Flavonoids and Nucleotide Analogs Show High Anity for Viral Proteins of SARS-CoV-2 by in silico Analysis: New Candidates for the Treatment of COVID-19.

The recent epidemic of COVID-19 caused by SARS-CoV-2 was declared by the World Health Organization as a public health emergency of international concern. The absence of an approved vaccine or a specic antiviral drug has made bioinformatic tools crucial for the identication of potential therapeutic targets and drugs for its control. As in other RNA viruses, the protease 3C-like and the RNA-polymerase are two of the SARS-CoV-2 targets to test drugs that can be analyzed in silico. In the present study, compounds derived from plants, fungi, and nucleoside 5'-triphosphate or uridine nucleotide analogs, with anti-DENV activity in vitro or in vivo, were analyzed by molecular docking as potential anti-SARS-CoV-2 drugs. Anthraquinone, with a DENV NS3 protease inhibitory activity; Balapiravir, Fisetin, Hyperoside, and Sofosbuvir, with a DENV NS5 RNA-polymerase inhibitory activity; and Quercetin, with both anti-NS3-NS5 activities, were tested against 3C-like protease and RNA-polymerase of SARS-CoV-2. All these drugs demonstrated a high anity for the corresponding SARS-CoV-2 proteins, representing excellent candidates for the treatment of COVID-19. Therefore, in vitro or in vivo studies should be carried out using these compounds on models for SARS-CoV-2 infection.


Introduction
Coronavirus disease 2019  is currently affecting millions of people worldwide. According to the World Health Organization (WHO), the COVID-19 pandemic has been declared as a public health emergency of international concern; causing more than 500,000 deaths worldwide [1,2]. Therefore; it is a priority to search antivirals that reduce or prevent COVID-19 mortality. COVID-19 is caused by the severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2), a spherical virus with a positive-sense RNA genome [3]. SARS-CoV-2 viral proteins include the 3C-like protease (NSP5) and the RNA-polymerase (NSP12), two of the proposed RNA virus targets tested for different drugs [4][5][6][7][8]. The NSP5 protein is a chymotrypsin-like protease; which processes the polyproteins ppa1a and pp1ab; generating unique viral proteins [9]. It consists of three domains: domain I (8-100 aa), II (101-183), and III (200-302 aa). The protease catalytic site is located in the ssure between domains I and II; with a catalytic dyad of His41 and Cys145 [10]. On the other hand; the NSP12 protein is the RNA-dependent RNA polymerase (RdRp or RNA-polymerase), responsible for the replication and transcription of the genomic RNA [6,11]. This enzyme has the preserved architecture of viral RNA-polymerases. It is composed of the ngers; palm; thumb; NiRAN (nidovirus RdRp-associated nucleotidyltransferase), present only in nidoviruses such as SARS-CoV-2; and interface domains ( Figure 1B) [6,11]. Therefore; it has been described that some antivirals, whose principal functions are to inhibit these two enzymes, may have a broad spectrum against RNA viruses [12,13].
Antiviral drugs have been developed to target either host cell proteins required for viral replication or viral proteins. The rst approach may provide a broader spectrum of activity and less chance of developing resistance; however, inhibiting host cell proteins has a more signi cant impact on cell viability. For this reason, the use of antiviral drugs that target viral proteins is more speci c and less toxic [14,15]. The recent closure of research laboratories worldwide as part of the containment measures to prevent the spread of the SARS-CoV-2 virus has hindered the in vitro and in vivo trials to test the effectiveness of drugs [4]. However, the integration of computational and experimental strategies in pharmacology has been of great value in the identi cation and development of new antivirals. Particularly, the molecular docking analysis is a widely used bioinformatics tool in modern drug design [16].
Currently, there is no speci c antiviral to content SARS-CoV-2 infection, and the expectation of an e cient vaccine continues to cause concern due to the time required for its approval and distribution. Therefore, testing the antiviral activity of FDA-approved drugs, using bioinformatic strategies, is today a promising tool for the selection of drugs that can potentially inhibit the replication of 8,17,18].
In recent years secondary metabolites of plants such as Quercetin, Fisetin, and Hyperoside avonoids, have demonstrated antiviral effect in vivo and in vitro [19,20], in RNA virus-like dengue virus (DENV) by acting on NS5 RNA-polymerase. Moreover, its antiviral effects have also been observed in other viral infections such as with Enterovirus, Rhinovirus, Ebola virus, Chikungunya, and Zika virus [19,[21][22][23][24]. On the other hand, Quercetin and Anthraquinone, secondary metabolites of plants and fungi, have demonstrated activity against the DENV NS3 protease [19,25]. Other examples of antiviral compounds are the nucleoside and nucleotide analogs such as Balapiravir and Sofosbuvir, respectively, alternative substrates for viral polymerases that competitively inhibit the synthesis of viral RNA in infections such as hepatitis C virus (HCV) and DENV [26,27] excellent candidates against the SARS-CoV-2.
In this study, we tested the a nity of Anthraquinone and Quercetin to the SARS-CoV-2 3C-like protease, and the a nity of Balapiravir, Quercetin, Fisetin, Hyperoside, and the previously tested Sofosbuvir for SARS-CoV-2 RNA-Polymerase using a molecular docking analysis. Our results showed that these drugs, previously used as antivirals for other RNA viruses, have an a nity for SARS-CoV-2 viral proteins and, therefore a potential inhibitory activity on viral replication.

