Natural Compounds as Non-Nucleoside Inhibitors of Zika Virus Polymerase through Integration of In Silico and In Vitro Approaches

Although the past epidemic of Zika virus (ZIKV) resulted in severe neurological consequences for infected infants and adults, there are still no approved drugs to treat ZIKV infection. In this study, we applied computational approaches to screen an in-house database of 77 natural and semi-synthetic compounds against ZIKV NS5 RNA-dependent RNA-polymerase (NS5 RdRp), an essential protein for viral RNA elongation during the replication process. For this purpose, we integrated computational approaches such as binding-site conservation, chemical space analysis and molecular docking. As a result, we prioritized nine virtual hits for experimental evaluation. Enzymatic assays confirmed that pedalitin and quercetin inhibited ZIKV NS5 RdRp with IC50 values of 4.1 and 0.5 µM, respectively. Moreover, pedalitin also displayed antiviral activity on ZIKV infection with an EC50 of 19.28 µM cell-based assays, with low toxicity in Vero cells (CC50 = 83.66 µM) and selectivity index of 4.34. These results demonstrate the potential of the natural compounds pedalitin and quercetin as candidates for structural optimization studies towards the discovery of new anti-ZIKV drug candidates.


Introduction
Zika Virus (ZIKV) is an arthropod-borne flavivirus that circulates globally and caused a worldwide concern due to its exponential spread in the Americas in 2015-2016 [1] and its association with severe congenital effects in pregnant women infected with the virus. The congenital ZIKV syndrome is characterized by neurological and neuropsychomotor complications, ophthalmological and hearing problems, craniofacial disproportion, epilepsy, cerebral palsy and microcephaly [2]. In adults, ZIKV can cause the Guillain-Barre syndrome [3]. Recently, researchers suggested that ZIKV strains with enhanced transmissibility and pathogenicity can reemerge [4].
ZIKV is constituted by a single-strand negative RNA which encodes three structural proteins, membrane (M), envelope (E) and capsid protein (C), arranged on a lipidic mem-ZIKV is constituted by a single-strand negative RNA which encodes three structura proteins, membrane (M), envelope (E) and capsid protein (C), arranged on a lipidic membrane, and seven non-structural (NS) proteins: NS1, NS2A, NS2B, NS3, NS4A, NS4B and NS5 [5]. Among the NS proteins, the NS5 RNA-dependent RNA-polymerase (RdRp is an essential protein, catalyzing the replication of viral RNA from the RNA template [6] and has been considered a promising target for ZIKV drug discovery. The nucleoside and nucleotide inhibitors (NI) of RdRp bind to the catalytic and RNA binding sites [7], whereas the non-nucleosides inhibitors (NNI) bind to the N-pocke (allosteric site) [8]. The NI antiviral drug sofosbuvir has been successfully used agains Hepatitis C virus (HCV) and depends on the activation by host kinases [9]. Sofosbuvi was also tested against ZIKV RdRp presenting an IC50 value of 0.38 ± 0.03 µM [10].
Computer-Assisted Drug Design (CADD) [11] techniques rationally promote the discovery, prioritization and optimization of drug candidates, using computationa resources, such as databases, algorithms, programs and web servers. Compared to experimental approaches, such as high-throughput screening (HTS), computationa techniques have been shown to be faster and presented higher success rates [12].
The present study aimed to discover new potential ZIKV NS5 RdRp inhibitor guided by computational and experimental approaches. DENV and ZIKV NS5 RdRp primary and tertiary sequences share high similarities. Due to this fact, DENV NS5 RdRp known inhibitors were used to search for new potential ZIKV NS5 RdRp hits. Docking calculations were performed to prioritize virtual hits, and enzymatic assays validated these computational predictions, showing that pedalitin and quercetin, two natura compounds, inhibited ZIKV NS5 RdRp. Moreover, both hits presented anti-ZIKV activity in in vitro antiviral assays, with low cytotoxicity. These results demonstrate tha integrated in silico and in vitro approaches can be used to accelerate the discovery of new ZIKV antiviral candidates.

Results and Discussion
A general workflow of the computational and experimental steps applied in thi study is presented in Figure 1.

