Polyphenols as Potential Inhibitors of SARS-CoV-2 RNA Dependent RNA Polymerase (RdRp)

An increasing number of studies have demonstrated the antiviral nature of polyphenols, and many polyphenols have been proposed to inhibit SARS-CoV or SARS-CoV-2. Our previous study revealed the inhibitory mechanisms of polyphenols against DNA polymerase α and HIV reverse transcriptase to show that polyphenols can block DNA elongation by competing with the incoming NTPs. Here we applied computational approaches to examine if some polyphenols can also inhibit RNA polymerase (RdRp) in SARS-CoV-2, and we identified some better candidates than remdesivir, the FDA-approved drug against RdRp, in terms of estimated binding affinities. The proposed compounds will be further examined to develop new treatments for COVID-19.

Since it was first identified in late 2019, SARS-CoV-2 has devastated the world, infecting more than 200 million people and killing more than 4.7 million [29]. In December 2020, the FDA approved the Pfizer-BioNTech vaccine for emergency use, resulting in more than 100 million vaccinations in the U.S. alone [30]. Despite the success of the vac- cines, COVID-19 cases continue to increase, making the discovery of safe antiviral drugs a significant concern. SARS-CoV-2 is an enveloped, positive-sense RNA virus closely related to SARS-CoV, the cause of the SARS pandemic in 2002-2003 [31,32]. The virus' genome consists of 14 ORFs, with ORF 1a and 1b being the most important for RNA replication [33,34]. After entry into a cell, ORF 1a and 1b are translated into two polyproteins that are further broken down into 16 non-structural proteins (nsps) [33]. These proteins assemble to form a replication complex that transcribes and replicates the virus' RNA [35]. This RNA is ultimately packaged to make new SARS-CoV-2 viruses.
As the enzyme complex responsible for generating RNA, RNA-dependent RNA polymerase (RdRp) is central to viral replication [32]. The main protein in the complex is nsp12 bound by two accessory subunits, nsp 7 and nsp8 [36]. Given that human cells do not use RdRp and that its role in the viral life cycle is critical, RdRp is a popular drug target [37,38]. Inhibitors of RdRp fall into two categories: nucleoside inhibitors (NI) and non-nucleoside inhibitors (NNI). After being incorporated, nucleoside inhibitors block new nucleotides from entering the RNA chain, so synthesis stops [36]. In an in vitro study using a newly developed CoV-RdRp-Gluc reporter assay, remdesivir and molupiravir, both NIs, inhibited RdRp with EC 50 values of 11.11 and 0.22 µM, respectively [32]. NNIs inhibit RdRp by changing its shape after binding to allosteric sites on the protein [36]. In a molecular docking study by Zijing et al., tegobuvir, an NNI drug in development for hepatitis C, bonded to the nsp12-7 and nsp12-8 sections of RdRp with docking scores ranging from −7.8 to −8.4 kcal/mol, showing an inhibitory effect [39]. Compared to NIs, which are vulnerable to exonuclease (nsp14) cleavage and competition with nucleotides present in the cell, NNIs have the potential to be highly potent drugs against SARS-CoV-2 [32,36].
Thus, to find potential therapeutic agents against COVID-19, this study screened 480 polyphenols to bind to SARS-CoV-2 RdRp. The structures of the 480 polyphenols were obtained from the Phenol-Explorer 3.6 database [40][41][42], and molecular docking was conducted using Maestro. Molecular mechanics with generalized Born and surface area solvation (MM-GBSA) scores were collected to quantify the affinity of the molecules for the proteins, and then absorption, distribution, metabolism, and excretion (ADME) and drug-likeness properties were analyzed for further screening. This study identifies three polyphenols with extremely low binding affinities to the SARS-CoV-2 RdRp as potential natural products for COVID-19 treatment.

