Case Study: Remdesivir analog GS-441524 and model human cell membrane

Molecular dynamics-based investigation of interactions between Remdesivir analogue GS-441524 and a model human cell membrane using LAMMPS

   As a case study for the Innovation Lab project we have decided to explore a topic somewhat related to the ongoing COVID-19 Pandemic. This said, in the current race against SARS-CoV-2, interest has been drawn to Remdesivir, an antiviral approved for emergency treatment of COVID-19 in many countries. Remdesivir is a prodrug to the pharmacologically active GS-441524 and it transforms into the latter promptly after reaching the bloodstream (Scavone et al., 2020).

    The purpose of this project was to predict whether GS-441524 is capable of entering a typical human cell, namely the cell of a lung wall. Since GS-441524 is a relatively hydrophilic molecule structurally resembling adenosine, we hypothesised that it might enter human cells through nucleoside transporters. To be more precise, through the human Equilibrative Nucleoside Transporter – 1, which transports purine and pyrimidine nucleosides, including adenosine (Beal et al., 2004). The first simulation was run using a cell membrane containing hENT1 and a certain number of GS-441524 molecules on one side of the membrane. In addition, a second simulation was run using the same membrane fragment and the same number of GS-441524 molecules, but without hENT1.            

    The first LAMMPS (Plimpton, 1995) data file was prepared using the membrane builder function (Wu et al., 2014) provided by CHARMM-GUI at www.charmm-gui.org (Jo et al., 2008). To create the membrane, a transmembrane protein hENT1 from www.rsbc.org (Berman et al., 2000) with ID 6OB7 was inserted into a lipid bilayer consisting of 102 POPC phosphatidylcholine, 10 DSPS phosphatidylserine, 30 DSPE phosphatidylethanolamine, 4 DSPA phosphatidic acid, 20 POPI13 phosphatidylinositol, 10 PSM sphingomyelin, and 4 PVCL2 cardiolipin lipids, as well as 20 cholesterol molecules. The membrane lipid composition was designed to approximately resemble a general eukaryotic cell (Vance, 2014). The aqueous solution was designed to contain 0.140 M KCl, with extra K+ ions to neutralize the charges in the system. Then, by applying the CHARMM-GUI ligand builder & modeler (Kim et al., 2017), a template of GS-441524 was created. Using the generated template, a LAMMPS data file containing 5 GS-441524 molecules was created using CHARMM-GUI solution builder. Then, both LAMMPS data files were combined in the LAMMPS input file. In this particular simulation, all GS-441524 molecules were concentrated bellow the membrane. After minimizing the energy and forces, the simulation was carried out in 310.15 K using the NPT ensemble. It was ran for 200000 1.0 fs timesteps and resulted in no molecules permeating to the other side of the membrane.

 

    The second LAMMPS simulation was carried out with the same number of GS-441524 molecules, using a model eukaryote membrane with the same lipid composition, but lacking hENT1. It was ran in 310.15 K using the NPT ensemble for 164690 1.0 fs timesteps and resulted in no GS-441524 molecules permeating to the other side of the membrane.

 

    To conclude our findings, GS-441524 was not able to permeate through the membrane with, and without hENT1. However, a more accurate investigation should be carried out to confirm our results. Our simulations had some limitations that are worth to consider. For instance, the membrane model might not be the most accurate in terms of its lipid composition, as well as its ionic makeup. Both simulations also had limited accuracy due to a lack of computation time. Although there were virtually no signs of GS-441524 dissolving in the lipid bilayer during the second run, some GS-441524 molecules seemed to be closely interacting with hENT1 in the first run. There is a large possibility that after extending the designated computation time, some GS-441524 molecules would have been able to pass through the membrane with the assistance of hENT1.

    To watch the whole simulation, click on the image below.

Sources:

Beal, P., Yao, S., Baldwin, S., Young, J., King, A. and Cass, C., 2004. The equilibrative nucleoside transporter family, SLC29. Pflügers Archiv European Journal of Physiology, 447(5), pp.735-743.

 

Berman, H., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T., Weissig, H., Shindyalov, I. and Bourne, P., 2000. The Protein Data Bank. Nucleic Acids Research, 28(1), pp.235-242.

 

Jo, S., Kim, T., Iyer, V. and Im, W., 2008. CHARMM-GUI: A web-based graphical user interface for

CHARMM. Journal of Computational Chemistry, 29(11), pp.1859-1865.

 

Kim, S., Lee, J., Jo, S., Brooks, C., Lee, H. and Im, W., 2017. CHARMM-GUI ligand reader and modeler for CHARMM force field generation of small molecules. Journal of Computational Chemistry, 38(21), pp.1879-1886.

 

Plimpton, S., 1995. Fast Parallel Algorithms for Short-Range Molecular Dynamics. Journal of Computational Physics, 117(1), pp.1-19.

Scavone, C., Brusco, S., Bertini, M., Sportiello, L., Rafaniello, C., Zoccoli, A., Berrino, L., Racagni, G., Rossi, F. and Capuano, A., 2020. Current pharmacological treatments for COVID‐19: What's next?. British Journal of Pharmacology.

 

Vance, J., 2014. Phospholipid Synthesis and Transport in Mammalian Cells. Traffic, 16(1), pp.1-18.

 

Wu, E., Cheng, X., Jo, S., Rui, H., Song, K., Dávila-Contreras, E., Qi, Y., Lee, J., Monje-Galvan, V., Venable, R., Klauda, J. and Im, W., 2014. CHARMM-GUIMembrane Buildertoward realistic biological membrane simulations. Journal of Computational Chemistry, 35(27), pp.1997-2004.