Binding Site Conservation Analysis
The binding-site conservation can provide an invaluable resource to understanding the affinity and binding mode of small molecules between homologs. In theory, proteins sharing a high similarity have the probability of sharing the same ligands [13]. Here, we employed the ConSurf analysis [14][15][16] to predict the evolutionary conservation profile of ZIKV RdRp amino acids based on the phylogenetic relations between homologous sequences such as DENV RdRp. All the polymerases resemble a right hand, with the three main regions (Figure 2a): fingers (residues 321-488 and 542-608), palm (residues 489-541 and 609-714), and thumb (residues 715-903). The RdRp domain is composed of three binding sites: the RNA site, the N-pocket (allosteric site) and the catalytic binding site [7,8,17]. The RNA site is a tunnel that single-stranded RNA enters and serves as a template for the formation of double-stranded RNA. The N-pocket is a tunnel through which the nucleotides enter. At this site, the initiation loop regulates template RNA binding and nucleotide entry. Finally, the catalytic site performs double-stranded RNA catalysis [18].
binding sites: the RNA site, the N-pocket (allosteric site) and the catalytic binding s [7,8,17]. The RNA site is a tunnel that single-stranded RNA enters and serves as a templ for the formation of double-stranded RNA. The N-pocket is a tunnel through which nucleotides enter. At this site, the initiation loop regulates template RNA binding a nucleotide entry. Finally, the catalytic site performs double-stranded RNA catalysis [1 Although the DENV and ZIKV RdRp proteins show 64.59% of sequential ident the evolutionary analysis of viral RdRps shows that ZIKV N-pocket is highly conserv (Figure 2b). These results suggest that the amino acid composition of the N-pocke strongly associated with its structural and functional importance. As we can see in Figu 2c, all DENV (highlighted in gray, PDB ID: 5I3Q [18]) and ZIKV (highlighted in cyan, P ID: 6LD4 [19]) N-pocket residues are conserved, except for Leu767 in ZIKV, replaced Met765 in DENV RdRp. It is important to point out that these two amino acids sh similar volumes and electronic properties, and thus should not promote significat changes in the binding of small molecules to the N-pocket. The high conservation state N-pockets corroborates with a high probability of ZIKV and DENV RdRps to share same ligands. Based on these findings, an unsupervised cheminformatics approach us known DEV RdRp inhibitors was performed to find ZIKV RdRp hits from an in-hou collection of natural and semi-synthetic compounds.  Although the DENV and ZIKV RdRp proteins show 64.59% of sequential identity, the evolutionary analysis of viral RdRps shows that ZIKV N-pocket is highly conserved (Figure 2b). These results suggest that the amino acid composition of the N-pocket is strongly associated with its structural and functional importance. As we can see in Figure 2c, all DENV (highlighted in gray, PDB ID: 5I3Q [18]) and ZIKV (highlighted in cyan, PDB ID: 6LD4 [19]) N-pocket residues are conserved, except for Leu767 in ZIKV, replaced by Met765 in DENV RdRp. It is important to point out that these two amino acids share similar volumes and electronic properties, and thus should not promote significative changes in the binding of small molecules to the N-pocket. The high conservation state of N-pockets corroborates with a high probability of ZIKV and DENV RdRps to share the same ligands. Based on these findings, an unsupervised cheminformatics approach using known DEV RdRp inhibitors was performed to find ZIKV RdRp hits from an in-house collection of natural and semi-synthetic compounds.
As shown in the t-SNE plot ( Figure 3A), 24 compounds from the in-house collection share the same chemical space of the known NS5 RdRp inhibitors. Most of them belong to the classes of acridones, diphenylamines, and flavonoids ( Figure 3B). In view of this, these compounds were prioritized for molecular docking to assess their binding modes in ZIKV RdRp protein [19]. Since the scaffolds of the compounds are different from the nucleoside-like structure, the analysis was focused on the N-pocket site.
As shown in the t-SNE plot ( Figure 3A), 24 compounds from the in-house c share the same chemical space of the known NS5 RdRp inhibitors. Most of them to the classes of acridones, diphenylamines, and flavonoids ( Figure 3B). In view these compounds were prioritized for molecular docking to assess their binding m ZIKV RdRp protein [19]. Since the scaffolds of the compounds are different f nucleoside-like structure, the analysis was focused on the N-pocket site.

