Development of a fluorescence-based method for the rapid determination of Zika virus polymerase activity and the screening of antiviral drugs

Zika virus (ZIKV) is an emerging pathogen that has been associated with large numbers of cases of severe neurologic disease, including Guillain-Barré syndrome and microcephaly. Despite its recent establishment as a serious global public health concern there are no licensed therapeutics to control this virus. Accordingly, there is an urgent need to develop methods for the high-throughput screening of antiviral agents. We describe here a fluorescence-based method to monitor the real-time polymerization activity of Zika virus RNA-dependent RNA polymerase (RdRp). By using homopolymeric RNA template molecules, de novo RNA synthesis can be detected with a fluorescent dye, which permits the specific quantification and kinetics of double-strand RNA formation. ZIKV RdRp activity detected using this fluorescence-based assay positively correlated with traditional assays measuring the incorporation of radiolabeled nucleotides. We also validated this method as a suitable assay for the identification of ZIKV inhibitors targeting the viral polymerase using known broad-spectrum inhibitors. The assay was also successfully adapted to detect RNA polymerization activity by different RdRps, illustrated here using purified RdRps from hepatitis C virus and foot-and-mouth disease virus. The potential of fluorescence-based approaches for the enzymatic characterization of viral polymerases, as well as for high-throughput screening of antiviral drugs, are discussed.


pair (Supplementary
). The amplification conditions used were as follows: 98 °C (3 min), 30 cycles of 98 °C (15 s), 51 °C (20 s) and 72 °C (4 min) each, and 10 min of elongation at 72 °C. The RdRp domain was amplified from a pcDNA vector containing the full-length NS5 gene. PCR amplification was similar to that described above but using the specific primers NS5 short_pET16_Fw and NS5_pET16_rv (Supplementary Table S1). After purification, the vector and insert were mixed in the presence of 2 × Gibson Assembly Master Mix and the assembly reaction was carried out following the recommendations of the manufacturer. The assembled product was transformed into E. coli BL21(DE3)-pRIL cells. After plasmid extraction from three independent bacterial colonies, nucleotide sequencing determined that two DNA samples contained the correct construct. The resulting plasmid pET16a-ZIKV-NS5RdRp encodes for ZIKV RdRp fused to an HHHHHHHHHHSSGHIEG amino acid tract in its N-terminus that is used for affinity purification using HisPur TM Ni-NTA resin The predicted molecular weight of this protein is 75 kDa.
A catalytically inactive enzyme was prepared by site-directed mutagenesis of the pET16a-ZIKV-NS5RdRp plasmid, encoding for substitutions D665N and D666N in the active site, which affects two catalytic Asp residues. The amplification reagents were the same as above with primers NS5_GNN_Fw and NS5_GNN_rv (Supplementary Table S1) and pET16a-ZIKV-NS5RdRp as template. PCR reaction conditions used were 98 °C (3 min), 30 cycles of 98 °C (15 s), 54 °C (20 s) and 72 °C (4 min), followed by an elongation step of 10 min at 72 °C. The resulting expression plasmid was termed pET16a-ZIKV-NS5RdRp-GNN.
The plasmid for the expression of HCV NS5B polymerase was prepared using the Gibson assembly method as described above. Briefly, the pET28a vector backbone was amplified by PCR using primers pET28a_Fw and pET28a_rv (Supplementary Table S1). An insert containing the HCV polymerase sequence was obtained by PCR amplification of plasmid Jc1FLAG2(p7-nsGluc2A) 37 , using primers NS5B_HCV_Fw and NS5B_Δ21_HCV_rv (Supplementary Table S1). Both vector and insert were assembled as described above. The resulting plasmid, termed pET28-HCV NS5bΔ21, encodes for HCV NS5b polymerase lacking the most C-terminal 21 amino acids and containing a C-terminal His-tag (LEHHHHHH). The predicted molecular weight of this recombinant protein is 65 kDa. For the construction of pET28-HCV NS5bΔ21-GNN (expressing a catalytically-inactive RdRp) we used Gibson assembly and plasmid Jc1FLAG2(p7-nsGluc2a)/GNN 37 to generate the insert. All the constructs were analyzed by sequencing to confirm the presence of the expected insert and the absence of undesired mutations.
