Interaction of hydroxychloroquine with SARS-CoV2 functional proteins using all-atoms non-equilibrium alchemical simulations

Piero Procacci *a, Marina Macchiagodena a, Marco Pagliai a, Guido Guarnieri b and Francesco Iannone *b
aDepartment of Chemistry “Ugo Schiff”, University of Florence, Via della Lastruccia 3-13, Sesto Fiorentino (FI), I-50019, Italy. E-mail:
bENEA, Portici Research Centre, DTE-ICT-HPC P.le E. Fermi, 1, Portici (NA), I-80055, Italy. E-mail:

Received 18th May 2020 , Accepted 30th June 2020

First published on 30th June 2020

Using a combination of enhanced sampling molecular dynamics techniques and non-equilibrium alchemical transformations with full atomistic details, we have shown that hydroxychloroquine (HCQ) may act as a mild inhibitor of important functional proteins for SARS-CoV2 replication, with potency increasing in the series PLpro, 3CLpro, RdRp. By analyzing the bound state configurations, we were able to improve the potency for the 3CLpro target, designing a novel HCQ-inspired compound, named PMP329, with predicted nanomolar activity. If confirmed in vitro, our results provide a molecular rationale for the use of HCQ or of strictly related derivatives in the treatment of Covid-19.

On 31 December 2019, the World Health Organization (WHO) was informed of cases of pneumonia with unknown etiology detected in Wuhan City, Hubei Province of China. By late January, the RNA of the a new virus (named SARS-CoV2) was identified and characterized,1 showing a large sequence identity to SARS-CoV, responsible for the SARS outbreak in 2003. On 12 March 2020, the WHO declared the Covid-19 outbreak a pandemic. The infection rapidly spread throughout the five continents involving, as of today, nearly 11 million of cases and causing the death of more than 500[thin space (1/6-em)]000.

From early February to March 2020, the X-ray structures of the two SARS-CoV2 3CL-like protease2 (3CLpro), and Papain-like protease3 (PLpro) and the cryo-EM 3.1 Å resolution structure of RNA-dependent-RNA-polymerase4 (RdRp) were deposited in the Protein Data Bank. These functional SARS-CoV2 proteins (PDB codes: 6LU7, 6W9C, 6M71), that are essential for virus replication,5 exhibited a sequence percent identity with their SARS-CoV homologues (PDB codes: 1UK4, 2FE8, 6NUR) exceeding 95%, with 3CLpro and RdRp of SARS-CoV2 differing by only 12 (out of 306)6 and 36 (out of 932) amino acids from their SARS-CoV counterpart. This striking similarity level leads naturally to believe that inhibitors of SARS-CoV functional proteins are very likely active for the SARS-CoV2 variants as well, and that, with appropriate funding in the mid-2000s, there would probably be a drug available for Covid-19 today.7 Unfortunately, no antiviral compound targeting 3CLpro, PLpro or RdRp for SARS was ever approved in the past decades due to a sharp decline in funding,7 probably relying on the wrong assumption that chance of a repetition of a new zoonotic viral transmission was extremely unlikely.

Several approved drugs, with hence known limited or mild side effects, are currently being administered for off-label or compassionate use against Covid-19. Among these drugs, we mention the anti-influenza virus Favipiravir (Avigan), tested in China and Japan on patients with mild symptoms, the anti-IL-6 biologic drug Tocilizumab, the adenosine-analog Remdesivir, the antiviral drug Kaletra, anti-HIV Lopinavir and antiretroviral Ritonavir, and the anti-malaria hydroxychloroquine (HCQ), approved by the FDA for the treatment of lupus and rheumatoid arthritis as well. This latter cost-effective drug is currently undergoing more than a hundred interventional clinical trials for the treatment of Covid-19 in several countries, including US, China and Italy.8

Back in the year 2005, Vincent et al. reported9 that chloroquine (CQ) had strong antiviral effects on SARS-CoV infection of primate cells. In a recent study,10 it was found that HCQ interfered with the entry mechanism of virions in Vero E6 cell cultures, by blocking the transport of SARS-CoV2 from early endosomes to endolysosomes. In a March 29 preprint,11 Italian researchers found evidence that the anti-SARS-CoV2 activity of a clinically achievable HCQ concentration was maximized when administered before but also after the infection of Vero E6 cells. These results suggest that HCQ may be active not only by blocking the entry of the virions through the host cell membrane, but also by way an anti-replication mechanism occurring after the release of the viral genome into the host cells.