Docking Calculations at the ZIKV NS5 RdRp (N-Pocket)
The prioritized compounds from chemical space analysis were submitted to calculations to rank the most promising hits as well as to predict the binding affin docking poses were analyzed according to the following parameters: (i) docking s intermolecular interactions at the N-pocket binding site; (ii) overlap and bindin similarity with the ZIKV RdRp NNI co-crystallized ligand 5-(3-fluorothiophen hydroxy-4-methoxy-N-[4-(trifluoromethyl)benzenesulfonyl]benzamide and (iii efficiency. The redocking of the co-crystallized ligand was performed to ve accuracy of the docking protocol in predicting the position of the ligands wi

Docking Calculations at the ZIKV NS5 RdRp (N-Pocket)
The prioritized compounds from chemical space analysis were submitted to docking calculations to rank the most promising hits as well as to predict the binding affinities. All docking poses were analyzed according to the following parameters: (i) docking score and intermolecular interactions at the N-pocket binding site; (ii) overlap and binding mode similarity with the ZIKV RdRp NNI co-crystallized ligand 5- Almost all ligands presented acceptable docking scores, close to the redocking score of the co-crystalized compound (docking score −8.73 Kcal·mol −1 ). Moreover, analyzing the binding modes and interactions, 16 ligands were prioritized for in vitro experimental validation. Thirteen of them had similar binding modes with the known ZIKV RdRp NNI. From them, 12 compounds presented ligand efficiency (LE) greater than 0.3 Kcal·mol −1 ·nonhydrogen atom −1 . LE is value that normalizes the binding affinity (∆G) or docking score with respect to the number of non-hydrogen atoms (n) [43][44][45]. The normalization of molecular weights influences the likelihood that a hit compound can be further optimized into prospective hit-to-lead investigations, as larger compounds tend to show greater docking scores due to the larger number of interactions [46,47].
A medicinal chemistry-based inspection was conducted [47,48], considering favorable scores for a higher number of hydrogen bonds between ligand and protein residues; salt bridges; π-cation and π-stacking interactions; and unfavorable scores for nonpolar regions of the ligand exposed to solvent. After this inspection, nine compounds were prioritized for the experimental evaluation (Table 1).
Four out of the nine virtual hits are naturally-occurring flavonoids (chrysin (6), sorbifolin (7), pedalitin (8) and quercetin (9)). Flavonoids have already been described in the literature as inhibitors of the RdRp domains of DENV and ZIKV [49]. Three compounds belong to the class of acridones, a class already described by some authors due to their antiviral activity and capability of inhibition of DNA and RNA viruses [50,51]. A potent activity of N-substituted acridones has already been demonstrated against DENV-2, blocking its multiplication in vitro [52]. ARORA and coworkers [53] demonstrated that compounds containing the diphenylamine subunit were able to inhibit the RdRp domain of DENV including the compound bis-chloro-diphenylamine, 2-aminoindan-2-carboxyl derivative NITD-434 (13) (Figure 4) that interacts with residues Thr795 and Thr796 of the N-pocket site. Three of the nine hits are diphenylamines.
RdRp inhibitors have been classified as NI and NNI. The NIs present a structural similarity to nucleosides and have to be converted into triphosphate forms by host kinases to be incorporated into viral DNA or RNA, acting as chain terminators [54]. On the other hand, the NNIs interact directly with viral polymerase and present different scaffolds, such as flavonoids, alkaloids, acetylenic acids, terpenes, steroids, benzothiazine 2,2-dioxide analogs, pyrazole-5-phenylamine analogs, thiophene-based analogs, N-sulfonylpyrazoles and Nsulfonylanthranilic acids, thiazolidinone-thiadiazole and pyridobenzothiazole analogs [49]. NNIs act into the RdRp allosteric site and, in general, display fewer side effects since they are more selective for viral than host polymerase targets [55]. Among the DENV NNIs, there are natural products including flavonoids 10, 11 and 12 ( Figure 4). Furthermore, another DENV RdRp NNI, the bis-chloro-diphenylamine, 2-aminoindan-2-carboxyl derivative compound (13) or NITD-434 ( Figure 4), occupies the template RNA site and performs interactions with conserved residues between the four serotypes of DENV and ZIKV [53]. The synthetic co-crystallized DENV RdRp NNIs acylsulfonamide derivatives compounds (14) and (15) [8] (Figure 4), occupy the N-pocket site and had IC 50 values ranging from 0.172 to 5.46 µM for compound 14 and 0.023 to 0.427 µM for compound 15 [10,56]. Among the ZIKV NNIs, there are few natural compounds such as chalcones and alkaloids, as well as synthetic compounds undecylenic acid compound 17 ( Figure 4) and thienylcarbonylpiperazinyl-benzothiophene (TBP), compound 16 ( Figure 4) that act to inhibit ZIKV NS5 RdRp [57]. docking score with respect to the number of non-hydrogen atoms (n) [43][44][45]. The normalization of molecular weights influences the likelihood that a hit compound can be further optimized into prospective hit-to-lead investigations, as larger compounds tend to show greater docking scores due to the larger number of interactions [46,47]. A medicinal chemistry-based inspection was conducted [47,48], considering favorable scores for a higher number of hydrogen bonds between ligand and protein residues; salt bridges; π-cation and π-stacking interactions; and unfavorable scores for nonpolar regions of the ligand exposed to solvent. After this inspection, nine compounds were prioritized for the experimental evaluation (Table 1). Anthranilic acid derivative −8.12 0.43 2 normalization of molecular weights influences the likelihood that a hit compound can be further optimized into prospective hit-to-lead investigations, as larger compounds tend to show greater docking scores due to the larger number of interactions [46,47]. A medicinal chemistry-based inspection was conducted [47,48], considering favorable scores for a higher number of hydrogen bonds between ligand and protein residues; salt bridges; π-cation and π-stacking interactions; and unfavorable scores for nonpolar regions of the ligand exposed to solvent. After this inspection, nine compounds were prioritized for the experimental evaluation (Table 1). further optimized into prospective hit-to-lead investigations, as larger compounds tend to show greater docking scores due to the larger number of interactions [46,47]. A medicinal chemistry-based inspection was conducted [47,48], considering favorable scores for a higher number of hydrogen bonds between ligand and protein residues; salt bridges; π-cation and π-stacking interactions; and unfavorable scores for nonpolar regions of the ligand exposed to solvent. After this inspection, nine compounds were prioritized for the experimental evaluation (Table 1). to show greater docking scores due to the larger number of interactions [46,47]. A medicinal chemistry-based inspection was conducted [47,48], considering favorable scores for a higher number of hydrogen bonds between ligand and protein residues; salt bridges; π-cation and π-stacking interactions; and unfavorable scores for nonpolar regions of the ligand exposed to solvent. After this inspection, nine compounds were prioritized for the experimental evaluation (Table 1). Four out of the nine virtual hits are naturally-occurring flavonoids (chrysin (6), sorbifolin (7), pedalitin (8) and quercetin (9)). Flavonoids have already been described in the literature as inhibitors of the RdRp domains of DENV and ZIKV [49]. Four out of the nine virtual hits are naturally-occurring flavonoids (chrysin (6), sorbifolin (7), pedalitin (8) and quercetin (9)). Flavonoids have already been described in the literature as inhibitors of the RdRp domains of DENV and ZIKV [49]. Three compounds belong to the class of acridones, a class already described by some authors Flavonoid −8.14 0.43 Four out of the nine virtual hits are naturally-occurring flavonoids (chrysin (6), sorbifolin (7), pedalitin (8) and quercetin (9)). Flavonoids have already been described in the literature as inhibitors of the RdRp domains of DENV and ZIKV [49]. Three compounds belong to the class of acridones, a class already described by some authors due to their antiviral activity and capability of inhibition of DNA and RNA viruses [ Four out of the nine virtual hits are naturally-occurring flavonoids (chrysin (6), sorbifolin (7), pedalitin (8) and quercetin (9)). Flavonoids have already been described in the literature as inhibitors of the RdRp domains of DENV and ZIKV [49]. Three compounds belong to the class of acridones, a class already described by some authors due to their antiviral activity and capability of inhibition of DNA and RNA viruses [ Four out of the nine virtual hits are naturally-occurring flavonoids (chrysin (6), sorbifolin (7), pedalitin (8) and quercetin (9)). Flavonoids have already been described in the literature as inhibitors of the RdRp domains of DENV and ZIKV [49]. Three compounds belong to the class of acridones, a class already described by some authors due to their antiviral activity and capability of inhibition of DNA and RNA viruses [50,51]. A potent activity of N-substituted acridones has already been demonstrated against DENV-2, blocking its multiplication in vitro [52]. ARORA and coworkers [53]  Recently, the flavonoids luteolin and quercetin were tested against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) RdRp and presented IC 50 s values of 4.6 µM and 6.9 µM, respectively [58]. The authors also performed docking and molecular dynamic simulations of both ligands at the N-pocket and RNA binding sites, suggesting that they may properly bind to both sites.

