Biophysical Properties of Zinc-Chloroquine Complexes from Multiscale Simulations
Local Project ID:
HPC Platform used:
SuperMUC-NG at LRZ
The pandemic of Covid-19, the viral infection caused by SARS-CoV-2, started in Wuhan in December 2019. Since then, it has spread leading to almost 633 million infected patients today, with more than 6.5 million deaths as reported by WHO of November 17, 2022.
Major routes to combat the virus include the development of vaccines and drugs. The application of vaccines has been shown to be highly effective in fighting the pandemic. Yet, apart from the difficulties associated with their worldwide distribution, the efficacy is not guaranteed for new viral mutants which may pose an even higher risk of infection. Therefore, it is imperative to foster the discovery of antiviral drugs. One strategy to quickly identify therapeutic agents that can be used in the clinics is to use U.S. Food and Drug Administration (FDA) approved drugs, whose toxicity is well known and controlled. Additionally, some of these drugs can be used in combination with other chemicals. One of them is the zinc ion, which is non-toxic at limited concentrations. Zinc ions have shown antiviral properties and inhibit in vitro proteins from specific members of the coronaviruses family .
Zinc ions have shown antiviral properties, but a key issue for their use for antiviral therapy is its difficulty, as a divalent metal ion, to cross the cell membrane and thus reach its targets inside the cell. A variety of ligands, including the FDA approved drug chloroquine (CQ), form complexes with these ions and have been proposed to assist zinc permeation , possibly promoting the combined beneficial action of both zinc ions and the drugs against the virus. Here, we studied the permeation of chloroquine and the interaction of the drug with zinc ions in aqueous solution. For the latter, we take advantage of highly scalable ab initio molecular dynamics simulations to explore the diverse coordination chemistry of zinc ions.
The permeation of chloroquine through a lipid bilayer was studied by well-tempered metadynamics enhanced sampling simulations. On SuperMUC-NG, ~180 ns/day could be achieved on 96 cores, exploiting the MPI parallelized GROMACS molecular dynamics package patched with PLUMED.
The simulations provided a detailed picture of the structure and the energetics of the permeating drugs. In addition, they turn out to be fully consistent with experimental data collected several decades ago. Finally, they led to a detailed understanding of the complex dehydration process of the drug upon permeating the membrane.
Next, we investigated how this molecule binds to Zn(II) ions, which display antiviral properties. The resulting zinc-chloroquine complexes in aqueous solution were studied by ab initio molecular dynamics simulations, employing the CPMD program package . The density functional theory-based electronic structure is evaluated within a plane wave basis set in combination with norm-conserving pseudopotentials, from which the system is propagated in time via a Born-Oppenheimer scheme. The CPMD implementation uses processor groups to efficiently perform this task. The parallelization within such processor groups is realized through MPI and OpenMP for inter-node and intra-node communication, respectively. Besides the usage of standard GGA functionals (BLYP), we aimed to study zinc complexes also at the more accurate hybrid functional DFT level (B3LYP). The inclusion of exact exchange is computationally extremely demanding, but also highly scalable. On SuperMUC-NG, we were able to efficiently use up to 6144 cores for these simulations, requiring ~15 min per MD step
NMR studies show that zinc binding to chloroquine occurs at its aromatic nitrogen. Thus, the drug’s aliphatic side chain bonded to the exocyclic nitrogen was modeled as methyl group to decrease dramatically the system size. The remaining coordination environment of the zinc ion can be composed of water molecules and/or chloride ions under physiological conditions. The ab initio molecular dynamics simulations provide insight on the coordination number and the nature of the ligands in water solution. This information may be useful in understanding how Zn(II) ions and chloroquine may act together.
Despite the high-performance applicability of ab initio molecular dynamics simulations, the sampling time is still limited in the picosecond timescale. A modern route to extend the simulation at quantum chemical accuracy is the usage of neural network potentials. Recently, machine learning potentials trained on ab initio data, like in the DeepMD framework, have been proposed , which can be used to parametrize the bonded energy terms of isolated molecules or extended systems based on ab initio reference data. We explore this methodology for zinc-chloroquine complexes, for which a huge set of configurations needs to be evaluated in parallel.
Two papers will be published on these projects.
 C. C. Lee, C. J. Kuo, M. F. Hsu, et al, FEBS Lett. 581, 5454–5458 (2007).
 J. Xue, A. Moyer, B. Peng, J. Wu, B. N. Hannafon, W.-Q. Ding, PloS one 9, e109180 (2014).
 CPMD, www.cpmd.org
 A. Krężel, W. Maret, Arch. Biochem. Biophys. 611, 3–19, (2016).
 L. Zhang, J. Han, H. Wang, R. Car, E. Weinan, Phys. Rev. Lett. 120, 143001 (2018).