Docking Analysis of Polyphenols against SARS-CoV-2 RdRp
To develop effective inhibitors from polyphenols against SARS-CoV-2 RdRp, 480 polyphenols were docked onto SARS-CoV-2 RdRp (PDB ID: 7BV2). Based on docking poses, the binding energies were calculated using Prime MM-GBSA in Maestro. As a result, the top three protein-ligand complexes, namely RdRp-cyanidin 3-O-rutinoside (−107.68 kcal/mol), RdRp-petunidin 3,5-O-diglucoside (−99.18 kcal/mol), and RdRp-delphinidin 3-O-rutinoside (−90.70 kcal/mol), have better estimated binding energies (Table 1). These three polyphenols belong to the group of anthocyanins [43][44][45]. Meanwhile, we selected four compoundsremdesivir-TP (−55.00 kcal/mol), theaflavin 3,3 -digallate (TF3) (−77.89 kcal/mol), swertiapuniside (−39.42 kcal/mol), and ATP (−57.83 kcal/mol)-as a control group. Here, remdesivir-TP is the best drug candidate, which was identified in our previous study [46]. TF3 and swertiapuniside were proposed as the top-ranked inhibitors of RdRp in the studies from Singh et al. [27] and Koulgi et al. [47], respectively. From Table 1, we find that the top three polyphenols show better binding energies than remdesivir-TP, TF3, and swertiapuniside. This result indicates that the top three polyphenols possess the potential to inhibit RdRp. Furthermore, the binding energies of the top three polyphenols are better than that of ATP, which also suggests that the top three polyphenols might exert strong competitiveness at the ATP binding site. The original ligand in 7BV2 and the best-scored potential drug identified by our previous study [46]. 2 The best-scored potential drug identified by a previous study [27]. 3 The best-scored potential drug identified by a previous study [47]. 4 The substrate of RNA synthesis.
By comparing the 2D ligand-protein interactions of the top three polyphenols bound to RdRp (Figure 1), we find that they all interact with U20, the terminal nucleotide that interacts with incoming ATP. Cyanidin 3-O-rutinoside interacts with U20 by forming one hydrogen bond, two π-π stackings, and one π-cation interaction. By forming two hydrogen bonds, two π-π stackings, and one π-cation interaction, petunidin 3,5-O-diglucoside interacts with U20. Meanwhile, delphinidin 3-O-rutinoside interacts with U20 by forming one hydrogen bond and one π-cation interaction. These interactions between the top three polyphenols and U20 suggest that these polyphenols might strongly bind to RNA and compete with incoming ATP for the binding site so that viral RNA elongation is blocked. the potential to inhibit RdRp. Furthermore, the binding energies of the top three polyphe nols are better than that of ATP, which also suggests that the top three polyphenols migh exert strong competitiveness at the ATP binding site. −57.83 1 The original ligand in 7BV2 and the best-scored potential drug identified by our previous study [46]. 2 The best-scored potential drug identified by a previous study [27]. 3 The best-scored potential drug identified by a previous study [47]. 4 The substrate of RNA synthesis.
By comparing the 2D ligand-protein interactions of the top three polyphenols bound to RdRp (Figure 1), we find that they all interact with U20, the terminal nucleotide tha interacts with incoming ATP. Cyanidin 3-O-rutinoside interacts with U20 by forming on hydrogen bond, two π-π stackings, and one π-cation interaction. By forming two hydro gen bonds, two π-π stackings, and one π-cation interaction, petunidin 3,5-O-diglucosid interacts with U20. Meanwhile, delphinidin 3-O-rutinoside interacts with U20 by forming one hydrogen bond and one π-cation interaction. These interactions between the top thre polyphenols and U20 suggest that these polyphenols might strongly bind to RNA and compete with incoming ATP for the binding site so that viral RNA elongation is blocked  Additionally, we summarized the number of hydrogen bonds between the top three polyphenols and the essential residues in Table 2. From Table 2, we find that these three polyphenols all interact with Arg555 by forming hydrogen bonds: one hydrogen bond for cyanidin 3-O-rutinoside, one hydrogen bond for petunidin 3,5-O-diglucoside, and two hydrogen bonds for delphinidin 3-O-rutinoside. Moreover, cyanidin 3-O-rutinoside interacts with Arg555 by forming a π-cation interaction, and delphinidin 3-O-rutinoside interacts with Agr555 through a salt bridge ( Figure 1A,C). This result suggests that Arg555 is a critical residue in the binding pocket. Furthermore, cyanidin 3-O-rutinoside and delphinidin 3-O-rutinoside both interact with Arg553 by forming one hydrogen bond each. In forming one hydrogen bond each, petunidin 3,5-O-diglucoside and delphinidin 3-O-rutinoside both interact with Asn691 and Asp760. Notably, the ligand delphinidin 3-O-rutinoside has a phenoxide ion interacting with Arg555 and Asn691 ( Figure 1C). Accordingly, we propose that cyanidin 3-O-rutinoside, petunidin 3,5-O-diglucoside, and delphinidin 3-O-rutinoside are the three best inhibitor candidates of RdRp among all 480 polyphenols based on these docking results.