Pedalitin and Quercetin Inhibits ZIKV RdRp Activity
Nine prioritized hits were submitted to endpoint assay at 20 µM to verify their inhibitory activity against ZIKV RdRp. Most of the compounds evaluated did not obtain significant inhibition results. Pedalitin and quercetin were the only ones with activity greater than 80%, with inhibitory activity of 97% and 99%, respectively, and consequently were selected for the concentration-response assays.
We then investigated ZIKV RdRp activity in the presence of pedalitin and quercetin. A concentration-response assay was performed at concentrations ranging from 80 µM to 0.156 µM to determine the inhibitory concentration of 50% (IC 50 ). From this range of concentrations, it was determined that pedalitin and quercetin had IC 50 values of 4.1 ± 0.3 µM and 0.5 ± 0.1 µM, respectively (Supplementary Figure S1). The enzymatic activities obtained are in agreement with those described for other flavonoids, as shown in Figure 4.

Pedalitin and Quercetin Binding Modes Predicted by Docking
From the enzymatic data, the two flavonoids pedalitin and quercetin were highlighted as promising ZIKV RdRp inhibitors. In Figure 5A,B we show the binding mode of quercetin and pedalitin, predicted by our docking calculations. Quercetin presented four interactions highlighted with an asterisk (*) ( Figure 5A), that are the same interactions performed by the co-crystallized 5-(3-fluorothiophen-2-yl)-2-hydroxy-4-methoxy-N-[4-(trifluoromethyl)benzenesulfonyl]benzamide compound. These interactions are hydrogen bonds with residues Ser712, Arg731, Trp797, Ser798 and Asp666 (catalytic triad residue). In the same way, pedalitin presented four hydrogen bonds, with Ser712, Ser798, Trp797 and Thr796 and a cation-π interaction with the residue Arg731. Moreover, it also interacts with Asp666 via a hydrogen bond.

Pedalitin and Quercetin Inhibits ZIKV RdRp Activity
Nine prioritized hits were submitted to endpoint assay at 20 µM to verify th inhibitory activity against ZIKV RdRp. Most of the compounds evaluated did not obta significant inhibition results. Pedalitin and quercetin were the only ones with activ greater than 80%, with inhibitory activity of 97% and 99%, respectively, and consequen were selected for the concentration-response assays.
We then investigated ZIKV RdRp activity in the presence of pedalitin and quercet A concentration-response assay was performed at concentrations ranging from 80 µM 0.156 µM to determine the inhibitory concentration of 50% (IC50). From this range concentrations, it was determined that pedalitin and quercetin had IC50 values of 4.1 ± µM and 0.5 ± 0.1 µM, respectively (Supplementary Figure S1). The enzymatic activit obtained are in agreement with those described for other flavonoids, as shown in Figu 4.

Pedalitin and Quercetin Binding Modes Predicted by Docking
From the enzymatic data, the two flavonoids pedalitin and quercetin w highlighted as promising ZIKV RdRp inhibitors. In Figure 5A,B we show the bindi mode of quercetin and pedalitin, predicted by our docking calculations. Querce presented four interactions highlighted with an asterisk (*) ( Figure 5A), that are the sam interactions performed by the co-crystallized 5-(3-fluorothiophen-2-yl)-2-hydroxy methoxy-N-[4-(trifluoromethyl)benzenesulfonyl]benzamide compound. The interactions are hydrogen bonds with residues Ser712, Arg731, Trp797, Ser798 and Asp6 (catalytic triad residue). In the same way, pedalitin presented four hydrogen bonds, w Ser712, Ser798, Trp797 and Thr796 and a cation-π interaction with the residue Arg7 Moreover, it also interacts with Asp666 via a hydrogen bond.

