane proteins that allow for selective transport of ions across the membrane in response to specific stimuli such as ligand binding or transmembrane voltage [1]. A large number of structural and functional studies on ion channels are available, especially on potassium channels such as the voltage-gated Shaker channel from the fruit fly Drosophila melanogaster and its relatives [2], and on the bacterial potassium channel KcsA from Streptomyces lividans, the first ion channel whose crystal structure was solved [3]. However, numerous questions remain open, such as the exact mechanisms by which these channels enter a so-called inactivated state after an initial activation [4], and the three-dimensional structures of many channels have not been solved so far.

One important example is the hERG (human ether-a-go-go related gene) potassium channel, which is responsible for the slow repolarization of human ventricular myocytes, i.e. heart muscle cells [5]. Its inactivation properties cause a prolonged depolarization phase of the action potential in the heart, which is crucial for optimal coupling of excitation and muscle contraction and also for avoiding premature re-excitation that can lead to possibly fatal arrhythmia. Several mutations, both gain- and loss-of-function, in the hERG gene can cause arrhythmia, and a wide variety of drugs has been found to block hERG, potentially causing arrhythmia, ventricular fibrillation and death. Consequently, there is great medical and pharmacological interest in understanding hERG structure and function and predicting whether a drug will interact with it. However, the structure of hERG could not be solved so far experimentally, highlighting the need for homology modeling and molecular dynamics simulations.

Fig. 1: The homology model of the hERG potassium channel obtained from Schrödinger, Inc. [6]. The voltage sensors were removed.

Molecular Dynamics Setup and Simulations

To obtain a simulation box for a membrane protein, it has to be inserted in a model lipid bilayer. The box is then filled up with water molecules. These tasks can be accomplished using the Gromacs [7] molecular dynamics simulation software suite (www.gromacs.org), which is later also used for the full MD simulations of the protein in a lipid environment. Additionally, for generating a hole in a model lipid bilayer, the "make_hole" patch for Gromacs available from the Gromacs website was used, as well as some custom scripts written in the Perl or Ruby languages for tasks such as removing surplus lipid molecules from a PDB file.

Fig. 2: Initial lipid bilayer solvated in water.

The first step was to install the Gromacs molecular dynamics software suite as well as FFTW, a library for discrete Fourier transformation used by Gromacs (www.fftw.org), on a Linux computer to be used for the simulation box setup. The PDB (Protein Data Bank) files of a hERG homology model [6] (Schrödinger, Inc., New York, www.schrodinger.com) and of a POPC lipid bilayer (obtained from the website of the group of Peter Tieleman at the University of Calgary) were used as starting points. To prepare the Gromacs run that actually generated the membrane hole, both protein and membrane were aligned to the same coordinates within their respective PDB boxes. The pore and lower cavity of hERG were filled with dummy atoms to prevent lipid molecules from slipping in. Additionally, a number of lipid molecules were removed from the center of the lipid bilayer, since Gromacs make_hole only pushes molecules away from the desired position of the inserted protein, leading to an undesired increase in membrane lateral pressure if the number of lipid molecules is kept constant. The space which is to be kept free of lipids is defined by the surface of the inserted protein, which was calculated for the hERG structural model including the dummy atoms by the program MSMS.

The membrane hole is generated by a restrained MD run on the prepared lipid bilayer PDB file using Gromacs. To do so, a number of files have to be assembled, most importantly: a topology file for the simulated system containing information on, e.g., the force field used, the number and type of molecules in the system, and the topology of these molecules; a simulation parameter file with the MD protocol of the simulation; and a restraint file to keep the z position of the lipid molecules approximately constant. The Gromacs command grompp then combines these files and the starting PDB file into a single .tpr file containing all information needed to do the actual MD run, which is executed by the mdrun command.

Fig. 3: Lipid bilayer with hole for hERG insertion.

To prepare the hERG protein for insertion into a solvated lipid bilayer, it has to be solvated itself, especially since hERG is an ion channel, i.e. it contains a (water-filled) pore. Solvation of hERG is accomplished using the Gromacs editconf and genbox commands, followed by a short energy minimization and a position-restrained MD run (using appropriately modified Gromacs em.mdp and pr.mdp simulation parameter files for .tpr file generation) to equilibrate the system. After that, all water molecules outside the ion channel pore in the resulting PDB file are removed to prepare it for membrane insertion.

A few problems were encountered during this process, most notably a pressure problem with the membrane patch in the make_hole MD run, caused by an initially too low number of water molecules for the size of the simulation box. Also, the position of the protein with respect to the membrane was initially not optimal, and the membrane patch originally contained water molecules modeled according to the SPC model, whereas the OPLS-AA force field used in our simulations demands tip4p water molecules. After correcting the protocol for all these issues, a membrane patch with a hole for the internally hydrated hERG molecule was obtained. The size of the membrane simulation box was slightly extended to accommodate hERG, and the hydrated protein was inserted by simply pasting its PDB file into that of the membrane patch. The resultant simulation box was filled completely with water molecules, excess water (in the membrane) was removed, a new topology file for the whole system was generated, and another energy minimization (em.mdp) and position-restrained MD run (pr.mdp) were performed to obtain the final energy-minimized simulation box of the hERG channel in a lipid membrane.

Conclusions

Fig. 4: Final simulation box with hERG inserted in the lipid bilayer.

A simulation box for the hERG potassium channel homology model in a lipid bilayer environment was successfully created using the Gromacs molecular dynamics simulation package. Thus, apart from learning the basics of MD simulation setup, I could contribute to the current work going on in the de Groot group by setting up a simulation system that, in the meantime, has already been used for full-scale MD simulations to further elucidate the function of hERG.

Acknowledgements

Many thanks to Dr. Ulrich Zachariae for his expert supervision, continuous availability for all my questions, and the very nice working atmosphere during this project week. I am also very thankful to Dr. Bert de Groot for suggesting this project to me and for allowing me to do my project week in his group.

References