Drug candidates were selected among antiviral datasets from published literature to identify the novel drugs which potentially interfere with the SARS-CoV-2 replication by inhibiting spike-ACE2 interactions (Ghahremanpour et al., 2020; Heiser et al., 2020; Jeon et al., 2020; Jin et al., 2020; Ku et al., 2020; Nguyenla et al., 2020; Riva et al., 2020; Touret et al., 2020; Weston et al., 2020; Yuan et al., 2020; Ahamad et al., 2021b; Chen et al., 2021; Dittmar et al., 2021; Ellinger et al., 2021; Ginex et al., 2021; Han et al., 2021; Mirabelli et al., 2021). The structure of the ACE2 receptor (PDB ID: 6M17) was downloaded from the Protein Data Bank (Goodsell et al., 2020).

The MD simulations were carried out on High Performance Computing (HPC) cluster of International Business Machines (IBM) Power9 CPU nodes (total 160 CPUs) with NVIDIA TESLA v100 32GB GPUs and Red Hat Enterprise Linux operating system.

To predict the preferred binding pocket on the ACE2 surface, molecular docking-based virtual screening was performed using Flare 5.0 and binding affinities calculated. Flare incorporates BioMolTech’s Lead Finder docking algorithm and combines its docking engine with genetic algorithm search containing local optimization procedures, enabling efficient sampling of ligand poses for refinement. The volume of the grid box was 287,154 A³, and the axis was set to be X: 116.403, Y: 97.474, and Z: 183.867 to cover all the amino acids in the box. It includes three different scoring functions (viz., LF dG, LF VSscore, and LF RankScore) for accurately predicting 3D docked ligand poses. The LF RankScore was selected for protein-ligand binding energy and rank ordering of active and inactive compounds in virtual screening experiments. The 2D plot was generated to study residue-ligand interactions, using the Schrödinger Maestro version 12.8.117, release 2021-2 suite (Schrödinger LLC, Cambridge, MA).

We used the general methodology to perform MD simulations of native ACE2 and the best-docked complexes using GROMACS (V5.18.3) (Abraham et al., 2015). For the MD simulation of the docked complexes, suitable force field parameters are required for the ligand/drug topology, which cannot be assigned using GROMACS. Hence, the PRODRG server was used to generate drug topologies and coordinate files (Schuttelkopf and van Aalten, 2004). We used the GROMOS9643a1 force field for native and drug-protein complexes, viz., ACE2-anidulafungin, ACE2-lopinavir, ACE2-indinavir, and ACE2-MLN-4670 (Van Der Spoel et al., 2005). Furthermore, systems were solvated using a Simple Point Charge (SPC) water model in a cubic box (Price and Brooks, 2004). 0.15 M counter ions of sodium (Na⁺) and chlorine (Cl⁻) were added to the simulation box for the system neutralization. All the neutralized systems were energy minimized using the steepest descent followed by conjugate gradient methods (50,000 steps for each). The system equilibration was achieved under the regulation of volume (NVT) and pressure (NPT) ensembles. The NVT ensemble was subjected to a constant temperature of 300 K and a constant pressure of 1 bar. The hydrogen (H) atoms were confined to equilibrium distances and periodic boundary conditions using the SHAKE algorithm. Additionally, the long-range electrostatic forces were defined using the Particle Mesh Ewald (PME) method (Lee et al., 2016). The cut-offs for Van der Waals and Coulombic interactions were set at 1.0 nm (Wang et al., 2016). The bonds and angles were constrained using the LINCS algorithm. Moreover, after a successful NPT ensemble run, the production run was performed for 100 ns. The energy, velocity, and trajectory were updated at a time interval of 10 ps. For the native ACE2 and the complexes, the MD trajectories were analyzed using GROMACS to calculate several parameters, namely, Cα-atom root mean square deviations (RMSD), root mean square fluctuations (RMSF) to investigate the relative fluctuations of each residue, radius of gyrations (Rg) to assess the protein compactness, solvent accessible surface area (SASA) to estimate the electrostatic contributions of molecular solvation, and free energy landscapes (FEL), as described in our previous publications (Ahamad et al., 2021a; Ahamad et al., 2021b).