Pedalitin and Quercetin Inhibits ZIKV Replication In Vitro
The anti-ZIKV activities of the pedalitin and quercetin were further investigated through the employment of Vero cells infected with ZIKV wild type (ZIKV BR ) ( Figure 6). For this, a concentration-response assay was performed to determine the effective concentration of 50% (EC 50 ) and cytotoxicity of 50% (CC 50 ), and to calculate the Selective Index (SI = CC 50 /EC 50 ). Vero cells were infected with ZIKV BR and simultaneously treated with pedalitin or quercetin at concentrations ranging from 200 µM to 0.005 µM for 72 h when viral replication rates were assessed ( Figure 6). Cell viability analysis was performed in parallel ( Figure 6).
5, x FOR PEER REVIEW 9 of 17

Pedalitin and Quercetin Inhibits ZIKV Replication In Vitro
The anti-ZIKV activities of the pedalitin and quercetin were further investigated through the employment of Vero cells infected with ZIKV wild type (ZIKV BR ) ( Figure 6). For this, a concentration-response assay was performed to determine the effective concentration of 50% (EC50) and cytotoxicity of 50% (CC50), and to calculate the Selective Index (SI = CC50/EC50). Vero cells were infected with ZIKV BR and simultaneously treated with pedalitin or quercetin at concentrations ranging from 200 µM to 0.005 µM for 72 h when viral replication rates were assessed ( Figure 6). Cell viability analysis was performed in parallel ( Figure 6).  Table 2).
Summarizing computational and experimental data ( Table 2), both pedalitin and quercetin bound to the N-pocket site of ZIKV RdRp, presenting good docking scores, compared to the redocking calculations, and binding site interactions similar with the cocrystalized ligand. Agreeing with docking calculations, enzymatic assays showed that both flavonoids inhibited ZIKV RdRp activity. Moreover, infection assays demonstrated that both compounds presented in vitro antiviral activity, and pedalitin presented a higher selectivity index (SI), representing a more promising hit.   (Table 2).
Summarizing computational and experimental data (Table 2), both pedalitin and quercetin bound to the N-pocket site of ZIKV RdRp, presenting good docking scores, compared to the redocking calculations, and binding site interactions similar with the cocrystalized ligand. Agreeing with docking calculations, enzymatic assays showed that both flavonoids inhibited ZIKV RdRp activity. Moreover, infection assays demonstrated that both compounds presented in vitro antiviral activity, and pedalitin presented a higher selectivity index (SI), representing a more promising hit. Other flavonoid compounds have already demonstrated anti-ZIKV effects on Vero cells [62], such as the flavones baicalein (18) and baicalin (19) (Figure 4), which showed an EC50 of 0.004 µM and 14 µM, respectively [59]. Baicalein was also tested against other flaviviruses, such as Japanese encephalitis virus (JEV) and DENV-2, displaying an EC50 values of 27 ± 4 µM [52] and 55.1 µM, respectively [60,61]. Quercetin, identified in our study, also demonstrated anti-viral activity against DENV-2 virus in the study of Zandi and coworkers [63]. They used DENV-2 infected Vero cells and tested different concentrations of quercetin. At a concentration of 165.4 µM, the replication was reduced by 67%. Concentration-response curves were performed with administration after viral adsorption to the cells, obtaining an EC50 value of 95.6 µM [63].
Alternatively, there is a crucial role in virus-host cell interactions that provide important targets for the development of non-specific acting antivirals [64]. Non-specific antivirals can interfere with viral infection by acting on cellular signaling pathways or by modulating the differentiation and function of several immune cells [65]. This antiviral  (Table 2). Summarizing computational and experimental data (Table 2), both pedalitin and quercetin bound to the N-pocket site of ZIKV RdRp, presenting good docking scores, compared to the redocking calculations, and binding site interactions similar with the co-crystalized ligand. Agreeing with docking calculations, enzymatic assays showed that both flavonoids inhibited ZIKV RdRp activity. Moreover, infection assays demonstrated that both compounds presented in vitro antiviral activity, and pedalitin presented a higher selectivity index (SI), representing a more promising hit.
Alternatively, there is a crucial role in virus-host cell interactions that provide important targets for the development of non-specific acting antivirals [64]. Non-specific antivirals can interfere with viral infection by acting on cellular signaling pathways or by modulating the differentiation and function of several immune cells [65]. This antiviral effect might contribute to a lower probability to develop viral resistance due to their reliance on host cell components [64]. Additionally, the intervention of virus-host interactions can include a broader range of activity with the immune system, especially for unknown emergent viral infections, where replication mechanisms are not elucidated [66].
Flavivirus polymerases have been reported to antagonize the interferon (IFN) signaling pathway via numerous mechanisms, including STAT2 degradation, inhibition of RIG-I, and suppression of IFNAR1 maturation [67][68][69]. Combating flavivirus infections by modulating the signaling pathway could be a factor to improve the infection outcome. In this case, non-specific antivirals are particularly desirable to be used combined with direct-acting antivirals and prepare the scientific community for future epidemics [70].

