Drugs repurposed for COVID-19 by virtual screening of 6,218 drugs and cell-based assay

Significance Recent spread of SARS-CoV-2 has sparked significant health concerns of emerging infectious viruses. Drug repurposing is a tangible strategy for developing antiviral agents within a short period. In general, drug repurposing starts with virtual screening of approved drugs employing docking simulations. However, the actual hit rate is low, and most of the predicted compounds are false positives. To tackle the challenges, we report advanced virtual screening with pre- and postdocking pharmacophore filtering of 6,218 drugs for COVID-19. Notably, 7 out of 38 compounds showed efficacies in inhibiting SARS-CoV-2 in Vero cells. Three of these were also found to inhibit SARS-CoV-2 in human Calu-3 cells. Furthermore, three drug combinations showed strong synergistic effects in SARS-CoV-2 inhibition at their clinically achievable concentrations.

Post-docking filtering with interaction similarity. Post-docking simulations were performed based on protein-ligand interaction similarity with the known active compounds to identify the accurate representation of docking poses. As such, the protein-ligand interactions of the binding poses obtained by the docking were analyzed with PLIP package (v1.4.5) (22). It returns a list of detected interactions between each compound and the amino acids of the target receptor, covering six interaction types (hydrogen bonds, hydrophobic contacts, pi-stacking, pi-cation interactions, salt bridges, and halogen bonds). The types of interactions with relevant amino acid residues can be used to generate interaction similarity as Tanimoto similarity by comparing the interaction patterns of the predicted hit compounds with those of the binding modes of known active ligands of the target proteins. The interaction similarity threshold was set to 0.3.