Results
2.1 Validation of the three-dimensional structures and sequences of the SARS-CoV-2 3C-like protease and RNA-polymerase, Previous to the molecular docking analyses, the crystalline structures of the SARS-CoV-2 3C-like protease (PDB ID: 6M2N) ( Figure 1A) and RNA-polymerase (PDB ID: 6NUR) ( Figure 1B) were validated by amino acid sequence comparison with those reported in the UniProtKB database (ID: P0C6×7). The comparison was performed using the EMBOSS Needle software (https://www.ebi.ac.uk/Tools/psa/emboss_needle/) [28], obtaining a sequence identity of 96.1% and 97.5%, respectively.
Furthermore, the quality of the protein structures was evaluated using ERRAT software (https://servicesn.mbi.ucla.edu/ERRAT/) which differentiates regions of protein structures based on atomic interaction, generating a structure quality factor in a percentage of 0-100% [29], and the Verify 3D software (https://servicesn.mbi.ucla.edu/Verify3D/) that determines the compatibility of the 3D atomic model with its amino acid sequence and its atomic coordinates, obtaining a high score when the structure was correct [30].
The 3C-like protease structure of SARS-CoV-2 showed a score of 90.43% with the Verify3D software and a structured quality factor of 94.5% with the ERRAT software. The structure of the RNA-polymerase showed a quality factor of 96.68% with ERRAT software and 83. 81% with Verify3D. Once the three-dimensional structures of both proteins and their amino acid sequences were validated, the molecular docking assays were carried out with the candidate drugs.
2.2 Screening of DENV NS3 protease inhibitors (Quercetin and Anthraquinone) against the 3C-like protease of SARS-CoV-2 First, the molecular docking assays were validated by recreating the experimental data on the binding of the drug 5,6,7-trihydroxy-2-phenyl-4H-chromen-4-one (Baicalein) found in the same structure as the PDB ID: 6M2N with the 3C-like protease of SARS-CoV-2. The analysis found that the drug in both traditional and computational experiments joined in the same protease structure position (Figure 2A), validating our conditions. The binding energy of the drug and protease obtained from the computational data was -7.95 kcal/mol −1 .
Quercetin and Anthraquinone inhibitors of DENV NS3 protease bound with a high binding a nity (-6.5 and -7.95 kcal/mol −1 , respectively) to the 3C-like protease of SARS-CoV-2 ( Figure 2B), both drugs bound to the active site of the protease, particularly to the HIS41 which belongs to the catalytic dyad. The a nity of the 3C-like protease for Quercetin and Anthraquinone was compared with the one for Baicalein, and the drug interactions with the highest a nity for the active site were graphed based on an RMSD <1 Å ( Figure 4A). Baicalein had the higher a nity for the active site of SARS-CoV-2 3C-like protease (-7,946 ± 0.005 kcal/mol -1 ) followed by Quercetin (-7,562 ± 0.243 kcal/mol -1 ), and Anthraquinone (-6.5 ± 0 kcal/mol -1 ).