Expression and purification of viral polymerases. For the expression of ZIKV NS5 RdRp (hereafter referred to as ZIKV RdRp), E. coli cells were transformed by electroporation with pET16a-ZIKV-NS5RdRp. Single kanamycin and chloramphenicol resistant colonies were cultured overnight in 10 mL of LB in the presence of antibiotics at 37 °C. Each culture was then inoculated into 200 mL of LB with antibiotics and incubated at 37 °C. When an optical density at 600 nm of 0.7 was reached, 500 μM IPTG, 50 μM MgCl 2 and 50 μM ZnCl 2 were added to the culture, which was incubated at 30 °C for 4 additional hours. Cells were then pelleted by centrifugation at 5,000 rpm for 15 min at 4 °C and stored at −80 °C until further use.
The bacterial pellets recovered from 200 mL cultures were resuspended in 20 mL of lysis buffer [50 mM Tris HCl, pH 8.0, 300 mM NaCl, 400 mM ammonium acetate, 4 mM MgCl 2 , 10%, glycerol, 10 mM imidazole, and 0.1% (v/v) Tween 20] and sonicated on ice for 6 cycles of 20 s alternating with 5 cycles of 10 s. Cell debris was pelleted at 11,000 rpm for 30 min at 4 °C and the supernatant mixed with 800 μL of Ni-NTA resin previously equilibrated with 20 volumes of lysis buffer without ammonium acetate (BWE buffer). The lysate was incubated with the resin in batch method with gentle mixing during 1 h at 4 °C. The unbound fraction was then removed by decantation and the resin was then loaded onto a column and extensively washed with 20 × column volumes of BWE buffer and 20 × column volumes of BWE buffer containing 25 mM imidazole. The resin was further washed with increasing concentrations of imidazole (successive one-column volumes of BWE buffer containing 50, 60, 70, 80, 90, 100 and 125 mM imidazole). Finally, the His-tagged protein was eluted in 400 μL of BWE buffer containing 400 mM imidazole. The sample was dialyzed for 3 hours at 4 °C against 200 volumes of dialysis buffer [50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , 10%, glycerol, 1 mM DTT and 0.05% (v/v) Tween 20]. Samples obtained from different purification batches were pooled, quantified, aliquoted and stored at −80 °C until further use. Expression and purification of the recombinant ZIKV NS5RdRp-GNN (with D665N and D666N substitutions) was carried out following the same protocol. Likewise, the expression and purification of HCV NS5bΔ21 and NS5bΔ21-GNN polymerases was carried out following the same protocol described for ZIKV NS5. The protocol for the expression and purification of FMDV 3D polymerases has been described previously 35,36 . Fluorescence-based activity assay for ZIKV RdRp. For the detection of RNA synthesis by ZIKV RdRp, we established a real-time assay based on the fluorescent dye SYTO 9, which binds dsRNA but not ssRNA template molecules. The fluorescence emitted was recorded in real-time using a Fluostar Optima fluorimeter (BMG Labtech) using excitation and emission filters at 485 and 520 nm, respectively. The assay records the synthesis of dsRNA in a reaction using a poly-U molecule as a template and ATP as the nucleotide substrate. This technique has been adapted from methods previously documented for the detection of DNA synthesis 38 .
Reactions were performed in individual wells of black 96-well flat-bottom plates. The standard reaction contained 50 mM Tris-HCl, pH 7.5, 2.5 mM MnCl 2 , 500 μM ATP, 20 μg/mL poly-U, 0.1 mg/mL BSA and 0.25 μM SYTO 9 (50 μM stock solution in TE buffer pH 7.5). The assay was initiated by the addition of 250 nM ZIKV RdRp and the fluorescence was recorded over 30 min at 30 °C.