HEK293T cells (ATCC CRL-3216) were cultured in Dulbecco’s modified Eagle medium (DMEM) with 1 g/L glucose (Life Technologies) supplemented with 10% fetal bovine serum (FBS) (Life Technologies) plus a final concentration of 100 IU/ml penicillin and 100 (μg/ml) streptomycin or without antibiotics were required for transfections.

Vero (WHO) Clone 118 cells (ECACC 88020401) were cultured in Dulbecco’s modified Eagle medium (DMEM, Life Technologies) with 1 g/L glucose (Life Technologies) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Life Technologies) plus a final concentration of 100 IU/ml penicillin and 100 (μg/ml) streptomycin or without antibiotics where required for transfection. Cells were incubated at 37°C, 5% CO₂.

Human ACE2 (Addgene #1786), pLVTHM/GFP (Addgene #12247), psPAX2 (Addgene 12,260), pMD2.G (Addgene #12259) were obtained from Addgene. pAAV-CMV-GFP was obtained from L. Zentilin (Molecular Medicine Lab, ICGEB). pAAV-spike-V5 and pAA-spike-d19-V5 SARS-CoV-2 spike expression vectors were used previously (Braga et al., 2021).

Antibodies against the following proteins were used: ACE2 (Abcam, ab15348), SARS-CoV-2 spike (GeneTex GTX632604), V5-488 (Thermo Fisher Scientific, 377500A488), α-beta-actin-HRP (Sigma-Aldrich), mouse-HRP (Abcam, ab6789), and rabbit-HRP (Abcam, ab205718).

Plasmid expressing human ACE2 reverse transfection was performed in a 96-well plate; 100 ng of plasmids were diluted in 25µl of Opti-MEM (Life Technologies) and mixed with the transfection reagent (FuGENE HD, Promega) using a ratio of 1 µg pDNA:3 µL FugeneHD. The transfection mixes were incubated for 25 min at RT and added to the 96 well plates (Cell Carrier Ultra 96, Perkin Elmer).

Vero cells (6.5 × 10³) or HEK293-ACE2 (8 × 10³) cells were seeded in each well. After 24 h of transfections, 100 ng of the pEC117-spike-V5 expression plasmid was transfected using a standard forward transfection protocol. After 24 h, cells were fixed in 4% PFA and processed for immunofluorescence.

After fixation in 4% PFA for 10 min at RT, cells were washed two times with 1xPBS and then permeabilized in same volumes of 0.1% Triton X100 (Sigma-Aldrich 1086431000) for 10 min at RT. Cells were then washed two times 1xPBS and blocked with 2% BSA for 1 h at RT. After blocking, the cells were stained according to the type of staining.

After blocking, a diluted primary antibody (1:500 in 1% BSA SARS-CoV-2 spike antibody or V5-488) was added to each well and incubated overnight at 4°. Cells were then washed two times with 1xPBS, and then a secondary antibody was added (45 µL/well, diluted 1:500 in 1% BSA) to each well and incubated for 2 h at RT. Cells were then washed two to three times in 1xPBS. Nuclear staining was performed using Hoechst 33,342 (1:5,000).

Image acquisition was performed using the Operetta CLS high content screening microscope (Perkin Elmer) with a Zeiss 20× (NA = 0.80) objective, a total of 25 fields were acquired per wavelength, well and replicate (∼10,000–15,000 cells per well and replicate).

Images were subsequently analyzed using the Harmony software (PerkinElmer). Images were first flat field corrected and nuclei were segmented using the “Find Nuclei” analysis module (Harmony). The thresholds for image segmentation were adjusted according to the signal-to-background ratio. The splitting coefficient was set to avoid splitting of overlapping nuclei (fused cells). The intensity of the green fluorescence (spike/GFP) was calculated using the “Calculate Intensity Properties” module (Harmony). All the cells that scored a nuclear area greater than four times (for manual quantification of syncytia, if fused nuclei >3, it counts as a syncytia) the average area of a single nucleus and were simultaneously positive for green (spike) signal in the cytoplasm area were considered as fused or syncytia. Data were expressed as a percentage of fused cells by calculating the average number of fused cells normalized to the total number of cells per well.

