Feasibility of Known RNA Polymerase Inhibitors as Anti-SARS-CoV-2 Drugs

Coronaviruses (CoVs) are positive-stranded RNA viruses that infect humans and animals. Infection by CoVs such as HCoV-229E, -NL63, -OC43 and -HKU1 leads to the common cold, short lasting rhinitis, cough, sore throat and fever. However, CoVs such as Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV), Middle East Respiratory Syndrome Coronavirus (MERS-CoV), and the newest SARS-CoV-2 (the causative agent of COVID-19) lead to severe and deadly diseases with mortality rates ranging between ~1 to 35% depending on factors such as age and pre-existing conditions. Despite continuous global health threats to humans, there are no approved vaccines or drugs targeting human CoVs, and the recent outbreak of COVID-19 emphasizes an urgent need for therapeutic interventions. Using computational and bioinformatics tools, here we present the feasibility of reported broad-spectrum RNA polymerase inhibitors as anti- SARS-CoV-2 drugs targeting its main RNA polymerase, suggesting that investigational and approved nucleoside RNA polymerase inhibitors have potential as anti-SARS-CoV-2 drugs. However, we note that it is also possible for SARS-CoV-2 to evolve and acquire drug resistance mutations against these nucleoside inhibitors.


Introduction
Human coronaviruses (CoVs) such as HCoV-229E, NL63, OC43 and HKU1 primarily infect the upper respiratory and gastrointestinal tract, causing mostly mild diseases. However, some CoVs such as SARS-CoV and MERS-CoV cause severe respiratory diseases (SARS and MERS) that result in~10% to~35% mortality [1][2][3][4][5]. SARS-CoV caused a pandemic in 2003 with 774 deaths worldwide [6][7][8]. MERS-CoV was first reported in June 2012, and according to a World Health Organization (WHO) report, the potential of remdesivir, the United States Food and Drug Administration (FDA) has granted an orphan drug status on 23 March 2020 so that it can be used in clinics, and a clinical trial has started at the University of Nebraska Medical Center, Omaha, NE (NCT04280705). In order to explore the feasibility of broad-spectrum nucleoside inhibitors of RNA polymerases as potential inhibitors of SARS-CoV-2, we used comparative molecular modeling, docking and bioinformatics to assess these compounds as potential inhibitors of nsp12. More specifically, we present the feasibility of remdesivir, 5-FU, ribavirin, and favipiravir (T-705) as anti-SARS-CoV-2 compounds.

Sequence Conservation Among SARS-CoV, MERS-CoV and SARS-CoV-2 nsp12 Proteins
Nucleoside analog inhibitors are administered as compounds containing a nucleic acid base with modified sugar moiety. These compounds are metabolized into their triphosphate (TP) form by cellular kinases, becoming the bona fide substrates of nucleic acid polymerases. The nucleic acid polymerases contain conserved motifs that participate in nucleoside-TP (NTP) binding [53]. First, we assessed sequence conservation in the NTP-binding motifs using available nsp12 sequences of SARS-CoV, MERS-CoV and SARS-CoV-2.
The modeled structure of SARS-CoV-2 nsp12 (Figure 2) superposed extremely well on to the cryoEM structure of SARS-CoV nsp12 (RMSD of <0.5 Å for 802 Cα atoms). Nine non-conserved residues in the polymerase domain are located at the surface of nsp12 distal to the polymerase active site (D621, D760 and D761). All conserved RdRp motifs (A-G) [13] were easily identifiable in the modeled structure of SARS-CoV-2 nsp12 (Figure 2b). One of the unusual features of modeled SARS-CoV-2 nsp12 is the partial β-strand structure of motif A that contains one carboxylate (D621) of the catalytic triad. In fact, a similar conformation is present in the cryoEM structure of SARS-CoV nsp12 [43]. Motifs A and C are known to form a three-stranded β-sheet composed of one strand from motif A and two strands from motif C in both RdRps and DNA-dependent DNA polymerases. However, the crystal structures of poliovirus RdRp (3Dpol) and enterovirus 71 RdRp (3Dpol) elongation complexes showed subtle conformational changes in the palm subdomain (called the 'active site closure') and that the presence of incoming substrate induces inter-β-strand hydrogen bonds required for classification as β-strand (reviewed by Peersen [54]). Therefore, the partial β-strand structure of motif A is expected to adopt a complete β-strand conformation in the complex consisting of primer-template (pt) and the nucleoside triphosphate (NTP). the crystal structures of poliovirus RdRp (3Dpol) and enterovirus 71 RdRp (3Dpol) elongation complexes showed subtle conformational changes in the palm subdomain (called the 'active site closure') and that the presence of incoming substrate induces inter-β-strand hydrogen bonds required for classification as β-strand (reviewed by Peersen [54]). Therefore, the partial β-strand structure of motif A is expected to adopt a complete β-strand conformation in the complex consisting of primer-template (pt) and the nucleoside triphosphate (NTP).