This observations lead naturally to speculate that HCQ can be an inhibitor of any of the three cited viral functional proteins 3CLpro, PLpro or RdRp. In order to shed light on this matter, we have performed an accurate in silico “measurement” of the dissociation constant of HCQ with 3CLpro, PLpro and RdRp, using a state of the art computational technique combined with reliable modern force fields for biological systems (Fig. 1). The computational approach is based on molecular dynamics (MD) enhanced sampling techniques (Hamiltonian Replica Exchange, HREM) and non-equilibrium (NE) alchemical MD simulations,12 encoded in an algorithm13 amenable to massive parallelization on high performing computational facilities. Such advanced technique has been recently used with excellent results for the calculation of the absolute binding affinity in drug–receptor systems14,15 and emerged as one of the top performing MD-based methodologies in the recent blind challenges rounds of Statistical Assessment of the Modeling of Proteins and Ligands,12,16,17 systematically yielding mean differences between calculated and experimental dissociation free energies of the order of 1 kcal mol−1 or less.

Initial pose assessment on the three viral proteins was performed using Autodock4.18 HREM and NE simulation were done using the program ORAC.13 Methodological details are extensively reported in the ESI, along with a step by step tutorial for reproducing our data. In the ESI, we also included a compressed archive containing the ORAC input files used for HREM/NE calculations on HPC systems. In Fig. 2 we show the probability distributions of the distance between the center of mass (COM) of HCQ and that of the three viral functional proteins as obtained from the HREM simulation with torsional tempering (see ESI) in the active site region. It can be seen that in all three cases, HCQ lingers in the active site in spite of large allowance volume (see ESI) due to the harmonic restraint potential. The dissociation constants Kd = eβΔG0 from the HREM-NE computed standard dissociation free energy ΔG0 are reported in the Fig. 1. HCQ is predicted to have micromolar activity for all three proteins, corresponding to ΔG0 values of 7.7 ± 0.9 kcal mol−1, 8.5 ± 1.1 kcal mol−1, and 9.1 ± 1.4 kcal mol−1 for PLpro, 3CLpro and RdRP, respectively. For the HCQ-3CLpro complex, in particular, by analyzing the HREM sampled configurations of the bound state (see details in ESI), we noticed that in the most probable conformation HCQ adopts a compact structure characterized by the ethyl moiety insisting on the chloroquinoline planar moiety, with the hydroxy group not engaging in any stable hydrogen-bond with the surrounding protein residues. We hence decided to modify HCQ so at to further stabilize this compact structure in bulk solvent as well, in order to reduce the penalty due to the conformational entropy contribution upon binding.

image file: d0cc03558k-f1.tif
Fig. 1 HCQ complexes with domain I + II of 3CLpro, PLpro and RdRp. HCQ is in van der Waals representation. The active site residues are in CPK representation. Reported dissociation constants are computed from the HREM/NE determined standard dissociation free energies ΔG0 (see text) as Kd = eβΔG.

image file: d0cc03558k-f2.tif
Fig. 2 Probability distribution of the ligand–receptor COM–COM distance for the complex of 3CLpro-HCQ, 3CLpro-PMP329, PLpro-HCQ, RdRp-HCQ, as obtained in 24 ns of HREM simulation.