For pseudotyped particle entry assays: mean intensities of the segmented nucleus in the 488 (green) channel and the Hoechst channel for each nucleus across all fields were extracted. Each assay plate included a negative control, DMSO. Briefly, nuclei were segmented based on Hoechst staining, and cells were then classified as positive or negative depending on the GFP signal. Data were expressed as a percentage of GFP + cells by calculating the average number of GFP + cells normalized on the total number of cells.

A HIV-1 based lentiviral system was used to produce SARS-CoV-2 spike pseudotyped particles in HEK293T cells by the co-transfection of pMD2.G or pAAV-spike (d19), psPAX2 (packing vector), and PLVTHM (GFP) as described previously (Ali et al., 2019). Viral supernatants were collected after 48 h of transfection and centrifuged at 3,000 rpm for 10 min at 4°. The supernatant was then filtered with a 0.45 µm pore size filter, aliquoted, and stored at −80°C. For pseudotyped particle entry, 1 h before spike pseudotyped particle transduction, HEK293-ACE2 cells were treated with selected drugs and control cells were treated with DMSO. After 36 h, cells were fixed; nuclei were labeled with Hoechst and assessed for pseudotyped particles transduction efficiency based on GFP positive cells.

After 20–24 h of drug treatment, Vero cells were processed for western blot analysis. Equal amounts of total cellular proteins (15 μg), as measured with the BCA (Thermofisher, 23,227), were resolved by electrophoresis in 4–20% gradient polyacrylamide gels (Mini-PROTEAN, Biorad) and transferred to nitrocellulose/PVDF membranes (GE Healthcare). Membranes were blocked at RT for 60 min with PBST (PBS + 0.1% Tween-20) and 5% skim milk powder (Cell signalling, 9.999). Blots were then incubated (4°C, overnight) with primary antibodies against ACE2 (diluted 1:1,000), and α-tubulin (diluted 1:10,000). Blots were washed three times (10 min each) with PBST. For standard Western blotting detection, blots were incubated with either anti-rabbit HRP-conjugated antibody (1:5,000) or anti-mouse HRP-conjugated antibody (1:10,000) for 1 h at RT. After washing three times at RT with PBST (10 min each), blots were developed with ECL (Amersham).

Functional ACE2-S interaction is essential for SARS-CoV-2 entry into host cells, as shown in (Figure 1A). Detailed structural analysis proved that both SARS-CoV-2 and SARS-CoV S proteins strongly bind to ACE2 receptors (Lan et al., 2020; Magro et al., 2021). Therefore, to identify drugs that could inhibit the ACE2-S interactions and potentially viral replication, we screened the 450 drugs (Supplementary Table S1) by exploiting the molecular docking approach (Figures 1B, 2A–C). Here, we found that several compounds show a strong affinity toward the ACE2 receptor, but we selected the top three compounds, namely, anidulafungin, lopinavir, and indinavir, showing more binding affinity than MLN-4670, which is a known enzymatic inhibitor of ACE2 (Dales et al., 2002a) (Figures 2B, C, Supplementary Figures S1A–C). Detailed molecular interactions and properties of the four selected docked complexes are shown in Table 1. In the ACE2-anidulafungin complex, anidulafungin forms two hydrogen(H) bonds with Arg518 and Thr371 amino acid residues (Figures 2B, C) and hydrophobic bonds with 32 amino acids (shown in Table 1) together with one Zn-ion at the binding pocket of the ACE2 receptor. Similarly, in the lopinavir-ACE2 complex, lopinavir forms 1-H bonds with Glu398 and hydrophobic bonds with 38 amino acids (shown in Table 1). Indinavir interacts with Gly395 and Glu402 through 2H-bonds, hydrophobic bonds with 32 amino acid residues and Zn ions (shown in Table 1). The docking and 2D plot are shown in (Supplementary Figures S1D–F). However, MLN-4760 interacts with ACE2 by forming one H-bond with Glu402 (Supplementary Figures S1C–F) and hydrophobic bonds with 25 amino acid residues and a Zn metal ion (shown in Table 1).