Screening of DENV NS5 polymerase inhibitors against SARS-CoV-2 RNA polymerase
In the case of the SARS-CoV-2 RNA polymerase, a crystal structure coupled to an inhibitor was not found in the PDB database. Therefore, it was not possible to recreate the data obtained by traditional methods.
However, Remdesivir has been described as an RNA-polymerase inhibitor and has been approved as a therapeutic drug against SARS-CoV-2 [31-33]. Therefore, to perform the Molecular Docking assays, we used Remdesivir as a control and then proceeded to test the drugs selected for our study. Fisetin, Quercetin, Hyperoside, Balapiravir, Sofosbuvir, and Remdesivir bound to two RNA-polymerase sites with high a nity, the RNA-polymerase active site, and the NiRAN-Fingers site ( Figure 3). On the other hand, Quercetin, Fisetin, Balapiravir, and Sofosbuvir showed a higher a nity for the active site of the RNApolymerase (-7.26, -7.08, -7.25 and -7.23 kcal/mol −1 , respectively), respect to the control drug Remdesivir (-6.81 kcal/mol −1 ) ( Figures 3A and 3B). Hyperoside showed a lower a nity (-6.68 kcal/mol -1 ); however, it is not ruled out as an RNA-polymerase inhibitor, since in predicting the number and type of non-covalent interactions using the PLIP software, it was found that Hyperoside has the maximum possible number of non-covalent interactions with RNA-polymerase (Table 2).
Besides, to the high a nity of the tested drugs to the active RNA polymerase site, a high a nity for the NiRAN sub-domain was also found. The NiRAN subdomain is present only in nidoviruses such as SARS-CoV-2, it has a nucleotidylation activity and is essential for viral replication; its mutation generates noninfective viral particles. It has also been involved as a protein primer for RNA synthesis, as the Vpg primer activity exhibited in Picornaviruses [34,35]; reasons why it might be an excellent target to inhibit viral replication.
Finally, to compare the a nity of drugs for the RNA polymerase, the drug interactions with the highest a nity for the active and NiRAN sites were graphed based on an RMSD <1 Å ( Figure 4B and 4C). Of the drugs analyzed, Fisetin and Quercetin have a higher a nity for the active site of the RNA polymerase (-7,042 ±0.035 and -7,038 ±0.304 kcal/mol -1 respectively) ( Figure 4B), showing that both can be excellent candidates for binding the SARS-CoV-2 RNA polymerase. Similar a nities of Fisetin and Quercetin to both the RNA-polymerase active and NiRAN sites were found since no signi cant differences were obtained ( Figure 4C).

Comparison of non-covalent interactions and ligand-binding energy, as well as drug legislation and their toxicity
Finally, a comparison of the different drugs analyzed as possible inhibitors of 3C-like protease and SARS-CoV-2 RNA-polymerase is shown in Table 1, including their PubChem access number, binding energy, toxicity reported by PubChem and the FDA legislation. Additionally, using the PLIP software that compares probable interactions according to distance thresholds and angles reported in the literature to give a probable report of non-covalent interactions between the target and the ligand [36]. We show the possible non-covalent interactions between the compounds and their targets (Table 2).