Remdesivir
Remdesivir (GS-441524) is a 1′-cyano 4-aza-7,9-dideazaadenosine C-adenosine nucleoside analog. It is a broad-spectrum RNA polymerase inhibitor that has been shown to inhibit human and mouse CoVs [52]. More importantly, remdesivir has been shown to inhibit SARS-CoV-2 in vitro [48,55]. A recent report showed that compassionate-use of remdesivir showed improvement in 68% of COVID-19 patients [56]. The antiviral activity of remdesivir against SARS-CoV-2 is not surprising as it is a nucleoside analog and expected to bind at the NTP-binding site, which is highly conserved among SARS-CoV, MERS-CoV and SARS-CoV-2 nsp12 polymerases ( Figure 1b). Except motifs D and G, all other motifs either directly participate in NTP binding/hydrolysis or are spatially located in close vicinity of the remdesivir-TP binding site. A molecular model consisting of enzyme, remdesivir-TP and RNA (pt) is shown in Figure 3a.
Motifs A and C harbor catalytic site carboxylates and motif B binds the base/sugar moiety of the NTP. Both are close to remdesivir-TP ( Figure 3a). Motif E, which is in the vicinity of the NTP binding pocket, is present only in RNA polymerases, and has been termed as 'primer grip' [6,7]. This motif is also in close proximity to remdesivir-TP. Motif F contains a highly conserved basic residue, which interacts with the TP moiety of NTP. We also identified R558 in the SARS-CoV-2 nsp12/pt/remdesivir-TP model as the conserved motif F basic residue, which interacts with the β-phosphate (Figure 3a).

Remdesivir
Remdesivir (GS-441524) is a 1 -cyano 4-aza-7,9-dideazaadenosine C-adenosine nucleoside analog. It is a broad-spectrum RNA polymerase inhibitor that has been shown to inhibit human and mouse CoVs [52]. More importantly, remdesivir has been shown to inhibit SARS-CoV-2 in vitro [48,55]. A recent report showed that compassionate-use of remdesivir showed improvement in 68% of COVID-19 patients [56]. The antiviral activity of remdesivir against SARS-CoV-2 is not surprising as it is a nucleoside analog and expected to bind at the NTP-binding site, which is highly conserved among SARS-CoV, MERS-CoV and SARS-CoV-2 nsp12 polymerases ( Figure 1b). Except motifs D and G, all other motifs either directly participate in NTP binding/hydrolysis or are spatially located in close vicinity of the remdesivir-TP binding site. A molecular model consisting of enzyme, remdesivir-TP and RNA (pt) is shown in Figure 3a.
Motifs A and C harbor catalytic site carboxylates and motif B binds the base/sugar moiety of the NTP. Both are close to remdesivir-TP ( Figure 3a). Motif E, which is in the vicinity of the NTP binding pocket, is present only in RNA polymerases, and has been termed as 'primer grip' [6,7]. This motif is also in close proximity to remdesivir-TP. Motif F contains a highly conserved basic residue, which interacts with the TP moiety of NTP. We also identified R558 in the SARS-CoV-2 nsp12/pt/remdesivir-TP model as the conserved motif F basic residue, which interacts with the β-phosphate (Figure 3a).