Out of all the tested compounds, three drugs show high binding affinities toward ACE2 compared to MLN-4760, a known enzymatic inhibitor of ACE2. So, potentially, these drugs might inhibit SARS-CoV-2 replication by interfering with the formation of functional ACE2-S interactions.

To investigate molecular interactions of the docked complexes further, we performed MD simulations of ACE2-native, four selected docked complexes (ACE2-anidulafungin, ACE2-lopinavir, ACE2-indinavir) and ACE2-MLN-4670 for 100 ns. The stability, interaction profile, and structural parameters including RMSD, RMSF, Rg, SASA, and free energy calculations were also evaluated throughout the simulation run time to select the most stable receptor-drug complex.

In RMSD analysis, native ACE2 showed steady RMSD and revealed a threshold of ∼0.44 nm toward the binding with ACE2 under given simulation conditions (Figure 3A). The docking complexes of ACE2 with lopinavir, indinavir, and MLN-4760 noticeably reached equilibrium with average RMSD values of 0.45, 0.41, and 0.41 nm, respectively. The compound lopinavir revealed a high drift in the average RMSD values. However, the average RMSD values of indinavir and MLN-4760 remained the same.

The RMS deviation of Cα-atoms remained stable throughout the simulation with a slight difference in the values but proposed one complex with anidulafungin, indicating strong binding due to polar interaction with Arg518 and Thr371 residues as well as various non-polar interactions. Anidulafungin displayed the least RMSD fluctuations at the ACE2 binding pocket compared to the other drug compounds. The overall results suggested that the anidulafungin was reliably stable among all the complexes.

Secondly, RMS-fluctuations play a crucial role in identifying the flexible and rigid regions of drug-receptor complexes. Hence, RMSF calculations were performed to measure the average atomic flexibility of the ACE2 receptor Cα-atoms alone and in complex with the tested compounds. The average RMSF values were recorded as —Native-ACE2 (0.18 nm), anidulafungin (0.14 nm), lopinavir (0.18 nm), indinavir (0.15 nm) and MLN-4760 (0.18 nm) (Figure 3B). Interestingly, we observed that the ACE2-anidulafungin complex showed a low degree of fluctuations compared to other docked complexes and native-ACE2. However, lopinavir and indinavir displayed the highest degree of fluctuations and hence comparatively less stable. The above-mentioned comparative analysis of Cα-RMSF confirms a high level of flexibility caused by the presence of drug molecules on the protein structure in comparison to the native.

Next, we analyzed the compactness of the native ACE2 and docked complexes by using the radius of gyration (Rg) calculations. The results showed that the Rg values of the native-ACE2 receptor and the complexes of anidulafungin, lopinavir, indinavir, and MLN-4760 compounds remained highly stable with ranges of 2.78, 2.86, 2.82, 2.76, and 2.77 nm, respectively, throughout the MD simulation period (Figure 3C; Table 2 ). The low oscillations in Rg and SASA values portrayed high stability for the anidulafungin-ACE2 complex compared to other complexes and ACE2 alone. Interestingly, the Rg results also revealed that the ACE2-anidulafungin complex is the most stable of all the tested complexes. Comparative analysis of the Rg values shows the folding behavior of ACE2 upon binding with anidulafungin, which indicates high compactness between the complexes. We also performed a SASA analysis to better understand the solvent behavior of native ACE2 and docked complexes. Here, we found an average value of native ACE2, anidulafungin, lopinavir, indinavir, and MLN-4760 complexes of 389.88, 392.60, 391.05, 389.90, and 392.10 nm², respectively (Figure 3D; Table 2). These results showed that the compound anidulafungin possessed more stable hydrophobic contacts than the other docked complexes, making most of the ACE2 receptor surface accessible to the solvent and other molecules.