Variations on this assay, for example, different concentrations of reagents and/or the presence of additional compounds, are specifically indicated in each corresponding section. For graphical representation, background fluorescence obtained at time point 0 was subtracted from each value.
To determine K m and V max constants for ZIKV RdRp binding to poly-U ssRNA, standard reactions were carried out in increasing concentrations of the template (0.5-50 μg/mL) in the presence of ATP at 500 μM. The kinetic parameters for ATP were obtained from assays in the presence of increasing concentrations of this nucleotide (200-2250 μM) and using 3 μg/mL of poly-U.
IC 50 values were obtained from standard reactions carried out in the presence of 3 μg/mL poly-U and 1500 μM ATP, and increasing concentrations of each inhibitor.
End-point fluorometric reactions were performed in black 96-well black-flat bottom plates at 30 °C in the presence of the same reagents as described above, but in the absence of dye. The reactions were quenched at 60 min by adding in 25 mM EDTA to the samples. Either SYTO9 or SYBR Green II dye was then added to the sample (0.25 μM or 1x, respectively) and the mix reaction was incubated at room temperature for 5 min to allow the stabilization of RNA-dye complexes and fluorescence emission. To determine background fluorescence levels, a negative control was assayed in parallel, where the reaction was quenched before adding ZIKV RdRp. The quenched control reaction was incubated for 1 h at 30 °C, and then 0.25 μM SYTO9 or 1 × SYBR Green II, respectively, was added to the sample and fluorescence recorded as described above.

Data analysis.
Fluorometric results were expressed as mean ± SD. Statistical significance was analyzed by two-way ANOVA using GraphPad Prism, version 7, as specified in the figure legends. K m determinations were obtained by plotting the velocity of the reaction as a function of nucleotide or ssRNA template concentrations using nonlinear regression. IC 50 values were obtained by fitting the velocity data to a four-parameter logistic equation. Kinetic parameters and IC 50 values were calculated using Sigmaplot, version 11. Z' factor was calculated according to Zhang et al. 39 where "c+" is the activity obtained in a standard assay and "c−" is the nonspecific activity obtained in a control performed in the absence of MnCl 2 .

Results
Purification and biochemical characterization of recombinant ZIKV RdRp. The RdRp domain of ZIKV NS5 and a catalytic inactive mutant (GNN) were purified as described in Methods. Recombinant proteins were ≥95% pure as judged by PAGE analysis and Coomassie brillant blue R-250 staining (Fig. 1A).
The overexpression and biochemical characterization of ZIKV RdRp under different experimental conditions has been previously published [40][41][42] . For the preliminary evaluation of ZIKV NS5 RdRp domain activity in vitro, we adapted a polymerization assay based on the detection of radioactive nucleotides incorporated by the polymerase. This method makes use of a homopolymeric ssRNA as a template in the absence of any primer, since it has been previously demonstrated that ZIKV NS5 can initiate RNA synthesis de novo 42 . The reactions were performed in the presence of radioactive-labeled nucleotides, and polymerization products were resolved by PAGE. RNA synthesis in the absence of primer was observed both in the presence of poly-U and [α-32 P]ATP as template and nucleotide substrates (Fig. 1B), and in the presence of poly-C and [α-32 P]GTP (Fig. S1). We observed polymerization activity de novo in the presence of Mn 2+ but not Mg 2+ , in agreement with a previous observation 41 (Fig. 1B). The same reaction in the presence of the catalytically-inactive mutant GNN showed no detectable signal (Fig. 1B,  lanes 7 to 9).
Previous studies suggested that, under certain circumstances, flaviviral polymerases can catalyze the terminal transference of nucleotides to RNA. However, this transferase activity has never been reported for ZIKV RdRp 42 . To rule out the possibility that the incorporation of nucleotides detected in our assay was due to the terminal transference of nucleotides and not to de novo RNA synthesis (as we expect), we performed the same assay but in the presence of radioactive nucleotides that were less competent for viral RNA synthesis: [α-32 P]GTP to poly-U and [α-32 P]ATP to poly-C. As shown in Fig. S2B, no elongation was detected under these conditions, supporting the notion that the activity detected was due to de novo RNA synthesis. Thus, these results show that both homopolymeric templates, poly-C and poly-U, can be used by ZIKV to initiate RNA replication, as has been previously documented 42 .