Resistance to remdesivir has been demonstrated in in vitro passage assays [57]. Two mutations (F476L and V553L) in MHV nsp12 appeared after 23 passages. Amino acid residues F476 and V553 counterparts are numbered F483 and V560, respectively in the cryoEM structure of SARS-CoV. Hence, hereafter we will refer to these residues according to their numbering in the cryoEM structure of SARS-CoV nsp12 (i.e., F483 and V560). Both F483 and V560 are absolutely conserved in α-, β-, and γ-CoVs, and belong to the fingers subdomain of nsp12. Their locations relative to remdesivir-TP are shown in Figure 3b,c, respectively. V560 is proximal to motif F. In our model of nsp12/pt/remdesivir-TP complex, V560 is close to the template nucleotide that is base-paired with remdesivir-TP (or incoming NTP). Topologically equivalent valine (V181) interacts with the templating base in the crystal structures Pathogens 2020, 9, 320 6 of 16 of foot-and-mouth disease virus (FMDV) 3Dpol consisting of E/pt/ATP [58] and Coxsackievirus 3B 3Dpol [54]. Therefore, mutation V560L in nsp12 may alter the position of the template nucleotide and reduce the binding of remdesivir-TP, thereby imparting resistance to remdesivir.
Hence, hereafter we will refer to these residues according to their numbering in the cryoEM structure of SARS-CoV nsp12 (i.e., F483 and V560). Both F483 and V560 are absolutely conserved in α-, β-, and γ-CoVs, and belong to the fingers subdomain of nsp12. Their locations relative to remdesivir-TP are shown in Figure 3b,c, respectively. V560 is proximal to motif F. In our model of nsp12/pt/remdesivir-TP complex, V560 is close to the template nucleotide that is base-paired with remdesivir-TP (or incoming NTP). Topologically equivalent valine (V181) interacts with the templating base in the crystal structures of foot-and-mouth disease virus (FMDV) 3Dpol consisting of E/pt/ATP [58] and Coxsackievirus 3B 3Dpol [54]. Therefore, mutation V560L in nsp12 may alter the position of the template nucleotide and reduce the binding of remdesivir-TP, thereby imparting resistance to remdesivir.   F483 is located adjacent to motif B and forms hydrophobic interactions with V640 and V696 of motif B. Mutation to L483 results in a shorter side chain yet maintains hydrophobic interactions with neighboring V640 and V696. It is possible that subtle changes in the hydrophobic interactions may assist in the known mechanism of active site closure in RdRps [29,30] to enhance fidelity of nsp12, i.e., preferential selection of NTP over remdesivir-TP.

5-fluorouracil (5-FU)
The polymerase domain of SARS-CoV nsp12 has a high structural homology with picornavirus 3Dpol [43]. Hence, we reasoned that the nucleoside analogs, known to inhibit the well-studied FMDV 3Dpol, might inhibit nsp12. The mechanism of inhibition of two major nucleoside analogs, 5-FU and ribavirin, has been structurally well studied [58][59][60]. Below, we discuss the feasibility of 5-FU and ribavirin for the inhibition SARS-CoV-2 nsp12. 5-FU is a pyrimidine analog that has been used in clinics as an anti-cancer drug for many years [61,62]. Additionally, it is a mutagen for several viruses [46,63]. Incorporation of 5-FU-monophosphate (5-FUMP) into the viral genome by RdRps leads to error catastrophe [64,65]. Efficient extinction of FMDV has been achieved by 5-FU in combination with guanidine hydrochloride and heparin [66]. Additionally, 5-FU after its conversion to 5-FU-triphosphate (5-FUTP) blocks initiation of FMDV RNA synthesis and therefore functions as an initiation inhibitor [67]. Mutations in RdRp enzymes under 5-FU pressure impart fitness loss in the absence of 5-FU, but confer a fitness gain in presence of 5-FU. Most RNA viruses do not possess a proofreading activity. Therefore, these viruses overcome the effect of mutagens by selecting resistance mutations that enhance the fidelity RNA synthesis [68].