Finally, the docked complexes were also subjected to the overall motion of all protein and drug atoms by using Free Energy Landscapes (FEL) analysis. The conformational stabilities of the native ACE2 and the docked complexes were examined by FEL analysis using PC1 (Principal Components) and PC2 values. The values of FEL ranged from 0 to 14, 12.9, 13, 12.9, and 14.4 kJ/mol for the native ACE2, anidulafungin, lopinavir, indinavir, and MLN-4760 docked complexes, respectively ( Figures 4A–E ). This analysis indicated that the complexes were stable and persistent energy minima, suggesting the amino acids of the ACE2 binding pocket-forming interactions with drugs are vital for the stability and interaction. The global free energy minima results showed that the docked complexes revealed stabilizing effect that lead to the observed folding behavior of ACE2 with anidulafungin. The analysis revealed that the anidulafungin (Figure 4B) has fewer basins compared to the native receptor with three basins (Figure 4A). Overall, the results for the anidulafungin complex revealed the presence of two basins in the conformational space, with distinct global free energy minima, which consequently lead to a more stable behavior of the protein.

The SARS-CoV-2 spike protein in the viral envelope is essential for virus entry into the target cells. The SARS-CoV-2 S protein induces cell–cell fusion and the formation of syncytia when it is ectopically expressed on the membrane of host cells and binds ACE2 receptors of adjacent cells (Bussani et al., 2020; Xia et al., 2020; Braga et al., 2021; Buchrieser et al., 2021; Gutmann et al., 2021). Therefore, we explored whether S-induced syncytia formation would be impaired in the presence of the selected drugs, such as anidulafungin, lopinavir, indinavir, and MLN-4760, which show high affinity toward the ACE2 receptor in an in-silico analysis (Figure 2; Supplementary Figure S1). First, we tested the effect of drugs on S-mediated syncytia formation in Vero cells, which our previous studies have shown to respond to S expression by fusion (Braga et al., 2021). After 6 h of S-protein expression, cells were treated with the indicated drugs at 10 μM (Workflow shown in Figure 5A). Niclosamide (2.5 μM) was used as a positive control because this drug is a potent inhibitor of S-mediated syncytia formation by acting on the cellular TMEM16F membrane protein (Braga et al., 2021). Interestingly, we observed that both anidulafungin and lopinavir treatment significantly reduced the S-mediated cell–cell fusion compared with DMSO-treated control cells (Figures 5B, C). As expected, niclosamide treatment significantly reduced syncytia formation (Figures 5B,C). Both indinavir (Vacca et al., 1994; Condra et al., 1996) and MLN-4760 (Dales et al., 2002b; Joshi et al., 2016) were ineffective in blocking S-mediated cell fusion (Figures 5B, C). None of the tested drug treatments interfered with ACE2 expression in the cells (Supplementary Figure S2). Additionally, we neither observed significant toxicity of the tested drugs (Supplementary Figure S3) nor any effect on S-transgene expression (Supplementary Figure S4). Together, these results are consistent with the conclusion that the observed effects of both anidulafungin and lopinavir are due to interference between ACE2-S interactions.

Entry of SARS-CoV-2 S-viral particles mimics the entry pathway of SARS-CoV-2 virions (Lu et al., 2020; Mykytyn et al., 2021). Therefore, we explored whether the entry of pseudotyped lentiviral vectors expressing S on their envelope would be impaired in the presence of the top selected drugs. For this purpose, HEK cells expressing the ACE2 receptor were treated with the indicated drugs 1 h before the addition of S-pseudotyped viral particles to the cells (Workflow shown in Figure 6A). Strikingly, we found that both anidulafungin and lopinavir treatment significantly impaired S-pseudotyped particle transduction (Figures 6B,C), while no significant effect was observed in the presence of indinavir and MLN-4760. In particular, anidulafungin, which showed the strongest affinity toward the ACE2 receptor in our in-silico analysis and formed the most stable complex with ACE2 throughout the MD simulation period, was also the most effective in blocking S-mediated virion internalization.

Collectively, these results indicate that anidulafungin impedes both S-mediated syncytia formation and S-viral particle entry into the target cells.

Mann-Whitney U significance test was used for the data analysis.