DENV and ZIKV NS5 RdRp Similarity Analysis
The 3D structure of ZIKV NS5 RdRp (PDB ID: 6LD4 [19]) was submitted to the ConSurf server [14][15][16] for estimating the evolutionary conservation of amino acids, based on their phylogenetic relations with homologues. Initially, 150 homologue sequences were imported from UNIREF-90 database [71]. The sequences with sequential identity <35% or >95% were ignored. A multiple sequence alignment (MSA) of the homologous sequences was built using the MAFFT-L-INS-I method [72] and the phylogenetic tree was built using the neighbor-joining algorithm [73]. Position-specific conservation scores were then computed using the empirical Bayesian method [74]. At the end of this analysis, the 3D structures and FASTA sequences of ZIKV (PDB ID: 5I3Q [8]) and DENV (PDB ID: 6LD4 [19]) were aligned using the PyMol v. 2.4 [75] and UniProt [76], respectively. The root-mean-square deviation (RMSD) was calculated for the distances of the conserved residues.

Chemical Space Analysis of RdRp Inhibitors
The chemical space of known DENV RdRp inhibitors and an in-house collection of natural and semi-synthetic compounds was performed using t-Distributed Stochastic Neighbor Embedding (t-SNE) [42]. The t-SNE dimensionality reduction was performed using scikit-learn v. 1.0.2 [81] and extended connectivity fingerprints (ECFP6) with 2048 bits available on RDkit package v. 2022.03.2 [82].

Molecular Docking
Molecular docking studies were performed using the DockThor VS webserver [88,89]. The grid box was centered at the x, y, and z coordinates of the co-crystallized ligand bound to the N-pocket site. The search algorithm precision mode was set up as the standard configuration of genetic algorithm parameters, and the soft docking mode was activated. At the end of the docking procedure, we used the PLIP server [90] to analyze the intermolecular interaction patterns of the docking poses (hydrogen bonds, hydrophobic interaction, cation-π, π-stacking, water and salt bridge interactions). Finally, the binding mode of the ligands obtained was compared to that of co-crystallized ligand (PDB ID: 6LD4). Then, the Pymol software v. 2.4 [75] was used for visual inspection and to render the pose images.

Quercetin and Pedalitin
Quercetin and pedalitin were obtained from Pterogyne nitens, a medicinal Brazilian tree, according to our previous phytochemical procedures [91].

Protein Cloning, Expression and Purification
ZIKV NS5 RdRp polymerase was cloned at pETTRX by the LIC method and expressed and purified according to the protocol described in [92]. Briefly, NS5 RdRp polymerase was expressed in ZYM 5052 auto-induction medium and purified in four steps: (i) a HisTrap HP 5.0 mL with a Ni Sepharose resin (GE Healthcare, Sao Carlos, Brazil); (ii) a buffer exchanged by dialysis and a concomitant TEV protease cleavage from 6His-TRX-tag; (iii) an inverse HisTrap HP 5.0 mL to separate protein from 6His-TRX-tag and (iv) a size-exclusion chromatography at a XK 16/60 Superdex 75 column (GE Healthcare, Sao Carlos, Brazil).