Physicochemical Properties Prediction
To explore the physicochemical properties of the top three polyphenols, we predicted the ADME and drug-likeness properties using Qikprop in Maestro. The results are shown in Table 3. Meanwhile, the physicochemical properties of three control compounds are also predicted and listed in Table 3. First, the molecular weight of the top three polyphenols is within the recommended range of 130.0 to 725.0, whereas the control compound TF3 falls out of the recommended range. QPlogS is another important property to predict aqueous solubility, and its recommended range is from −6.5 to 0.5. All the compounds in Table 3 are in that range. In addition, the Lipinski's rule of five (RO5) and the Jorgensen's rule of three (RO3) are essential descriptors for evaluating drug-likeness. The top three polyphenols all pass the RO5 and RO3, which further suggests the drug-likeness of these polyphenols. In summary, Qikprop predictions indicate that cyanidin 3-O-rutinoside, petunidin 3,5-O-diglucoside, and delphinidin 3-O-rutinoside can be considered as viable drug candidates worthy of further research.  [49]. The recommended range: maximum is 3.

Molecular Dynamics (MD) Simulation Analysis
To further analyze the stability of the complexes, we conducted MD simulations to calculate RMSD and energy for the top three protein-ligand complexes. First, RMSD can be used to assess the stability of a protein-ligand complex. As shown in Figure 2A-C, the RMSD for the 7BV2-ligand complexes stabilized at around 0.25 nm after 5 ns, which suggests that the protein-ligand complexes are stable during the simulation process. Moreover, the total energies of these three complexes are shown in Figure 2D-F. The energies of 7BV2-ligand complexes stabilized at around −1.98 × 10 6 kJ/mol after 2 ns, further revealing the good stabilities of these three systems.  [49]. The recommended range: maximum is 3.

Molecular Dynamics (MD) Simulation Analysis
To further analyze the stability of the complexes, we conducted MD simulations to calculate RMSD and energy for the top three protein-ligand complexes. First, RMSD can be used to assess the stability of a protein-ligand complex. As shown in Figure 2A,B, and C, the RMSD for the 7BV2-ligand complexes stabilized at around 0.25 nm after 5 ns, which suggests that the protein-ligand complexes are stable during the simulation process. Moreover, the total energies of these three complexes are shown in Figure 2D-F. The energies of 7BV2-ligand complexes stabilized at around −1.98 × 10 6 kJ/mol after 2 ns, further revealing the good stabilities of these three systems. To compare the stability of RdRp with the three best polyphenols and the control group, we conducted MD simulations on the control complexes: RdRp-remdesivir-TP and RdRp-ATP. The RMSD values were then calculated and shown in Figure 3. As a result, the RMSD values of RdRp-remdesivir-TP and RdRp-ATP were higher than those of the complexes of RdRp with the three best ligands, which suggests that these three topscored polyphenols, bound to RdRp, show better stability. Figure S1 shows that the RMSD To compare the stability of RdRp with the three best polyphenols and the control group, we conducted MD simulations on the control complexes: RdRp-remdesivir-TP and RdRp-ATP. The RMSD values were then calculated and shown in Figure 3. As a result, the RMSD values of RdRp-remdesivir-TP and RdRp-ATP were higher than those of the complexes of RdRp with the three best ligands, which suggests that these three top-scored polyphenols, bound to RdRp, show better stability. Figure S1 shows that the RMSD values of 7BV2 before and after docking all stabilize at around 0.25 nm from 5 ns to 100 ns. Interestingly, the RMSD value of 7BV2 before ligand docking is slightly higher than the others, which indicates that the protein structure of RdRp becomes more stable after ligand-protein docking. Therefore, we conclude that the top three polyphenols bound to RdRp promote the stability of the protein structure. After MD simulations, we created three 2D interaction diagrams between RdRp and the three top-scored polyphenols ( Figure S2). Compared to the docking results, the number of interactions (hydrogen bonds and π-π stacking) was reduced (Figure 1 vs. Figure S2A-C) because the fluctuation of the structures is the nature of MD simulations. However, we also found that the most significant interactions between the ligands and the protein remained stable ( Figure S2). For example, the hydrogen bond, π-cation, and π-π interactions between the top-scored polyphenols and Arg555 and U20, the terminal nucleotide, remained stable during the MD simulations. As the control, the complex of RdRp and remdesivir-TP had fewer interactions after the MD simulation, while remdesivir-TP still interacted with U20 but not with Arg555 ( Figure S3).
Molecules 2021, 26, x FOR PEER REVIEW 6 of 12 three 2D interaction diagrams between RdRp and the three top-scored polyphenols (Figure S2). Compared to the docking results, the number of interactions (hydrogen bonds and π-π stacking) was reduced (Figure 1 vs. Figure S2 A-C) because the fluctuation of the structures is the nature of MD simulations. However, we also found that the most significant interactions between the ligands and the protein remained stable ( Figure S2). For example, the hydrogen bond, π-cation, and π-π interactions between the top-scored polyphenols and Arg555 and U20, the terminal nucleotide, remained stable during the MD simulations. As the control, the complex of RdRp and remdesivir-TP had fewer interactions after the MD simulation, while remdesivir-TP still interacted with U20 but not with Arg555 ( Figure S3).