CoVs also encode nsp14, which acts as a proofreading enzyme. Deletion of nsp14 renders SARS-CoV sensitive to 5-FU [69]. Furthermore, mapping the mutations affecting fidelity in Coxsackievirus B3 onto the MHV nsp12 molecular model, and introducing these mutations into MHV with [nsp14-ExoN(+)] or without [nsp14-ExoN(−)] exonuclease activity, two mutations (V560I and M618F) were identified that conferred resistance to 5-FU [46]. Mutation at nsp12 codon 560 (V560L) has also been reported for remdesivir (discussed above). Therefore, resistance to 5-FU by mutation at V560 may occur through the repositioning of templating nucleotide, which, in turn may alter the selectivity of the enzyme for UTP over 5-FUTP.
We have previously reported that mutation V173I in FMDV 3Dpol enhances selectivity of UTP over 5-FUTP [70]. Using pre-steady state kinetics, we showed that V173I mutation in FMDV 3Dpol enhances the selectivity of UTP over 5-FUTP by~3.2-fold compared to the wild-type enzyme. The selectivity of UTP over 5-FUTP by V173I 3Dpol was primarily due to the increase in the dissociation of 5-FUTP from the elongation complex, which resulted in restricted 5-FUMP incorporation [70]. FMDV containing V173I survived the mutagenic activity of 5-FU by compensating for the increase in A→G and U→C transitions that the wild-type virus endures in the presence of 5-FU [70]. Compensation in the mutant virus entails an increase of G→A and C→U transitions in the presence of 5-FU, which approximates the mutational pattern to that of the wild-type virus replicating in the absence of 5-FU [70]. Due to the fact that CoVs contain an exonuclease enzyme, the change in NTP selectivity may be a primary mechanism of 5-FU resistance, since the misincorporation of 5-FUMP would most likely be corrected by the nsp14 exonuclease.
5-FU resistance mutation position M618 belongs to the conserved motif A. As described above, the active site closure mechanism of RdRps serves as a fidelity control [71]. A comparison of RdRp palm domains suggests that all (+) strand RNA viruses use this active site closure mechanism to optimize the fidelity of RNA synthesis [14]. As shown in Figure 4, M618 is clustered among hydrophobic residues emanating from motifs A (dark-red), C (red), and D (green). Mutation M618F will result in the introduction of a bulky side chain (phenylalanine), which is also more hydrophobic than methionine. This may lead to a subtle change in the palm-based closure mechanism of RdRps (in the case of nsp12) and therefore enhance the fidelity of RNA synthesis. M618 is topologically equivalent to position I230 in Coxsackievirus and F230 of poliovirus. Mutation at this position has been shown to affect the fidelity of the RdRp [14]. Hence, it is possible that selection of M618F in the presence of 5-FU is related to the fidelity of nsp12.
Resistance to ribavirin in different RNA viruses is achieved by mutations in the RdRp coding gene. In poliovirus, FMDV and enterovirus 71, mutations in the 3Dpol (RdRp) confer resistance to ribavirin [84][85][86][87]. HCV develops resistance to ribavirin (when used in combination with pegylated interferon) by blocking downstream signaling actions of STAT1, STAT2, IRF9 and JAK-STAT pathways [88,89], and by mutation in the RdRp gene [90,91]. Mutations G64S and L420A in poliovirus 3Dpol, M296I in FMDV 3Dpol, and F415Y in HCV NS5B have been reported to impart resistance to ribavirin. Mutations at G64 and L123 in enterovirus 71 3Dpol have also been reported to confer ribavirin resistance [87].