NS5 RdRp Activity Assays
ZIKV NS5-RdRp activity assays were performed as described by Fernandes and coworkers [93]. The endpoint assays were performed at 20 µM, and the compounds that inhibited more than 80% of activity in this assay were submitted to a concentration-response test. The concentration-responses assays were performed as described in [94]. In all cases, the percentage inhibition values were calculated based on a control reaction, containing only DMSO in the same concentrations used for the tested compounds. The results were analyzed and plotted using the GraphPad Prism v. 8.0 program [95].

Virus Rescue and Titration
A wild-type ZIKV isolate from a clinical patient in Brazil (ZIKV BR , PA, Brazil) was provided by the Evandro Chagas Institute in Belém, Pará [96]. The virus was amplified employing Vero cells in a 175 cm 2 flask. To determine viral titers, 1 × 10 4 Vero cells were seeded in each of 24 wells plate 24 h prior to the infection. Cells were infected with 10fold serially dilutions of ZIKV BR for 1 h at 37 • C and then supplemented with medium containing 1% penicillin, 1% streptomycin, 2% FBS and 1% carboxymethyl cellulose (CMC). Infected cells were incubated for seven days in a humidified 5% CO 2 incubator at 37 • C, followed by fixation with 4% formaldehyde and staining with 0.5% violet crystal. The viral foci were counted to determine viral titers which were expressed in plaque formation unit per milliliters (PFU/mL).

Cell Viabillity
Cell viability was measured by the MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide] (Sigma-Aldrich) method. Vero cells were seeded in a 96-well plate at a density of 5 × 10 3 cells per well and incubated overnight at 37 • C in a humidified 5% CO 2 incubator. A drug-containing medium at concentrations ranging from 200 to 0.005 µM was added to the cell culture. After 72 h at 37 • C, the media was removed and a solution containing MTT at the final concentration of 1 mg/mL was added to each well and incubated for 30 min at 37 • C in a humidified 5% CO 2 incubator, after which media was replaced with 100 µL of DMSO to solubilize the formazan crystals. Absorbance was measured by the optical density (OD) of each well at 490 nm, using a spectrophotometer. Cell viability was calculated according to the equation (T/C) × 100%, where T and C represent the mean optical density of the treated group and vehicle control group, respectively. The cytotoxic concentration of 50% (CC 50 ) was calculated using Graph Pad Prism v. 8 [95].

Antiviral Assays
Vero cells were seeded at density of 5 × 10 3 cells per well into 96-well plates 24 h prior to the infection. ZIKV-WT BR at a multiplicity of infection (MOI) of 0.1 and compound at concentrations ranging from 200 to 0.005 µM were simultaneously added to cells. 72 h post-infection (h.p.i.), cells were fixed with paraformaldehyde 4%, washed with PBS and blocking buffer (BB) containing: 0.1% Triton X-100 (Vetec Labs, PR, BR), 0.2% bovine albumin (BSA) and PBS for 30 min. Then, cells were incubated with primary rabbit polyclonal anti-NS3 antibody diluted in BB for 1 h. Alexa Fluor 488 conjugated anti-rabbit IgG was used as secondary antibody (Abcam, Cambridge, UK). Images were analyzed by EVO cell imaging systems fluorescence microscopy (Thermo Fisher Scientific, OH, USA) and foci of infection were counted. The antiviral activity was calculated according to the equation (T/C) × 100%, where T and C represent the mean of the treated group and mean of the last concentration, respectively. The effective concentration of 50% inhibition (EC 50 ) was calculated using Graph Pad Prism v. 8. The values of CC 50 and EC 50 were used to calculate the selectivity index (SI = CC 50 /EC 50 ).

Conclusions
Despite the severe neurological consequences caused by ZIKV infection, there is still no antiviral for the treatment of ZIKV, and only a few ZIKV NS5 RdRp inhibitors have been described in the literature. In our study, guided by known DENV NS5 RdRp inhibitors, through binding site conservation and chemical space analysis as well as docking calculations we prioritized and identified the flavonoids pedalitin and quercetin as new inhibitors of ZIKV NS5 RdRp. Enzymatic assays reinforced the computational results, and both compounds presented antiviral activity against ZIKV in infected cell cultures. Therefore, quercetin and pedalitin may be promising candidates for hit-to-lead optimization, boosting the discovery of new anti-ZIKV drug candidates.