Discussion
The World Health Organization has declared the recent epidemic of COVID-19 caused by SARS-CoV-2, as a public health emergency of international interest [2]. Approximately 17 million cases and more than 650,000 deaths for COVID-19 have been reported until mid-2020. America, the continent most affected had reported 7 million new cases in 24 hours [1]. Given this, it is crucial to nd drugs as antivirals to inhibit viral replication. However, the closure of research laboratories to prevent the contagion and spread of the virus has made it di cult to conduct trials to test the effectiveness of drugs [37]. Therefore, an alternative strategy is a search for drugs against SARS-CoV-2 using computational techniques, whose in vitro and in vivo action could be later corroborated by traditional experimental methods.
Viral proteins such as 3C-like protease and RNA-polymerase of SARS-CoV-2 are two of the proposed targets for preventing viral replication [4,5]. 3C-like protease (NSP5) is a chymotrypsin-like protease, which processes the polyproteins ppa1a and pp1ab, generating unique viral proteins involved in the rst steps of the virus replication [9]. The RNA-polymerase (NSP12) is responsible for the replication and transcription of the genomic RNA, together with the NSP7 and NSP8 cofactors required to stimulate its polymerase activity [11]. Given the critical functions of both proteins, they have been widely proposed as targets for pharmacological action.
Remdesivir is a drug that inhibits the action of different viral RNA polymerases, and Inhibitory activity against SARS-CoV-2 has been demonstrated. It has been tested as a treatment option in phase III clinical trials with patients, showing an improvement; however, there is still mortality of 13% of the patients, and in 60% of those treated, side effects such as liver problems, diarrhea, skin rash, kidney failure, and hypotension occurred [38,39]. Therefore, the search for new effective treatments for COVID19 disease with fewer side effects, which can be used in severe patients and which are available to the entire population, must be studied.
In this work, we analyzed utilizing molecular docking techniques, drugs with a previous in vitro, or in vivo anti-DENV activity. Anthraquinone with a DENV NS3 protease inhibitory activity, and Balapiravir, Fisetin, Hyperoside, and Sofosbuvir with a DENV NS5 RNA-polymerase inhibitory activity. Also, Quercetin with anti-DENV NS3 and NS5 activity was tested [40,41]. Here, the ability of these compounds to bind to SARS-CoV2 NSP5 protease and RNA-polymerase NSP12 were tested.
Fisetin, Quercetin, Hyperoside, and Anthraquinone are drugs derived from plants and fungi. Fisetin and Quercetin are the drugs that exhibited the highest a nity for the SARS-CoV-2 RNA-polymerase active site (-7.08 kcal/mol -1 and -8.1 and -7.26 kcal/mol -1 respectively) and the NiRAN-Fingers domain, and concerning Remdesivir control (-6.81 kcal/mol -1 for the active and 7.67 kcal/mol -1 and the NiRAN-sites), suggesting that so they may be good anti-SARS-CoV-2 drug candidates.
Fisetin and Quercetin are avonols and secondary metabolites from a diversity of plants (e.g., Rhus cotinus and Quercus spp.), with an antibiotic, antiviral, antioxidant, anti-in ammatory, and anti-cancer activities [42,43]. The DENV2 activity of Quercetin was demonstrated in experimental studies with a halfmaximal inhibitory concentration (IC50) of 28.9 µg/ml -1 and the half-maximal cytotoxic concentration (CC50) of 252.6 µg/ml -1 in Vero cells [19]. In DC-SIGN infected cells, Quercetin showed an IC50 of 24.5 µM and a CC50 of 340 µM. It was also found that both Quercetin and Fisetin inhibit DENV2 and DENV3 infection in the absence or presence of antibodies (anti-DENV) and negatively regulate the production of pro-in ammatory cytokines induced during severe DENV infection [44]. Quercetin has been described as a candidate in the prevention and treatment of COVID-19, due to its e cacy in other antiviral diseases, low cost, bioavailability, and low toxicity [45]. Our study would reinforce the experimentally use of this drug in patients with  Hyperoside showed the greatest number of interactions with the NiRAN domain and the active site of RNA-polymerase (Figure 3 and Table 1B); this metabolite comes from the extract of Houttuynia cordata, which is consumed in Thailand. It has shown an anti-DENV activity by slowing down viral RNA synthesis and inactivating infection by blocking the virus entry into HepG2 cells. In DENV-infected LLC-MK2 cells, an EC50 = 0.8 µg/ml and a CC50 of 1.24 mg/ml were obtained [46,47] In this study, the drug with the lowest binding score to the NSP5 protease was Anthraquinone with -6.5 kcal/mol -1 ; however, this binding energy is comparable to that exhibited by Remdesivir (-6.4 kcal/mol -1 ) in other studies, using Autodock Vina [48]. Therefore, its use and effectiveness for Covid-19 cannot be ruled out. Anthraquinones are plants (Rheum palmatum, Cassia obtusifolia, Morinda panamensis, etc.) [49] and fungal (Monascus spp.) metabolites with pharmaceutical, food, and clothing dye use [50]. Experimental studies with both computational and traditional DENV2 have shown that it reduces viral replication by a concentration of 4.2 µM (EC50) and a CC50 of 69 uM in LLC-MK2 cells [25]. Anthraquinone also has been tested on DENV infected Huh7 cells, where an EC50 was of 2.69 µM, and a CC50 of 106.6 µM were found, in which there was an absence of cytotoxicity during three days of incubation [51]. Besides, anthraquinone derivatives have been shown to have anti-Herpetic, anti-DENV, anti-human immunode ciency virus (HIV), and antineoplastic effects [41,52,53].
The RNA-polymerase inhibitors Balapiravir and Sofosbuvir demonstrated a high a nity for the active and the NiRAN sites of SARS-CoV-2 RNA-polymerase, even higher than the drug Remdesivir, suggesting they can be excellent candidates for the treatment of COVID-19.
Balapiravir is a nucleoside 5'-triphosphate analog and Sofosbuvir is a uridine nucleotide analog, which are alternative substrates for viral polymerase and competitively inhibits viral RNA synthesis [54,55].
Balapiravir was developed for the treatment of hepatitis C virus (HCV) and was tested in humans with chronic phase 1b HCV infection, exhibiting a viral suppression of 3.7 log(10) in 14 days of treatment at its highest dose of 4500 mg, where reversible hematological changes occurred in the patients. The tolerable dose of 300 mg reduced the infection by 2.6 log(10) [56]. Because of its similarity between the NS5 polymerases of both HCV and DENV, Balapiravir was tested in phase II human trials, in which it showed no effect on viremia or fever elimination [57]. However, another study reported that Balapiravir in its triphosphate form acts during DENV infection, but cell activation may decrease the conversion of the drug by requiring a host enzyme and losing its activity [58]. Therefore, the conversion and effectiveness of the drug in infections with SARS-CoV-2 should be analyzed.
As Balapiravir, Sofosbuvir is a prodrug that must be triphosphorylated in the cell for its action. Clinical studies have been done in phases I and II, and it currently has an FDA approval for the treatment for chronic HCV infections [55]. In Huh7 cells infected with DENV2, Sofosbuvir has been shown to have an a nity for NS5 polymerase, with an EC50 of 4.9 µM and a CC50> 100 µM [59]. Currently, it has been proposed as a drug for the treatment of COVID-19, using a comparative analysis of the binding site in the The structure of 3C-like protease (X-Ray diffraction, 2.20 Å resolution) and RNA polymerase (Electron Microscopy, 3.10 Å resolution) of SARS-CoV-2 were obtained from the Protein Data Bank library (RCSB PDB) with PDB ID: 6M2N and 6NUR, respectively. The quality of the protein structures and the sequence was evaluated using ERRAT [29], Verify 3D [30], and EMBOSS Needle [28] software.