Structurally, ribavirin resistance mutation sites in RNA polymerases do not appear to be in absolutely equivalent positions. A structural alignment showed that G64 in poliovirus 3Dpol is ~17 Å away from the active site, whereas M296 in FMDV 3Dpol is part of the NTP binding site (i.e., within 12 Å). While Y415 in HCV NS5B and L420 in poliovirus 3Dpol are at topologically equivalent positions, they are ~22 Å away in the thumb subdomain. In poliovirus 3Dpol, resistance to ribavirin is achieved by mutations at G64 and at L420 ( Figure 5). These two mutation sites are almost posterior to the active site. Residue D868 of nsp12 is topologically equivalent to L420 of poliovirus 3Dpol, whereas N462 (nsp12) can be tentatively assigned as the equivalent to G64 (3Dpol). Both Y415 (of NS5B) and L420 (of poliovirus 3Dpol) interact with the RNA primer strand near the active site [92,93]. Mutation G64S in poliovirus 3Dpol and M296I in FMDV 3Dpol change the fidelity of the two enzymes [94,95], whereas mutation L420A facilitates RNA recombination [86]. These examples suggest that resistance to ribavirin can be achieved by more than one mechanism. It is possible that resistance to
Resistance to ribavirin in different RNA viruses is achieved by mutations in the RdRp coding gene. In poliovirus, FMDV and enterovirus 71, mutations in the 3Dpol (RdRp) confer resistance to ribavirin [84][85][86][87]. HCV develops resistance to ribavirin (when used in combination with pegylated interferon) by blocking downstream signaling actions of STAT1, STAT2, IRF9 and JAK-STAT pathways [88,89], and by mutation in the RdRp gene [90,91]. Mutations G64S and L420A in poliovirus 3Dpol, M296I in FMDV 3Dpol, and F415Y in HCV NS5B have been reported to impart resistance to ribavirin. Mutations at G64 and L123 in enterovirus 71 3Dpol have also been reported to confer ribavirin resistance [87].
Structurally, ribavirin resistance mutation sites in RNA polymerases do not appear to be in absolutely equivalent positions. A structural alignment showed that G64 in poliovirus 3Dpol is 17 Å away from the active site, whereas M296 in FMDV 3Dpol is part of the NTP binding site (i.e., within 12 Å). While Y415 in HCV NS5B and L420 in poliovirus 3Dpol are at topologically equivalent positions, they are~22 Å away in the thumb subdomain. In poliovirus 3Dpol, resistance to ribavirin is achieved by mutations at G64 and at L420 ( Figure 5). These two mutation sites are almost posterior to the active site. Residue D868 of nsp12 is topologically equivalent to L420 of poliovirus 3Dpol, whereas N462 (nsp12) can be tentatively assigned as the equivalent to G64 (3Dpol). Both Y415 (of NS5B) and L420 (of poliovirus 3Dpol) interact with the RNA primer strand near the active site [92,93]. Mutation G64S in poliovirus 3Dpol and M296I in FMDV 3Dpol change the fidelity of the two enzymes [94,95], whereas mutation L420A facilitates RNA recombination [86]. These examples Pathogens 2020, 9, 320 9 of 16 suggest that resistance to ribavirin can be achieved by more than one mechanism. It is possible that resistance to ribavirin in SARS-CoV, MERS-CoV and SARS-CoV-2 can develop through mutation at D868 or through yet another unknown mutation and/or mechanism.
Pathogens 2020, 9, x FOR PEER REVIEW 9 of 16 ribavirin in SARS-CoV, MERS-CoV and SARS-CoV-2 can develop through mutation at D868 or through yet another unknown mutation and/or mechanism.
Favipiravir was discovered by chemical modification of a pyrazine analog (T-1106) [96]. After entering the cell, favipiravir is metabolized into the triphosphate form (T-705-RTP) that is recognized by RdRps. T-705-RTP competes with ATP and GTP [101], suggesting that it is recognized as a purine analogue. In contrast to many nucleoside inhibitors, favipiravir does not have a sugar moiety when administered. Human hypoxanthine guanine phosphoribosyltransferase converts T-705 into its ribose-5'-monophosphate (RMP) prior to formation of T-705-RTP [102]. Mechanisms of inhibition by T-705 have been demonstrated by the chain termination of nascent RNA [103] and by induction of lethal mutagenesis [104,105]. Currently, favipiravir is being evaluated for COVID-19 treatment and the results are awaited.
Resistance to T-705 by RdRps is achieved by mutation of a conserved lysine of motif F. Thus, in chikungunya RdRp, K291R exerts low-level resistance to T-705 [101], and mutation K159R in Coxsackievirus B3 (CVB3) 3Dpol resulted in a nonviable virus [106]. The replication competence of K159R virus was restored by the A239G mutation. Biochemical results suggested that K159R reduced the processivity of CVB3 3Dpol, and the double mutant (K159R/A239G) had low fidelity [106]. The CVB3 K159 equivalent in nsp12 is K548. Currently, it is not known if the mutation of K548 will have effect on favipiravir. However, considering its conserved position, the resistance to favipiravir by SARS-CoV-2 is quite possible.