Virus and cells.
Vero cells were obtained from the American Type Culture Collection (ATCC CCL-81) and maintained at 37°C with 5% CO2 in Dulbecco's modified eagle medium (DMEM; Welgene), supplemented with 10% heat-inactivated fetal bovine serum (FBS) and 1× antibiotic-antimycotic solution (Gibco). For Calu-3, cells were seeded at 2.0 × 10 4 cells per well in Eagle's minimum essential medium (EMEM), supplemented with 20% FBS, 1× MEM Non-Essential amino acid and 1× antibiotic-antimycotic solution (Gibco) in black, 384-well, μClear plates (Greiner Bio-One), 24 h prior to the experiment. SARS-CoV-2 (βCoV/KOR/KCDC03/2020) was provided by Korea Centers for Disease Control and Prevention (KCDC) and was propagated in Vero cells. All experiments involving live SARS-CoV-2 followed the guidelines of the Korea National Institute of Health (KNIH) using enhanced biosafety level 3 (BSL3) containment procedures at Institut Pasteur Korea approved for use by the KCDC.
Immunofluorescence assay of SARS-CoV-2 infection. Infected Vero and Calu-3 cells were subjected to evaluation of antiviral activity using an immunofluorescence-based imaging assay, labeling viral N protein of the SARS-CoV-2 within infected cells. In each assay detailed below, including dose-response assays and drug synergy assays, Vero cells were seeded at 1.2 × 10 4 cells per well in DMEM, supplemented with 2% FBS and 1× antibiotic-antimycotic solution (Gibco), in black, 384-well μClear plates (Greiner Bio-One) 24 h prior to the experiment. Ten-point DRCs were generated, with compound concentrations ranging from 0.1 to 50 μM. For the viral infections, plates were transferred into the BSL3 containment facility, and SARS-CoV-2 was added at multiplicity of infection (MOI) of 0.0125. Before validation experiments with Vero cells, we examined both cell viability and cell infectivity by changing the MOI of SARS-CoV-2. MOI of 0.0125 was chosen as the best experimental condition based on the best cell viability (91.73%) and the highest virus infectivity (75.75%). For Calu-3, cells were seeded at 2.0 × 10 4 cells per well in EMEM, supplemented with 20% FBS, 1× MEM Non-Essential amino acid and 1× antibiotic-antimycotic solution (Gibco) in black, 384well, μClear plates (Greiner Bio-One), 24 h prior to the experiment. Ten-point DRCs were generated with compound concentrations ranging from 0.1 to 50 μM. Omipalisib was analyzed at concentrations ranging from 0.031 to 15.63 nM due to cytotoxicity. For viral infection, plates were transferred into the BSL-3 containment facility, and SARS-CoV-2 was added at MOI of 0.5. The Vero and Calu-3 cells were fixed at 24 hours post-infection with 4% PFA and permeabilized with Triton-X100 to promote entering antibodies into cells. The acquired images were analyzed using in-house software to quantify cell numbers and infection ratios, and antiviral activity was normalized to positive (mock) and negative (0.5% DMSO) controls in each assay plate. DRCs were fitted by sigmoidal dose-response models, with the following equation: Y = bottom + (top − bottom)/[1 + (IC50/X) Hillslope ], using Prism7. IC50 values were calculated from the normalized activity dataset-fitted curves. All IC50 and CC50 values were measured in duplicate, and the quality of each assay was controlled by Z'-factor and the coefficient of variation in percent (%CV). For drug synergy quantification, drug combinations were evaluated using a checkerboard assay at eight points with a 2-fold serial dilution from 4 × IC50, where the IC50 values were determined in separate single-drug experiments. Synergy analysis was performed using synergyfinder R-package (v2.4.0) (23) using Bliss independence and ZIP models. . However, even remdesivir triphosphate (RDV-TP) as a positive control drug did not show activity even at high concentrations (50 M), which suggests that there is a problem with the enzyme assay kit; Profoldin's RdRp enzyme assay kit was the only available kit at the time of our revision. Thus, we instead performed additional binding free energy calculation.
Binding free energy calculation. The protein-ligand complexes derived from docking simulations were subjected to MD simulations by GROMACS 4.5 package (3). The CHARMM36 force field was assigned to the protein. Ligand parameterization was performed using CHARMM General Force Field (24). Each system was immersed in a dodecahedron box of TIP3P water. The Na + or Clwas applied to neutralize the system. The systems were first minimized using the steepest descent methods. After minimization, the systems were heated from 0 to 300 K over 100 ps using the NVT ensemble with a weak restraint on the enzyme and ligand. Following this, the systems were equilibrated over 200 ps at a constant pressure of 1 bar and temperature of 300 K using the NPT ensemble. Finally, the 3 ns production run was performed. Based on the 3 ns MD trajectory, binding free energy was calculated with molecular mechanics Poisson-Boltzmann surface area (MM/PBSA). The MM/PBSA calculations were performed using g_mmpbsa (25). The binding free energy was calculated according to the following equation: ∆Gcal = ∆H-T∆S = ∆Evdw + ∆Eele + ∆Gpb + ∆Gnp -T∆S, where ∆Evdw and ∆Eele refer to van der Waals energy and electrostatic terms, respectively. ∆Gpb and ∆Gnp refer to polar and nonpolar solvation free energies, respectively. The entropy term (T∆S) was not calculated in this study. Since the multiple ligands were compared based on the same target, it is reasonable to ignore the entropy. To reduce false positives often obtained by performing docking simulation alone, pre-docking and post-docking simulations were performed to filter drug candidates. In the pre-docking filtering process, compounds with similar shapes to the known active compounds for each target protein were selected. In the post-docking filtering process, the chemicals identified through docking simulations were evaluated considering the docking energy and the similarity of the protein-ligand interactions with the known active compounds.                          Note S1. Instantaneous inhibitory potential (IIP) from the dose-response curves of the antiviral drugs.
The dose-response curve of a single antiviral drug can be analyzed based on the median-effect equation (equation (1) Here, fu is the fraction of infection events unaffected by the drug (i.e., 1 − equals the fraction of drug-affected events). D is the drug concentration, IC50 is the 50% antiviral concentration, and m is the dose-response curve slope. Dose-response curves for compounds with higher m values show higher antiviral activity at the same normalized concentration so long as the concentration is higher than IC50.
The antiviral activities of compounds can be expressed as the IIP (equation (2)) (26,41,42): Here, fu is the fraction of infection events unaffected by the drug, D is the drug concentration, IC50 is the 50% antiviral concentration, and m is the dose-response curve slope. If a drug reduces SARS-CoV-2 replication by 1 log then fu = 0.1 and its IIP = 1, whereas if it reduces viral replication by 2 logs, i.e. 100-fold, its IIP = 2. Importantly, IIP focuses on the remarkable effect of the slope parameter on antiviral activity.
Note S2. Drug combinations with synergistic antiviral activity assessed by the DI values.
Drug combinations can be characterized by two fundamental indices, the Loewe additivity and Bliss independence. We evaluated the drug combinations for Loewe additivity and Bliss independence, because there have been successful cases using these two fundamental indices to evaluate the combined effects of antiviral drugs for HIV-1 and HCV (26,42). The Loewe additivity is based on isobolograms and assumes similar mechanism or competition for the same binding site. For positive inhibitory slopes, Loewe additivity is described by equation (3) Equation (3) is numerically solved for 1+2 to predict the additive effects of the drug combinations.
Bliss independence assumes that each drug acts on different target, and is defined as equation (4) where , and denote the logarithmic drug effects (log [(1 − 1+2 )/ 1+2 ]) of experimental data, Bliss independence and Loewe additivity, respectively. Note that this index incorporates both Bliss independence and Loewe additivity, and categorizes the experimental data of combination effects. From the DI values calculated by equation (5), the anvi-SARS-CoV-2 effects of drug combinations can be assessed (SI Appendix, Fig. S10).