Drugs candidates tested
Six drug candidates with inhibitory activity against 3C-like protease and SARS-CoV-2 RNA-polymerase were selected through literature reviews of drugs with inhibitory activity of viral proteins against DENV [40,41]. They were selected based on in vitro and/or in vivo experimental background, in addition to previous Food and Drug Administration (FDA) development and approval processes.

Visualization of molecular docking results
The molecular docking results were analyzed with the AutoDockTools 1.5.6, PyMOL 2.3.3, and UCSF Chimera 1.14 software [62,64,67], to visualize the best target-ligand bond formation with lower ΔG. The number and type of non-covalent interactions between the drug and the viral protein were obtained using the PLIP software [36].

Statistical analysis
The higher a nity interactions between the candidate drugs directed to the active sites or the NiRAN domain of 3C-like protease and the RNA-polymerase of SARS-CoV-2, with RMSD <1 Å, was statistically analyzed using ordinary one-way ANOVA with Dunnett's multiple comparisons test to determine signi cant differences among ΔG (kcal/mol) means ± standard deviation among the control drugs (Remdesivir and Baicalein) and the candidate drugs (Fisetin, Quercetin, Hyperoside, Balapiravir, Sofosbuvir, and Anthraquinone) using the GraphPad Prism software version 6. The results were considered statistically signi cant when p values were < 0.05.

Declarations
Author Contributions: LADJ-G, JFO-R, and JMR-R collaborated in generating the experimental data. LADJ-G, JFO-R, JMR-R, CNF-M, SNP-R, CDC-R, AMH-M, ALG-E, and RMdÁ, collaborated equally in the manuscript writing. ALG-E and RMdÁ, coordinated and edited the manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This research was supported by CONACYT (Mexico), grants: CB-220824 and A1-S-9005 from RMDA and CB-250696 from ALGE, and Fundación Miguel Alemán. The funders had no role in study design, data collection, and analysis, decision to publish, or preparation of the manuscript.