Discussion
Polyphenols possess a wide range of health benefits and biological activities, including antioxidant [50], antitumoral [51], anti-inflammatory [52], and antiviral properties [53], the latter of which suggests that polyphenols may be useful in the battle against multiple viruses, even SARS-CoV-2 [27,[54][55][56][57][58][59][60][61][62]. In fact, numerous studies have shown that polyphenols exhibit antiviral effects by inhibiting DNA/RNA synthesis [63][64][65][66][67]. Our previous study also demonstrated that miquelianin, a flavonol glucuronide, can compete for the binding site of dATP on HIV-1 reverse transcriptase, inhibiting the viral DNA synthesis by interacting with the last nucleotide of the RNA chain and the binding residues of DNA polymerase [68]. Competing for the binding site of the incoming dNTP on the 3′ end of the elongating chain of DNA or RNA on the DNA/RNA polymerases is a significant mechanism through which polyphenols or other artificial compounds such as remdesivir exert

Discussion
Polyphenols possess a wide range of health benefits and biological activities, including antioxidant [50], antitumoral [51], anti-inflammatory [52], and antiviral properties [53], the latter of which suggests that polyphenols may be useful in the battle against multiple viruses, even SARS-CoV-2 [27,[54][55][56][57][58][59][60][61][62]. In fact, numerous studies have shown that polyphenols exhibit antiviral effects by inhibiting DNA/RNA synthesis [63][64][65][66][67]. Our previous study also demonstrated that miquelianin, a flavonol glucuronide, can compete for the binding site of dATP on HIV-1 reverse transcriptase, inhibiting the viral DNA synthesis by interacting with the last nucleotide of the RNA chain and the binding residues of DNA polymerase [68]. Competing for the binding site of the incoming dNTP on the 3 end of the elongating chain of DNA or RNA on the DNA/RNA polymerases is a significant mechanism through which polyphenols or other artificial compounds such as remdesivir exert their inhibitory effects. Therefore, to virtually screen potential inhibitors against DNA/RNA polymerases, the DNA or RNA molecule has to be considered and located precisely in the docking box. However, compounds were docked onto the empty active site of RdRp of SARS-CoV-2 without the RNA chain involved in many previous studies. This should be a reason why many drug candidates identified in in silico experiments are false positives.
In this study, we found that the binding affinities of the top-ranked polyphenols all showed much better results than remdesivir in terms of binding affinity. To further discuss the docking pose of the best drug candidate, we overlapped the docking pose of 7BV2-cyanidin 3-O-rutinoside with the original structure of 7BV2 ( Figure 4). We found that cyanidin 3-O-rutinoside occupies the position of the original ligand remdesivir-MP, but that it also interacts with Arg555 and the last nucleotide of the RNA chain in addition to the interactions between RdRp and the original ligand remdesivir-MP. Cyanidin 3-O-rutinoside interacts with Arg555 by forming one hydrogen bond and one π-cation interaction; however, remdesivir-MP interacts with Arg555 only through one π-cation interaction. Cyanidin 3-O-rutinoside interacts with U20, the last nucleotide, through a π-π interaction as well. These findings indicate that cyanidin 3-O-rutinoside, the polyphenol with the best scores, should compete for the binding site to block viral RNA synthesis.
R PEER REVIEW 7 of 12 7BV2-cyanidin 3-O-rutinoside with the original structure of 7BV2 ( Figure 4). We found that cyanidin 3-O-rutinoside occupies the position of the original ligand remdesivir-MP, but that it also interacts with Arg555 and the last nucleotide of the RNA chain in addition to the interactions between RdRp and the original ligand remdesivir-MP. Cyanidin 3-Orutinoside interacts with Arg555 by forming one hydrogen bond and one π-cation interaction; however, remdesivir-MP interacts with Arg555 only through one π-cation interaction. Cyanidin 3-O-rutinoside interacts with U20, the last nucleotide, through a π-π interaction as well. These findings indicate that cyanidin 3-O-rutinoside, the polyphenol with the best scores, should compete for the binding site to block viral RNA synthesis.