Detection of ZIKV RdRp polymerization activity by fluorometric assays in real time. Based on the above results, we next sought to detect polymerization activity using a fluorescence-detection method. For this aim, we attemped to establish an assay to quantify RNA synthesis activity as the relative increase in fluorescence emitted by SYBR Green II dye after binding to dsRNA. This procedure was adapted from methods previously described to detect dsDNA synthesis by the human primase-polymerase PrimPol 38 . We anticipated that binding of this intercalating agent to dsRNA generated by ZIKV RdRp polymerization activity would lead to an increase in the emitted fluorescence.
Preliminary real-time assays, involving the addition of SYBR Green II to the sample before initiating the reaction, showed an undetectable (using poly-C) or barely detectable (using poly-U) increase in fluorescence. In contrast to real-time experiments, we found significant increases in polymerase activity in an end-point experiment where the dye was added after the reaction was completed (Fig. S1). Previous studies have documented that an excess of SYBR Green I, chemically related to SYBR Green II, can inhibit other polymerase activities, such as those of Taq polymerase 43 or human PrimPol 38 . Our resulted suggested that SYBR Green II acts as an inhibitor of ZIKV RdRp activity. Thus, we decided to test other fluorescent dyes for the real-time detection of newly synthesized dsRNA. It has been reported that SYTO 9 dye shows lower interference on polymerization assays when binding to dsDNA 44,45 . In contrast to the assays with SYBR Green II, we found that both end-point and real-time polymerization assays resulted in similar increases in fluorescence when using poly-U as template (Fig. S1, compare A with D). The relative increase in emitted fluorescence (ratio between the values obtained after a 60 min reaction and the background value observed at time 0) was similar using both approaches. This result suggested that SYTO 9 does not inhibit ZIKV RdRp, and thus can be used for real-time detection of activity. We also (2019) 9:5397 | https://doi.org/10.1038/s41598-019-41998-1 www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ observed high reproducibility among different experimental samples, as reflected in the modest standard error values in different experiments (Fig. S1D). According to our assay, dsRNA synthesis was linear up to 60 min, and then reached the maximum accumulation of product at 150-180 min ( Fig. 2A). ZIKV RdRp activity is dependent on the presence of Mn 2+ in agreement with our data in radioactivity-based assays. The disruption of the catalytic site (RdRp GNN) also led to complete loss of polymerase activity (Fig. 2B). Likewise, terminal transferase activity was not detected in assays using poly-U as template and either GTP or UTP as substrate.
To further investigate the possible use of SYTO 9 dye to detect activity in the presence of other templates, we used poly-C. However, we detected increases in fluorescence only in end-point reactions, when SYTO 9 was added after the recation was complete, and not in a continuous reading assay when it was added before initiating the reaction (Fig. S1B). These results suggest that poly-C is not a suitable substrate for real-time assays.
Optimization of the fluorescence-based assay. To improve the detection of RNA synthesis, we examined how changes in the concentration of reagents (i.e., NaCl, DTT, MnCl 2 and enzyme) affected RdRp activity. The presence or absence of DTT and NaCl in the assay had little effect on RNA synthesis, which was only slightly impeded at high concentrations (Fig. S3A,B). As expected, no increase in fluorescence was detected when using MgCl 2 (0 to 20 mM), whereas maximum activity was recorded with 2.5 mM MnCl 2 (Fig. S3C). The increase in the velocity of reaction correlated linearly with increases in RdRp concentration along the 10-250 nM range. The maximum velocity was reached with 750 nM RdRp in the assay (Fig. S3D). From these assays we obtained a K m for poly-U of 3.3 ± 0.5 μg/mL (~31 nM) and a K m for ATP of 561 ± 38 μM (Fig. S4).

Fluorescence-based activity can be inhibited by broad-range antiviral compounds.