This was achieved by moving the OH group away to the penthyl moiety (yielding an R stereogenic center), thereby enhancing the hydrophobic character of the OH-depleted propyl moiety and its interaction with the planar group (transformed in a methyl–naphthyl moeiety) in both the bound and unbound states. As shown in the ESI, our modifications leading to PMP329 did preserve, to a large extent, the favorable ADME-Tox profile of HCQ. In the Fig. 3 we show in (a) the structure and the SMILES code of the HCQ analog, named PMP329, and in (b) a typical compact conformation in the bound state with the hydrophobic interaction between the methyl–naphthyl moiety and the propyl group. A detailed comparison of HCQ and PMP329 binding pattern in 3CLpro can be found in Section VI of the ESI. The ligand is surrounded by mostly hydrophobic residues with a single transient H-bond between the hydroxy group of PMP329 and the carboxy group in SER46. The ΔG0 of PMP329 for 3CLpro, computed using the previously described HREM/NE protocol, is found to be of 9.8 ± 1.4 kcal mol−1, translating in the nanomolar activity of Kd ≃ 70. PMP329 is not commercially available according to the ZINC19 database. However its synthesis should be straightforward starting from inexpensive an readily available commercial precursors.

image file: d0cc03558k-f3.tif
Fig. 3 (a) Chemical structure and SMILES code of PMP329; (b) typical PMP329 compact conformation in the binding site of 3CLpro.

The results presented in this communication, obtained using advanced molecular dynamics techniques with full atomistic details, show that HCQ may act as a mild to moderate ligand of important functional proteins for SARS-CoV2 replication, with potency increasing in the series PLpro, 3CLpro, RdRp. While clinical trials are underway worldwide,8 retrospective or anecdotal evidence on HCQ efficacy against Covid-19 is still contradictory and, by any means, not conclusive.20–22 By analyzing the HREM bound state configurations, we were able to improve the potency for the 3CLpro target, designing a novel HCQ-inspired compound, named PMP329, with predicted high nanomolar activity. These results call for a rapid in vitro measurement of the activity of PMP329 on 3CLpro as a possible antiviral agent for Covid-19. If confirmed, our prediction shows the potential of our computational methodology in the design of specific protein binding agents.

The computing resources have been provided by CRESCO/ENEAGRID High Performance Computing Infrastructure, see for information.

Conflicts of interest

There are no conflicts to declare.