Ligand Preparation
The structures of 480 tested polyphenols were retrieved from Phenol-Explorer 3.6 (http://phenol-explorer.eu/) (accessed on 1 December 2020). All the tested compounds were prepared using Ligprep in Maestro 12.4 (Schrödinger). The process for Ligprep includes adding hydrogens, computing correct partial charges, and generating possible conformations. The force field is OPLS3e by default [69]. . The yellow-colored compound represents remdesivir-MP on 7BV2 (yellowcolored); the green-colored compound represents cyanidin 3-O-rutinoside docked on 7BV2 (blue-colored). The yellow dash line represents a hydrogen bond, and the green dash line represents a π-cation interaction.

Ligand Preparation
The structures of 480 tested polyphenols were retrieved from Phenol-Explorer 3.6 (http://phenol-explorer.eu/) (accessed on 1 December 2020). All the tested compounds were prepared using Ligprep in Maestro 12.4 (Schrödinger). The process for Ligprep includes adding hydrogens, computing correct partial charges, and generating possible conformations. The force field is OPLS3e by default [69].

Protein Preparation
The protein structures of RNA-dependent RNA polymerase (PDB ID: 7BV2) from RCSB's Protein Data Bank (https://www.rcsb.org/) were prepared for use by Maestro in three steps: preprocessing, optimization, and minimization [70,71]. The OPLS3e force field was applied in both the optimization and the minimization steps [69].

Ligand-Protein Docking
To estimate the interactions between target proteins and polyphenols, we conducted ligand-protein docking by using the Ligand Docking panel in Maestro. Before running docking jobs, a receptor grid box was generated based on the existing ligand remdesivir in the protein structure. The size of the receptor grid box was set as default (20 Å). Ligandprotein docking was performed in extra-precision (XP) mode.

MM-GBSA Calculation
To predict the binding energies of polyphenols bound to RdRp, we performed Prime MM-GBSA (molecular mechanics generalized Born surface area) in Maestro. In the MM-GBSA panel, the pose viewer files of the docked complex were uploaded into the MM-GBSA panel. The force field was OPLS3e [69].

ADME and Drug-Likeness Properties Prediction
Qikprop module in Maestro was applied to predict the absorption, distribution, metabolism, and excretion (ADME) and drug-likeness properties for further screening [72]. For Qikprop, the top-ranked polyphenols were prepared by using Ligprep. Finally, descriptors such as RuleOfFive (RO5) and RuleOfThree (RO3) were applied to analyze the candidates.

Molecular Dynamics Simulation
To further investigate the dynamic interactions between RdRp and the top three polyphenols, we conducted molecular dynamics (MD) simulations by using GROMACS version 2018.1 and CHARMM36 force field [73]. The starting coordinates of the proteinligand complex were obtained from a ligand-protein docking study. Then, we used CHARMM-GUI to build the MD simulation solution box, a cubic box with a length of 125 Å, which was then filled with water [74][75][76]. Next, the minimized structures were equilibrated using an NVT ensemble (constant Number of particles, Volume, and Temperature) and NPT ensemble (the Number of particles, Pressure, and Temperature). The target equilibration temperature was 300 K. Finally, MD simulations were performed for 100 ns. After the MD simulations, we calculated the root-mean-square deviation (RMSD) and the potential energies.

Conclusions
In summary, this study identified three polyphenols out of 480 as the best drug candidates for COVID-19 treatment. They all showed better estimated binding affinities than control compounds such as remdesivir. This can offer inspiration for new drug development.
Supplementary Materials: The following are available online, Figure S1: The RMSD of 7BV2 before and after ligand-protein docking. Figure

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.