We hypothesized that this fluorescence-based method could be exploited for the development of high-throughput screening methods to identify polymerase inhibitors. To test this, we used several broad-spectrum nucleoside and non-nucleoside polymerase inhibitors. Addition of polymerase NNI heparin 46,47 to the reaction completely abrogated fluorescence-associated activity (Fig. 3A). To further confirm the sensitivity of our assay to inhibitors, we tested two nucleoside analogs: cordycepin 5′-triphosphate (3′dATP; a chain terminator analog of ATP [48][49][50] and ribavirin 5′-triphosphate (RTP; a purine analog that inhibits but does not terminate RNA elongation during viral replication [51][52][53][54]. Both compounds reduced the polymerase activities (Fig. 3A). We calculated the IC 50 values of these compounds (Fig. S3): as expected, the most potent inhibitor was the NNI heparin (IC 50 = 81 ± 21 nM), followed by the NAIs, 3′dATP (54 ± 7 μM) and RTP (946 ± 46 μM). To confirm that the decrease in fluorescence was linked to the inhibition of RNA synthesis, we repeated these experiments using radioactive-labeled nucleotides. We found a reduced polymerization activity in the presence of inhibitors (3′dATP, RTP and heparin) that correlated with the aforementioned IC 50 values (Fig. 3B). Similar inhibitory activities were observed when poly-C or poly-U were used as template molecules in radioactive-based activity assays (Fig. S2).
The robustness and suitability of this assay as a prospective, high-throughput method to screen polymerase antiviral compounds was examined by calculating the Z′ value, a standard statistical measure to evaluate the quality for high-throughput platforms 39 . The relative activity of both positive and negative controls was calculated as the average value obtained from 8 independent experiments (see Methods). Each experiment was carried out in triplicate and on independent days. Relative activity values were determined as the velocity of polymerization www.nature.com/scientificreports www.nature.com/scientificreports/ recorded during the first 10 min of the reaction. The mean Z' value obtained was 0.62, which according to published standards, qualifies our method as an excellent assay for high-throughput screening application 39 .
Fluorescence-based activity assay can be adapted to monitor different viral RdRps. To investigate whether the assay can be adapted to other viral polymerases, we used two unrelated RdRps from FMDV (3Dpol) and HCV (recombinant NS5B) (Fig. 4A). We found that recombinant HCV polymerase can synthesize RNA de novo using radiolabeling (Fig. 4B) and the fluorescence-based approach (Fig. 4C), which is in agreement with the mechanism of genome replication for this virus 52,53 . Again, an increase in fluorescence was only observed with a catalytically-active HCV RdRp (NS5bΔ21) but not an inactive mutant, and the activity was dependent on the presence of Mn 2+ (Fig. 4C). We also found that FMDV 3D polymerase can catalyze RNA synthesis in vitro, in an assay primed by the viral protein-primer VPg. It has been previously demonstrated that FMDV 3D catalyzes the addition of a uridine-monophosphate residue to Tyr3 in VPg. Once uridylylated, VPg can act as a competent primer to initiate viral genome replication 55,56 . Vpg protein-primed polymerization in vitro can be achieved using a polyadenylic acid as template, UTP as substrate, Mn 2+ as catalytic metal and the VPg1 peptide synthetically produced 35 . Real-time experiments showed an increase of fluorescence as a function of time when using active 3D, whereas no activity was monitored in the presence of a catalytically-inactive protein, or in the absence of Mn 2+ or VPg1 (Fig. 4D). Overall, our data show that this 96-well format assay can be exploited to characterize different viral RdRps in vitro, as it permits real-time monitoring of replication, allowing the characterization of small-molecule libraries in a cost-effective and rapid manner.