  1. P. Zhou, X.-L. Yang, X.-G. Wang, B. Hu, L. Zhang, W. Zhang, H.-R. Si, Y. Zhu, B. Li, C.-L. Huang, H.-D. Chen, J. Chen, Y. Luo, H. Guo, R.-D. Jiang, M.-Q. Liu, Y. Chen, X.-R. Shen, X. Wang, X.-S. Zheng, K. Zhao, Q.-J. Chen, F. Deng, L.-L. Liu, B. Yan, F.-X. Zhan, Y.-Y. Wang, G.-F. Xiao and Z.-L. Shi, Nature, 2020, 579, 270–273 CrossRef CAS PubMed.
  2. Z. Jin, X. Du and Y. Xu, et al. , Nature, 2020, 582, 289–293 CrossRef CAS PubMed.
  3. J. Osipiuk, R. Jedrzejczak, C. Tesar, M. Endres, L. Stols, G. Babnigg, Y. Kim, K. Michalska and A. Joachimiak, The crystal structure of papain-like protease of SARS CoV-2, RSCB PDB, 2020 Search PubMed , pdbode: 6W9C.
  4. Y. Gao, L. Yan, Y. Huang, F. Liu, Y. Zhao, L. Cao, T. Wang, Q. Sun, Z. Ming, L. Zhang, J. Ge, L. Zheng, Y. Zhang, H. Wang, Y. Zhu, C. Zhu, T. Hu, T. Hua, B. Zhang, X. Yang, J. Li, H. Yang, Z. Liu, W. Xu, L. W. Guddat, Q. Wang, Z. Lou and Z. Rao, Science, 2020, 368, 779–782 CrossRef CAS PubMed.
  5. R. Hilgenfeld, FEBS J., 2014, 281, 4085–4096 CrossRef CAS PubMed.
  6. M. Macchiagodena, M. Pagliai and P. Procacci, Chem. Phys. Lett., 2020, 750, 137489 CrossRef CAS PubMed.
  7. S. K. Burley, Nature, 2020, 580, 157 CrossRef PubMed.
  8. N. J. DeVito, M. Liu and J. K. Aronson, COVID-19 Clinical Trials Report Card: Chloroquine and Hydroxychloroquine, CEMB reports, 2020, accessed May 15.
  9. M. J. Vincent, E. Bergeron, S. Benjannet, B. R. Erickson, P. E. Rollin, T. G. Ksiazek, N. G. Seidah and S. T. Nichol, Virol. J., 2005, 2, 69 CrossRef PubMed.
  10. J. Liu, R. Cao, M. Xu, X. Wang, H. Zhang, H. Hu, Y. Li, Z. Hu, W. Zhong and M. Wang, Cell Discovery, 2020, 6, 16 CrossRef CAS PubMed.
  11. N. Clementi, E. Criscuolo, R. A. Diotti, R. Ferrarese, M. Castelli, R. Burioni, M. Clementi and N. Mancini, bioRxiv, 2020 DOI:10.1101/2020.03.29.014407.
  12. P. Procacci, M. Guarrasi and G. Guarnieri, J. Comput. Aided Mol. Des., 2018, 32, 965–982 CrossRef CAS PubMed.
  13. P. Procacci, J. Chem. Inf. Model., 2016, 56, 1117–1121 CrossRef CAS PubMed.
  14. F. Nerattini, R. Chelli and P. Procacci, Phys. Chem. Chem. Phys., 2016, 18, 15005–15018 RSC.
  15. P. Procacci, J. Chem. Theory Comput., 2018, 14, 3890–3902 CrossRef CAS PubMed.
  16. A. Rizzi, S. Murkli, J. N. McNeill, W. Yao, M. Sullivan, M. K. Gilson, M. W. Chiu, L. Isaacs, B. C. Gibb, D. L. Mobley and J. D. Chodera, J. Comput.-Aided Mol. Des., 2018, 32, 937–963 CrossRef CAS PubMed.
  17. A. Rizzi, T. Jensen, D. R. Slochower, M. Aldeghi, V. Gapsys, D. Ntekoumes, S. Bosisio, M. Papadourakis, N. M. Henriksen, B. L. de Groot, Z. Cournia, A. Dickson, J. Michel, M. K. Gilson, M. R. Shirts, D. L. Mobley and J. D. Chodera, J. Comput.-Aided Mol. Des., 2020, 34, 601–633 CrossRef CAS PubMed.
  18. G. M. Morris, R. Huey, W. Lindstrom, M. F. Sanner, R. K. Belew, D. S. Goodsell and A. J. Olson, J. Comput. Chem., 2009, 30, 2785–2791 CrossRef CAS PubMed.
  19. J. J. Irwin and B. K. Shoichet, J. Chem. Inf. Model., 2005, 45, 177–182 CrossRef CAS PubMed.
  20. E. S. Rosenberg, E. M. Dufort, T. Udo, L. A. Wilberschied, J. Kumar, J. Tesoriero, P. Weinberg, J. Kirkwood, A. Muse, J. DeHovitz, D. S. Blog, B. Hutton, D. R. Holtgrave and H. A. Zucker, JAMA, 2020 DOI:10.1001/jama.2020.8630.
  21. W. Tang, Z. Cao, M. Han, Z. Wang, J. Chen, W. Sun, Y. Wu, W. Xiao, S. Liu, E. Chen, W. Chen, X. Wang, J. Yang, J. Lin, Q. Zhao, Y. Yan, Z. Xie, D. Li, Y. Yang, L. Liu, J. Qu, G. Ning, G. Shi and Q. Xie, medRxiv, 2020 DOI:10.1101/2020.04.10.20060558.
  22. B. Yu, D. W. Wang and C. Li, medRxiv, 2020 DOI:10.1101/2020.04.27.20073379.


Electronic supplementary information (ESI) available. See DOI: 10.1039/d0cc03558k

This journal is © The Royal Society of Chemistry 2020