Sequence Retrieval and Phylogenetic Analysis
The nsp12 protein sequences of Bat CoV, SARS-CoV, MERS-CoV and SARS-CoV-2 were retrieved using BLASTp (protein-protein BLAST) algorithm with BLOSUM62 matrix. Multiple sequence analysis was performed in AliView software. The ML tree was inferred using RAxML v8.1.20 [107]. The branch supports were computed out of 100 bootstrapped trees. The tree was visualized in FigTree v1.4.4 (http://tree.bio.ed.ac.uk/). The CIRCOS plot was created using Circos
Favipiravir was discovered by chemical modification of a pyrazine analog (T-1106) [96]. After entering the cell, favipiravir is metabolized into the triphosphate form (T-705-RTP) that is recognized by RdRps. T-705-RTP competes with ATP and GTP [101], suggesting that it is recognized as a purine analogue. In contrast to many nucleoside inhibitors, favipiravir does not have a sugar moiety when administered. Human hypoxanthine guanine phosphoribosyltransferase converts T-705 into its ribose-5'-monophosphate (RMP) prior to formation of T-705-RTP [102]. Mechanisms of inhibition by T-705 have been demonstrated by the chain termination of nascent RNA [103] and by induction of lethal mutagenesis [104,105]. Currently, favipiravir is being evaluated for COVID-19 treatment and the results are awaited.
Resistance to T-705 by RdRps is achieved by mutation of a conserved lysine of motif F. Thus, in chikungunya RdRp, K291R exerts low-level resistance to T-705 [101], and mutation K159R in Coxsackievirus B3 (CVB3) 3Dpol resulted in a nonviable virus [106]. The replication competence of K159R virus was restored by the A239G mutation. Biochemical results suggested that K159R reduced the processivity of CVB3 3Dpol, and the double mutant (K159R/A239G) had low fidelity [106]. The CVB3 K159 equivalent in nsp12 is K548. Currently, it is not known if the mutation of K548 will have effect on favipiravir. However, considering its conserved position, the resistance to favipiravir by SARS-CoV-2 is quite possible.

Molecular Modeling
Homology-derived molecular model of SARS-CoV-2 nsp12 was generated using 'Prime' of Schrödinger Suite (Schrödinger LLC, New York, NY, USA) and the cryoEM structure of SARS-CoV [43] (PDB file 6NUR). To generate ternary complex containing enzyme/pt/NTP or enzyme/pt/nucleoside-TP, the pt and RTP from the crystal structure of FMDV 3Dpol [58] (PDB file 2E9R) were extracted and docked into the modeled structure of nsp12. The templating nucleotide was modified as required for the complementarity of the incoming substrate. All the complexes were energy minimized using the Jaguar program of Schrodinger Suite.

Conclusions
In conclusion, here we show that the nucleoside inhibitor binding pocket is largely conserved among diverse RNA-dependent RNA polymerases, and that the broad-spectrum nucleoside inhibitors discussed may have potential in COVID-19 treatment. While it is possible that SARS-CoV-2 may evolve with drug resistance mutations against these nucleoside inhibitors, knowledge of potential escape mutants may aid in the development of more specific SARS-CoV-2 inhibitors with a higher resistance barrier. The emerging genomic sequences and structures of SARS-CoV and SARS-CoV-2 nsp12 also offer increasing insight into the design and identification of novel nucleoside inhibitors or small molecules that are specific to SARS-CoV-2 nsp12, and could be used against the current COVID-19 pandemic or in future CoV outbreaks. The in vitro analysis that is presently on-going in the laboratory will provide a better picture of their comparative in vitro potency and resistance profile.
Additionally, the use of these antivirals has an added benefit, as a significant wealth of knowledge already exists regarding their administration, efficacy, toxicity and side effects in humans, which can speed up clinical trials in COVID-19 patients. Inhibitors targeting nsp12 will block the replication of both (+) and (-) strand of the viral genome, which is essential for the formation of a mature and infectious virus.