Discussion
There is an urgent need to develop new treatments for ZIKV infection and to control its rapid geographical spread. Different approaches for the discovery of potential small-molecule inhibitors include the screening of chemical libraries, molecular modeling and virtual screening 29,57-59 . Although promising developments in this direction have been achieved (reviewed in 32 ), there are as yet no antiviral agents licensed against ZIKV at the www.nature.com/scientificreports www.nature.com/scientificreports/ clinical level. Owing to significant differences in the mechanisms of replication between cellular DNA and viral RNA genomes, the latter involving the synthesis of RNA molecules templated by RNA, RdRps are attractive targets for the development of specific antiviral treatments. The use of antivirals against non-RdRp viral polymerases, such as human immunodeficiency virus and hepatitis B virus reverse transcriptases, and herpes virus DNA polymerase, supports the suitability of this group of enzymes as therapeutic targets 60 . Accordingly, the development of a fast and reproducible method for the screening of compounds with anti-ZIKV properties is a promising advance.
Several methods for high-throughput drug screening against virus RdPs have been described and validated [61][62][63][64][65][66] . However, there are several practical limitations to these approaches, such as the requirement for radioactive substances, entailing additional biosafety measures 63 , or an arduous experimental setup 65,66 when compared with fluorescence-based methods 61,62,64 . Indeed, an advantage of fluorescence-based methods over traditional approaches is the absence of radioactive compounds, which facilitates their broad use in different laboratory settings without specific facilities or training requirements. An additional advantage of our strategy is that it allows the time-resolving determination of the polymerase activity, which we believe gives further insight into the mechanisms of replication and inhibition.
There have been previous attempts to develop high-throughput activity assays to identify drug inhibitors specifically against ZIKV RdRp. These methods include the use of either radioactive nucleotides 41 , or costly fluorescent-labeled RNA substrates 40 . Our method has the advantages of being a more economically affordable alternative, as it uses inexpensive homopolymeric RNA as a template substrate instead of labeled synthetic heteropolymeric RNA. Our approach would also allow for assay scale-up to high-throughput formats (e.g., 96 or 384-well formats, automation, etc.), for the rapid testing of small-molecule compound libraries 61,62,64 . www.nature.com/scientificreports www.nature.com/scientificreports/ Fluorometric measurements of polymerase activity in the absence/presence of three representative inhibitors (heparin, 3′ATP and RTP) revealed a positive correlation with the results using a traditional assay based on radiolabeled nucleotides, further validating the fluorescence-based approach for the screening of antiviral compounds. The possibility of visualizing dsRNA synthesis in real-time increases the sensitivity of the assay, allowing an accurate determination of replication kinetics in the presence or absence of the drug and the predicted affinity constants of the compound tested. This method also permits the characterization of inhibitory molecules in vitro, which is of use in the identification of prospective antivirals. As part of our studies we have shown that RTP elicits a mild inhibition on ZIKV RdRp polymerization in vitro. We posit that the observed inhibition might be a consequence of reduced polymerase efficacy to elongate molecules where an RTP residue is incorporated [52][53][54]67 . RTP is not a chain terminator and its incorporation into the viral RNA has been linked to an increase in transition mutation rates in vitro and in cell culture for different RNA viruses, including ZIKV [68][69][70][71] .
Under different experimental conditions, including the use of different concentrations of SYBR Green II and SYTO 9, we found that poly-C RNA was not a suitable substrate for the detection of polymerase activity in real-time. Conversely, both poly-A and poly-U homopolymers were effective as template molecules in real-time assays. We hypothesize that the inhibition of polymerase activity is produced by the interaction of RdRp with the dye during dsRNA synthesis. It is possible that poly-G synthesis as a result of replicating poly-C can lead to non-canonical G-quadruplex structures 72 , which in the presence of a fluorophore can further increase their inherently elevated stability 44 . Thus, these RNA-dye complexes might impede the effective elongation by the viral RdRp.
In conclusion, we have demonstrated that the procedures developed here can be easily adapted to measure polymerization activity of several viral RdRps, strengthening our method as a universal procedure for the development of high-throughput tools to characterize viral polymerases (e.g., enzymology, polymerase variants of interest) and to screen small-molecule libraries to identify antiviral drugs. In particular, we believe that this platform can be a useful tool for the development of therapeutics against ZIKV and other flaviviruses, which are currently unavailable.